Éditeurs scientifiques Marie-Hélène Durand Philippe Cury Roy Mendelssohn Claude Roy Andrew Bakun Danlel Pauly Global versus Local Cha.nges in UpweIIing Systems Global versus Local Changes in Upwelling Systems Edited by Marie-Hel&ne DURAND, Philippe CURY, Roy MENDELSSOHN, Claude ROY, Andrew BAKUN and Daniel PAULY Editions de IfOrstom INSTITUT FRAN~AIS DE RECHERCHE SCIENTIFIQUE POUR LE DEVELOPPEMENT EN COOPERATION Collection Colloques et seminaires Paris, 1998 Mise en page Marie-Christine Pascal Index Francisco Torres jr. Fabrication Catherine Plasse Maquette de couverture Michelle Saint-Léger Maquette intérieure Catherine Plasse Photo de couverture J. N. Vinter : Bois gravé de René Quivillic, a en pleine mer >,, 1921. Musée de la Faïence, Quimper. La loi du ler juillet 1992 (code le la propriété intellectuelle, premibre partie) n'autorisant, aux termes des alinéas 2 et 3 de l'article L. 122-5, d'une part, que les copies ou reproductions strictement réservées à l'usage du copiste et non des- tinées à une utilisation collective * et, d'autre part, que les analyses et les courtes citations dans le but d'exemple ou d'illustration. G< toute représentation ou reproduction intégrale ou partielle faite sans le consentement de l'auteur ou de ses ayants droit ou ayants cause, est illicite 13 (alinéa ler de l'article L. 1224). Cette représentation ou reproduction, par quelque procédé que ce soit, constituerait donc une contrefaçon passible des peines prévues au titre III de la loi précitée. O Orstom éditions, 1998 To the memory of Itaf Deme Gningue Itaf est morte accidentellement le 26 octobre 1997 sur la route entre Rosso et Saint-Louis en revenant d'une mission de travail. Elle avait 41 ans. Itat' avait démarré sa carrière au CRODT en tant qu'ingénieur chimiste. Elle avait su surmonter de nombreux obstacles pour ré~iiser une thèse, devenir chercheur, puis diriger la section Environnement du CRODT. Tous ceux qui l'ont connue et côtoyée dans son travail saluent avant tout sa volonté, son courage et sa gentillesse. Les auteurs de ce livre n'oublieront pas non plus son rire et sa bonne humeur. En 1994, lorsque Itaf étaient venue à Monterey participer au colloque CEOS, eue était enceinte de sa petite fille Khadidja. Notre pensée va à sa famille et à ses deux petites filles Ndeye Anta et Khadidja. Ku ko guissone sope na ko, ku ko xamonegnak na ko, yala na ko Allah yeureum te mre ko adiana Anzine TABLE OF CONTENTS .................................................. Preface and Acknowledgments .xi The Editors The Climate and Eastern Ocean Systems Project (CEOS) ............................. .1 Andrew Bakun, Philippe Cury, Marie-Hélène Durand, Roy Mendelssohn, Daniel Pauly and Claude Roy ... The CEOS Comparative Analysis Framework: Motivations and Perceived Opportunities .7 Andrew Bakun PART 1 - GLOBAL AND LOCAL ENVIRONMENTAL CHANCES The Development and the Use of a Climatic Database for CEOS using the COADS Dataset ................................................ .27 Claude Roy and Roy Mendelssohn How to Detect a Change both on Global and Local Scale in Oceanographic Time Series . .45 Marie-Hélène Durand and Roy Mendelssohn Long-term Variability in the Seasonality of Eastern Boundary Current (EBC) Systems: an Example of lncreased Upwelling .from the California Current .................... .79 Franklin B. Schwing and Roy Mendelssohn Recent Trends in the Spatial Structure of Wind Forcing and SST in the California Current System ........................................... .IO1 Franklin B. Schwing, Richard H. Parrish, and Roy Mendelssohn Freshwater Yields to the Atlantic Ocean: Local and Regional Variations from Senegal to Angola ................................................. .127 Cil Mahé Spatial Dynamics of the Coastal Upwelling off Côte-d'Ivoire ...................... .139 Angora 4man and Siaka Fofana Spatial and Temporal Dynamics of the Upwelling off Senegal and Mauritania: LocalChangeandTrend ................................................. 149 Hervé Demarcq Variability and Trends in Some Environmental Time Series along the lvoirian andtheGhanaianCoasts ................................................. 167 Kwame A. Koranteng and Olivier Pezennec Trends and Variability of Environmental Time Series along the Senegalese Coast ....... .179 ltaf Deme-Cningue Food, Transport and Anchovy Recruitment in the Southern Benguela Upwelling System of South Africa ................................................. .195 Alan.). Boyd., 1.). Shanno-n, Fritz.H. Schülein, andlohn Taunton-Clark Sardine and Other Pelagic Fisheries Changes Associated with Multi-Year Trade Wind Increases in the Southern Canary Current ............................... .211 Denis Binet, Birane Samb, Mahfoud Taleb Sidi, lean-)acques Levenez andlacques Servain Climate Dependent Fluctuations of the Moroccan Sardine and their Impact on Fisheries .. .235 Souad Kifani Marine Environmental Conditions and Fishery Productivity in the Black Sea .......... .249 Ceorgi Daskalov and Kamen Prodanov The Recruitment of the Chilean Sardine (Sardinops sagax) and the "Optimal Environmental Window" ......................................... .267 Rodolfo Serra, Philippe Cury and Claude Roy ............ Pelagic Fish Stocks and Environmental Changes in the South-East Pacific .275 Eleuterio Yaiiez, Miguel Garcia and Maria-Angela Barbier; Sardinella aurita Population Dynamics Related to Environmental Parameters in the Southern Caribbean (Venezuela) ..................................... .293 )eremy Mendoza, Pierre Fréon, Ramon Cuzman and Rubén Aparicio Global and Local Change: Penaeid Stocks in French Guyana ..................... .311 Christophe Béné and Philippe Moguedet ... VIII Table of Contents Changes in the Dynamics and Biology of Small Pelagic Fisheries off Côte d'ivoire and Ghana: .................................................... an Ecological Puzzle .329 Olivier Pezennec and Kwame A. Koranteng Stock Assessment of Sprat and Whiting in the Western Black Sea in Relation to Global and ............................................. Local Anthropogenic Factors .345 Kamen B. Prodanov, Ceorgi M. Daskalov, Konstantin Mikhailov, Konstantin Maxim, Emin Ozdamar, Vladislav Shljakhov, Alexandr Chashchin and Alexandr Arkhipov ............................. Variability of Fish Catches in Different Ecosystems .359 Konstantinos 1. Stergiou .. Desperately Searching for Natural Eutrophication: the Case of the NE Mediterranean .371 Konstantinos 1. Stergiou and E. D. Christou Pelagic Fisheries and Environmental Constraints in Upwelling Areas : How Much Is Possible ? ................................................. .391 Valérie Faure and Philippe Cury ..... Clupeoids Reproductive Strategies in Upwelling Areas: a Tentative Generalization .409 Yunne-)ai Shin, Claude Roy and Philippe Cury Comparative Modelling of Trophic Flows in Four Large Upwelling Ecosystems: Global versus Local Effects ............................................... .423 ktridlarre-Teichmann and Villy Christensen PART 3- HUMAN ACTlVlTlES FACINC SHORT AND LONG TERM CHANCES Comparative Study of the Dynamics of Small-Scale Marine Fisheries in Senegal and Ghana . .447 locelyne Ferraris, Kwame A. Koranteng and Alassane Samba Fishmeal Price Behaviour: Global Dynamics and Short-Term Changes .............. .465 Marie-Hélène Durand Modeling Fishery Activity Facing Change: Application to the Senegalese Artisanal Exploitation System ............................................. .481 lean Le Fur Dome or U-Shaped Physiological Responses of Populations, and Ecosystems ......... .503 Cary D. Sharp Table of Contents ix Life History Strategies for Marine Fishes in the Late Holocene . . . . . . . . . . . . . . . . . . . . .525 Richard H. Parrish Fisheries Resources as Diminishing Assets: Marine Diversity Threatened by Anecdotes . . .537 Philippe Cury and Orlane Anneville The CEOS Network: a Brief Summary of Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . .549 Philippe Cury, Marie-Hélène Durand and Daniel Pauly Authorslndex* ........................................................ 559 Specieslndex* ......................................................... 585 * prepared by F. Torres Ir. (ICLARM) x Table of Contents Preface and Acknowledgments This volume is comprised of 33 contributions, most onginally presented at an international conference held on 6-8 September 1994 in Monterey, California, USA under the auspices of the Climate and Eastern Ocean Systems Project (CEOS). The CEOS was a collaborative project linking a variety of research institutions, notably ORSTOM, NMFS and ICLARM and devoted to a study of the potential effects of global change on the resources of upwelling systems through identification of global and local effects impacting on these systems. This volume is one of the two major products of CEOS, the other being a set of CD-Roms that will empower more researchers, especially in developing countries, to study oceanographic processes (see Roy and Mendelssohn, this vol.). We apologize for the delay in publishing this volume, due mainly to the inherent difficulty of coordinating a vast cast of authors in different parts of the world, and most with first languages other than the English we have chosen to present the result of CEOS. We thank our respective institutions, ORSTOM, NMFS and ICLARM for their support of the effort that led to this volume. Also we take this opportunity to thank Ms Marie-Cliristine Pascal for the typing and layouting of this book and also Ms Nathalie Richard, Ms Barbara Chollet, Ms Sandra Gayosa, Mr Emmanuel Suisse de Sainte-Claire and Mr Francisco Torres Jr. for their assistance in the production of this volume. Also we thank the PNDR (Programme National sur le Déterminisme du Recrutement, France) and SEAH (Systèmes Ecologiques et Actions de l'Homme, CNRS-France) for their support of our effort. The Climate and Eastern Ocean Systems Project (CEOS) 1. CEOS: AN INTERNATIONAL NETWORK WORKING ON CLIMA-TE AND FISHERIES The injection of millions of tonnes of greenhouse gases into the earth's atmosphere may be viewed as a gigantic experiment aimed at exploring the earth's reaction to such challenge. Unfortunately, this experiment is run without proper 'controls', and hence the heated debates about the actual impact of those gases may last too long, beyond the time where the 'experiment' should be called off. The international scientific community is forced, however, to address this problem in spite of the lack of scientific controls. One way to address this is through the comparative method, a major tool in those disciplines in which experiments are hard to perform, e.g. evolutionary biology (Mayr, 1982), fisheries science (Bakun, 1985, 1996). Given the importance of the four major upwelling systems off Pem, Chile, California, Northwest and Southwest Africa both as sources of fish and as CO2 'pumps', scientists from the Pacik Fisheries Environmental Group (PFEG), of the National Marine Fisheries Service (NMFS), the Institut Français de Recherche Scientifique pour le Développement en Coopération (ORSTOM), and the International Center for Living Aquatic Resources Management (ICLARM) and partners from other institutions, teamed up to investigate these systems in the context of global changes, through a project called CEOS (Climate and Eastern Ocean Systems), funded by NOAA and ORSTOM. Several national research laboratories working on similar systems were associated with this project through a cooperative agreement with ORSTOM. The African research institutes associated to the CEOS project were the Institut Scientifique des Pêches Maritimes (ISPM) in Morocco; the Centre de Recherches Océanographiques d'Abidjan (CROA) in Côte-d'Ivoire; the Fisheries Research Utilization Branch (FRUB) in Ghana; the Centre de Recherches Océanographiques cle Dakar-Thiaroye (CRODT) in Senegal, and the Sea Fisheries Research Institute (SFRI) in South Africa which focus on regional case studies of climatic variability, coastal ecosystem dynamic and associated human responses. In Latin America, the Instituto del Mar del Peru (IMARPE) in Peru; the Instituto de Fomento Pesquero (IFOP) and the Universidad Catolica de Valparaiso in Chile are also involved in this project. The collaboration of scientists from these institutes with the CEOS project was also partly funded by the Scientific Committee on Dynamics and Use of Renewable Resources (DURR) of ORSTOM and PNDR (Programme National sur le Déterminisme du Recrutement). The great stocks of sardines and anchovies, and other small pelagic fishes, account for about one third of the world's yield of marine fish and are of key economic importance in many nations. Production from these great stocks depends upon a delicate balance of physical ocean processes. The optimal environmental window for small pelagic fish depends upon a triad of physical factors (Bakun, 1996): enrichmentprocesses that lead to the production of the zooplankton upon which the young stages depend for food; concentration processes that aggregate foods and thereby increase their availability to growing larvae; and retention processes that keep the young in their favored nursery habitats. Without a doubt, global heating will alter this 'triad' of physical processes. These processes are functions of atmospheric forcing, ocean dynamics, and fresh water inflow; al1 of which are expected to be altered by climate change. The most immediate response to greenhouse warming would occur within the atmosphere rather than within the ocean, affecting the wind field over the ocean, and hence, patterns of upwelling. Bakun (1990) presents evidence that this is already occurring over the past several decades. Thus global climate change could substantially alter these factors that determine favorable reproductive habitat long before ocean temperature changes due directly to greenhouse warming may be evident. Some initial scenarios are already available. For example, Bakun (1990) has argued that one consequence of increased greenhouse effects that can be confidently expected is that temperature gradient between the ocean and the continents will increase during the Spring-Summer upwelling seasons in these systems. This would be reflected in increased alongshore wind and enhanced sea breeze circulation, which would impact recruitment (Mendelssohn and Mendo, 1987. Evidence exists for an 'optimal environmental window' (Cury and Roy, 1989; Cury et al., 199j; Serra et al., this vol.) with respect to wind effects such that changes in characteristic wind speed may disrupt finely tuned reproductive strategies of the small pelagic fishes. A related project on climatic change and pelagic fish stock dynamics was recently concluded in West Africa (Mauntania, Senegal, Côte-d'Ivoire and Ghana) (Cury and Roy, 1991). One important result of this project was to establish the existence of recurrent patterns between the environment and the pelagic fish dynamics. The aim of the CEOS project was thus to build on this analysis through a comparative approach and to generalize previously obtained results. Upnreiiing ecosystems in the Pacific are dominated by large interyear variability whereas in the Atlantic they are characterized by a more seasonal variability. To compare the dynamics of two different functioning upwelling systems would help to 2 The Clirnate and Eastern Ocean Systerns Project (CEOS) unclerstand the impact of environmental constraints on pelagic fish dynamics. Through the CEOS network it was possible to assemble and compare knowledge and data that were collected during several decades in the different upwelling areas. Eastern ocean upwelling ecosystems present certain advantages that may make the study of effects of climate change on marine ecosystems particularly tractable; thus the study may serve an even wider purpose as an illustration of the sorts of impacts that could affect a variety of more complex marine ecosystems. As environmental changes may affect fisli population dynamics in many ways and at different time and space scales, local case studies and comparative global studies are presented for the different upwelling systems. A better understanding of the links between environmental changes and fish population response is expected using various approaches. By analyzing time series data from similarly functioning regional ecosystems distributed over the globe, we hope to tease out the significant global trends from within the 'noise leifel' of naturally-occurnng regional climatic vanability. Disentangling global versus local environmental changes appears to be a major challenge when analyzing environmental time senes. New statistical techniques are developed and applied to environmental indices relevant for fish population dyriamics in order to extract trends and sometime changing seasonal patterns. Another approach to separating global from local ecosystem processes is constructing trophic models of the ecosystems, then computing the values of indices expressing their emergent properties. The Ecopath software of Christensen and Pauly (1992) was used for construction of se~eral models for each of the investigated systems and to compare their properties. Human activities facing local and global changes are also studied. The exploitation of marine renewable resources in the different upwelling areas appears to be a real challenge due to the fact that these resources are unstable. Here, new insights are presented e.g., on how markets cope with variability of supply (Durand, this vol.). The general theme copes with variability and instability: instability of the environment in which the resources evolve, inriinsic instability of the pelagic fish stocks and uncertainty which govern the economic exploitation of a natural resource. Al1 these dynamics are intrinsically linked. The CEOS project was, therefore a multi-disciplinary project, where physical, bicilogical and econometric methods could be focussed on this common theme. The CEOS project was an international collaborative study of potential effects of global climate change on the living re5ources of the highly productive eastern ocean upwelling ecosystems and on the ecological and economic issues directly associated with such effects. A major focus of the study were the clupeoid fishes (anchovies, sardines, etc.) that are heavily exploited in the world's large marine ecosystems and which have recently been exhibiting episodes of collapse, rebound, or switches in dominance. The major objectives of the CEOS project were thus: (1) to assemble, summarize, and analyze the data record of the past four decades regarding the four eastern ocean boundary upwelling ecosystems mentioned above and other upwelling areas, (2) to apply the comparative method to identify key physical processes and ecosystems rezponses, (3) to resolve underlying global-scale trends that in each individual regional system may be obscuretl by local interyear and interdecadal variability, (4) to investigate the relationship of these global trends to accumulating greenhouse effects, (5) to construct scenarios for future consequences of global climate change on upwelling resources, and (6) to analyze and project ecological and social impacts on associated human activities and values. More generally CEOS aimed at promoting scientific exchanges on the theme of the environment, the resource and the fisheries in upwelling areas and consequently to: - promote relevant scientific themes on the environment, on marine resources and on fisheries; - develop research and to promote scientific exchanges between developed and developing countries; - promote the multidisciplinary approach in the management of the marine renewable resources; - exchange data, methods and models in order to improve scientific knowledge; - promote comparative ecosystem analyses in order to generalize process or to identify specificities in the environmental, the ecological or the economical dynamics; - consider news ways for managing fisheries that take into account social and economic processes that are involved in fisheries. 4. LINKAGES OF THE CEOS PROJECT The CEOS project addressed most of the strategic and integrating priorities listed in the U.S. 'Global Research Program Priority Framework', especially the 'Ecological Systems and Dynamics' category, and also addressed in some way most of the issues listed under that category: e.g. '(assernbly and analysis of) Long-Term Measurements of Structure/Function', 'Response to Climate and Other Stresses', 'Interactions Between Physical and Biological Processes', 'Models of Interactions, Feedbacks, and Responses', 'Productivity/Resource Models', etc. The project was designed within the general framework of the International Program of Ocean Science in Relation to Living Resources (OSLR), CO-sponsored by the Intergovernmental Oceanographic Commission and the Food and Agriculture Organization of the United Nations (Bakun et al., 1982). It can be considered an initial effort in the newly proposed subprogram of OSLR and Ecosystem Dynamics and Living Resources (EDLR). Elements of CEOS directly rehted with 'recruitment' constitute contributions to the Sardine-Anchovy Recruitment Project (SARP), a major component of the International Recruitment Program (IREP) of OSLR. It also interacted with the GLOBEC/SPACC initiative and in the Programme National pour le Déterminisme du Recrutement (PNDR GLOBEC-FRANCE, France). The present, multi-authored volume presents the results of an international conference held in Monterey (California) at PFEG (Pacific Fisheries Environmental Group) on September 6-8, 1994 where fifty CEOS participants met to exchange their results. The book is composed of thirty three papers nnging from descriptive and comparative analyses of the major upwelling ecosystems, presentation of new statistical analyses and modeling techniques, to the analysis and modeling of human activities exploiting renewable resources. We hope that the CEOS network reached, through the present book and its other products and activities, at least some of the ambitious objectives that were shared at the beginning. 4 The Climate and Eastern Ocean Systems Project (CEOS) Bakun A. 1985. Comparative studies and the recruitment pro- blems: searching for generalizations. CalCOFI Rep., 26: 30-40. Bakun A. 1990. Global climate change and intensification of coas- ta1 ocean iipwelling. Science, 247: 198-201. Bakun A. 1996. Patterns in the ocean: oceari processes and niarinepopulation dvnamics. Calif. Sea Grant College Syst. Univ. of Calif., La Jolla, 323p. Bakun A., J. Beyer, D. Paulv,J.G. Pope and G.D. Sharp. 1982. Oct!an science in relation to living resources. Can.J Fish. Aquat. SC! ,39(7): 1059-1070. Christensen V. and D. Pauly. 1992. The Ecopath II - a software for balancing steady-state ecosystem models and calculating net- work characteristics. Ecological nzodelling. 61: 169-185. Cury P. and C. Roy. 1989. Optimal environmental window and pelagic fish recruitment success in upwelling areas. Can.J Fish. Aquat. Sci ,46: 670-680. Cury P. and C. Roy. (eds.). 1991. Pêcheries ouest-africaines : variabilité, instabilité et changement. Paris, ORSTOM, 52 jp. Cury P., C. Roy, R. Mendelssohn,A Bakun, D.M. Husby and R.H. Parrish. 1995. Moderate is better: exploring nonlinear climatic effect on Californian anchovy (Engraulis nio~dm). Climate chan- ge and Eish population. RJ. Beamish (ed.). Can. Spec. Publ. Fish. Aquat. Sci., 12 1: 4 17424. Mayr E. 1982. The growth of biological thought. Harvard Univ. Press, Canibridge, Mass., 974p. Mendelssohn R. and J. Mendo. 1987. Exploratory analysis of anchoveta recruitment off Peru and related environmental series. In: D. Pauly and 1. Tsukayania (eds.). The Penivian anchoveta and its upwelling ecosystem: three decades of change. ICLARVl Studies anà Reviez~a, 1 5: 294-306. A. ~AKUN ET AL. 5 The CEOS Comparative Analysis Fra-mework: Motivations and Perceived Opportunities ANDREW BAKUN Foc )d and Agriculture Organization of the United Nations (FAO) Fishery Resources and Environment Division Viale delle Terme di Caracalla 001 00 Rome ITA Y A number of features of the CEOS (Climate and Eastern Ocean Systems) scientific analysis Eramework are citecl and discussed: (1) the specific focus on application of the comparative method, (2) the use of nonlinear empirical techniques, (3) inclusion of both biological-ecologicd ancl socioeconomic aspects within a common investigative design, (4) installation of integrative conceptual bases (e.g., the 'triad') for organizing multidisciplinaiy research nctivity. Radical interdecadal variability may be intrinsic to riiany important fish populations and may introduce serious difficulties with respect to certain conventional tools of fisheries science. If, however, the apparent global synchrony in interdecadal-scale fluctuations reflects true mechanistic linkages, it may signifi some substantial simplifications in the problem of developing scientific predictive capability. Un certain nombre de canctéristiques du résenu scientifique CEOS (climat du bord est des océans) sont énumérées et discutées : (1) l'application de la méthode comparative, (2) l'utilisation de techniques d'exploration dans le domaine non-linéaire, (3) l'intégration des aspects biologiques-écologiques et socio-économiques au sein d'une même a'pproche, (4) mise en place de concepts de base (par exemple, la "triade'!) pour organiser une recherche pluridisciplinaire. D'une décennie à l'autre des changements drastiques peuvent être observés pour de nombreuses populations de poisson et cela est difficilement explicable avec les outils classiques développés en halieutique. Si, cependant, la synchronie apparente de ces fluctuations décennales reflète l'existence d'un véritable mécanisme causal, cela peut singulièrement simplifier le problème de la capacité prédictive. The oceans cover nearly four-fifths of the earth's surface and more than a billion people rely on fish as their main source of animal protein. In some countries, fish are nearly the sole source of animal protein. Demand for food fish rincl various other useful attributes obtainable from the sea has been accelerated by population growth and by the global trend toward population migration toward coastal areas. Fisheries and fish products provide employment to nearly 200 million people. Globally, the bulk of the people employed in fisheries are poor and many are without acceptable alternative sources of work and sustenance. In addition, fish and fishing are enormously important to the cultural life of many coastal societies, and may often define a 'quality of life' for people having a cultural tradition of harvesting the sea. Hence, maintenance of viable fishery resources mav be extremely important to presening traditional ways of life, associated economic activities, tourism, etc. In addition, fish represent the fastest growing food commodity entering international tracle. Accordingly, fih and fish products represent an extremely valuable source of foreign exchange to many countries, in some cases providing as much as half of total available foreign exchange income. The methodologies of fisheries science are intended to ensure sustainable resource populations to support productive and profitable fisheries. Unfortunately, the conventional methodologies are not working vely well. Over and over again, extremely important fish stocks around the world have been collapsing, causing economic dislocations and persona1 suffering to people whose livelihood depends on fishing or on related commercial activities. Some would say that it is not the methodologies that are at fault, but that the problem lies in imperfect application of the methodologies; there is always somewhat of an adversary relationship between conservation-minded fishery scientists and the fishing industn~, and at any tirne there are always some who Say less fish should be taken. But it seems fair to Say that the sudden onset of most of the coliapses comes as a relative surprise to the 'mainstream' of fishery scientists involved. Figure 1 is an illustration of an aspect of the conventional conceptual basis for scientific management of an exploited fishery. 'fie concept is one that may apply well to many terrestrial systems, such as to the management of buffalo on a prairie grassland or of wild deer in a natural forest area. The idea is that at high population sizes (to the far right of the 8 CEOS Comparative Analysis Framework diagram) the population may be using up its available resources of food, habitat space, etc. In such a case, if the population size is reduced by harvesting, the productivity of the stock may actually be improved (Le., in the diagram the system moves to the left toward the 'peak' of the stock-recruitment curve where the rate of production of new members of the population is highest). This has the comforting implication that accumulation of new individuals that would tend to increase the population to stock sizes that lie to the right of the peak, represents a 'surplus production' that can be safely (and, in fact, beneficially) harvested. Spawning stock size The fact that the methodologies based on a conceptual mode1 that reflects experience with wild teri.estria1 populations are not generally working very well for marine fish populations has led some to the notion that marine ecosystems m;iy represent 'chaotic systems', in which unexpected results are a feature intrinsic to the system's dynamics. Certainly, the coupled nonlinear equations that describe the dynamics of predator-prey systems represent classic examples of mathematical chaos (May, 1979). Even very simple models of biological population interactions tend to exhibit chaoric behavior. But, the idea that one is working with chaos in attempting to manage Fisheql resource populations seems defeating. If tiny changes in initial conditions could indeed lead to widely differing outcomes, then there would appear to be little hope of beneficially affecting the outcome by intelligent and prudent adjustments of the rate of exploitation. Happily, there are some recent developments that seem to contradict such a point of view. 2. RADICAL INTERDECADAL POPULATION VARlABlLlTY It is only recently becoming widely recognized by fishery scientists that extreme variability may be an intrinsic feature of many fish populations. Contributing to the developing change in outlook has been the evidence deriveci from cleposits of fish scales in sediments (Soutar and Isaacs, 1974; DeVries and Pearcy, 1982; Baumgartner et al., 1992). Tliese sedimentary records indicate that radical fluctuations in fish population abundance have been common occurrences long before the advent of large-scale fishing. Moreover, in the period for which data records from fisheries have been available, abrupt population declines have been recurrent features. The decadal period from the mid-1970s to the mid-1980s was particularly remarkable in this respect. A pattern of population increases during that interval, followed by population declines after its end in the mid-1980s, seems to have been extremely widespread and consistent (Lluch-Belda et al., 1989,1992; Bakun, in press). For example, this mid- 1970s to mid-1980s penod was one of phenomenal productivity and growth of the major groundfish populations of the Subarctic North Pacific which sustained the massive expansions of the fisheries of that region through the periotl (FAO, 1993). Conversely, since the mid-1980s these populations are in decline, in spire of continuing elaborate stock assessment activities and state-of-the-art fishery management efforts. Alaskan salmon stocks also increased dramatically during the mid- 1970s to mid-1980s period. Total chlorophyll in the water column appears to have increased north of Hawaii (Venrick et al., 1987). Lobsters, sea birds, seals, and coral reef fishes in the northwestern Hawaiian Islands al1 seem to have experienced increased production (Polovina et al., 1994); conversely, since the period ended in the mid-1980s, lobster landings from tliis area have dropped by two thirds (Anon., 1993) and other biological populations are in a downward trend. On the other hand, North Pacific Albacore tuna appear to have suffered a steep population decline during the mid-1970s to mid-1980s period (FAO, 1993). Many salmon stocks of the California Current also declined during this period (Pearcy, 1992; Francis and Hare, 1994). Remarkably, the very large populations of anchovies and sardines that dominate the fish biomass in the majoi eastern ocean upwelling regions of the ~vorld, as well as the northwestern Pacific offJapan, seem to have been rising :ind hlling somewhat in phase (Kawasaki, 1983). Both the Californian and Japanese fisheries grew during the 1920s and early 1930s to peak in the mid to late 1930s. (There were no corresponding landings off western South Amel-ica because no significant fishing occurred.) Both populations remained at extremely low numbers for some three decades. The sardine fislieries in both regions then commenced sudden rapid growth near the mid-1970s, the same pend in which enormous numbers of sardines appeared off South America initiating a massive fishery in that region. Now, toward the latter part of the 1980s, the above-mentioned additional simultaneous reversais in trend have occurred (Lluch-Belda et al., 1989, 1992). Since the advent of substantial fisheries, anchovy populations have been generally out of phase with the sardine populations in the three regions. In the California Current, after a time lag of about a decade following the sardine collapse, the anchovy population increased to the point that over 340 000 tonnes were taken in 1981. Off Japan, the anchovy catches grew during the period of low sardine abundance following the initial sardine collapse, attaining maximum levels of nearly half a million tons during the late 1950s and the 1960s. The anchovy catches then gradually declined as the sardine population proceeded in its rebuilding phase. More recently, as the Japanese sardine population declined, extremely large shoals of anchovy were reported (Lluch-Belda et al., 1992) In the Pem-Humboldt Current sysrem, the fishery for anchovy (anchoveta) peaked in 1970 at more than 13 million tonnes, constituting by far the largest single fishery that has ever existed on earth. It then collapsed to less than 1 million tonnes aFter the 1972 El Niiio, rebounded briefly to about 2 million tonnes for several years, and then fell back following the 1976 El Niiio to below 1 million tonnes and remained at this relatively low level during the period of sardine abundance up to the mid-1980s (see contributions in Pauly et al., 1989). Now as the sardines are plunging, once again the anchoveta population is explosively building to the point that it promises to return Pem to its former position as the world's 'number one' fishing nation (Mendo, 1991), at least until the next El Niiio event. These radical alternations between sardines ancl anchovies occurring on interdecadal time scales have been given the name 'regime change' by Lluch-Belda et al. (1989). 1 O CEOS Comparative Analysis Framework Thus, we have a situation in which fish populations distributed in widely distant parts of the Pacific, ancl in other oceans of tht world as wel (Bakun, 1996), appear to be fluctuating in a degree of synchrony. The populations aie certainly far too witlely separated to interact in any direct way. For example, the populations of sardines in the different 'corners' of the Pacific were considered until very recently (Parrish et al., 1989) to be separate distinct species. But as one views the variability of more and more populations, and notes the same apparent rhythms, the idea that the variabilities are sornehow interlinked becomes compelling. This presents us with somewhat of a 'bad news' versus 'good news' situation. In large part, the paramerei estimations for the various conventional models used in fisheries science depend in one way or the other on equilibiium assumptions. Without the assumption that the various data points can be regrirded as reflections of an identical process, albeit with a substantial random noise component superimposed, the degrees of freedom available to produce estimates vanish. Thus one can't obtain the needed parameters even if the methodologies should be appropriate. However, radical variability on interdecadal scales may also introduce serious problems in the methodologies themselves. For example, consider the following heuristic example, taken from Bakun (1996). Let us artificially construct a stock- recruitment 'history' (Fig. 2) for some hypothetical fish stock that followed the dome-shaped productivity pattern dis~.:ussed above, using only the concept that good recmitments contribute to population growth and poor recruitments to population decline. Let us Say, for example, that in the early 1970s recmitment and stock size both remained low. Thrn suddenly, near the mid-1970s, a series of good recmitments caused the stock to progressively build (i.e., the points move progressively to the right in the diagram). Then, suppose that near the micl-1980s the recruitment le\.els fell back to a lower mean state and so stock size progressively decreased (Le., the points move back toward the origin of tlie diagram) . If a 'stock-recmitment curve' is fitted to these points (see Fig. 1 which was purposely consrmctecl from these same 'data' po.nts), the fit is really quite good compared to many examples one sees in the fisheries scientific literature (e.g., see examples in Rothschild, 1986). Following the usual logic, the fact that the fitted curve turns sharply downwards at the high biomass side of the diagram would imply srrong compensatory density-dependence and therefore surplus production that could be freely exploited with no damage to stock productivity (in fact improving average reprocluctive success) as long as the stock biomass is not allowed to fa11 to the left of the peak of the curve. Under such a logic, a11 management would need to do would be to keep the stock size somewhere toward the middle of the cui-ve and a11 should be well. But of course, this is a delusion. The concept used to generate the data points underlying this curve was not density- dependence, but simple addition: a group of good recmitments adds up at some later time to large spawning biomass (see for example, Sharp et al., 1983). It is recruitment controlling population size, not population size controlling recruitment. An; apparent density-density dependence in this particular example is illusory. There is nothing in the example to support an implication of 'surplus production', and a methodology based on such an implication may be comforting to those trying to balance exploitation and conservation, but totally wrong in this case. Decade of strong recruitment Fig. 2: Hypothetical stock recruitment diagram for a fish population that has exhibited a 'dome-shaped' productivity curve, with particularly high recruitments in the period from the mid-1970s to mid- 1980s. Numbers next to each point indicate the year ('71' refers to 1971, etc,), Dorted lines connecting points indicate the temporal sequence (modified from Bakun, 1396). Spawning stock size 4. THE GOOD NEWS But on the other hand, there may also be some good news in the situation. If radical interdecadal variability is indeed occurring synchronously in distantly separated fish populations distributed over the world's oceans, there would nppear to be some very hopeful implications for the future of our science. One would think that the separation between the populations must in most cases be Far too great for an! significnnt population exchanges or other purely biological interactions that might provide linkage mechanisms which could explain the synchrony. And it seems unlikely that a set of separate autonomous ecosystems, each dominated by its own interna1 chaotic dynamics, could somehow 'self-organize' themselves to generate mutual synchronous variability on n global scale. Thus, if these populations are indeed varying synchronously, the conclusion seems to be that they must not be hnctioning primarily as chaotic systems, at least on the time scales of the synchronized variability. There would, in such a case, be no implied tendency toward unstable responses to minute changes in initial conditions and therefore no reason foi the earlier-mentioned pessimistic viewpoint concerning potential benefits of carehl, skillful fishery management. It would seem that such synchronous behavior would have to be generated by some type of very large-scale external forcing, most probably through climatic teleconnections acting through the atmosphere. Actually, interdecadal modulations of the El Niiic-Southern Oscillation (ENSO) system appear to this writer to be the most likely type of linkage mechanism (Bakun, 1996; in press) (Fig. 3). Moreover, since biological models representing anything but the simplest of marine trophic interactions are characterized by chaotic behavior, the implication of global synchrony would seem to be 7 2 CEOS Comparative Analysis Framework 1 1 Period of generally "enhanced" El Niiio characteristics < 70 72 74 76 78 80 82 84 86 88 90 92 Year thar the biological dynamics involved must be very simple. The synchrony must be a rather direct effect of the external physical forcing acting either on the fish themselves, or very directly on a primay food source, at some sensitive life stage. It must not be, for example, an effect working through a complex food web. If si.), then there would appear to be a realistic hope of success in gaining a real scientific understanding of the factors determining reproductive success and population dynamics of fishery resource stocks. If the large-scale linkage is principally through the atmosphere, as seems most likely, it must probably be transferred through the local sea suiface, leaiing various 'signatures' in the boundary layers of the ocean and atmosphere which exist on either side of the air-sea interface. Temperature trends have not been consistent in the various regions (for example, during the mid-1970s to mid- 1980s period when the eastern Pacific was in a definite warm phase, the northwestern Pacific, where sardines were expanding equally dramatically, was in, if anything, a cool phase). Thus it seems most likely that the effect must be 3 mechanical one. Wind stress acting on the sea surface is the predominant mechanism for transfer of momentum and mechanical energy between the atmosphere and the ocean. Accordingly, we should expect that the causal mechanism Mie are looking for would be a process, or more likely a sum of processes, driven by the action of the wind on the sea surface. The experimental method and the comparative method have been called "the two great methods of science" (Mayi-, 198i). Drawing valid scientific inference requires multiple realizations of the process of interest, preferably over 3 range of diffcring conditions, in order to separate causality from happenstance with a reasonable degree of confidence. The most direct approach to assembling the needed suite of realizations is the experimental method, wherein experimental controls are imposed that allow the scientist to systematically vary conditions of interest while holding other factors constant. But marine ecosystems are hardly amenable to experimental controls. Fortunately, the comparative method presents an alternative. And potentially it is a powerful one. For example, Mayr (1982) credits the comparative method for nearly al1 of the revolutionary advances in evolutionary biology. The comparative method assembles the separate realizations needed for scientific inference by a process of recognition of informative patterns of naturally-occurring temporal and spatial variations in existing conditions and phenomena. That is, different sets of seasonal and/or geographical settings, encompassing a range of natural variability in conditions and mechanisms, substitute for controlled experimental 'treatments'. CEOS was designed to apply this approach in a collaborative, multilateral manner, using different regional fish stocks and ecosystems as sources of the multiple realizations needed to draw scientific conclusions (Bakun et al., 1992). Two different general approaches have been used to apply the comparative method to addressing the fish recruitment problem (Bakun, 1985). The first is to compare the seasonality and geography of spawning to the environmental climatologies of several regions in order to try to resolve patterns of correspondence that can point out the dominant common factors appearing to determine the temporal and spatial aspects of reproductive activity. The studies of Parrish et al. (1983), Roy et al. (1989, 1992), and Bakun (1993) are examples of this type of approach. The second type of approach involves comparative time series modeling, where empirical mode1 formulations are compared among similar species and ecosystems. Inter-regional consistency can then enhance confidence in empirical relationships. The "optimal environmental window" (Cury and Roy, 1989) is a prime example of an ernpirical relationship that bears greatly enhanced credibility and influence due to its high degree of inter-regional reproducibility. 6. GOINC NONLINEAR Fish stocks would have a natural tendency to adapt their spawning habits to represent choices of seasonality and geography that would most often yield the most favorable combinations of the principle factors controlling recent reproductive success. That is how natunl selection works peiner, 1994). Accordingly, fish populations would tend to be adapted to, and therefore fare best under, conditions which are rather typical of their habitua1 spawning habitats. Therefore it would seem that highest success should be associated more with typical conditions than with atypical conditions on the spawning grounds (unless, for example, the atypical conditions represented favoiable circumstances which were not normally available elsewhere within the range of the population). Consequently, one would generally expect 'dome-shaped' relationships, with highest success at intermediate values of a crucial factor and lower success at more extreme values on either the high or low sides. For example, temperature can either be too high or too low, with the optimum for a given species at some intermediate value. Over the recent period of development of fishery-environmental science, reliance on linear statistics and empirical methods has been very much the fashion. This is in spite of the fact that one would intuitively expect 'dome-shaped' relationships rather than linear ones. Thus it may not be surprising that empirical studies of environment-recruitment linkages have often yielded inconsistent, and therefore intellectually non-satisfactory, results. For example, if an empirical study addressed a situation where in most instances conditions were on one flank of such a 'dome-shaped' relationship (e.g., near an extreme end of species range, etc.), then linear analysis might pick up a significant relationship. Likewise, in another situation where most of the data were on the other flank, an equally significant result, but having opposite 'sign', could be found. In such a case, comparison of the two situations would yield directly opposing results, even though the underlying dome-shaped relationship held consistently in al1 cases. And of course, if data were distributed on both flanks of such a 'dome-shaped' relationship, linear methods would probably fail completely to pick up any significant empirical relationship at all. 7 4 CEOS Comparative Analysis Framework Recently, effective nonlinear methods of empirical analysis have been introduced to marine ecology and fisheries science (Mendelssohn and Cury, 1987; Mendelssohn and Mendo, 1987; Cury et al., 1995). A problem with introducing the poskibility of nonlinear relationships is that it is much easier to fit data when one has an indefinite choice of functional fornis. Without the discipline of a single a priori choice, such as linearity, the problem of spurious fits becomes even worse thari usual. In such circumstances, the discipline of comparative interregional consistency in functional form offers a veiy useful alternative. Optimal environmental window Wind intensity - c ? + .- 2 U a, a: In tlie preeminent example that we have available up to this rime, nonlinear methods were applied in a compai-ative context to an empirical investigation of effects of intenlear variations in mean wind intensity on the population djnamics of various stocks of small pelagic fish in several coastal upwelling regions. The result (Fig. 4) nas the famous, domed-shaped, 'optimal environmental window' relationship (Cury and Roy, 1989). 'ïhis finding, and its follow-on extensions (manv of which appear in this volume), finally provides some tangible empirical support for certain concepts (e.g., the 'triad' framework which is precented in the next section) that have been emerging 'inferentially' over the past decade within the context of the international SARP Project (IOC-FA0 Sardine-Anchow Recmitment Project), and more recently, within the CEOS context. It also provides empirical support to arguments such as presented above in Section 4, Le., that the drivirig mecl~anism for synchronized 'regime'-scale population variability must most probably be simply and directly linked to inter-decadal-scale globil climatic variability, probably transmitted locally through the sea surface by action of the wind. - 6 m s-1 \1 Weak Moderate Stronq 7. THE 'TRIAD' FRAMEWORK Comparative studies of geographical climatology of fish reproductive habitats (Le., the first type of approach intrc)duced in Section 5) have tended to identify a 'fundamental triad' (Bakun, 1993, 1996, in press) of three major classes I Lower Density Water (fresher andior warmer) I I The control on the 'right Bank', Le., the 'high wind' side, could come about either through (1) increased offshore transport leading to excessive offshore loss of pelagic lanrae from the favorable coastal habitat (Parrish et al., 1981; Sinclair, 1988) or through (2) overly intense turbulent mixing which could disperse fine-scale concentrations of appropriately sized food particles needed for successful first feeding (Lasker, 1978, 1981a, 1981b) as well as inhibit basic photosyntlietic production by mixing phytoplankton cells beyond their 'critical depth' (Sverdmp, 1953; Steele, 1974). Stiong turbulence mignt also impair a larva's ability to physically capture prey (MacKenzie et al., in press). As a framework for research activity, the triad concept has the advantage that it may appeal to both physical and biological scientists, and so provide a common basis for interdisciplinary studies. Moreover it avoids the defeating complexities of small-scale trophic processes that are impossible to observe directly on population scales and thus to formulate as indisator time series for empirical analysis and verification. (These of course must be extremely important processes in the overall trophic economy upon which fish populations depend. But in terms of influence on radical interdecadal population variability, if the 'good news' arguments presented in Section 4 are valid, one can reason that these trophic complexities muit be operating somewhat in the background and not directly controlling the fluctuations. Thus, in tliis respect they would seem to be less important than the direct linkage of the physical climatic system, through the triad processes, immediately to the fish themselves). Fisheries scientists would like to think that if the industry would only do what the scientists advise, even~hing would be ;il1 nght: populations would remain prolific and productive and fisheries would remain profitable and sustainable. Tliis might not be true, and the notion may lead to a false sense of security that contributes to the disastrous level of overcapitalization of world fisheries (FA0 estimates that world fisheries operate at an annual loss of US$ j3 billion, which must be offset by government subsidies). It seems rather likely that, for many highly variable populations, there may be no mariagement system that would succeed in maintaining them continuously at population sizes approaching a large fraction of h.storica1 peak levels. Even so we should be able to learn from experience to avoid some of the economic losses and social dislocations. In the cases of inherently variable classes of fishery resource populations, rather than clinging to the hope of managing the populations so as to provide secure bases for stable fisheries, it may be necessary to shift the focus toward directly managing the fisheries and to developing 'robust' strategies for economic viability under conditions of radically varying resources. For one thing, a holistic view of risk and uncertainty might help avoid the disastrous overcapacity that has made fisheries a net burden on other economic sectors. This in turn would tend to ameliorate the overfishing problems, and would certainly promote better economic return to those earning their livelihoods by fishing and associated activities. In order to 'learn from experience' in dealing with such issues, CEOS incorporates both environmental biology and socioeconomics as major components of its comparative research framework. In fact, CEOS is the only international scientific program addressing small pelagic fisheries that includes both biology and socioeconomics as substantial components. Of course development of robust strategies is one thing and specific scientific prediction is quite another. Taking an analogy (and jargon) from the American stock market, 'dollar cost averaging' is a robust strategy, Le., one that may serve to maximize earnings by maximizing the likelihood that, on average, more shares are bought when the markets are in a lower price phases than when they are in higher price phases. This may a good type of strategy when no specific prediction is possible and when the markets are reasonably well-behaved. But every stock market player knows that 'dollar cost averaging' is no match for 'insider information' (Le., specific foreknowledge of market events). In the sense of this analogy, one might Say that the goal of CEOS is to generate some real, tangible 'insider information' on fishery resource variability and attendant socioeconomic consequences. So, one might ask what level of prediction one might hope for. Well, maybe the least we might expect on the near term would be to be able to reasonably identify when we may be in a period of transition or 'regime change'. At such a time, for example, a particular level of precaution might be appropriate. One might be warned that the experience accumulated during a recent period of relative stability may not hold in the new conditions, or that equilibrium models that may have been valid to some level of approximation over the previous period may no longer be sol or that a recent high catch per unit effort may not indicate that the population is doing particularly well, but only that conditions have changed (recall the Peruvian experience of very high catch/effort occurring just before the disastrous collapse of the anchoveta fishe~? in the early 1970s). These may represent very valuable pieces of foresight. And that level of prediction indeed seems a realistic hope, particularly in view of the relative simplification (see Section 4) of the system response that mai be implied by the apparent interregional synchronies. The retrospective analytical approach, making use of the new analysis techniques applied in a comparative context, has predominated in the first years of CEOS. The most extensive application has been on the biological-ecological side of the spectrum of issues. Here, the 'triad' idea has been implicit in the choice of independent variables (Le., in the extensive use of wind-related indices of transport, turbulence generation, upwelling, etc.). There would seem to be room for expanded use of a similar approach on the socioeconomic side. Of course, advances on an underlying conceptual framework (in analogy to the ecological side where we have the 'triad', for example) on which to 18 CEOS Comparative Analysis Frarnework structure empirical analysis activities will be extremely useful. To this end, some additional very basic interregional coriiparative 'pattern recognition' among the available histories, anecdotes, and informational fragments might be warranted. The conventional management approach of trying to keep the size of a resource population continually at a relatively high level by managing the level of fishery removals, and thus to provide a basis for a relatively static fisheries industry, is simply not working in a substantial number of cases. Fisheries science badly needs an alternative, but it must be one which can stand up to rigorous evaluation (Sissenwine, 1993). Although in many situations it may not be possible to maintain a desired population level bv adjusting fishing pressure, it most clearly will always be possible to destroy a population by too much fishing. In a situation in which it is obviously in no one's interest (except perhaps in the ve1-y short-term interest of very selfish entities) to utterly destroy a resource, the conventional methodologies at least provide a formal basis for saying 'stop fishing now'. Sol until there is available a specific well-founded alternative, it is clearly unwise to dismantle or discredit what is in place. On the ecological side there seem to be some promising opponunities to become more process-oriented. The train of logic developed earlier suggested the probability that the interregional linkages are transferred from the atmospheric teleconnections to the ocean ecosystem through the sea surface. Many of the triad processes involve transfers through, or changes in properties of, the sea surface skin. Consequently they may leave 'signatures' that might be identified from satellites. This is a timely consideration because there will be shortly up to five separate ocean-specialized satellites in orbit and active; previously there has been no more than one at any one rime (Kieffer, pers. comm.). This new generation of ocean-oriented satellites will also provide new tools. For exarnple, useful direct estimates of ocean primary pi'oductivity from satellites are a strong possibility. Sat::llite images represent a wealth of spatial detail that may be linked to triad processes. A key problem will be in cori~erting this information to the longer time scales for which recruitment information, or other net population-scale outputs, may be obtainable. The standard method for transferring satellite information to longer time periods has been by making longer-term averages of the data. However, such an averaging process degrades the spatial detail which is the strc)ngest feature of satellite-derived information. The trick wiU be to find a more intelligent way to deal with the short time-scale data flow. Thr: tnad framework rnay be of use in identifying features or qualities that rnay in some way be quantifiable to the estent tha: index time senes could be developed. Advanced statistical techniques such as spectral EOFs (Mendelssohn and Ro!r, 19815) may be useful in this regard. Perhaps the techniques of 'artificial intelligence', using neural network computer sofiware, etc., might be enlisted to make use of ever more available and less expensive computer ponler to deal with finding and resolving the pertinent features from within the enormous data flo~vs produced by satellite-mounted sensors (i.e , to allow the computer itself, through shear computing power, to 'learn' to recognize the relevant attributes). Another technique to rationally carry the spatial detail provided by satellite observation systems to the longer rime scales associatecl with population-scale recruitment success is to use the satellite information to drive coupled dynamic coastal ocean models that rnay correctly incorporate pertinent triad mechanisms. Testing of this approach is currently being implemented in a CEOS-associated study off Senegal (C. Roy, pers. comm.). Thtre appear to be some key biological-ecological questions involved in the development of greater understanding of the nature interdecadal regime changes and how best to cope with them. These are often questions that are particulaily difficult to address. 1 have often felt (Bakun, 1996) that there are two types of questions in our business. One type are questions for which we can get answers but don't need them and the other type are questions for which we need answers but can't get them. One major one of the second type is: 'What is the nature of the interaction between anchovies and sardines?'. The classic example of regime change involves a shift of dominance between anchovies and sardines within the extremely imponant small pelagic fish component of the trophic structure of an ocean boundary current ecosystem. But it is unclear b!r what mechanism these two groups might interact to effect displacement of one by the other. Some level of 'competition' seems to be implied if indeed there is a tendency for mutual exclusion. But if the dominant mechanism were some kind of predatory exclusion such as a very major effect of predation of adults of one group on the eggs of the other, etc., it would seem to be counter to the earlier arguments about global synchrony in populations (i.e., lack of major predatory interactions, corresponding absence of chaotic dynamics, etc.). One evident difference between the two is that sardines are larger, correspondingly stronger swimmers, and more adapted to migration, while anchovies appear to be more adaptable, being able to utilize a variety of habitat configurations yielding appropriate 'triad' tradeoffs (Bakun, 1993). Since sardines have done well in several systems which evidently experienced intensification of dynamic aspects during the mid-1970s to mid-1980s period, Bakun (1996; in press) speculated tliat sardines could perhaps deal best with an intensified system (stronger flows, more intense ocean turbulence, etc.). But there is no information presently available as to exactly how such an effect might act. Moreover, this particular argument seems to imply a mechanism acting at the adult stage. On the other hand, one might think it more likely that the linkage mechanism might be acting at the early life stages through differential effects of climate-mediated alterations in characteristic configurations of triad processes. In such a case, if one could define variability in characteristic triad structures through analysis of satellite imagery, it should be possible to design quite simple and feasible programs to sample spatial distributions of larvae in relation to such structures, and to find out where and if anchovy or sardine larvae were abundant and also where they were growing well; for example, larval growth could be gauged by measuring RNIVDNA ratios (Nakata et al., 1995; Buckley, 1984). Similarly, temporal patterns in suivival and growth of the respective species, in relation to temporal variability in the triad physical structures, might be investigated by analyzing daily marks on larval or juvenile otoliths (Campana and Neilson, 198j; Guitierrez and Morales-Nin, 1986) collected at some later time. Clearly, it will be well to continue to try to find innovative investigative approaches to key issues. One might expect that much of the type of results that can issue from standard conventional sampling programs, 'shotgun' approaches to data gathering, etc., might already be largely in hand and that innovation and ingenuity should be the key to getting the answers we need. CEOS has a number of relative innovations incorporated in its analytical framework: (a) use of nonlinear empirical techniques, @) a specific focus on application of the comparative method, (c) promising conceptual bases for organizing collaborative interdisciplinary activity (e.g., the 'triad', which integrates both physical and biological aspects in n simple, relatively comprehensive framework), (d) incorporation of socioeconomic aspects together with ecological aspects. This framework has already led to certain salient results, such as the demonstration of the remarkable robustness of the estimated optimal environmental windows. According to the arguments presented earlier in this paper, the apparent tendency for a degree of global svnchrony in interdecadal population variability and 'regime changes' may allow certain conclusions to be drawn, pointing the wvay to narrowing the range of possible processes and degrees of complexity in the problem. If indeed the biologicd complexity may be minimal, as appears to follow from the argument, substantial progress on issues that have been resistant for mnnJ1 decades, seems a distinct possibility. Thus, if the apparent global synchrony is real, it is indeed a 'gift'. There is no reason that such synchrony would have Iiad to occur. But if indeed it does occur, it is surely rife with significance. We should carefully decipher the full range of implications in order to make good use of the information in directing our research activities along particularly cogent 20 CEOS Comparative Analysis Framework piithways. In particular, it could be very meaningful as regards fisheries science itself. For example, it would appear to iniply that the key problems in our science are problems we can address, without being lost in endless complexities. Ultimately, it could mean that fishenes management need not forever be a sort of operatioml craft, based on 'rules of ti-umb' and aphorisms, but rather have at its disposa1 real prognostic power based on understandable meclianisms and sound scientific laws. More immediately, it could mean that, after decades of frustration, real, tangible progress on tlie fisheries-environment problem may be within our reach. That would be very good news indeed. Arion. 1993. Why lobster boom is over? Fishing Neti~slnter-iza- Campana S.E. and1.D. Neilson. 198j. Microstnicture of fish oto- tional, March, 4. litlis. Can.J. Fish. Aquat. Sci., 42: 1014-1032. B.ikun A. 1985. Comparative studies and the recniitment pro- blem: searching for generalizations. CalCOFl Rep., 26: 30-40. Bakun A. 1993. The Califomia Current, Benguela Curreiit, and sou- thwestem Atlantic shelf ecosystems: Acomparative approach to identifyiiig factors regulating biomass yields. In: K. Sherman, L. hl. Alexander, aiid B. Gold (eds.). Large marine ecosystena: stress, nzitigation, andsustainabili&. Am. Assoc. for the Advan- cernent of Science: 199-224. B.ikun A. 1996. Patterns in the ocean: ocean processes and nlnlinepopulatio?~ dvnanzics. Centro de Investigaciones biolo- gicas del Nordeste, La Paz, Mexico and University of Califoriiia St~a Grant, San Diego, USA, 323p. Eikun A. In press. Adynamical scenario for simultaneous "regi- mi:-scale" marine population shifts in widely separated LMEs of the Pacific.1n: The large marine ecosystems of the Pacific Ocean. Blackwells, London. Bakun A., V. Christensen, C. Curtis, P. Cury, M.H. Durand, D. Husby, R. Mendelssohn, J. Mendo, R. Parrish, D. Pauly and C. Roy. 1992. The climate and Eastern Ocean Systems project (CEOS), Naga, the ICLARM Quarterly 15(4): 26-30. BaumgartnerT.R., A. SoutarandV. Ferreira-Bartrina. 1592. Recons- tniction of the past histol7 of Pacific sardine and northernancho- vy ~~opulacions over the past two milleiiia from sediments of the Santa Barbara Basin, Califorriia. CalCOFIRep., 33: 24-40. Buckley L.J. 1984. RNA/DNA ratio: an index of larval fish growth in the sea. Mar. Biol., 80: 291-298. Cury P. and C. Roy. 1989. Optimal environmental window aiid pelagic fish recmitment success in upwelling areas. Can.J. Fis,',. Aquat. Sci., 46: 670-680. Cury P., C. Roy, R. Mendelssohn,A. Bakiin, D.M. Husbyand R.H. Parrish. 1995. Moderate is better: exploriiig nonlinear cliiiiatic effects on the Californian northern anchovy (Engraulis ~101.- dax). In: R,. Beamish (ed.). Proceediiigs of the 1nternation;il Syniposium on Climate Chaiige and Northern Fisli Popiihtions. Can Spec.Pub1. Fish. Aquatic Sci., 12 1: 417-424. Cushing D.H. 1969. Upwelling and fisli productioii. E40Fish. Tech. Pap., 84,40p. DeVries T.J. arid W. G. Pearcy. 1982. Fish debris in sediiiieiits of the upwelling zone off central Pem: a late Quarternaiy record. Deep-Sea Res., 28:87-109. FAO. 1993. Reviewof the stateofworld marine fisheiyresoiirces. FA0 Tech. Pap., 335, 136p. Francis R.C. and S.R. Hare. 1991. Decadal-scale regiiiie shifts in the large marine ecosystems of the iiorth-east Pacific: a case for tiistorical science. Fish. Oceanogr., 3: 279-291. Gu tierrez E. and B. Momies-Nin. 1986. Time series of daily growtli iii Dicent~archuslabmL oto1iths.J Exp. Ma?: Biol. Ecol. 103: 163-179. Kawasaki T. 1983. Why do some pelagic fishes have wide fliic- tuations in their numbers? Biological basis of fliictuatioii froiii theviewpoint ofevolutionaryecologi.1~z: G.D. Shnrp andJ. Csir- ke (eds.) Repom of the Expert Consultation to htiniiie Changes in Abundance and Species Composition ofNeritic Fish Resources. FA0 Fish. Rep., 291,l: 106 5-1080. Iasker R. 1978. The relation between oceanographic conditions and larval ancliovy food in the California Current: Identification of the factors leading torecmitment h~lure.~app. P.-v. Réun.Cons. int. Explor. Mer, 173: 212-230. Lasker R. 198la. Factors contributing to variable recniitment of the northern anchovy (Etzgraulis nzordm) in the California Cur- rent: contrasting years 1975 through 1978. Rapp P.-v. Réun. Cons. int. E~plor. Mer, 178: 37 5-388. Lasker R. 1981b. The role of a stable ocean in larval fish sunival and subseqiient recniitment. ln: R. Iasker (ed.) Marine fish lar- vae: morphology, ecology and relation to fisheries. Univ. Wasliing- ton Press, Seattle: 80-87. Legendre L. and S. Demers. 1985. Ausiliary energy, ergoclines and aq~iatic biological production. Naturaliste can. (Rev. Ecol. Syst.), 112: 5-14, Lluch-Belda D., R.J.M. Crawford, T. Kawasaki, AD. MacCall, R.H. Parrish, R.A. Schwartzlose and P.E. Smith. 1989. \X1orld-wide fluc- tuations of sardine and anchovy stocks: the regime probleni. S. Mar. Sci., 8: 195-205. Lluch-Belda D., R.A. Schwartzlose, R. Serra, R.H. Parrish, T. Kawasaki, D. Hedgecock and R.J.M. Cnwford. 1992. Sardine and anchovy regime fluctuations of abundance in four regions of the world oceans: a worksliop report. Fish. Oceanogr., 1: 339-347. MacKerizieB.R., T.1. Miller, S. Cyrand W.C. Leggett. In press. Evi- dence for a dome-shaped relationship between turbulence and larval fish ingestion rates. Linmol. Oceanogr. May R. M. 1979. Bifurcations and dyliamic complexity in ecolo- gical systenis. Atzn. New York Acad. Sci., 3 16: 517- j29. Mayr E. 1982. Thegroivth of biological thougl~t. Harvard Univ. Press, Cambridge, Mass., 97411. Mendelssohn R. and C. Roy. 1986. Environmental influences on the French, Ivory Coast, Senegalese and Moroccan tuna catches in the Gulf of Guinea. 67: Special Publication ofthelnterna- tional Skipjack Year Progranz, ICCAT, Madrid: 170-188. Mendelssohn R. and P. Cury. 1987. Forecastirig a fortniglitly abun- dance index of the Ivoirian coastal pelagic species and associated environmental conditions. Can.J. Fish, Aquat. Sci, 44: 408-421. Mendelssohn R. and J. Mendo. 1987. Exploratory analysis of anchoveta recruitmeiit off Pem and related environmelital series: 294-306. ln: D.Pauly and 1. Tsukayarna (eds.) The Penivian ancho- veta and its upwelling ecosystem: tliree decades of change. ICLARV Studies and Reuie~us ,15: 294-306. Mendol. 1994. The Peruvian anchouetafishe~y a17d enuilan- inetztal changes. Presented at the first international CEOS Mee- ting, Monterey, California, 6-8 Septeniber, 1994. Nakata K., H. Zenitani and D. Inagake. 199 5. Differeiices in food availability forlapanese sardine lame between the frontal regioii and the waters on the offshore side of Kuroshio. Fish. Oceanog: : 4: 68-79. Parrish R.H., C. S. Nelson and A. Bakun. 1981. Transport riiecha- nisnis and reproductive success of fishes in the California Cur- rent. Biol. Oceatîogi: , 1: 17 5-203. Parrish R.H., A. Bakun, D.M. Husby and C.S. Nelson. 1983.Coii-i- parative climatology of selected environmental processes in reln- tion to eastern bounday current pelagic fish reproductioii. 117: G.D. Shaq arid J. Csirke (eds.). Proceedings of the Expert Consul- tation to Examine Changes in Abundance and Species Cornpo- sition of Neritic Fish Resources. FA0 Fish. Rep., 291: 731-778. Pmisli RH., R Semarid W.S. Grant. 1989. The nionotypic sardines, sardina and sardinops: their taxonomy, distribution, stock stmc- ture, and zoogeography. Cat2.J. Fish. Aquat. Sei, 46: 2019-2036. Pauly D., P. Miick, J. Mendo and 1. Tsukayarna (eds.). 1989. The Pwuvian upt~dling ecoqistenz: d~~nalizics and internctio17s. ICLARhI Conf. Proceedings 18,468~. Pearcy W.G. 1992. Ocean ecologii ofNoi?h Pacific s~~lii~onids. Washington Sea Grant Prograni. Univ. Washington Press, Seattle and London, 179 JI. Poloviiia J.J., G.T. Mitclium, N.E. Graharii, M.P. Craig, E.E. De Martini and E.N. Flint. 1994. Physical and biological coiisequeiices ofa clirnate event in the centnl North Pacific. Fish. Ocea170gl:! 3: 15-21. Rothschild B.]. 1986. Dynanzicsof i~zaiine~shpopulatiora. Iiar- vard liniversity Press. Cambrige, Mass., 277p. Rothschild, B.T. and T.R. Osborn. 1988. The effects of turbulen- ce on planktonic contact rates.J. Planbton Res., 10: 46j-471. Roy C., P. Cu., P. Fontana and H. Belvèze. 1989. Stratégies sp;i- tio-temporelles de la reproduction des clupéidés des zones d'upwelling d'Afrique de I'ouest.Aquat. LiuingResoul:. 2: 21-29. Roy C., P. Cury and S. Kifani,1992. Pelagic fisli reproductive suc- cess and reproductive stntegy in upwelling areas: eiivii.oniiierital compromises.ln: A.I.L. Payne, K.H. Brink, K.H. Maiiii and R. HiI- born (eds.). Benguela trophic fii[ictioning. S. Aji: ./. 1l1ni-. Sei., 12: 135-146. 22 CEOS Comparative Analysis Framework Sharp G. D., J. Csirke and S. Garcia. 1983. hlodellirig fisheries: Wliat was the question? In: G. D. Sharp and J. Csirke (eds.) Pro- ceedings of the Expert Consultation to Examine Changes in &)undance and Species Composition of Neritic Fish Resources. R40 Fish. Rep., 291: 1177-1224. Siiiclair M. 1988. Marinepopulations. An essay onpopulation regulation and speciation. Washington Sea Grant Prograrii. Uiiiv. Washington Press, Seattle and London, 252 p. Sissenwine, M.P. 199'3. Comment. In: D.D. Piatt (cd.) Thesys- teni in the sea, appllling eco~~~stenz principles to nzarinefishe- I-ics, 2: Conference proceedirigs. The Island Institute, Rockland, Miine: 142. Soutar A. and J.D. Isaacs. 1974. Abundance ofpelagic fisli d~iriiig the 19th and 20th centuries as recorded in anaerobic sediriient off the Californias. Fish. Bull., US., 72: 2 j7-273. Stee1eJ.H. 1974. The structure of~ilarine ecosysterlls. Haivnrd Univ. Press, Cambridge, Mass., 128 p. Sverdrup H.U. 193. On conditions for vernal blooiiiiilg of pli!.- top1ankton.J Cons., Cons. int. Explor. Mer, 18: 287-29j. Venrick E.L., J.A. McGowan, D.R. Cayan and T.L. Hayard. 1987. Cliniate and chlorophyll a: long-terrn treiids in tlie ceritr;il North Pacific Ocean. Science, 238: 70-72. WeinerJ. 1994. The heak oftheJinch. Alfred A. Kiiopf, New York, 332 p. PART 1 Global and Local Environmental Changes The Development and the Use of a Climatic Database for CEOS using the COADS Dataset CLAUDE ROY* ROY MENDELSSOHN* * * Sea Fisheries Research lnstitute Private Bag X2 Rogge Bay 801 2 Cape Town 5 )UTH AFRICA *" Pacific Fisheries Environmental Group (PFEG) 1352 Lighthouse Avenue Piicific Grove CA 93950-2097 USA The Comprehensive Ocean-Atmosphere Dataset (COAûS) was selected by CEOS for use in analyzing the climatic variability of the world upwelling ecosystems during the past four decades. The COADS database summarizes over 100 million surface meteorological observations collected by ships of opponunity and other platforms over the world. This dataset lias world-wide coverage and the earliest data dates back to 1854. A preliminary investigation of the climate variability of the world coastal upwelling regions was performed using a reduced version of the COADS dataset (the 2" by 2" monthly summary files). It appeared that these files are subject to numerous biases and are not suitable for performing the retrospective analysis of the climatic vanability planned by CEOS. It was necessaiy to set up a database using the individual observations instead of tlie summary files. A version (CMR-5) of the 100 million individual records available in the COAûS was provided by NCAR (Iqational Centre for Atmospheric Research - USA). The CMR-5 version of the COADS dataset was reorganized in a fashion that allowed for rapid access to the dataset using a micro-computer, and software for processing and summarizing the data was developed. The reorganized dataset was then put ont0 CD- Roms for distribution. The software and the five CD-Roms represent the core of the climatic information used by the CEOS program for the retrospective and comparative analyses. Some important biases encountered when using the COADS data are reviewed. Changes through time in the measurement procedures are the most common source of systematic errors in COADS; these biases occur in particular in sea surface temperature and wind. The sudden change that occurs after the Second World War due to the use of insulated buckets and of engine-intake measurements is thought to be responsible for the abrupt change in the SST time series that occurs at the same time. Wind data reported by ships are either measured with an anemometer or are estimated from sea-state. Estimated wind data predominate before the Second World War. Today, wind data coliected using measurement devices are predominant. Several examples are given in order to illustrate the potential biases that can affect the seasonal cycle and the long term trend of the wind intensity due to the gradua1 shift through time from estimated wind data to measured wind data. The potential biases introduced by merging buoys data with ship data are illustrated by looking at the wind data off California during the last twenty years. Finally, a discussion on the reality of the positive trend existing in the wind data off West Africa is presented. La base de données COADS (Comprehensive Ocean-Atmosphere Dataset) a été sélectionnée par CEOS pour analyser la variabilité climatique des zones d'upwelling mondiales au cours des quarante dernières années. Cette base de données rassemble 100 millions d'observations météorologiques de surface collectées par les navires marchands et d'autres plates-formes sur les océans du monde entier. La couverture est mondiale et les premières données remontent à 18 54. Un premier travail sur la variabilité climatique fut réalisé en utilisant une version réduite de COADS (moyennes mensuelles par carrés de 2"par 2") Il est apparu que les données contenues dans ces fichiers sont affectées par de nombreux biais qui rendent hasardeuse leur utilisation pour l'étude de la variabilité climatique sur le long terme. Pour pallier ces difficultés, il fut nécessaire de bâtir une base de données rassemblant les données originales au lieu de données déjà pré-traitées. Une version de la base de données COADS (CMR-5) contenant les 100 millions d'enregistrements originaux fut acquise auprès du NCAR (National Centre for Atmospheric Research - USA). Cette base de données fut restructurée afin de permettre un accès rapide à partir d'un micro-ordinateur, et un logiciel d'extraction et de traitement fut -- 38 Climatic database for CEOS using COADS développé. L'ensemble de la base de données fut ensuite transféré sur un jeu de j CD-Roms. Les principaux biais rencontrés lors de l'utilisation de COADS sont passés en revue. Les changements au cours du temps des méthodes de mesure sont à l'origine des erreurs les plus courantes, les variables concernées sont plus particulièrement la température de surface de la mer et le vent. Des modifications des modes de prélèvement et l'introduction de nouveaus instruments de mesure à la fin de la Seconde Guerre Mondiale sont à l'origine d'un changement abrupt dans les séries de température de surface de la mer à la fin des années quarante. Les données concernant le vent, récoltées par les navires marchands, sont de deux types : "estimées" visuellement i partir de l'état de la mer ou "mesurées" i l'aide d'un anémomètre. Jusqu'aus années cinquante, I'essentiel des données de vent était du type "estimées", aujourd'hui les données "mesurées" à l'aide d'un anérilomètre sont prédominantes. Plusieurs exemples sont donnés afin d'illustrer les biais introduits dans les séries temporelles de vent, construites i l'aide de COADS, par l'augmentation régulière du nombre de données "mesurées" au cours des quarante dernières années. Les biais introduits par la prise en compte de données provenant de bouées météorologiques sont illustrés à partir de séries extraites devant les côtes de Californie. Les séries temporelles de vent obtenues à partir de COADS devant les côtes ouest-africaines sont comparées à d'autres séries, et une discussion concernant la réalité de la tendance observée dans les données COADS dans cette région est réalisée. One of the major objectives of the CEOS program is to assemble, summarize and analyze the climatic data record of the four eastern ocean boundary upweüing ecosystems during the past four decades, as weii as the record for the other upwelling areai (Bakun et al., 1993). For this purpose, a climatic data base covering the last fih years with a worldwide coverage is neetled. This also should be homogeneous among areas and over time and thus aiiow the comparative analyses planned by the CEOS prognm. The Comprehensive Ocean-Atmosphere Dataset (COADS) is one of the few datasets that meet these criteria (see Slutz et al. (198 5) and Woodruff et al. (1987) for an extensive description of the source data and of the procedures used to create the C:OADS dataset). The COADS database summarizes over 100 million surface meteorological observations collected by ship of opportunity and other platforms over the world oceans which have been quality controlled and put into a consistent format. This dataset is the most complete record of surface marine climate to date. It has a world-wide coverage and che earliest records date back to 1854. The early studies of the CEOS program have used Release 1 of the COADS for the period 1854 to 1979; and for the period from 1980 to 1990, an interim release has been used, which is compatible in format and organization to Release 1, but is constructed using simplified procedures and preliminary input data. Besides the individual observations, which are available in several formats containing different amounts of the data, reduced or summarized versions of the COADS dataset are available in several forms, including monthly or decadal summaries average over a spatial grid of 2" by 2" for the entire ocean. Due to the massive number of individual reports (100 million), the majocity of researchers have used the monthly summaries. A preliminary investigation of the climate variability of the world coastal upwellings was performed using the 2" by 2" rnonthly summary files. Except for a trirnming procedure which removes extreme outliers, the summary files are computed using al1 the available data. Biases in the long term behaviour of the time series derived from these pre-processed files rapidly arose: data from meteorological buoys are merged with ship data (see Section 3); wind data are aggregated without distinction benveen the measurement procedure used (estimated using the Beaufort scale or measured using an anemometer, see Section 3). The f~ved 2" by 2" spatial grid is also a strong limitation for coastal studies where the spatial grid has to be flexible in order to be adapted to the shape of the coastline. Mer this first attempt, it appeared that the 3" bv 2" degree summaries files were not suitable for performing the retrospective analysis of the climatic variability planned by CEOS. To avoid some of the bias existing in the summaries files, it was necessary that the retrospective analyis be performed using the individual observations instead of the 2" by 2" summary files. Producing useful time series from the 100 million individual records available in COADS is a much more complicated task than working with the 2" by 2" monthly summaries files; this is even more so as the CEOS program needs to run on micro- computers rather than mainframes or workstations in order to allow dissemination of the database throughout the CEOS network. At the time the project started, large capacity optical storage devices became available which allowed for the storage and processing of large amounts of data with a micro-computer. A version of the 100 million individual records available in the COADS database which uses a compressed binary format (CMR-5) was provided by NCAR (National Centre for Atmospheric Research - USA) and transferred to a set of optical disks connected to a micro-computer. The CMR-i version of the dataset was reorganized in a fashion that allowed for rapid access to the dataset by area, and software foi processing and summarizing of the data was developed. The reorganized dataset was then put ont0 CD-Roms for distribution. The software and the five CD-Roms represent the core of the climatic information used by the CEOS program for the retrospective and comparative analyses. 1. DATA SOURCE USED FOR THE DEVELOPMENT OF THE CEOS CLlMATlC DATABASE The primary data source for the CEOS climatic database is the Release 1 of COADS which covers the period 18 54-1979. An interim data product is used to cover the period from 1980 to 1990 (Woodruff and Lubker, 1986). During the design and development phases of the database, Our main goal was, despite the massive volume of information, to preserve access to the raw individual observations and to allow access with a microcomputer. The CMR-5 format provides a good compromise between keeping access to the raw data and a significant reduction of the total volume. It consists of approximately 100 million individual observations from 1854 to 1990. It is a packed binary format containing the most 30 Climatic database for CEOS using COADS frequently used information, designed as a compact alternative to the original reports. The variables included in each report under the CMR-j format are given in Table 1. The following observed quantities are available: - air temperature; - sea surface temperature; - dew point depression; - zona1 and meridional wind components; - sea level pressure; - cloudiness; - present weather. Each record in the CMR-j format contains data on the measurement procedure (fields BI, Wi and HI in Table l), on the pri:cision and units (fields TI and DI in Table l), on the type of observing vesse1 (field ST in Table 1) and on the origin of thc record (field CD in Table 1). Quality control indicators (noted as 'flags' in Table 1) for each observed quantity area also av.;ilable. The COADS documentation (Slutz et al., 1985) gives a detailed description of the quality control procedures. A br::f summary of these procedures is given below. A riiultiple step statistical method was set up during the development of COADS in order to identify outliers for sut variables: sea surface temperature, air temperature, sea level pressure, zona1 wind, meridional wind and humidity. The result of this process is the definition of the smoothed lower and upper median deviation (sl, s2) around the smoothed mfan 01) for each of the six variables; these calculations are performed with a monthly time step and a 2" spatial step. These means and limits are used to create trimming bounds for the variables. The value assigned to the quality control flags for a given observation is set according to Table 2. The entire CMR-j format version of the COADS dataset was reorganised and sorted in order to have an homogeneous structure over the whole time period (1854-1990). 648 folders (or subdirectories) were created, each one corresponding to ;i 10" by 10" geographical square. A folder (or subdirectory) contains 25 data files, one for each 2" geographical square within the 10" square. This structure allows quick access and retrieval of data for a specific ocean location. A cornputer program, the Comprehensive Ocean Data Extraction or CODE was developed in order to quickly access, process ancl summarize using a microcornputer, the COADS dataset in the CMR-5 format. The advantage of CODE over the use of the 2" by 2" monthly pre-processed COADS fies is that aii calculations are performed by going back to the individual observations. This is ~:~articularly important for the calculation of derived parameters such as wind stress or wind speed cubed which are provided by 1CODE. It also aiiows a flexible spatial grid (0.1" of latitude or longitude step in the data, 0.5" in the CODE program), and makes it possible to select between several types of platform (buoys, merchant or research ship, ocean station , etc.) or betureen thr methods used to collect the data (estirnated wind using the Beaufort scale or measured wind using an anemometer). 2. BIAS IN THE COADS DATA Some important biases encountered when using the COADS data are reviewed. Changes through tirne in the me:isurement procedures are the rnost common source of systematic errors in COADS; these biases occur in particular in Sea Surface Temperature (SST) and wind. Air tempenture and dew point temperature are also affected by systematic Field Description Value Units BOX10 * MONTH BOX2 * I'EAR DAY HOUR LAT LONG SST BI AT DP n u v DI WI SLP C NH CL H HI CM CH ST pw CD (e*) LF SF AF RF WF PF Location 10" box number 2" box number (from BOX2 SW corner) (idem) Ternperature Sea Surface Temperature Bucket Indicator Air Ternperature Dew Point Depression Teniperature Indicator Wind Eastward component Northward component Direction indicator Wind speed indicator Pressure and clouds Sea Level Pressure Total cloud amount y*) Lower cioud amount (''7 Low cloud type (**) Cloud height (") Cloud height indicator (**) Middle cloud type (+*) High cloud type (**) Miscel. Ship type Present weather (") card deck ?*) Flags hndlocked flag SST flag Air temperature flag Relative humidity flag Wind flag Pressure flag Table 1 : List of ail variables available under the CMR-5 format. (*) : this refers to the NCAR 10" and 2' box coordinate systerns (see Slutz et al., 1985) for detail (**) : parameters cannot be extracted using the CODE prograrn. 32 Clirnatic database for CEOS using COADS Flag value Triiiirning limits - O X - 2.8~~ 5 x 2 X + 2.8~~ 1 X-3.jsl ix?X+3.js2 2 xX+3.js2 missing missiiig Table 2: Quality control flags values in the CMR-5 format; x is the value of an individual observation, X is the smoothed median and sl and s2 are the smoothed lower and upper median deviation. errm but they are not docuniented here (see Isemer and Hasse, 1987; Folland et al., 1984; Kent et al., 1983 and 1983b foi- details). Along the US coastline, data coming from moored meteorological buoys have been incorporated into the COADS database during the 1980s; the effect of such changes is documented in an area where buoy data have become prcdominant after 1980. Comparisons between atmospheric pressure data from COAûS and several other independent atmosphenc pressure data have been made; the results are generally good and yield some confidence in the usefulness of the COAûS pressure data (Iones et al., 1986; Ward and Hoskins, 1996). 2.1. Biases in SST The method of collecting SST samples has changed through time. Before the Second World War, SST was measured on a sea water sample collected using a bucket; engine intake measurements became predominant for the later period (Iones et al., 1986). Due to evaporative cooling, measurements performed using an uninsulated bucket is thought to be about 0.5"C lower than a measurement from an engine intake (Ramage, 1984). The sudden change that occurs after the Second World War due to the use of insulated buckets and of engine-intake measurements is thought to be responsible for the abrupt change in the SST time senes that occurs at the same time (Folland and Parker, 1990). This sudden shift of the mean value of SST is illustrated in Figure 1. Se\ eral models has been developed to correct bias in bucket observations prior to the Second World War (Folland et al., 1984; Folland and Hsiung, 1987; Folland and Parker, 1990; Jones and Wigley, 1992), however these corrections cannot be applied to individual observation because of the unreliability of the buckethntake indicator and of the lack of information on the kind of bucket used in COAûS. 25 - F - Fig. 1: Time series of Sea Surface Tempera- 2 3 24 ture in the tropical Atlantic (1OoN-3O0N, a, 3O0W-5O0W). A shift in the mean value of a SST occurred in the late 1940s. $ '- 23 1850 1900 1950 2000 Year - - C. ROY AND R. MENDELSSOHN 33 Since the beginning of the 19jOs, insulated bucket or engine intake measurements are thought to be the predominant method used for measuring SST. Thus, even if some bucket measurements have been used after 1945, the biases rire probably considerably less than those occurring before. 2.2. Biases in the wind data Wind data reported by ships are either measured with an anemometer or are estimated from sea-state. Estimated wind data predominate before the Second World War. The percentage of measured wind data started to increase in the late 1940s (Fig. 2). Today, wind data collected using measurement devices is the predominant way of collecting wind data (Fig. 2). The CMR-j format contains a flag for each wind record which indicates whether the data were estimated or measured. Several extractions were made in order to compare the characteristics of the wind signals calculated using either estimated, measured or the combination of estimated and measured wind data in several areas. These comparisons illustrate the potential biases that can affect the long term trend of the wind intensity due to the gradua1 shift through time from estimated wind data to measured wind data. Measured wind data are measured in meters per second. Estimated wind data are reported using the Beaufort scale and have to be converted to an equivalent wind speed. Conversion from the Beaufort scale to wind speed expressed in m/s is done using a Beaufort equivalent scale. The scale used in COADS is the CODE 1100 (or old Wh40 scale). This scale is known to underestimate wind speed for Beaufort number less than 6 and to overestimate wind speed for Beaufort number greater than about 6 (WMO, 1970; Cardone, 1969; Kaufeld, 1981; Isemer and Hasse, 1991). The calculation of wind speed using measured wind data will therefore give different results than the same calculation performed using estimated wind data. Examples of the differences obtained in the determination of the mean monthly seasonal cycle of the wind speed for several areas located in different upwelling ecosystems are given in Figure 3. From those examples, it appears that the mean monthly values of wind speed calculated using measured wind data are from O to 0.8 m/s higher than wind speed calculated using estimated wind; the phase of the seasonal cycle does not appear to be strongly affected. When both estimated and measured wind data are merged to calculate wind speed, the variation through time of the ratio benveen the numbers of estimated and measured wind observations introduces an artificial variabili?; in the time series of wind intensi?;. Because estimated wind data tend to be underestirnated for Beaufort scale lower than 6, wind speed will tend to be 34 Clirnatic database for CEOS using COADS 80 $- a 5 60 u u C .- S 40 u 2? $ 20 $ 2 Canary current California current - u r_d C .- - u 2? - $20 a i l II l 2 1920 1940 1960 1980 1920 1940 1960 1980 Year Year Fig. 2: Percentage of measured wind data in COADS from 1920 to 1990 in two eastern boundary current ecosystems. Guinea current Humboldt current w w California current Canary current - 8 - 8 73 6 U a, a, a a V) E V) u 4 u C C i5 2 i5 4 J FMAMJ JASOND J FMAMJ JASOND Fig. 3: Mean rnonthly seasonal cycle of the wind speed calculated in four areas located in the major eastern boundary current ecosysterns using estimated and measured wind data (calculated for the period 1950-1 330). Measured lowcr during time periods characterized by a high number of estimated wind data than during time periods when measured wind data are predominant (al other things remaining constant). The time series presented in Figure 4 illustrate the effect of the increasing percentage of measured data wind during the last 40 years on the long term variability of the wind speed. The percentage of measured wind data starts to increase significantly around the late 1960s. As a consequence, the mean annual wind speed calculated using di the available wind data @oth estimated and measured, E&M time series) becomes significantly higher than the mean annual wind speed calculated using only estimated wind data (E time series). V) L V) L Figure 5 shows that the difference between the annual mean of the E&M and the E wind time series is mostly explained by the nterannual variability of the percentage of measured wind data. The difference reaches O. j m/s in the California Current and in the Guinea Current when the measured wind comprises 70% of the E&M time series. Measured The consequence of the increasing percentage of measured wind data in the E&M fie series is the introduction of an artificial positive trend superimposed over the long term tendency given by the E wind time-series (Fig. 4). This artificial increase of the winti intensity is particularly noticeable in the Caiifomia Current and the Guinea Current regions. When a linear trend is fitted froni 1970 to 190 to the time series of the Caiifornia and Guinea Current wind speed, the slope of the trend of the E&M wind speed tirne series is 60%, respectively 80% higher than the slope of the trend given by the E wind speed time series. A way to avoid such biases would be to use measured wind data only, but for many areas, the data density of the measured wind data in COADS is not high enough to allow the calculation of a reliable monthly time series of measured wind data before the early 1970s. Time series of wind speed based on estimated wind data can go back in some areas as Far as the late nineteenth century, but the mean value of the wind speed will be affected by the bias introduce with the use of the CODE 1100 conversion table. Changes in the observation practices that were used to estimated the wind from sea-state, California current 26-28"N 1 - All wind data - Percentage of measured wind data Guinea current 4-6"N - All wind data V) È -- Percentage of measured wind data 8 40 O Year Canary current 12-1 6"N - All wind data 1950 1960 1970 1980 1990 80 - Percentage of measured wind data $40 - O - Humboldt current 10-1 4"s - All wind data Estimated wind data 6 V) . E 5 Year series of wind speed calculated using al1 the ble wind data (E&M tirne series) and using only estirnated wind data (E time series) in four upwelling areas. For each region, the annual percentage of rneasured wind data is given. The long-term trend of the wind speed for each tirne series is also presented. the increase of the mean size of the ships over the centuiy and many other factors may also have introduced biases in the long term behaviour of estimated wind time series. 2.3. Meteoroiogicai buoys off the US coast The operation of a network of moored buoys by the US National Data Buoy Centre (NDBC) started in the early 1970s. Hourly buoys data have been incorporated into the COADS interim release for the period 1980-1990 (Moodruff and Lubker, 1986). Some of the biases introduced by merging buoy and ship data are documented here using data extracted off the California coast. 36 Clirnatic database for CEOS using COADS California Current Canary Current 0.50 0.50 -0.25 4-74 -0.25 O 40 80 O 30 60 O/O measured wind data % measured wind data Guinea Current Humboldt Current m 0.75 0.50 FI. 5: Plok of the difference between the 2 1::: 1 2 025 E&M and M wind speed time series against the percentage of measured wind O .%mm O data in four different upwelling areas. -0 25 -0 25 O 40 80 O 40 80 % measured wind data % measured wind data Se~eral buoys were instailed off the Califomia coast during the 1780s. Hourly SST, wind and atmospheric pressure data from these buoys have been incorporated into COADS. The example given in Figure 6 shows that between 36"N and 40°N the number of wind observations stays relatively constant during the 1970s (between 3000 and 4000 observations per year) ancl that it suddenly increases by a factor of 5 in 1981 to reach 22 920 observations. This sudden increase of the number of observations in 1981 is the consequence of the introduction of the buoy data into COADS. The reason for the decrease of the number of observations after 1984 is unknown to us. The buoys are located in the nearshore area and the ship trafic hes are located several miles offshore. Therefore, we can suspect that the dominance of data coming from the buoys during the 1780s introduces important changes in the value of the surilce atmospheric parameters in the tirne series. Unfortunately, the ship type indicator is missing for the interim release and therefore it is impossible to discriminate between buoy and ship observations from 1780 to 19%. 'fie buoys are equipped \vit11 an anemometer and the buoy wind data were recorded in COADS as 'measured' data. In the COADS interim release, this is the onlv way to investigate the change introduced by the buoys data into COADS. The time senes of the scalar wind speed calculated usicg 'measured' wind data in an ara located off the Califomia coast shows that a sharp decrease of the scalar wind speed ccciirred in 1981 when the data from the buoys are incorporated (Fig. 7). This apparent relaxation of the wind speed is likely the result of the coastal location of the buoys where the wind field presents a loal minima (when compared to the offshore domain). The time series of the scalar wind speed calculated using both measured and estimated wind is also affected by the implementation of the buoys (Fig. 7). In this series, the dominance of data coming from the buo. after 1780 results in a spurious relaxation of the wind. When wind variables are calculated using 'estimated' wind data, the buoy dara are not taken into account. The time series of scalar wind speed calculated using 'estimated' wind data shows that the 1981 wind speed relaxation is an artifact resulting from the dominance of buoy data after 1780 (Fig. 7). The COADS monthly 2Ox2" summary files do not discriminate between either ship or buoy data and it is likelv that wind time series off California extracted from these files will be affected by the spurious relaxation of the wind after 1980. 1950 1960 1970 1980 1990 Year Fig. 6: Annual number of wind data available in the COADS database along the California coast between 36"N and 40°N. Measured -)- Estimated El Number of measured IIIIIIIIIIIIIIIIIIIIII Fig. 7: Scalar wind speed aiong the California coast between 36"N and 40°N calculated using 1) measured wind data, 2) al1 wind data (estimated and measured) and 3) estimated wind data. The annual nurnber of rneasured wind data in the area is also presented. 38 Clirnatic database for CEOS using COADS A dominant feature of the time senes of the COADS wind data off West Africa since the late 1740s is a continuous positive trend. Between 12"N and 22"N, wind stress (proportional to the square of the wind) increased by rilmost 50% from 1746 to 1%)O. During the upwelling season, it is expected that such an important increase of the wind stress would intensify the upwelling process and the upward flux of cold subsurface water to the surface, resulting in colder SST. Off West Africa, the magnitude of the wind stress increase from 17 j0 to 1770 is significant enough that drastic changes would have been introduced in the physical and biological properties of the region. However, the validity of the trend in the COADS wind data has been the subject of intense debates (Ramage, 1787; Cardone et al., 1770; Isemer, 1775). In the foiiowing paragraphs, an attempt is made to examine the validity of the trend in the COADS wind off West Africa, between 12"N and 24"N. 1950 1960 1970 1980 1990 Year 1950 1960 1970 1980 1990 Year eries of the Coastal Upwelling Index (CUI) averaged over the upwelling season (January through O to 1330 along the Coast of West Africa between 12"N and 24"N. C. ROY AND R. MENDELSSOHN 39 Monthly time senes of wind stress using the estimated wind data and of SST are extracted from the COAûS database along the coast of West Africa, from 12"N to 24"N with a 2" latitudinal step. A Coastal Upwelling Indices (CUI) is calculated using the wind stress data following Bakun (1973). CUI is the offshore component of the total wind induced Ekman transport. With an upwelling favounble wind, the Ekman transport is directed toward the offshore direction and CUI is positive. From 12"N up to 24"N, monthly CUI and SST are averaged over the upwelling season (Fig. Sa and Sb). In each area, the long term behaviour of the CUI time series is characterized by a continuous positive trend (Fig. Sa). Tlie long term trends of the SST time series are rather stable over the whole time period with a slight cooling in the 1970s following by a warming in the 1980s (Fig. Sb). This suggests that there is no apparent relationship between the long term variability of the upwelling process and SST. A comparison of the COAûS data with other information would be extremely valuable but veq fe~v wind data exist with such a long term coverage. The wind data routinely collected eveq 3 hours at the Dakar-Yoff airport (14'40' N) is one of the few time series available in the region to perform such a comparison. The airport is located on the tip of the Cap-Vert peninsula Year Year Fig. 8b: Time series of the Sea Surface Temperature (SST) averaged over the upwelling season (Ianuary through lune) from 1950 to 1990 along the coast of West Africa between 12"N and 24"N. 40 aimatic database for CEOS using COADS an$ the wind data are thought to be representative of the wind over the offshore domain (Roy, 1989). The annual values dunng the upwelling seasons and the trend of the CUI derivedfrom the Dakar-Yoff airport wind data from 1964 to 1990 are compared with the CUI time series derived from the COADS estimated wind data in the same region (14"N-16"N) (Fig. 9). The interannual variability of both time series are quite similar but there is no apparent positive trend in the CUI derived from the Dakar-Yoff wind data (Fig. 9); during the same time period, the trend in the CUI derived from the COADS wind da[:[ shows an increase of about j0% (from 1. j to 2 m3.s-1.m-1) of the upwelling intensity (Fig. 9). This comparison between the COADS data and the airport data gives two different pictures of the long term variability of the wind in the region. Previous studies of the interannual variability of wind and SST in the region had shown that variability of the upwelling favourable wind accounted for a significant part of the variability of SST (Arfi, 1985; Portolano, 1986; Roy, 1989; Nikjaer and Van Camp, 1994). As an intensification of the upwelling favourable wind results in an increase of the CUI and in an intensification of the upward flux of cold subsurface water, a negative correlation is expected between CUI and SST. With the COADS data, the correlation coefficients between the mean values of CUI and SST during the upwelling seasons from 1950 to 1990 are low (between -0.22 and -0.4 except for the northern most area where it reaches 0.j8) (Fig. 10). This Fig. 9: Coastal Upwelling Index (CUI) averaged over the upwelling season (January through lune) frorn 1964 to 1990 calculated using a) the wind data at the Dakar Yoff airport and b) the COADS wind data in the corresponding square (14"N-16"N, 19"W-16"W). Fig. 10: Correlation coefficient between CUI and SST during the upwelling season Uanuary through lune) along the Coast of West Africa between 12"N and 14"N using a) the original CUI data and b) the detrended CUI data. 1 0.5 1962 1972 1982 1992 Year 1. . >,, II 0, , \ , , :: Dakar-Yoff - 11 1,1111,1111 111 -0.8 W Detrended data Original data 13 15 17 19 21 23 Latitude (ON) suggests that the increase of the upwelling favourable wind in the COADS data over the whole time period has had little effect on the interannual variability off SST. Given the magnitude of the wind intensification (j0% over 40 years) and previous studies, this is a surprising result that raises some doubt about the reality of the trend in the COADS wind data off West Africa. The use of a detrended CUI time series gives a totally different picture. The correlation coefficients between the detrended CUI and SST time series varies between -0.18 and -0.70 (Fig. 10). Except for the northern-most area, using the detrended CUI time series results in a significant increase of the correlation between the strength of the upwelling and SST. To obtain a significant correlation between the fluctuations of upwelling and SST is in accordance with our knowledge of the dynamics of the region and with previous studies, which have shown similar results. This questions the validity of the trend in the COADS wind data off West Africa. The comparison benveen the wind data from the Dakar-Yoff airport and the C0.4DS wind data also raises some doubt about the validity of this trend. If this trend is not a real phenomenon, its origin remains unknown to us. The wind data used to calculate the CUI are restricted to the estimated wind data and the origin of the trend cannot be explained by an increasing fraction of anemometer type data in the wind time series used to calculate the CUI. The COADS database represents a vecy important source of information for studying the long term variability of the climate over the oceans. In many areas like in the tropics, the COADS data are the only information available regarding the variability of the environment. For fishery oceanography, it represents a vecy valuable tool. However this dataset is not exempt of important limitations. Time series of surface meteorological parameters constructed using the COADS data can be strongly affected by several biases and lead to erroneous interpretation. For the wind variables, the main difficulty results from the difference between estimated wind data using the Beaufort scale and measured wind data using an anemometer. It is shown that the calculation of wind speed using the COhDS dataset can be strongly biased depending on the kind of data used. Wind speeds calculated using estimated wind data will tend to be underestimated. The calculation of the mean monthly cycle of the wind speed in several areas shows that difference between estimated and measured values can be as high as 0.8 m/s but the phase of the signal does not seem to be strongly affected. The consequence of the difference benveen estimated and measured wind data is that time series of wind speed can be severely biased by the progressive shift from estimated wind data to measured wind data. When both measured and estimated wind data are merged, the resulting time series is affected by an artficial positive trend. In some areas, the slope of the trend in the wind speed time series calculated using both measured and estimated wind data can be 60% 080% higher than the slope of the trend calculated on a wind speed time series using estimated data only. Recently, Ward and Hoskins (1996) made a comparison between the reported winds in COADS (from the COADS 2Ox2" monthly summaries files) and a wind derived from seasonal mean sea level pressure gradients over the world oceans. Their results show a considerable disagreement between the long term trend in the reported wind and the trend in the pressure derived wind. Globally averaged over the world oceans, there is no trend of any substance in the pressure derived wind, whereas there is an upward trend in the reported wind data of about 14% from 1949 to 1988. They conclude that the difference benveen the reported wind and the pressure derived wind was mainly due to the growing percentage, over time, of measured wind data in the COADS records. 42 Climatic database for CEOS using COADS For the calculation of a reliable wind time series over a long time period, it is strongly recommend to use either estimated or measured wind data and to avoid the use of time series where both wind data type are merged. This represents a strong limitation for the use of the COADS 2Ox2O rnonthly summaries files to study long term climatic variability. In those summary files, both estimated and rneasured wind data are used to calculated a mean monthly wind speed. The introduction of the buoy data along the California coast during the 1980s is another exarnple of the bias introduced by merging data from different origins in COADS. Along the US California coast, the wind tirne series given by the COADS 2"x.l" monthly summaries files are strongly affected by the dominance of the buoys data in the data records after 1980, resulting in an artificial relaxation of the wind. Obviously, the use of these summary files can lead to a completely erroneous interpretation of the long term vanability of the wind in the region. The positive trend in the COADS wind time series off West Africa between 12"N and 22"N is a striking feature. Using estimated wind data, the magnitude of the trend results in an increase of about 50% of the wind stress over the last 40 years. This positive trend cannot be explained by an increasing fraction of anemometer type data in the wind time series as the anemometer wind data were not included in the calculation. The comparison with an independent wind time series and the dramatic increase of the correlation between CUI and SST when the trend is removed mises some doubt as to the validity of this trend. However, one should notice that a positive trend exists in the pressure derived wind in the region (sel: Fig. 6 in Ward and Hoskins, 1996). The trend of the Nonh Atlantic Oscillation (NAO), which is an indicator of the strength of the atmosphenc circulation in the Atlantic Ocean related to the Acores high (Hurrell 1995, 1996), is also in accordance with the long rem trend that exists in the COADS wind data in the 14"N-16"N area (Fig. Il), but with a less pronounced upward trend during the last 30 years. Given these contradictory results, the validity of the trend in the COADS wind over West Africa rernains an open question. From our experience with the use of the COADS database to study long terrn clirnatic variability over the oceans, we conclude that extreme care should be taken in the interpretation of the results. It is strongly recommended to get access to ~he individual records rather than using the COADS 2Ox2" monthly summaries files. The five CD-Roms produced by CEOS provide an easy to use alternative to these summary files. -1.2 1945 1955 1965 1975 1985 1995 1945 1955 1965 1975 1985 1995 Year Year Fig. II: Time series of the standardized anomalies of the North Atlantic Oscillation (NAO) and of the scalar wind speed during the upwelling season in an area off West Africa (1 4"N-1 G"N) from 1946 to 1990. MIR. 198 j. Variabilité interannuelle d'un indice d'intensité des remontées d'eau dans le secteur du cap Blanc (Mauritanie). Can. .I. Fish. Aquat. Sci ,42 (12): 1969-1978. Bakun A. 1973. Daily and weekly upwelling indices, West coast of North Anierica 1946-71. US. Dep. Conznz., NOAA Tech. Rep. IVVFS SSRF-67 1, 103 p. Bakunh,V. Christensen, C. Curtis, P. Cury, M.H. Durand, D. Husby, R Mendelssohn,J. Mendo, R. Parrish, D. Pauly and C. Roy. 1993. The Cliiizate and Eastern Ocean Svstenzs Project. Naga, The ICLARM Querterly lj(4): 26-30. Cardone V.J. 1969. Specification of the wind distribution in the nzarine bolrndary .)fer for zuave forecasting. Report TR69-1, New York University, New York, NY, 131p. CardoiieV. J., J.G. Greenwood and M. A. Cane. 1990. On trends in historical marine wind data.J Clinzate, 3: 113-117. Folland C.K., D.E. Parker and F.E. Kates. 1984. Worldwide niari- ne temperature fluctuations, 1856-1981. Nature, 310: 670-673. Folland C.K and J. Hsiung. 1987. Correction of seasonally varying biasesin uninsulateci bucket seasuhce temperaturedata usinga phy- siai mdel. Met. O@ceS]moptic Clinzatology Branch Menw., 1 j4. Folland C.K and D.E. Parker. 190. Observed vxiations ofseasurfa- ce temperature. In: M.E. Schlesinger (ed.). Climateaean intenc- tion, NATO workshop, Oxford, Kluwer hdemic hblishers: 21-52. HurrellJ. W. 1995 Decadal trends in the North Atlantic oscillation: regional temperatures and precipitation. Science, 269: 676-679. Hurrell J. W. 1996. Influence of variations iii extratropical win- tertiine teleconnections on northern heniisphere temperatures. Geophys. Res. Lett., 23: 665-668. Isemer H,]. 1995. Trends iii the marine surface wind speed: ocean weather stations versus volontary observing ships. In: H.F. Diaz and H.J. Isemer (eds.). Proceedingsof the International COADS Winds Workshop, Kiel Germany, 31 May-2 June 1994. US dept. of Coniiiierce-Institut fur Meereskunde, Kiel: 68-84. Isenier H.J. and L. Hasse. 1987. The bunker atlas of the North Atlantic Ocean. 2: Air-sea interactions. Springer-Verhg, 252p. Isemer H.J. and L. Hasse. 1991. The scientific Beaufort equiva- lent scale: effect onwind statistics and climatological air-sea flux estimates in the North Atlantic 0cean.J Clirriate, 4: 819-836. Jones P.D., T.M.L. Wigley and P.B. Wright. 1986. Global tempe- rature variations, 1861-1984. Nature, 322: 130-434. Jones P.D. and T.M.L. Wigley. 1992. Corrections to pre-1941 SST measurements for studies of long-terni changes in SSTs. In : H.F. Diaz, K. Wolter and S.D. Woodruff (eds.). Proceedings of the International COADS Workshop, Boulder, Colorado, 13-1 j Janua- ry 1992. US Department of Conimerce, NOAA. 227-237. Kaufeld L. 1981. The developnient ofa new Beaufort equivalent scale. kIeteoi: Rundsch., 34: 17-23. Kent E.C., P.K Taylor, B.S. Tniscott and J.S. Hopkins. 1983a. The accuncy ofvoluntaiy observing ship's meteorological obseivations - Results of the VSOFNAJ. Airrios. & Oceanic Tech., 10: j91-608. Kent E.C., R.J. Tiddy and P.K. Taylor. 1983b. Correction ofmari- ne daytime air temperature observations for radiation effects. J. Atnios. & Oceanic Tech., 10: 900-906. Nykjaer L. and L.V. Van Camp. 1994. Seasonal and interannual variability of coastal upwelling along northwest Africa and Por- tugal from 1981 to 1991. J. Geophys. Res., 99 (C7): 14197-14207. Portolano P. 1986. Analyse des séries vent-tenipératures de la nier en surface le long des côtes sénégalaises. Océanogi: trop., 21 (2): 2Oj-227. Raniage C. S. 1984. Can shipboard measurements reveal secu- lar changes in tropical air-sea heat flux ?J Clilnate Appl. ~VIeteor:, 23: 187-193. Ramage C.S. 1987. Secularchange in reported wind speeds over the ocean. J, Clinl. Appl. Met., 26: 52 5-528. Roy C. 1989. Fluctuations des vents et variabilité de I'upwelliiig devant les côtesdu Sénégal. Oceanologica Acta, 12 (4): 361-369. Slutz R.J., S.J. Lubker, J. D. Hiscox, S.D. Woodruff, R. L. Jenne, D.H. Joseph, P.M. Steurer and J.D. Elms. 1985. Conzpi.ehensiue Ocean- atiizosphereData Set; Release 1. NOAA Environmental Researcli Laboratories, Climate Research Prognam, Boulder, CO, 268p. Ward M.N. and B.T. Hoskins. 1996. Near surface wind over the global ocean 1949-1988.J of Clinfate, 9: 1877-189j. Woodruff S.D. and S.J. Lubker. 1986. COADS 1980-85 update. In: S.D. Woodruff (ed .). Proceedings of a COADS Woi.kshop, Boulder, Colorado, January22-24,1986. US Departnient of Coni- merce, NOAA. Technical Memorandum, ERL ESG-233: 36-48. Woodruff S.D., R.J. Slutz, R.L. Jenne and P.M. Steurer. 1987. A Comprehensive Ocean-atmosphere Data Set. Bull. Airlm ,l.leteoi: Soc., 68: 1239-12 j0. World Meteorological Organization (WlO). 1970. Repoi-ts on tzarine science affairs. 3: The Beaufort scale of wind force. WMO Geneva, Switzerland, 22p. 44 Climatic database for CEOS using COADS How to Detect a Change both on Global and Local Scale in Oceanographic Time Series MARIE-HÉLÈNE DURAND* ROY MENDELSSOHN* * * OKSTOM Laboratoire Halieutique et Ecosystèmes Aquatiques BP 5045 34032 Montpellier Cedex 1 FRANCE ** Pacific Fisheries Environmental Group (PFEG) 1352 Lighthouse Avenue Pacific Crove CA '!3950-2097 USA Analyzing change in climatic time series requires a more complete understanding of the different way that time series can Vary over time, and the different kinds of 'trends' and 'pseudo-trends' that can occur. Time series can Vary in the mean in either a deterministic or stochastic manner, they can Vary in the seasonal component in either a deterministic or stochastic manner, and the underlying 'innovations variance' can also Vary with time. Each of these types of changes in a series can produce similar kinds of behavior over a short period of observation, but are driven by very differing processes. Multiple time series, particularly those that are spatially distributed, may exhibit these types of behavior in a seemingly unrelated manner, yet can be shown to be driven by common underlying components. These types of dynamics will be vet? different from those that would be estimated by EOF's or other techniques based on principal component analysis. Ire discuss recent advances in time series analysis and econometrics that allow us to more fully explore how time series may vary and to test for and to estimate the underlying components of change. Pour analyser les changements dans une série climatique, il ne suffit pas de « repérer )) les différentes sortes de « tendances )) ou « pseudo-tendances )) qui peuvent apparaître. Sous peine de profondes erreurs de compréhension, il est nécessaire de pouvoir caractériser correctement les différents processus d'évolution dans le temps. Une série temporelle peut évoluer de façon déterministe ou stochastique dans sa composante moyenne et/ou dans sa composante saisonnière. Les processus résiduels peuvent également avoir une variance qui évolue dans le temps. Ces différents processus d'évolution ont des propriétés très différentes et la caractérisation que l'on peut faire des changements temporels en dépend grandement. Malheureusement, sur de courtes périodes de temps, ces différents processus apparaissent très semblables. Il est donc primordial de pouvoir distinguer correctement les différentes composantes d'une série temporelle pour en caractériser les propriétés. Lorsqu'on examine conjointement un groupe de séries temporelles, par exemple des séries de température distribuées sur différentes latitudes adjacentes, il est possible de mettre en évidence des processus d'évolution sous-jacents qui leur sont communs. Ce type de dynamique est très différent de celui qui pourrait être estimé'par des modèles EOF ou autres méthodes basées sur l'analyse en composante principale. Des avancées récentes en analyse des séries temporelles et en économétrie permettant d'explorer le type d'évolution d'une série temporelle, de tester et d'estimer les composantes sous-jacentes de ces changements temporels sont présentées. The earth's climate has been changing over the centuries, and coupled with these changes have been change in the ocean environment and in marine resources. Over the last several decades, concern has risen that anthropogenic effects, particularly in the atmosphere, may be altering or accelerating the natural pattern of climate change. However, the observed anthropogenic changes in the atmosphere are not necessarily producing concomitant changes in the ocean, or may be producing changes that are more complex than those observed in the atmosphere. Even without human influence, this natural progression of climate could have profound effects on the ocean and its resources. Understanding how the ocean has been changing, how the resources in the ocean have been changing, and detecting anthropogenic effects in 46 Change in Oceanographic Time Series eithcr, are difficult problems that require careful thinking about what we mean by change, about the ways change can occur in the data we analyze, and whether even a global process of change could produce varying effects in different regions of the world. The types of dynamics we are presently witnessing in the ocean and the atmosphere raise questions that typically can not be answered by the usual methods of analysis. The term 'climate change' has been used loosely in the oceanographic literature, referring to almost any variation in the ocean environment. What is meant by 'climate change' must be more precisely defined if we want to characterize it properly. Changes in the ocean environment can be transitory or relatively permanent. Even such large shocks as major El Ninos can be transitory in nature when analyzed properly over a longer time-scale. Roy and Mendelssohn (this vol.) have shown that the major El Ninos of 1972 and 1982 in the Humbolt Current, though large, were short-lived in their effects while the 1956-57 El Nirio had long-lasting effects on the California Current system. ENS0 events may be part and parce1 of the long-term trend in the ocean, or they may be relatively frequent shocks to the ocean that in fact obscure the long-term signal. The upwelling regions of the world are of particular interest because of their high productivity. The upwelling process, which is local in scale, rnay be influenced by changes more global in scope. Bakun (1990), for example, has hypothesized that global warming would increase upwelling, making the upwelling regions cooler during the upwelling months, and perhaps affecting the oxygen content and other properties of the local ocean waters important to fish stocks. These changes, however, would not be reflected in the ocean as a whole. A change obsemed in the ocean environment at a particular area can be unique to that area, or part of a process obsemed over an entire region, or part of a process occurring over the entire globe. Moreover, there could be regional or global processes that are causing the obsemed change in the local environment, but with a different manifestation depending on the site it is occurring. It is then necessary to analyze vari~.bility at different spatial scales. Separating out common trends or common seasonal components from local trends and season is important to handle this type of question. Changes in the upwelling process can occur not only in the intensity of the upwelling, but also in the timing of crucial events such as the seasonal cycle, the onset of the spring transition in the California Current system for example. Identifying the proper source of these changes needs to be able to separate what pertains to a trend and what pertains to some cycle around this trend. To properly deal with these types of problems requires precise definitions of the types of changes that might be observed and carehl thinking about how change can be manifested in the data we analyze. Similar problems arose in economics wht:re highlighting common features among variables, isolating source and timing of changes among any of numerous and a1w;iys moving explicative factors, separating out a proper trend from some cycle and distinguishing a shifting trend from 0thr.r cyclical fluctuations are crucial questions. Great advances have been made in econometrics and time series analysis to handle such a challenge. These new statistical techniques, which at least begin to deal with some of these issues, must be iised also to analyze changes in oceanographic data, or else the results could be misleading. The aim of tl-iis paper is to describe some of these concepts and techniques; examples of their use are also presented. From the previous section, it is clear that change, particularly in the context of climate change, can be very complex in nature. Yet, when the term 'change' is mentioned, it is firstly as opposed to some sort of relatively stable state and change is therefore associated with some sort of non-stationary state. Climate change is often used in contexts such as 'global M. H. DURAND AND R. MENDELSSOHN 4 7 warming', meaning that the weather worldwide is progressively becoming warmer. In other words, the worldwide mean temperature has a positive time trend in the long-term. In this sense, 'change' and 'trend' in a series are almost synonymous. This is a limited view of what constitutes a 'change', particularly in regards to the climatic changes in the ocean. Only some of the examples presented in the Introduction correspond to this notion of change. Models that focus on the long-term change in the mean will not lend insight to the other problems discussed. Even restricting ourselves to 'trends' in the data, the problem of defning the trend is more complex than finding the change in the overall mean. If a large region is being examined, this assumes that the entire region is changing in a homogeneous, uniform manner. In the ocean, due to the complex interactions txtween circulation processes and atmospheric forcing, the obselved changes in different areas could be in different directions even if the underlying process is global in scope. Aggregate methods, or even disaggregate methods that assume a uniform trend wil be misleading in identifying the change. Shifting trends are also changes that are very dificult to put to the fore if the trends are not correctly measured. 'Trend' can occur in the mean, the usual way people think of trend, but 'trend' can also occur in the seasonal component, and in the variance of the series. Properly identifying a trend will help to better characterize a change. More formally, a time series random variable is said to be stationary if its distribution does not depend on time. A time series is said to be weakiv stationay if the mean and variance of the senes do not depend on t'me. A tirne senes is non-stationary, or has a 'trend', if either the mean or the variance of the time series (or both) are functions of time. Changes in the mean and the variance of a series can come about in several ways, for example, changes can be either deterministic or stochastic. Changes at intermediate frequencies can occur due to changes in the seasonal component. For non-stationary series, a senes that has a trend in the mean has different properties than a series that has a trend in the variance. Wule over the shon run the dynamics can appear similar, this difference has imponant consequences both for mode1 building and for our understanding of the behavior of the process. 1 .l . Deterministic and stochastic trends The common understanding of trend is a changing mean level that varies deterministically. Let y, be an observed tirne series, then a deterministic trend in the series y,would be given for example by: y =,u+m+E where~,-+N(O,cr~) t t (1) The mean or expected value of this series is E @J = p + a t. This mean evolves with time while the variance is constant, VarCyJ = a2. Such a series is said to have a deterministic trend or to be "trend stationary" since a simple regression on time will detrend or stationarize the series, the resulting detrended series will have a constant mean p. The mean level of a deterministic trended series increases by some f~ed amount every period (Fig.1~). However, a time series can behave in the short-run as if it has a deterministic trend in the mean, yet be generated bv a different mechanism. The simplest case of such a series is the so-called "random walk" which is of the form: YI = YI-1 +El (2) where the E, are independent, identically distributed random variables with mean of zero and variance a2 (Fig.lb). Tlie series y, can be rewritten as: 48 Change in Oceanographic Time Series Such a series has a constant rnean, ECyJ = yO, and a variance which increases towards infinity over tirne: Val-@,) = cs2t If the E, are zero rnean stationary but autocorrelated, the senes y, is no longer a pure random walk but the trend of the y, series will still behave as a random walk. These types of trends are called stochastic trends. In the econornetric literature, thel are also referred to as 'unit root', since the random walk model is equivalent to an autoregressive model with a root of moilulus one in the autoregressive polynornial (Le. roots of q(B) = O in a series such as: NB)jI = $(B)E~), see Hatanaka (1996) for details. In the above example (Eq. 2), it can be noticed that the coefficient in lag term is equal to 1. Differencing the series will remove the stochastic trend, leaving the series stationary. Processes that becorne stationary when diffcrenced are called 'integrated' or 'difference stationary'. Let the backshift operator B be defined as: Then differencing a series can be expressed as Ay, = (1 - B)y,. Sometimes a series needs to be differenced d times to become stationary. It will be then necessary to calculate ddy, = (1 - B)~~,. These series are said to be 'integrated of order d', (lenoted by I(d). A random walk process is a very simple case. Generally, observed series exhibit more complex behlvior than that of a pure random walk. A randorn \valk with drift process can be given by: where E, is given as in Equation 2. Rewriting yt as: t=o we see that this series has both a deterministic trend (i.e. a trend in mean given by ECyJ = vo + w) and a stochastic trend (Le. a trend in variance given by VarCyt) = o2t). Since differencing will cancel out both trends, ECy, - yb1)=p and Val,?, - = o 2, such a series with both a deterministic and a stochastic trend is also an integrated series (Fig.ld). A randorn walk with drift will change in each period by some f~ed amount on average. The change in each period will be by a predictable arnount p, which is called the drift, plus an unpredictable random amount. For this reason it is referred to as a 'stochastic trend'. Thi. different behaviors produced by these types of time series can be understood by examining artificial time series gen'rrated from closely related equations (Fig. la,b,c,d). The SST series at 36-38"N was whitened and the resulting residuals were used as the innovations (errors) in each of the simulations. The first series (Fig.la) is a stationary autoregressive series with one lag and an autoregressive parameter equal to 0.5. The series has a constant mean and a constant variance. When the autoregressive parameter is set to 1, a unit root is introduced in the series. The randorn walk natiire of the series is evident (Fig.lb). Adding a deterministic trend to the stationary series (Fig.1~) and adding an intttrcept to the random walk series, which becomes a random walk with drift (Fig.ld), highlights the problem of differentiating between deterministic and stochastic trend. The behavior of the two last series is ve? sirnilar yet generated by rwo very different processes. Most people would say that the random walk with drift has a deterministic trend and wo~ld wrongly detrend this series by a regression on time. In the case of a Trend Stationary senes, it is only the mean of the series which will bring information on the long-tem evolution of the process. The variance of forecast errors is finite, the uncertainty attached to these forecast is then bounded. It is indeed not the case for a Dfference Stationary senes since the variance of a senes with a unit root is infinite. The best forecast of the future value of an integrated senes we can make is its present value since the prediction error is going to infinity with time. The important dfference between these two classes of non-stationanty lies on the fact that a 'shock' at any given penod on a trend stationary process wili only have a transitory effect while it will have a lasting effect on a difference stationaly process. integrated series are 'long memory' processes since any short-lived event wili influence definitively the future level of such senes. a: Stationary Series AR(1) b: Random Walk Series c: Trend Stationary Series d: Difference Stationary Series -10 1950 5 1955 1960 1965 1970 1975 1960 1985 1990 Year -10 1950 5 1955 1960 1965 1970 1975 1960 1985 1 Year Fig. 1: Four artificial series generated {rom the same white noise q (t is the time term trend). a : Stationary : Xt = 0.5Xt_1 + q b : Random walk : Xt = Xtm1 + q c : Trend Stationary : Xt = 0.1 T + 0.5Xt_1 +q d : Difference Stationary : Xt = 0.2 + Xt-? +q 1.2. Stationary and non-stationary seasonal component Oceanographic data, particularly series such as SST, are cliaracterized by their strong seasonality. In the oceanographic literature, seasonal components are ~~sually estimated by t~king monthly means or by fitting one or several Iiarmonics ro the data. Another method, more recently adrocated, to deseasonalize montlily oceanographic time series is to transform the series to 12th-differences (Bskun, 1996, p. 165). Eacli of these tnro deseasonalizing metliods suppose very different seasonal processes. The first one assumes that the season21 component i3 deterministic and stationary, while the second one assumes a stochastic and non stationsry sexonal component. To use n method in an inappropriate case will lead to spurious results. Prior to any deseasonaliza~ion or to any study of the seasonal pattern, it is then impoilant to be able to 50 Change in Oceanographic Time Series decide which seasonal process is suitable for the observed time series. Some testing procedures have been recently developed for that purpose. The concept of seasonaiity is, unless we have a precise and formal definition, as vague as that of a trend. Yet finding a precise definition for seasonality is not as simple as it appears. Despite a long history in analyzing seasonality, there is no geneelly agreed upon definition, nor is there a widely accepted view about how seasonality should be treated in empirical work. The maiil definitions offered by the literature (see for example Hyleberg et al., 1990; Franses, 1996) are: - A detert?zinistic seasonulpmcess : This seasonal process is stationary with a mean that varies by season. It is modeled by a regression on seasonal dummy variables such as the following quarterly series : Yt = mo + mlSlt + m2SZt + nz3S3t + E~ where E~ is stationary. Seasonal dummy models imply a regular and non-changing seasonal pattern which can be perfectly forecast though some changes in phase and amplitude may appear in raw series due to the error process and to the existence of an autoregressive structure. - A stationary stochastic seasonal process is a process generated by an equation such as : cp (B) yt = E~ where E~ are iridependent and identically distributed with al1 the roots of cp (8)= O lying outside the unit circle and with peaks in its spectrum at seasonal frequencies, as for example: Yt = pYt, + E~ with Ipl V) V) a, 1920 1940 1960 1980 Year : ig. 2: Time series plots of 'upwelling' scries (April-July averagcs frorn seasonal rnodel cornponents) for poleward wind stress (solid linesl and SST (dashcd lines), for COADS 2" boxes. Vertical axes denotc 3rnjls? and 0.lo, rcspectively. IATITVDE T 6. YEAH (Apr:liin) SST w. YMR (Apr:liiii) SSTis. .r (Alir:liiii) SST 1,s. 7 (OCL-J;II~) r b(10.') I99%C1 r b(10.~) I99%LI r b(10.') r99%CI r h(l0.1) r99acl 46.48 44-46 42.44 .;O-42 38-40 36-38 34-36 Bight so-52 28-30 26-18 24-26 22.24 -1oIlllllllllill 30 40 Latitude ("N) 88 Seasonality of Eastern Boundaty Current Systerns 2.2. A cornparison of shore-based seasonal series to the COADS series AI coastal upwelling (April-July seasonal) SST and sea level series decreased over time and al1 SSS series increased over timrt, implying an increase in upwelling (Fig. 4, Table 4). The magnitude of their changes corresponds to about 0.07- 0.14"C1 0.01-0.1 ppt, and 0.1-0.7cm over the past 45 years. Correlations versus time are al1 highly significant (p < .01). Excc-pt for Crescent City, where the adjacent COADS stress (41N) shows an increasing seasonal trend, SST and sea level (SSS) are highly positively (negatively) correlated with local wind stress (Table 3). The patterns occurring during the pericxl covered by the COADS data (1946-90) are consistent with those seen in the full-length shore series (57-78 years) (Tai-de 5). Regressions between coastal SST, SSS and sea level (the series shown in Fig. 4) are highly significant and of the sign consistent with that expecred if upwelling is the controlling process (Tables 56). This is consistent with the fact that up elling is a dominant process off much of the West Coast during April-July. SSS /---/-- NEA AH BAY -- / ' - ,/' 1 L ,/" / SST SST a 7 O 1 SST-LLA] ..~ .~ .. -- .~. ... SL---. L---- \ --- SSS ----.---------d-------- 1 1920 1940 1960 1980 Year STATION SST vs. 7 SSS vs. T SL vs. T r b (10.~) r b (103) r b (10.~) - - - - - ~~~~~ NEAH BAY .656 3.41511.542 -.764 -72.178I23.951 .715 4.60111.766 CRESCENT -327 -4.047I1.082 ,936 64,96119,596 -390 -4,200I.844 FARALLON .944 11.2812.729 -.966 -1.9201.202 .971 2.6382.256 LA JOLLA .892 5.77711.150 -.645 -4.26411.983 .966 ,918I.097 STATION SST vs. k'EAR SSS vs. YEAR SLvs. YEAR r b (103) r b (10~) r (IO-*) NEAH BAY -.972 -1.3621.129 .959 24.41612.820 -.919 -1.591t.269 CRESCENT -.898 -1.5002.289 .981 23,23521,794 -.974 -i.j68+.i43 FARALLON -.958 -3.009I.355 ,976 10.8702.955 -.gai -1.493+-,116 LA JOLLA -.566 .1.589I.910 .775 2.2191.712 -.728 - .300+.111 An examination of the April-July seasonal averages at other coastal stations reveals very similar patterns (Fig. 5). Note the high degree of visual correlation between coastal SST series in adjacent boxes and along the entire coast. Several stations along the central and southern California coast (e.g., Avila, Bodega), over the 32-40°N range of increased upwelling suggested by the COADS data, feature decreasing SST and increasing SSS. These time series agree quantitatively with the other coastal sites discussed previously, both in the linear tendency and the decadal period fluctuations. Up~velling SST and SSS series generated for coastal stations north of 50°N (e.g., Cape St. James, British Columbia; Seward, Alaska) display no statisticai change during April-July, suggesting the patterns of increased upwelling noted in the center of the CCS are not evident in the subarctic Pacific region inîluenced by the Alaskan Gyre. The correlations between COADS and coastal SST are significant; although the slopes of the linear regressions between shore-based SST and COADS stress series are larger than with the COADS SST (Tables 3, 6); Le., shore SST changes are 90 Seasonality of Eastern Boundary Current Systems F C ST JAMES ----;-1-. 1 / SEWARD \<------- / /- NEAH BAY Y i CHARLESTON y-- / _-_---- CRESCENT CITY BODEGA t FARALLON __/--- - - a C O HUENEME STATiONS (ON) 47 46-48 124 - 127 Neah Bay (48"22') 1935-1992 45 44-46 123.5- 127 - - 43 4244 124 - 127 Charleston (43"2 1') 1966-1992 4 1 40-42 123 - 127 Crescent City (41°45') 1933-1992 39 3840 122 - 127 Bodega Bay (38"19') 1957-1992 37 36-38 122 - 126 Fanilon (37"2 5') 192 j-1992 35 34-36 120.5- 124 Avila (3j010') 194 5-1992 33 32-34 120 - 122 - - BIGHT 32-34 116 - 120 Hueneme (34"09') 1919-1987 La Jolla (32" 52') 1916-1992 3 1 30-32 116 - 120 - - 29 28-30 114 - 119 - - 27 26-28 113 - 118 - - 25 24-26 111 - 116 - - 23 22-24 110 - 114 - - - - - Station P (5O0N,145~ 19 50-1992 - - - Station N (3OoN,140'W) 1954-1974 Table 1 : Dimensions of COADS boxes containing der hly averaged equatorward wind stress and SST (for period 1946-1 990), and selected shore stations wi OADS boxes. Dates for monthly averaged time series of SST, salinity and sea level shown. To estimate a time-varying (Le. non-stationary) trend component for each monthly-averaged time series, we assume that each monthly average y(t) is the sum of four components y(t) = T(t) + S(t) + I(t) + e(t), t=l,T (1) where, at time t, T(t) is the unobserved rime-dependent mean-level (trend), S(t) is the seasonal component, I(t) is the irregular term (stationary but autocorrelated), and e(t) is the stationary uncorrelated component, which can be viewed as 'observation' or 'measurement' error. A non-parametric and non-linear trend is estimated for the monthly-averaged time series using a state-space model solved by using a combination of the Kalman filter and maximum likelihood methods (Kitagawa and Gersch, 1984, 1988). The trend term in Equation 1 can be viewed as an unknown function of time, and parameterized as Vk T(t) - N(0, ozT). (2) For k = 1 and 02T = O, Equation 2 reduces to a linear fit; i.e. T(t) = a + bt, rather than the discrete equivalent of a k-th order smoothing spline. Figures 2 and 3 show examples of observed time series of wind and SST, respectively, along with their respective trends from the state-space models. The models work much better in accounting for SST variability; the difference between the SST trend and the residual of the observed, less the seasonal and AR series, (equal to the trend plus model error) is negligible. The errors in the wind stress models are higher, presumably because the response rime of SST to atmospheric forcing acts to 'smooth' month-to-month variability occurring in wind forcing. These examples also demonstrate that vanabiliry in long-term trends is smail relative to that in the seasonal cycles of wind stress and SST, pointing to the need for a method that wiü extract long-term variability for analysis of climate change. Schwing and Mendelssohn (this vol.) and Durand and Mendelssohn (this vol.) discuss in detail the statistical techniques applied here. 7 06 Spatial Structure of Wind and SST in California 50 1 a) OBSERVED, TREND (BOLD) I b) SEASONAL COMPONENT I c) AUTOREGRESSIVE COMPONENT 1 d) TREND (BOLD) 50 OBSERVED - SEASONAL - AUTOREGRESSIVE (GRAY) 1946 1956 1966 1976 1986 Year Fig. 2: Comparison of COADS 37"N (36-38"N) poleward wind stress rnonthly observations to model results, for period 1946-90. a) overlay of trend model component (bold line) and observed monthly series; b) seasonal model component; c) autoregressive (AR) model component; d) overlay of trend model component (bold line) and observed series minus seasonal and AR model components (gray line). The y-axes of panels a) and d) have been condensed to ease presentation of one large negative observation. 18 , a) OBSERVED, TREND (BOLD) A b) SEASONALCOMPONENT C) AUTOREGRESSIVE COMPONEIVT d) TREND (BOLD) 15 OBSERVED - SEASONAL - AUTOREGRESSIVE (GRAY) Year Fig. 3: Cornparison of COADS 37"N (36-38"N) SST m model results, for period 1946-90. a) overlay of trend model component (bold line) and es; b) seasonal model cornponent; C) autoregressive (AR) model cornponent; d) overlay O ent (bold line) and observed series minus seasonal and AR model cornponents (gray line). The y-axis of panel D has been expanded for clarity. 7 08 Spatial Structure of Wind and SST in California Time series of poleward wind stress trends for the COADS 2" boxes display the spatial and temporal variability of the CC5 wind field (Fig. 4). The wind separates into three distinct geographical regions; 22-32"N (south), 32-40°N (central), and 42-t8"N (northern), based on a visual comparison of the time senes, and the clustering of statistical correlations between the senes (Table 2). Wind stress trends in the southern region (dashed-dotted lines) became increasingly equatonvard (negative) over time in a relatively monotonic pattern, as noted by the highly linear fits to the series (Table 3). Stress also strcngthened from 22" to 30°, but was weaker in the southern California Bight; local maximum equatonvard stress \vas seen over 26-30°N. The Bight featured weaker stress and more variability on 5-10 year scales, relative to adjacent boxes, and is weakly correlated with the other wind time series in the CCS, particularly those off central California. . . 1940 1950 1960 1970 1980 1990 Year Fig. 4: Time series of poleward wind stress trends for COA t time series from the southern region (22-32%). Solid lines represent 4OoN). Bold dotted lines represent time series from the northern region ( e 39"N time * series. 33N .57 .57 .57 .68 .67 .3j .7j .37 .12 .28 .73 .j1 .4j BIGHT .61 .62 .51 .55 .60 .78 .74 .41 .13 .22 .70 .61 .47 31N .80 .88 $2 .88 .98 .87 .78 .43 .11 .28 .70 .63 .jO 29N .90 .91 .94 .97 .95 .84 .66 .27 -.O8 .1 j .70 . j2 .3 j 27N .90 .96 .98 -96 .95 .77 .67 .33 .O0 .17 .75 .57 .41 25N .96 .96 .93 .94 .93 .78 .77 .49 .22 .38 .97 .76 .62 23N .97 .87 .83 .88 .87 .72 .77 .56 .35 .46 .91 .82 .73 Table 2: Correlation S 2" box trend time series. Upper-left half between poleward wer-right half shows correlations bet correlation values in bo e groupings of highly coherent series in three geographic regions. Spaces between rows an regions within CCS, based on subjective examination of time series. The 0.01 level of significance is 0.1 1 (n=540). The central region (solid lines) displayed the strongest equatorward stress in the CCS (Fig. 4). Stress trends became increasingly equatorward over time (Table 3), but exhibited much more interannual variation compared to the soutllern region. The stress series at these latitudes are less correlated with those in the other regions as a result of this interannual variability (Table 2). A period of stronger than normal stress in the 1750s was followed by a period of decreasing equatorward stress in the mid-1760s. The center of this region (3g0N, bold dashed line in Fig. 4) featured the greatest change over time (Table 3), shifting from the site of the region's weakest stress in the 1760's to strongest stress in the 1980's. While stress in the Bight appears relatively uncoupled with this region (Fig. 4, Table 2), winds immediately offshore of the Bight (33"N in Fig. 4) are similar to the central region's stress series. In contrast to wind stress off California and Baja, the region north of 44"N featured a mean poleward stress that strengthened over time (Fig. 4, Table 3). These series are negatively correlated with virtually the rest of the CCS, due to their opposing series-long trend. 4044"N was a transition area between the central and northern regions; equatorward stress decreased rapidly north of 40°N. In about 1776 the 43"N COADS series shifted from the pattern seen in the northein region to that of the central region. This results in the 43"N winds being poorly correlated with most of the other stress series. Two temporal phenomena are notable for their absence in the stress trend series. ENSO events (e.g. 1757, 1783) are not apparent in the series. Instead the model allocated their variance into the AR series (Fig. 2c), and model error (difference in gray and black lines in Fig. 2d), presumably because the wind field responds rapidly to a developing and decaying ENSO. I IO Spatial Structure of Wind and SST in California The well-documented regime shift in 1976 (Trenberth, 1990; Ebbesmeyer et al., 1991; Miller et al., 1994, Trenberth and Hur:ell, 1994) is not seen in the wind trends either, despite its clear presence in north Pacific atmospheric pressure indices. A likely possibility is that the CCS is out of the region where winds were directly affected by the intensification of the .4leutian Low beginning in 1976. However a substantial increase in equatorward stress did occur in about 1983 in the central region and nonh to about 44"N. One interpretation of this intensification is that the transition zone between the central and northern regions has broadened south; another is that the area off northern California and Oregon developed its own distinct wind regime after 1983. In either case, the net effect is that the alongshore gradient in poleward stress has streiigthened greatly over the last 45 years. 23N 25N 27N 29N 31N BIGHT 33N 3jN 37N 39N 41N 43N 45N 47N slope -17.7 -21.7 -21.3 -16.j -10.7 -7.9 -7.9 -17.0 -19.3 -48.3 -11.9 3.3 19.4 14.7 for COADS 2" box trend time series. First line shows correlations between stress and box. Second and third lines show correlation between COADS stress and time, and slope of linear fit between stress and time (linear rate of change, 10-2 m2/s2/year), respectively. Positive (negative) sign denotes linear trend for increasing poleward (equatorward) stress. Fourth and fifth lines show correlation between SST and time, and slope of linear fit between SST and time (linear rate of change, 10-3 OCJyear), respectively. Positive (negative) sign denotes warming (cooling) trend. Spaces between columns separate three geographic regions within CCS, based on subjective examination of time series. The 0.01 level of significance is 0.1 1 (n=540). Coniours of COADS poleward wind stress trend anomalies, relative to the long-term mean at each latitude, reveal additional temporal and spatial patterns in the wind field (Fig. 5). For reference, the zero contour of the stress trends (daslied line) is included. Pnor to the mid-1950s, stress anomalies were negative (more equatorward) north of 40°N and positive (less equatorward) south of 40°N. A particularly strong spatial contrast occurs in the 38-44"N region at the beginning of the series. A penod of negative stress anomalies north of 30°N began near the onset of the 1957 ENSO anrl continued into the early 1960s, contrasting with a contemporaneous penod of weak positive anomalies south of 30°N. Stress was anomalously strong throughout the entire CCS from about 1964 to 1970. Since 1970, anomalies have become increasingly negative (more equatorward) south of 40°N, and negative at higher latitudes since about 1983. The COADS SST trends separate visually (Fig. 6) and statistically (Tables 2 and 3) into essentially the same geographic regions as wind stress. For example, the relatively poor correlation of SST series off northern California with those in tlie south (Table 2) imply a different pattern of interannual ocean vanability in the centrril CCS. However SST is more correlated in space than wind stress on interannual (1-5 year) scales (Fig. 6). SST decreased consistently with htitude frorn about 22ON to 4(:1°N (fine lines), coincident with a region of equatorward wind stress. SST nonh of 40°N (bold lines) was nearly uniform with latitude, an area where stress was either generally poleward or becoming less equatorward over time. While ENSO wind events are relatively ephemeral, and thus are relegated to the model AR term, warmer conditions associated with ENSOs are obvious in the SST trend series, and appear to gradually dissipate long after the ENSO atmospheric signal (compare also Fig. 2 and 3). However some ENSO signals appear in the SST AR component as well (Fig. 3c). Interannual variations in the series (e.g. the 1957 ENSO event) are greatest off central and southern California (Fig. 6). In contrast to the 1957 event, the 1982 ENSO is reflected in SST as a smaller local maximum relative to adjacent years. 1950 1960 1970 1 980 1990 Year Fig. 5: Contours of poleward wind stress trend anomaly series for COADS boxes. Anomalies are with respect to series means for each 2" box. Shading denotes positive anomalies; hatching denotes negative anomalies. Contour interval is 3 rn2Is2. Broken line denotes location of zero wind stress. The obvious warming beginning in 1976, a climate shift not reflected in stress, suggests that decadal-scale SST variability in the CCS is controled by the large-scale pressure and wind fields, rather than local wind forcing. The shift in 1976 rivals the degree of warming associated with the 1957 ENSO. Since the mid-1980~~ SST south of 32"N increased slightly, but cooled substantially north of 34"N. The series-length linear tendencies (linear regressions of time series, Table 3) are significantly positive (increasing SST) south of 34"N and in the 43"N box, but significantly negative north of 36"N (save 43"N). The alongshore SST gradient off southern California strengthened over time (Fig. 6). Simultaneously the SST gradient along northern California decreased. SST at 39"N (fine broken line) decreased gradualiy over time, cooling to values seen at higher latitudes. At the sarne time, 43"N SST (bold broken line) warmed relative to nearby boxes. The convergence of SST in the 38- 42"N region coincides geographically with the strong temporal intensification in stress at these latitudes (Fig. 4 and 5). 7 72 Spatial Structure of Wind and SST in California Con tours of SST trend anomaly series for the COADS 2" boxes (Fig. 7) reinforce the idea that decadal-scale variability in SST is generally coherent throughout the CCS. Anomalously cool periods prior to 1956 and during 1968-1977 contrast with warrn events during 1956-1968 (following the 1957 ENSO) and since 1977 (following the 1976 regime shift). These warm and cool penods are not coherent with any periods of anomalous wind stress (Fig. 5). Years featuring ENSO events were charactenzed by rapid warming. Warm conditions remained several years after the 1957 event. However the warming and cooling associated with the 1983 ENSO was more symmetnc. Since about 1985, conditions have remained warm south of 34"N, but have been cooler in the north. The strong series-long cooling trend at 39"N (Table 3) can be seen in the anomalies (particularly when contrasted with regions to the south), which corresponds to strengthening equatomard stress Year Fig. 6: Time çeries of enote time series south of 40°N. Bold lines denote time çerieç no note tirne series for 39"N and 43"N COADS boxes, respeaively. 1960 1970 1980 1990 Year Fig. 7: Contours of S$T trend anomaly series for COADS boxes ("C). Anomalies are with respect to series means for each 2" box. Shading denotes positive anomalies; hatching denotes negative anomalies. Contour interval is 0.2"C. (Fig. 4 and j). As mentioned earlier, however, the SST (Fig. 7) and wind stress (Fig. 5) anomalies do not agree othei-wise, suggesting that local winds are not the primary factor controlling decadal-scale temperature variations in the CCS. Shore-based SST trend time series along the western North American Coast (Fig. 8), derived with the state-space model, provide an independent assessment of SST variability, extended in space and time. These series also allow a finer (along- and across-shore) spatial scale look at temporal variability. As with the COADS SSTs (Fig. 6 and 7), the shore series demonstrate well-defined latitudinal regimes in ocean temperature on decadal scales. ENSO events (Quinn et al., 1987) are shown by the shaded vertical lines in Fig. 8; their large-scale warming influence is seen. Coastal warming associated with ENSOs was greatest along central California. The SST signal of some ENSO events (e.g., 1972-1973) appears to be weak and constrained to southern stations. 'The shore series (Fig. 8) show that many of the temporal features of CCS SST extend north into the Gulf of Alaska and well offshore (Ocean Stations P and N). However there also are obvious regional differences between the CCS and coastal locations off British Columbia and Alaska, particularly during ENSO events and on decadal scales. Numerous other regional differences in the trends are evident. Most coastal sites feature series-long warming trends that are highly non-linear and display considerable variability on annual-to-decadal scales. Most series have 10-20 year penods of relatively level temperature, foiiowed by similar periods of rapid temperature increase. Some series have multi-year periods of decreasing temperature. Higher latitude sites have a well-defined cycle of about 20 years, which Royer (1993) attributes to the 18.6-year tide. 7 7 4 Spatial Structure of Wind and SST in California Year Fig. 8: Trend time series of SST 1°C) from several coastal sites along North American West Coast (locations shown on map in inset). Time series for Ocean Stations P and N also shown. Bold lines denote time series from central California locations. Shaded vertical bars denote ENS0 periods, based on Quinn et al. (1 987). F. SCHWINC, R. PARRISH AND R. MENDELSSOHN 7 7 5 The shifts to a cool regime in 1941 and to a wam regime in 1976 (MacCall and Prager, 1988) are evident in the shore series as well. The 1976 shift occurred rapidly along southem California. However it was less dramatic and occurred over several years at more northem sites, thus appearing more as a gradua1 warming. The 1941 transition is more difficult to document, but appears less dramatic relative to the 1976 shift. The regime shift concept may be valid for some regions (e.g., southern California), but displays a considerable degree of inter-regional difference. In addition to alongshore differences in CCS SST, a comparison of selected COADS SST and nearby coastal SST trend time series (Fig. 9) suggests there also are significant differences between nearshore and offshore temperature trends. SST in the Southern California Bight agrees well with shore-based SST at La Jolla. Further north, however, offshore and shore SST series at corresponding latitudes are poorly correlated. Although the COADS SST trends cooled over time north of about 36"N, the coastal SST trend series warmed over time along the entire U.S. West coast. Shore series accentuate interannual- to-decadal changes as well. Like the COADS SSTs, shore SST trend series are poorly correlated with COADS poleward wind stress. A particularly interesting feature is the close positive correlation berween Crescent City and adjacent COADS SST from about 1970-1983, followed by an apparent inverse relationship since. The adjacent wind stress also appears to be negatively correlated with shore SST, and positively correlated with COADS SST, since 1983. Another intriguing relationship is the close correspondence between an approximately five-year period of weakening equatonvard stress and warming shore SST in the late 1960s, centered along central California (e.g., Farallon, Avila). This was followed by a similar length period when wind stress and SST returned to values seen in the early 1960s. This warminglcooling event is not reflected in the adjacent COADS boxes. Both of these events attest to the considerable decadal-scale variability inherent in the CCS, which is not always associated with ENS0 events, and merit a closer analysis. The results also suggest the CCS may experience a substantial degree of cross-shelf variability as well. Although the COADS observations do not have sufficient wide-spread density of coverage to produce monthly time series at 2" resolution for the greater northeast Pacific, or to resolve the spatial variability of the CCS on finer scales, we can examine spatial differences in the area's SST and wind fields at a 1" resolution by averaging the observations for two separate decades (1966-1975 and 1977-1986) to contrast conditions prior to and after the 1976 climate shift. Differences in SST between these periods (Fig. 10) corroborates the monthly time series' patterns. SST in recent years south of Oregon has been warmer (gray shades) in, and well offshore of, the CCS. The region north of California also appears to have been warmer at the coast and to about 1 j0 km offshore after 1976. However SST in 1977-1986 in an offshore area off Oregon and Washington was cooler (hatched areas) relative to the earlier period. A closer analysis of these differences by season (not shown) suggests the cool anomaly in the northern CCS is connected to a much larger cool anomaly that covers much of the central North Pacific. Furthermore the eastem edge of this large cool anomaly moves eastward from February to September. Negative SST anomalies nearly reach the coast of Oregon and Washington by summer. Like the relationship between the wind and SST time series, however, the distribution of SST anomalies over the northeast Pacific since 1976 is not coherent with changes in local wind stress or mixing. For example, most of the CCS south of 40°N, which was warmer in 1977-1986 relative to 1966-1975, featured greater equatonvard stress and turbulent wind mixing during the latter period, which should lead to cooler SST. Thus it does not appear that interannual to decadal changes in SST, and more generally upper ocean conditions, are dominated by climate shifts in local wind forcing. We conclude that a complex interaction of local and remote Ekman advection, wind mixing and direct heating is responsible for the long-term fluctuations in SST in the CCS and northeast Pacific. Regional-scale differences in the relative importance of the mechanisms responsible for decadal-scale climate change in the northeastern Pacific is subject of ongoing research. 7 7 6 Spatial Structure of Wind and SST in California 47N STRESS ,-%-A ---.\---. ,--/----A 1 O 1 O SST (48'22') 0 - CRESCENT CITY -20 1 -30 -40 -20 a; - aï - -30 - V) V) - - -40 2 - V) - - -50 7 SST (38'1 9') . - I I I cn 3 - -20 - -30 - -40 '. STRESS -50 AVlLA SST (35'1 0') 35N STRESS -30 Year Fig. 9: Trend time series of COADS SST and poleward wind stress, compared with selected nearby coastal SST series. Bold solid lines represent COADS SST series. Fine solid lines represent coastal SST series. Broken lines represent wind stress series. Location of the time series are shown in each plot. F. SCHWINC, R. PARRISH AND R. MENDELSSOHN 1 17 Longitude ("W) Fig. 10: Difference map of SST for northeast Pacific comparing annual means in 1" by 1" boxes frorn two ten- year periods (1 977-1 986 less 1966-1 975). Positive values (shaded areas) denote relatively warrner SST in 1977- 1986; negative values (hatched areas) denote relatively cooler SST in 1977-1 986. Contour interval is 0.25"C. Range of contours is -0.75 to +1.25"C. The different regions of the CCS and the likely relationships between wind forcing and SST on long time scales are illustrated in Figure 11. The Figure shows the slope (+/- 99% confidence intervals) of a linear fit to poleward stress and SST trend time series in each COADS box; Le., the linear tendency of each series (Table 3). A statistically significant tendency of increasing equatorward stress coincides with a cooling trend in much of the central region (34-42"N). A reasonable explanation for this pattern is that increasing stress leads to greater offshore Ekman transport and more coastal 7 7 8 Spatial Structure of Wind and SST in California upwelling, which cools the surface waters of this portion of the California Current. This region coincides with the geographic range of the upwelling maxima along the West Coast (Parrish et al., 1981). However shore-based SSTs feature a long-term warming tendency, countering the argument for greater coastal upwelling. A more likely explanation is that SST in the CCS is responding to large-scale atmospheric forcing that is changing in a way that leads to cooler surface conditions in [lie CCS at these latitudes. It is also possible that an increase in the positive wind curl occurring off nonhern California (Bakun and Nelson, 1991) could have accompanied the greater equatomard stress, and led to cooler SST through intensified Eknian pumping. Increased equatomard stress also would contribute to cooler SST through greater turbulent mixing of the upper ocean, and may be associated with increased southward transport of cool water by the California Current. COOLER SST WARMER SST SST (1 OC Jyear) ) and SST (open circles) time series. 99% SST). Dashed lines denote boundaries of Outside of the central region the trends in stress and SST are negatively correlated. Increasing equatorward stress coincides with warming SST south of 34"N, while greater poleward stress accompanies a cooling trend north of 44"N. 'i'he cooler SSTs off Oregon and Washington may be due to greater wind mixing. Greater poleward stress could be linked to a larger (gyre) scale pattern in atmospheric circulation, such as the documented intensification of the Aleutian Low, which could lead to fundamental changes in the region's ocean circulation. However, a more rapid cyclonic transport of the Subarctic Current, typically associated with a deeper low, should increase SST. An examination of long-term changes in Ekman divergence due to vanability in the wind field is beyond the scope of this study, but this could impact SST as well. We suspect that the transitional region between the Subarctic and California Currents could be extending southward, based on the expanding region of spatially uniform SST (Fig. 6). This would be consistent with cooler SSTs. We also believe, based on preliminary analysis of data from the entire north Pacific, that major shifts in the magnitude, position, and composition of the West Wind Drift feeding into these boundary currents may have changed over time. This is currently an active area of investigation. lie relationship between wind stress and SST in the southern region (22-34ON) is more dificult to explain. It is difficult ro arrive at a scenano that is consistent with a long-term tendency for greater equatorward stress and warmer SST. Roemmich and McGowan (1995) present results for southern California consistent with those here, and speculate that increasing stratification over the last 45 years (due to an increasing heat flux from the atmosphere into the upper ocean) has more than compensated for greater wind stress, leading to shallower upwelling and a less effective 'cooling' from upwelling- favorable stress. Extending this idea to the regions farther north implies a different dynarnical balance, such that the SST signal associated with greater stratification, a consequence of atmospheric warming, is ovenvhelmed by the role of changing wind patterns in driving more cool water into the surface layer of the CCS, through a combination of lateral advection from the central north Pacific, and upward movement by greater upwelling and vertical mixing. In any case, it is apparent that over the last several decades surface waters in the CCS south of 34"N have experienced a different set of forcing conditions from those farther north. Not only are the tendencies of wind and SST different in these regions of the CCS (Fig. Il), but the spatially changing relationship between stress and SST implies that the primary mechanisms driving variability in SST, and probably the general circulation, on decadal time scales are fundamentally different in these various regions. It is not possible to develop a more firm conclusion about the mechanisms leading to these tendencies without a more thorough analysis of the temporal variability of wind vector fields over the north Pacific Ocean. However it is important to recognize that, over the last j0 years, wind and SST over the extent of the CCS has not fluctuated in a uniform manner (Tables 2 and 3). Thus using a single time series to represent climate variability in this ecosystem will give an incomplete, if not inaccurare, view of decadal changes in the CCS. 3.1 . Ecologicai impacts of environmental variability The temporal and spatial variability of the physical environment of the CCS described here must be considered ivhen analyzing changes in the biological structure of this and other EBC ecosystems. Parrish et al. (1981) concluded that the larval assemblages and reproductive strategies of coastal fish species can be divided into regions that approximate the three areas defined from the long-term wind and SST patterns described here. Roemmich and McGowan (1995) have found that zooplankton biomass off southern California has decreased by 80% since 1951, while the surface layer has warmed by more than l.j°C over the same time. They suggest this is due to an increase in stratification of the surface 720 Spatial Structure of Wind and SST in California layer, which has led to a shallower, hence lower nutrient, source of upwelling. However their observations are restricted to south of 34"N. The results reported here show that regional differences in environmental variability have existed in the CCS over the last five decades, therefore it is conceivable that analogous differences in biological productivity have occurred as weU. Specifically Roemmich and McGowan (1995) note a correlation between declining zooplankton biomass and increasing near-surface temperature and stratification. If this mode1 is correct, then could biomass be increasing in the norrhern portions of the CCS in association with the cooling tendencies of the COADS SST series? This cannot be imniediately concluded, since SST variability may be linked with other variations (e.g., mixed layer depth) that influence production more directly. Ir h,is been shown here that decadal-scale climate variability has considerable regional differences as well. Considerable eviaence suggests that environmental conditions in the northeast Pacific in recent years can be described in terms of two climte phases, separated by a transition occurring in about 1976. From the early 1940s to about 1976, the eastern north Pacii-ic was under the influence of increased equatomard stress (associated with a weaker than normal Aieutian Low in winier and a southward flowing jet stream), resulting in a cool pool of water along the entire North American West Coast, and a warm pool in the central north Pacific. Beginning in 1976, the Aieutian Low deepened and shifted eastward each winter, which altered large-scale wind patterns, Storm tracks and ocean currents. The resulting ocean thermal structure featured a warm water mass off central and southern California, but cooler water in the northern portion of the CCS and over much of the north subtropical Pacific (Fig. 10). Anchovies were the dominant small pelagic in the CCS prior to 1976 (MacCall, 1986); however sardines have shown evidence of increasing biomass since (Bames et al., 1992). The decadal-scale fluctuations of the two dominant pelagic fishits appear to be associated with the larger scaie variations in the CCS. From the mid-1920s to the early 1740s sardines dominated the entire CCS. This population collapsed during the 1950s and did not show any evidence of recovery until aftei. the environmental shift in 1976; since then the population has shown a very large increase in its nite of growth (Mai:Call, 1979; Barnes et al., 1992). Anchovies were not abundant in the southern region during the 19jOs, but the pop~~lation increased during the 1960s and reached a short term peak in the enrly 1970s. The California anchovy fishen peaked in 197j and then declined sharply; whereas the Mexican anchovy fishery expanded after the mid-1970s Peak in population, reached a maximum in 1981 and then totally cohapsed in 1990 (Ainley et al., 1993). These regional differences ma)- be linked to spatial differences in the rate and magnitude of environmental changes. Based on the fishery information available since the 1920s, it appears that the CCS sardine stock rises during warm water periods and the southern region's anci.iovy stock rises during cold water peiiods. Conditions prior to the early 1940s, when sardines dominated the CCS, were similar to those since 1976. Baumgartner et al. (1992) analyzed anaerobic sediments off southern California and concluded that altemating dominance between sardines and anchovies has occurred back to at least 300 AD. Aithough they are similar in many other respects, it appears that the anchoiy and sardine utilize the CCS in quite different ways. Sardines, which are larger and more mobile than anchovies, migrate from the southern region into the central and northern regions for feeding and then return to the southern region to spawn. Anchovies, although they occur over the entire CCS, appear to be much more localized in their movements. They have separate genetic stocks in the northern and southern regions and, within the histoncal record, have not been very abundant in the central region. Thus it appears that anchovy populations may rise and fall based on differences in the environmental conditions in the three CCS regions; whereas the single sardine population may be affected by environmental conditions in any or al1 of the three CCS regions. On shorter tirne scales, studies in the CCS demonstrate that biological production decreases dramatically during ENS0 everits (McGowan, 1985). ENSOs also are responsible for poor salmon survival off the northwest U.S. (Pearcy et al., 198 j), redi~ced recruitment, growth and condition of groundfish (Lenarz et al., 1995), and dramatic range extensions for fish (Radovich, 1961). While ENSOs have a large-scale influence, it is shown here that their relative environmental signal is highly variable over the length of the CCS ecosystem. In addition, individual ENSOs appear to have unique characteristics of timing, intensity and extent, superimposed on their canonical signal (1957, 1983). The corresponding biological effects presumably Vary as well. It is important to note that we observed substantial anomalies in the CCS that are unrelated to these well-defined climate events (e.g., 1965-1975, Fig. 9). Nevertheless, these periods of unusual environmental conditions are likely to have significant consequences for the ecosystem. 'Warm' years, when conditions off California are similar to those during ENSO events despite the absence of an equatorial ENSO signal, are linked to poor recruitment of central California iockfisli, while 'cool' years feature enhanced recruitment (S.V. Ralston, Tiburon, NMFS, pers. comm.). Extreme year classes of several species of fish over large geographical areas tend to occur in association with unusual environmental conditions (Hollowed and Wooster, 1992). The results presented here clearly demonstrate the highly variable nature of the CCS environment in time and space, and argue against oversimplifying EBC climate change as a constant linear trend, or in terms of the climate record from a single location. The distinct latitudinal regionalization and cross-shelf variability of the CCS wind and SST fields has key implications for fisheries. For example, which time series or regions are more important in terms of defining a stock's environment? Regional differences also mean that widespread stocks, or stocks that are highly migratory over their life history, face a spatially heterogeneous changing climate. Widespread stocks also may display a very different long-term variability from species whose domain is limited to the homogeneous regions of the CCS described here. Fisheries scientists must evaluate the relative environmental differences in each region, as they pertain to the climate signal and its variability, and compare them to a species' distribution and behavior as a function of life stage, to fully understand the consequences of climate change on populations. State-space statistical models are applied to long environmental time series of monthly COADS northward wind stress and sea surface temperature (SST) from the California Current System (CCS) off the West coast of North Arnerica (22-48"N) for the period 1946-1990. The models estimate a non-parametric and non-linear trend, a non-stationary and non- deterministic seasonal signal, and an autoregressive (AR) term. The models are applied to long time series of SST from selected coastal sites as well, for comparison to the COADS series. Based on a visual and statistical comparison of the mode1 trend series, the CCS can be divided into three distinct geographical regions, which roughly correspond to the biological regions defined by Parrish et al. (1981) from mean meteorological and oceanic relationships and coastal fish reproduction patterns. The northern region (42-48"N) features a strong transition from strongly equatorward to poleward with distance north (compare 41°N, 43"N and 4 jON stress time series in Fig. 4). The mean stress north of 44"N is poleward and has become increasingly poleward over time (Table 3). The transition zone in wind stress has expanded southward over time, strengthening the zona1 gradient in pole~vard stress. The CCS north of 40°N features spatially uniform mean SST (Fig. 6), and SST trends show a series-length cooling tendency (Table 3). This region of uniform SST has expanded southward over time as well. 722 Spatial Structure of Wind and SST in California Ainley D., B. DeLong, S. Herrick, L. Jacobson, E. Konno,J. Lan- Lenarz W.H., D. VenTresca, W.M. Graham, F.B. Schwing and F. fersieck, M. Lowry, R. Parrish, C. Thomson, and P. Wolf. 1993. Chavez. 1995. Explorations of El Niiios and associated biologi- Draftfishery description for the Pacific Coastal Pelagic Spe- cal population dynamics off central California. Cal$ Coop. Ocea- cies Fishe?~~Ma?îagement Plan. NMFS, SWFSC, Admin. Rep., nicFish. Invest. Rep., 36: 106-119. Bakun A and C.S. Nelson. 1991. The seasonal cycle ofwind-stress curl in subtropical eastern boundary current regions..l. Phys. Oceanogr., 21: 1815-1834. Barnes J.T., L.D. Jacobson, AD. MacCalland P. Wolf. 1992. Recent population trends and abundance estimates for sardine (Sardi- nopssagax). California. Cal$ Coop. OceanicFish. Invest. Rep., 33: 60-75. Baumgartner T.R., A Sou tar and V. Ferreira-Bartrina. 192. Recons- truction of the history of Pacific sardine and northern anchovy populations over the past two millennia froni sediments of the Santa Barban Basin. California. Cal$ Coop. Oceanic Fish. Invest. Rep., 33: 24-40. CardoneV.J., J.G. Greenwood and M. Cane. 1990. On trends in the historical marine wind data..[ Clinzate, 3: 113-127. MacCa1lA.D. 1979. Population estimates for the waning Fats of the Pacific sardine fishery. Cal$ Coop. Oceanic Fish. Invest. Rep., 20: 72-82. MacCa1lA.D. 1986. Changes in the biomass of the California Cur- rent ecosystems. In: K. Sherman and L.M. Alexander (eds.) Valin- bilipand nzanagenzent oflalge nzarine ecoqIstenls. M Selec- ted Symposia 99, Westview Press, Boulder, CO: 33-j4. MacCa1lA.D. and M.H. Prager. 1988. Historical changes in abun- dance of six fish species off southern California, based on Cal- COFI egg and larva sarnples. Cal$ Coop. Oceanic Fish. Invest. Rep., 29: 91-101. McG0wanJ.A. 1985. El Nino 1983 in the southerri California Bight. In: W.S. Wooster and D.L. Fluharty (eds.). ElNino noith: El Nirio effects in the eastern subarctic Pacific Ocean. Univer- sity of Washington Press, Seattle, WA: 166-184. Ebbesnieyer C.C., D.R. Cayan, D.R. McLain, F.H. Nichols, D.H. Mendelssohn R. and C. Roy. 1996. Comprehensive Ocean Data Peterson and K.T. Redmond. 1991. 1976 step in the Pacific Extraction Users Guide. U.S. Dep. Coninz., NOAA Tech. ~PIenzo. climate: Forty environmental changes between 1968-1975 NOAA-TM-NMFS-SWFSC-228, h Jolla, CA, 67p. and 1977-1984. In: J. Betancourt andV.L. Sharp (eds.). Proc. 7th Ann. Pacific Clinzate Workshop, Cal$ Dept. of Wuter MillerA..I., D.R. Cayan, T.P. Barnett, N.E. ~raham and J.M. 0be- Resources, InteragenqiEc0l, stud. prog Tech, ~ep,, 26: 129- rhuber. 1994. The 1976-77 cliniate shift of the Pacific Ocean. 141. Oceanographi!, 7: 21-26. Hollowed AB. and W.S. Wooster, 1992, Variabilityof ocean Parrish R.H., C.S. Nelson and A. Bakun. 1981. Transport niecha- collditions and strongyear classes of northeast Pacifie fish stocks, nisms and reproductive success of fishes in the California Cur- ICESMar. Sci Synfp., 195 433-444. rent. Biol. Oceanogt:, 1: 175-203. Isemer H.-J. 192. Cornparison ofestimated and measured mari- Pearcy W.,J. Fisher, R. Brodeur and S. Johnson. 198 j. ~ffects of ne surface wind speed, In: H.F. Diaz, K. Walter and S,D. ~00- the 1983 El Niiio on coastal nekton off Oregon and Washingtoii. druff (eds.). Proceedings ofthe Intemational COADS Work- In: W.S. Wooster and D.L. Fluharty (eds.). ElNino north: ElNino shop, Boulder, Colorado, 13.1 j January 1992. U.S, Dep, Comm, effecfs in the eustel-n subarctic Pacqk Ocean. University of 142-1 58. Washington Press, Seattle, WA: 188-204. Kitaga~va G. and W. Gersch. 1984. A sliioothness pnors - state Quinn W.H.,V.T. Neal and S.E. Antilnezde May010.1987. El Niiio space modeling of time series with trend and seasonality..J Am. OCCU"enCeS OVerthe Pas[ four and a half centuries.J GeOPhlls~ Statist. Assoc., 79: 378-389. Res., 92: 14449-14161. fitagan$a G, and W. Gers&, 1988, Smoothness priors il, rime Radovich J. 1961. Relationships of some niaririe organisms of series. In: J, Spa/\ (ed.), Ba?~esj~n ana&is ofLinle selies and the northeast Pacific to n8ater tempentures: particularly during dynatnic nzodels. Marcel Dekker, N.Y. 1957 through 1959. Calif. Dep. Fish GanzeFish. Bull., 112,62 p. 724 Spatial Structure of Wind and SST in California Rotbmmich D. and J. McGowan. 1995. Climatic warining and tlie decline of zooplankton in the California Current. Science, 267: 13;j-1326. R0perT.C. 1993. High-latitude oceanic variability associatedmitli the 118.6-year nodal tide.J Geophlls. Res., 98: 4639-4644. Skillman J.B. 1993. ENSOforced 11ci1-iations ofthe sea siuface teriiperatureand adjustedsea leuel cilong the zuest conrt ofthe Un!ted States. M.S. thesis, Naval Postgraduate School, Monte- rey. CA, 92 p. Slu~z R.J., S.J. Lubke1,J.D. Hiscox, S.D. \Yrocclniff, R.L.Jenne, D.H. Jostiph, P.M. Steurer and J.D. Eliiis. 1985. Coinprehensiue Ocealz- At~irosphereData Set; Release 1. I\'OAAEn~~ironmental Researcli Lat~:~ratories, Climate Researcli Prooram, Boulder, CO, 268 p. Trc iberth KE. 1330. Recent obsenred interdecadal clirn~te changes in tie northem hemisphere. BI!![. ai lie^ Anleteor. Soc., 71: 938-993. Trenbertli K.E. and J.W. Hurrell. 1994. Decadal atmospheric- ocean variations in the Pacific. Clinl. Diin., 9: 303-319. Walker P.M., D.M. Newton and hW. Mantyla. 1993. Su!-face~lwter teinpemtzires, salinities anddensities at shore stations, United States \Vest Coast, 1992. University of California, San Diego, Scripps Institution of Oceanography, LaJolla, CA, SI0 Ref., 93 (18), 46 P. Woodruff S.D., R.J. Slutz, R.L. Jenne and P.M. Steurer. 1987. A Comprehelisive Ocean-Atmosphere Data Set. Bull. Aniei: iVlet~1: SOC., 68: 1239-1250. Wu Z. and R.E. Nen~ell. 1992. Tlie mind problem iii COADS aiid its influence oii the niater balance. In: H.F. Diaz, K. Wolter and S.D. \Voodruff (eds.). Proceedings ofthelnternational COADS Wolbshop, Boulder, Colorado, 13-15 January 1992. U.S. Dep. Comm : 189-200. F. SCHWING, R. PARRISH AND R. MENDELSSOHN 7 25 Freshwater Yields to the Atlantic Ocean: Local and Regional Variations from Senegai to Angola OKSTOM Bf'î528 Biirnako M ILI The total annual yield of freshwater flowing into the Atlantic Ocean from West and Central Africa is computed over the 19 51- 1989 period. Interannual variations are presented for 34 rivers. Water yields are computed by decades using a data base composed of 57 rivers, and also considers water originating from coastal areas. These coastal areas represent 20% of the 7.7 million km2 of total surface area studied and contribute approximately 25% of the water yield to the sea, due to high runoff coefficients. The total annual water yield to the Atlantic Ocean is about 2700 billion m3. The study of the interannual variations of water yield shows two periods, namelv wetter before 1970 and drier afterwards. During the period 1981-1989, total yield fell by 28% as compared to the mean during the period 19 51-1980. Le débit total d'eau douce arrivant dans l'océan Atlantique de l'Afrique de l'Ouest et Central est calculé pendant une période allant de 1951 à 1989. Les variations d'une année sur l'autre sont présentées pour 34 rivières. Les débits sont calculés par décennie en utilisant une base de données composée de 57 rivières, mais aussi en utilisant les débits des stations côtières. Ces zones côtières représentent 20 % des 7,7 millions de km2 de la surface totale étudiée, et en raison de leurs forts coefficients de débit, contribuent à 2 5 % des débits partant en mer. Le débit total arrivant dans l'océan Atlantique est de 2 700 milliards de m3. L'étude des variations interannuelles des débits montre l'existence de deux périodes, humide avant 1970 et plus sèche après. Pendant la période 1981-1989, les débits ont chuté de 28 % par rapport à ceux de 1951-1980. 1 NTRODUCTION On the coasts of the Atlantic Ocean from Senegal to Angola, and in particular along the Coast of the Guinea Gulf, fishing is an important social and economic activity for the region's populations. For many years researchers have sought to understand the reasons for stock variations of coastal fishes. Studies have focused on environmental factors such as salinity, wind, and coastal freshwater yields. From the mouth of the Senegal River, at the Senegalese-Mauritanian border, to the mouth of the Cunene River, at the Angolan-Namibian border, tributary rivers which flow into the Atlantic Ocean drain a basin of about 7.7 million km2 (Fig. 1). Data regarding this runoff have been recorded for 57 rivers. For 33 of these rivers, it is possible to use or extrapolate from annual series for the period 1951 to 1989, which will be studied. The surface area drained by these 33 rivers occupies 81% of the total surface area considered here, with the Congo river basin alone contributing 46% (Table 1). SENEGAL frica, with numbers presented in Table 1 7 28 Freshwater Yields to the Atlantic Ocean - -- RIVER STATION SENEGAL GAMBIE CASAMANCE CORUBAL FATALA KONKOURE KOLENTE ROKEL PAMPANA SEWA MAN0 ST PAUL ST JOHN CESTOS SASSANDRA BANDAMA COMOE BiA TAN0 VOLTA MONO OUEME NIGER BENOUE (NIGER) WOURI MUNGO SANAGA NYONG KIENKE NTEhl OGOOLE NïAtG.4 KOUILOU CONGO Bakel Gouloumbo Kolda Saltinho Amont Bindan Ailiaria Tassin Bumbuna Matatota Jaijma Mano Mines Mount Coffe St John Falls S3wolo Gaoulou Daboitié Mbasso Ayamé Alenda Seiichi Halcrow Tététou Pont de Savé Onitslig Makurdi Yabasi blundnlne EdCa Deliniie Ikibi i\')il.bessan Lnmbaréné Tcl!iba~iga Soundn Frnzznville BASIN SURFACE KM~ % OF TOTAL SURFACE 218 O00 2.83 42 200 0.jj 3 700 0.0 j 23 800 0.31 5 100 0.07 16 200 0.21 6 600 0.09 4 O00 0.0 j 2 400 0.03 6 900 0.09 5 500 0.07 21 400 0.28 11 400 0.lj 4 600 0.06 70 600 0.92 60 O00 0.78 70 500 0.92 10 O00 0.13 16 O00 0.21 394 O00 5.11 20 j00 0.27 23 600 0.31 1 100 O00 14.28 300 O00 (3.89) 8 200 0.11 2 400 0.03 132 O00 1.71 26 400 0.34 1 100 0.C1 26 300 0.34 203 O00 2.64 12 400 O. 16 j6 600 0.73 3 550 000 46.08 7able 1: The 34 rivers studied durino, tie period 1951-1989: stations, surface areas, and percent of basin surface compared to total surface area of 7.7 million km' (sec Fig. 1 for location of river basins, by number). 'rli:is, it appears ili.:!t for tlie 19% of [!le tût71 suifice, t!:e 'non-controllecl zones' of about 1.5 million km2, it 1s inipossible to generate time series of annu11 ii!r.nJ \i?lues for the period 1951-1989. 'The non-controlled zones are Found particularly along the coastal bsnds (Fig. l), wliere gauging stations are nrely located clue to the anomalies cf water Iieight mrnsurements caused by rides and sedimentation near river moutlis Honrever, in coastal regions, rainfall is often more abundant than on the continent. With only few exceptions, such rain exceeds 1500 mm per year, and even reaches 10 meters on average at the foot of Mount Cameroon at Debundsha Station. Using the example of the nvers of Cameroon, Olivry (1986) showed that in the zones of average annual rainfall greater than 1 jOO mm the runoff coefficients npidly increase up to 70% for rains of 3.5 meters per year. Due to these higher runoff coefficients, as compared to continental regions, these non-controlled zones represent a considerable yield of freshwater, which must be included in estimates to total yields. For Angola, we have not been able to gather information more detailed than the average runoffs presented in a reference book (Angola, 1974) published for the international decade of hydrology. Hence, for Angola, we only present water yields variations by decades. The freshwater yields to coastal manne waters directly influence salinity and retention of nutrients. Hence, variations of freshwater yields impact marine life. Numerous authors (e.g., Binet, 1983; Cavenvière, 1991) have reported correlations between variations in coastal fish stock sizes and the variations of freshwater yields. Since the creation of ORSTOM, 50 years ago, its Hydrologic Service, integrated today with ORSTOM's Department of Continental Waters (DEC), has been at the forefront of developing networks of runoff measures for numerous African rivers. Not surprisingly, results regarding freshwater yields were collected and compiled at ORSTOM, by the former Research Unit A7, now UR22. Thus, we can benefit frorn ORSTOM's extensive expenence in Africa, rnost notably from runoff results that represent years of uninterrupted work by generations of hydrologists working in Africa. However, in the interest of thoroughness, in addition to the data from ORSTOM, we have gathered runoff data from non-francophone countnes, namely Nigeria, Angola, Equatonal Guinea, Guinea Bissau, Ghana, Sierra Leone and Liberia. The runoff values used in this study come from different sources. For most of the rivers the yearly runoff values are taken from daily or monthly runoffs measured at gauging stations. Some gaps in certain series have forced us to reconstruct or extrapolate certain daily, monthly, or yearly values, or even to establish correlations between data from several stations or several rivers near the stations, for periods greater than ten years. In certain cases, we have been unable to establish correlations between different runoff data, and, thus, have had to establish rainfall/runoff correlations both on a rnonthly basis (at Saint John, Saint Paul, and Ogooue) or, more generally, on an annual basis. Details regarding such calculations are found in Mahé (1993). Time series are analyzed in two ways. First an annual representation is given for the period 1951-1989 (Fig.2 and 3). This concerns 33 rivers (Fig. 1, Table 1) as well as the Benoue, the principal tnbutary of the Niger River. This tributary represents the runoff of a vast hydroclimatic region which is very different frorn the rest of the Niger River basin. Secondly a decadal representation is given for the four decades studied, 19 j1-1960, 1961-1970, 1971-1980, 1981-1989 (only nine years). By chance, these decades correspond to four different climatic periods. Space is used in the same way as time. First, regarding the interannual study, the hydrologic units used are individual basins. Second, regarding the study by decades, we have grouped the rivers, as well as the non-controlled zones, into eight regional entities (Fig. 2, Table 2). Among these eight zones, two are individual basins of the largest rivers in Africa, as measured by runoff, narnely the Congo (second largest river in the world after the Amazon) and the Niger. 130 Freshwater Yields to the Atlantic Ocean SENEGAL NORDGOLF Equator Fig. 2: French names and locations of the eight maj esternlcentral Africa. The non- controlled zones are shaded in grey. REGION BASIN SURFACE % OF TOTAL NON CONTROLLED % OF TOTAL (103 m2) SURFACE SURFACE (103 m2) SURFACE SEh EGAL-FOUTA 590 7.7 230 3 O GUINEE 182 2.4 120 1.6 NO1üIGOi.F 900 11.7 210 2.7 NIGER 1 100 14.3 O O ADAMAOUA 355 4.6 2 15 2.8 EQlJATEUR 475 6.2 1 j0 2.0 CONGO 3 550 46.1 O O ANGOLA 551 7.0 551 7.0 n, non-controlled 2. VARIATIONS IN FRESHWATER YIELDS TO THE ATLANTIC OCEAN 2.1 . Interannuai variations 2.1.1 - Normalized runoffs Figure 3 illustrates variations of normalized mnoffs (= yearly/total mnoff) of 34 rivers for the period 1951-1989, following from left-to-right the geographic location of the river's mouths as one moves northward from the Equator. 'ille black rectangles correspond to a normalized mnoff greater than 1.05. The white rectangles correspond to a normalized mnoff less than 0.95. The rectangles with a dash indicate a normalized mnoff benveen 0.95 and 1.05. The choice of rivers depended on the length of the original data series, as well as on the size of the basins, and on their location on the Coast. Runoff of small coastal rivers were not reconstmcted; it can be assumed that their mnoff variations are similar to the overvall pattern presented here. Fig. 3: Normalized r 89, and corresponding hydroclimatic regions. The black rectangles corre les correspond to H < 0.95; and the dashes mean that H is between 0.95 an 732 Freshwater Yields to the Atlantic Ocean Tb.0 periods are distiguishable in Figure 3: before 1970, with high values and after 1970, with low values. This pattern is veiy visible to the west of the mouth of the Niger River, but becomes less clear to the south of the Sanaga River. In this regard, it is appropriate to mention that runoff variations are rarely the same for al1 rivers at the same rime. In fact, 1783 is th<.: only year where every values of normalized runoff are negative, while 1962 is the only year where al1 normalised ruiioffs are positive. Likewise, al1 rivers experienced a runoff minimum between 1771 and 1973. Also, during the last decade, the rivers in the Adamaoua (Sanaga, Mungo, and Wouri) region here displayed different normalized ninoffs than otlier rivers in the same area. Runoff variations are more similar to those of rivers in more tropical climates which flow frclm the Pampana River of the Guinean Mountains, to Senegal. Figure 3 also presents contrasts between nvers in the north and coastal rivers of the littoral of the Gulf of Guinea. For example, in 1958, runoff deficits were only observed to the northward of the Sewa River in northern Sierra Leone, at 10°N latitude. An almost completely opposite case occurred in 1968. In 1987 numerous regions displayed greater-than-average mi-ioffs, except for the Mungo, Wouri and Sanaga River basins in Cameroon, the Senegal-Fouta region, and the northern Giiinea region. Since 1970, the only rivers which have experienced at least four years of greater-than-average iunoffs were thi: Kouilou, the Nyanga, the Ogooue, the Kienke, and the Nyong in the Equator region; and more north\vards, the Berloue, the Oueme, the Mono, the Comoe, and the Sr. John Rivers. Finally, during the last decade, the only rivers which hal,e not been in continua1 deficit have been the equatorial rivers the Nyong and the Kouilou, 3s illustrated in Figure 4. Fig. 4: Years of occurrence of maximum and minimum annual runoffs for each of the 34 rivers studied from 1951 -1 989. Primary maximum and secondary maximum are represented by big and small black circles, respectively; primary minimum and secondary minimum are represented by large and small white circles, respectively. 2.1.2- Extreme runogs As illustrated in Figure 4, years of maximum and minimum runoffs for the period 1951-1989 are indicated respectively by black dots and white circles. In the Equator region, primary maxima are observed during 1980-1989 only for the Nyanga, the Kienke, and the Nyong. The Mungo and the Wouri present secondary maxima in 1982. Almost al1 rivers experienced a minimum between 1982 an4 1987, except the Sanaga (data extrapolated from rainfall), the Nyanga, and the Kouilou. The runoffs were so weak during the last decade in this study (1981-1989) that the minima of the 1970 decade are only secondary minima. In fact, during this last decade only the Sanaga and the Corubal expenenced primary minima. From the Equator region to the Nordgolf region, 1958 is a year of marked deficits, and of occurrences of primary minima (Kouilou, Nyanga, Mono, Oueme). Finally, minima for the Volta in 1965 and 1966 coincide with the filling of the Lake Akosombo, a reservoir. The maxima are divided between two decades, 1950-1959 and 1960-1969. For the Senegal Fouta region and for the Niger, the main period for maxima is the decade 1950-1959. In contrast, for the Equator and the Nordgolf regions from the Oueme to the Comoe, the maxima are concentrated in the decade 1960-1969. The years with the greater number of maxima are 1955,1963, and 1968. For purely informative purposes, and without drawing any definitive conclusions, we remark finally a progressive time lag in the timing of maxima from West to east. In the middle of the period of deficits of the last twenty years, we obseive that the Senegal and the Gambia have higher-than-average runoffs in 1974/75. This period of higher-than-average runoffs moves in 1975/76 to the Kolente and the Konkoure, in 1977/79 to the St. Paul and the St. John, in 1979/80 to the nonli coast of the Gulf of Guinea, and in 1981/82 to the Mungo and the Wouri. 2.2. Decada.1 freshwater yields variations This section of the study fulfills two objectives : - to synthesize the results from every river basins by grouping them into larger hydroclimatic regions ; - to integrare the contribution of non-controlled zones into our calculations of freshwater yields to the Atlantic Ocean We do not have complete runoff data for the period 1951-1989 for the non-controlled coastal zones. Nevertheless, we can obtain information on the order of magnitude of specific iunoffs for these zones from the results of studies undenaken periodically on small coastal rivers (24 in all, representing 5% of the total surface area). To estimate these lunoffs from non-controlled zones, the simplest solution is to first calculate an average ninoff value from known values for nearby rivers in the same zones, tlien apply annual coefficients of variation also derived from the same rime-series of runoffs of nearby rivers. The problem with this method is th3t it does not consider specific coastal runoffs and rainfall variations. Tlie solution that we have chosen for reconstructing runoff time series, taking into consideration coastal specificity, was inspired by Olivry (1986) mho, after studying rainfall and runoff of the rivers of Cameroon, proposed a ielationship between total annual rainfall and specific ninolf values. This relation demonstrated a correlation between rainfall and specific runoffs, applicable to cases of annual ninfall of at least 1 j00-2000 mm/year. Tlie transposition of Olivry's relation to regions other than those of the Cameroon coast is problematic, due to differences in climatic conditions. On the Cameroon coast, the climate can be classified as equatorial transition, witli a small dry 734 Freshwater Yields to the Atlantic Ocean season and signifiant cloudiness. We use this relationship for the coastal zones from the Guinean Mountains to the Gabclnese and Congolese coasts. In these littoral zones, rains reach and often exceed 1500-2000 mm/year, with the length of the dry season as the only factor that makes such zones different from the Cameroon coast. The conditions for optimal utilization of the relation between rain and runoff are thus largely met, and specific runoffs can be deduced from the relationship described by Olivry. Then the runoff is calculated from the surface of the areas. Hence, Olivry's relationship requires calculation of average rainfall on the surface area of the non-controlled zones. The annual rainfall values for each zone during the period 1951-1989 are calculated automatically through spline interpolation, based on data on annual rainfiill at about 900 rainfall stations covering the total surface of the basins under study. Details of the method are described in Mahé and L'Hôte (1992) and Mahé et al. (in press). Thus, we have precipitation time-series, wliich, with Olivry's equation, can be transposed into time-series of annual runoff, with the specific variability of coastal precipitation being presemed in the constructed data series, as well as the nuances of specific runoffs along the coast. However, the accuiacy of the calculation of annual runoff from these zones obtained through graphical methods and the accuracy of the runoi'f coefficients (ratio of runoff to rainfall, in percentages), are not identical for the past four decades of study (Mahé, 1993 I. Specifically, annual runoff values from non-controlled zones are only qualitative, and runoff averages by decades are mucli more reliable. For each hydroclimatic region we can present freshwater yields and estimated freshwater yields from non-controlled zone:;. Figure 5 displays interannual variations of annual runoff, by hydroclimatic regions, with average runoffs by decades represented with thin horizontal lines. For Angola, due to the lack of information concerning runoffs, it is not possible to recorkstruct annual values. During the last decade, the runoffs of the Senegal-Fouta and the Nordgolf regions were more than ilialf as weak than dunng the decade 1950-1959. For the Niger and the Guinea regions, the decrease is only one-thii-d. It is also noticeable in the Adamaoua region, but much less perceptible in the Congo Basin and the Equator region, where, howcver, vely strong runoffs are obsemed dunng the 1960s. To il.ustrate the variations of freshwater yields, Table 4 presents the volumes of freshwater yields by decades and by region. As illustrated, each year approximately 2.7 1012 m3 of freshwater enters the Atlantic Ocean between Senegal and Angola. About half of this amount comes from the Congo River. The decrease in freshwater yields during the last decade is considerable, with a 28% annual average decline dunng 1981-1989 as compared to 1951-1980 This decline is also not equal throiighout Central and West Africa. Specifically, in Central Africa the decrease in freshwater yields is small in the Adan,aoua region, and very small in the Equator region and in Angola, where periods of drought affected only slightlv the runo!k. In West Africa, the decline in freshwater yields becomes greater as one advances from Cameroon to the northwest (Guiriea, Mali, Senegal, Mauritania). During the decade 1981-1989, the yields from the Senegal-Fouta region fell by more than half in comparison to their average value during the 1951-1960 decade. As a result, runoff changed, with a fa11 in grourld water resemes, and in river recharge during dry seasons. This phenomenon is the intensification of depletion, as described by Olivry et al. (1993). One i.xample should suffice to illustrate the observed magnitude of the runoff deficits. In 1983, the year of maximum rainfa tl decline both during the 1951-1989 period and since the beginning of the century (Sircoulon, 1989); for al1 regions, freshwater yields to the Atlantic Ocean have been 34% less than the average from 1951-1989, which represents a decline of about 900 billion m3. This value corresponds to total annual runoff of al1 of West Africa from Senegal to the Cameroon MOUP tains in a normal year. Guinea t 2 2 50 Zaire i=: 2 400 2 350 130 Adamadoua 135 & 120 I: 125 ,E IIO + 2 IO0 rF '15 w 2 IO5 t 90 - 80 & 95 O 2 70 $ 85 60 75 Y ear Year Fig. 5: Annual runoff by hydroclimatic region, in rn3s-1, , (Angola not included). Decadal rneans are indicated by horizontal Iines. For each region, the runoff values represent the surn of yields in controlled and non-controlled zones. 736 Freshwater Yields to the Atlantic Ocean MEAN ANNUAL MELD (m3 109, BY PERiOD REG11 IN 1951 - 1960 1961 - 1970 1971 - 1980 1981 - 1989 MNN SENEGAL - FOUTA 127 119 73 62 96 GUIKEE 2 78 2 69 211 165 233 NOHIGOLF 173 171 Il j 98 140 NIGER 2 14 211 170 146 186 mMIAOUA 294 300 241 22 j 266 EQUATEUR 321 362 303 312 325 CON( ;O 1280 1530 1310 1200 1335 ANG( )L4 87 94 80 83 86 TOTN. -. 2780 2870 2 500 2290 2670 ield to the Atlantic Ocean, by hydroclimatic region and decade, in billions of rn3 89 are presented in the right column. ()f the 2.7 1012 m3 of freshwater reaching the Atlantic Ocean each year from West and Central Africa, approximatel!~ half, I .e., 1.34 1012 m3, come from the Congo River. The freshwater yields from the non-controled zones (19 % of the total surface) where rainfali is stronger and hydrologic studies rare, represent about 25% of total runoff to the ocean. Estimation of this part of the runoff, which has not been previously calculated, constitutes an addition to previous estimates. In relation to the average runoff for the period 1951-1989, maximum runoffs occurred between 1951 and 1970. Since 1970, fresi- water yields have decreased. This study constitutes a first synthesis of freshwater yields on the Atlantic side of the African continent. Results presented, however, can be further refined. For example, complementary information on runc,ffs from non-francophone countries such as Angola and southern Nigeria would improve on the results of this study. \Ve would like to thank the National Hydrologic and Meteorological Service of West and Central Africa for assistance with the creation of our data base. Angola. 1974. Collectanea. de estudos hidrologicos. Direcçao- General de Obras publicas e Communicaçoes. Grupo de tra- balho para O decenio hidrologico internacional. Lisboa, 400 p. Binet D. 1983. Phytoplancton et production primaire des régions côtières à upwellings saisonniers dans le golfe de Guinée. Océa- nographietropicale, 18 (2): 331-355. Caverivière A. 1991. L'explosion démographique du baliste (Balistes carolinensis) en Afrique de i'ouest et son évolution en relation avec les tendances climatiques.ln: P.Cury et C.Roy (eds.). Pêcheries ouest-africaines: variabilité, instabilité et changement. ORSTOM: 354-367. Mahé G. 1993. Les écoulementsjluuiaux sur la façade atlan- tique de L'Afrique. Etude des élénzents du bilan hvdrique et variabilité interannuelle, analyse de situations lydroclinza- tiques niovennes et extrêit~es. Thèse, Université Paris-Sud- Orsay/ORSTOM. Coll. Etudes et Thèses, ORSTOM, 438 p. Mahé G. andY. L'Hôte. 1992. Utilisation de la nzéthodedu uec- teur régionalpour la descliption des uatiationspluuiométri- ques interannuelles en Afrique de l'ouest et centrale de 1951 à 1989. Huitièmes Journées hydrologiques de I'ORSTOM, hlont- pellier, Septembre 1992,14 p. Mahé G., F. Delclauxand A. Crespy. In press. Elaboration d'une chaîne de traitement pluviométrique et application au calcul automatique de lames précipitées (bassin versant de l'Ogooué au Gabon). k!vd~-ologie continentale, 9 (2). Olivry J.C. 1986. Fleuueset riuikesdu Canzeroun. Thèse d'Etat. MESRES-ORSTOM. Coll. Monographies hydrologiques ORSTOM, Paris, 9,733 p., 360 tab., 2 cartes. Olivry J.C.,J.P. Bricquet and G. Mahé. 1993. Vers un appauvris- sement durable des ressources en eau de I'AFrique humide '112 : J.S. Gladwell (ed.). Hydrology of warm humid regions. 4ènie Assemblée AISH, Yokohama, AISHPubl., 216: 67-78. SircoulonJ. 1989. Bilan hydropluviométrique de la sécheresse 1968-84 au Sahel et companison avec des sécheresses des années 1910 à 1916et 1940 à 1949.h: Leshotilmes faceawrséche~.esses. Nordeste brésilien. Sahel africain: 107-1 14. 138 Freshwater Yields to the Atlantic Ocean Spatial Dynamics of the Coastal Upwelling off Côte-d'Ivoire ANGORA AMAN SIAKA FOFANA [lépartement de Physique de la Faculté des Sciences et Techniques tle l'université d'Abidjan (FAST) fiP 582 /\bidjan 22 (:ÔTE-D'IVOIRE Coastal SST measurements and images derived from Meteosat Satellit TIR channel are used to studv the spatio- temporal evolution of the minor upwelling off Côte-d'Ivoire. The TIR images derived from Meteosat provide a comprehensive view of the spatial dynamics of the upwelling off Côte-d'Ivoire. The cooling water was detected by Meteosat sensors because there is generally a clear sky situation and a sufficient spatial extension of the upwelling each time it takes place. The use of the Meteosat data seems to be sufficient to localize the surface cooling off Côte-d'Ivoire. The minor upwelling appears to be more persistent in the western side than in the eastern side off Côte-d'Ivoire. In spite of its local aspect, the use of the coastal SST measurements can be considered as a good tool to detect the presence of the coastal upwelling. Les mesures de SST côtières et les images dérivées du canal IR du satellite Météosat sont utilisées pour étudier l'évolution spatio-temporelle du petit upwelling devant la Côte-d'Ivoire. Les images TIR dérivées de Météosat donnent une image interprétable des dynamiques spatiales des upwellings devant la Côte-d'Ivoire. Les refroidissements sont détectés par les capteurs de Météosat parce qu'en général un ciel clair et une extension suffisante au large se produit. L'utilisation de Météosat apparaît satisfaisante pour localiser les zones de refroidissement devant la Côte-d'Ivoire. Le petit upwelling apparaît plus persistant dans la partie ouest que dans la partie est de la Côte-d'Ivoire. En dépit de son aspect local, l'utilisation des mesures de SST côtières peuvent être considérées comme un bon moyen pour détecter la présence de I'upwelhg côtier. 1 NTRODUCTION The temperature of the uppermost layer of the ocean determines the heat content and affects biological activity. Observed large-scale and persistent anomalies in sea surface temperature are of great interest in the context of global climate change monitoring (McClain et al., 1985; Bakun, 1990). Sea surface temperature (SST) is also known to affect the spatial distribution of marine species ( Brown and Winn, 1989). Thus, knonrledge of the spatio-temporal evolution of SST can be related to annual indices of the abundance of fishes and to the impact of climate changes on the spatial distribution of pelagic fishes (Mendelssohn and Cury, 1989; Pezennec and Koranteng, this vol.). Coastal SST measurements have sho~vn that, twice a year, there is a moderate, then a great decrease of the SST between January and March (minor upwelling) and between July and September (major upwelling) along the coasts of Côte-d'ivoire and Ghana. Upwellings-induced plankton production maintain large stocks of pelagic fishes. Indeed, there has been an increase of catches of Sardinella au& in Côte-d'ivoire and Ghana dunng the last decade and a new spatial and seasonal distribution of the stock (Binet and Servain, 1993; Pezennec and Bard, 1992). Generally, the large-scale behaviour of coastal upwelling can be better observed from geostationary satellites imagery than by coastal SST measurements. Satellite data provide uniform and continuous coverage of the SST when there is a clear sky situation. The Meteosat data derived from the Thermal Infra Red (TIR) channel have been shown to be appropriate for these studies because they provide the kind of spatial and temporal coverage required for fisheries related investigations. Since 1991, the Meteosat high resolution transmission can be received by a Pnmary Data User Station (PDUS) installed at the University of Abidjan (Côte-d'Ivoire). This ensures that we receive regular half hourly Visible, Water Vapor and Thermal Infrared images data with a spatial resolution of 5km for the TIR channel, only a few minutes after they have been scanned. In its initial form, the PDUS system has been installed to satellite estimation of rainfall over Côte-d'Ivoire. However, images data derived from TIR channel have been archived for the purpose of sea surface temperature monitoring. In the present paper, coastal SST measurements ancl TIR images from Meteosat are used to study the spatio-temporal evolution of the SST off Côted'Ivoire. This preliminary study is focused on the minor upwelling of the year 1993. 7 40 Coastal Upwelling off Côte-d'Ivoire 1.1. General climatology along the coastline of Côte-d'lvoire A knowledge of the Ivoirian and Ghanaian coastal upwelling mechanism requires to study the climatology of the Gulf of Guinea and to pay attention to the local climate of Cape Palmas and Cape Three Points, because it is recognized that surface cooling is strongest on the eastem sides of these two capes (Ingham, 1970). The ciimate of South Côte-d'Ivoire and Ghana is charactenzed by four seasons: a) The long dry season, which begins in December and ends in March; b) Th:: long rainy season, which starts in May and lasts three months, with rain Storm events from April to May; c) Au {ust and September, which correspond to the short dry season; d) October and November, which correspond to the short rainy season. Accorijing to Cautenet (1979), the precipitations are more important at Axim (1969 mm) than at Takoradi (1068 mm) and at Ac( ra (659 mm), whatever the season. Axim and Takoradi are j6 km faraway and are symmetrical in relation to Cape Thret~ Points (Fig. 1). The same trend is also observed at Tabou near Cape Palmas. The effective level of rainfall records at Taboii is about 2100 mm/year. The levels of rainfall recorded at San Pedro and Fresco are less important than the levels of rainfal at Tabou. 6 4 2 Longitude ("W) Fig. 1 : The study area, along the coasts of Côte-d'Ivoire and Ghana. A. AMAN AND 5. FOFANA 7 4 7 In these two cases, there is a decrease of the level of rainfalls in the eastern sides of the capes (Cautenet, 1979). The monsoon is the main provider of precipitations during the long rainy season along the Gulf of Guinea, while the sea breeze circulation is the only factor which determines the precipitations during the dry season (Cautenet, 1979). The sea breeze circulation is due to the difference (AT) between the air temperature on the ground and the sea surface temperature. During the minor upwelling, AT is positive and favours the development of the sea breeze circulation. Thus, SST can be considered as an important factor for determining rainfall during the minor upwelling period. 2. THE DATA SET AND THEIR PROCESSING We have two data sources for this preliminary study: a) The five coastal stations sampled by the Centre de Recherches Océanologiques (CRO) of Abidjan. These stations are located along the open Ivoirian gulf (Fig. l), from Tabou (near Cape Palmas) to Assinie (near the border of Ghana). The available ground SST measurements are carried out from January to March 1993; b) The Meteosat data, which cover the Coast of Ghana and Côte-d'Ivoire for the year 1993. These data consist of TIR images over the area from 10%' to 0" and 0" to 10°N (from Cape Palmas to Takoradi). The TIR images acquired cover the two upwelling periods. Scanning of the earth nominally takes place every half-an-hour, providing images in al1 three spectral channels. The infrared and water vapour images are composed of 2500 lines, each of 2500 pixels, whilst the visible image consists of 5000 lines of 5000 pixels. The corresponding spatial resolution at the sub-satellite point are 5 km and 2. j km, respectively. The TIR images used for this study are from the formats AI and AIVH. The latter is a combination of visible and infrared images. In the AMI format, visible channel data is reduced to the spatial resolution of the infrared channel. The Thermal InfraRed image, the administrative message and the calibration coefficients are extracted from the raw data. The raw TIR is converted to radiance image by using: Radiance = (NC - SPC) x Ircal. (1) where NC is the numencal count, and SPC and Ircal are calibration coefficients disseminated with the raw image The radiance data, in the form of digital counts for each pixel, are produceci by transcription to temperature through transcription table (inversion of Planck's Law). The spatial resolution of the images is 5 km and the temperature resolution is O. j°C. 47 digital TiR half- hourly images are received, processed and archived every day by the Department of Physics of the University of Abidjan. A daily synthesis image is obtained from the 47 half-hourly images of temperature brightness maxima. The presence of cloud cover leads to a severe reduction of our ability to observe daily oceanic structures. So, for every period of five days, a period synthesis of temperature brightness maxima is routinely carried out. No correction for atmospheric attenuation is carried out, but the method of processing the synthesis image minimizes this attenuation effect. In our study, the image of January 12th was used to hide the land. Meteosat images data from January lst, 1993 to September 31th, 1993 have been analyzed at the time of writing. 7 42 Coastal Upwelling off Côte-d'Ivoire Table 1 presents the percentage of cloudless daily images (56% of total) archived from January to March. The observational stiidies based on Meteosat TIR images have shown that the diurnal cycle of convective activity can be described by a beginning of cloud growth around 12h and a maximum of cloudiness 5 or 6 hours later; this process occurs in similar fashion along the West African coastline, from Côte-d'Ivoire to Nigeria. - Loi ation Tabou San Pedro Fresco Abidjan Assinie Clciudless images (%) 57 53 60 56 53 lable 1 : Pcrcentage of cloudless images from January 1993 to March 1993 for the five coastal stations. Figure 2 shows the temporal evolution of ground SST measurements in January 1993. This figure documents a decrease of the SST from January 4 to January 9, 1993 for the al1 sites studied, except Assinie, where there was a weak increase of SST dcring t his same penod. 22 J 1 1 6 11 Days 16 21 26 3 1 Fig. 2: Temporal evolution of ground SST rneasurcments in January 1993 at different coastal stations off Côte-d'ivoire. A. AMAN AND S. FOFANA 143 Fig. 3: Daily synthesis images derived from Meteosat TIR Channel off Côte-d'Ivoire and Ghana, January-July 199 3. 7 44 Coastal Upwelling off Côte-d'Ivoire in order to better detect and analyze the spatial extension of this phenomenon, three daily SST images derived from Mcteosat TIR channel were carried out from January 7 to 9 (Fig. 3a,b,c). They confirm the cooling observed with the coastal SST measurements and show that the eastern sides of the two capes were more affected by the decrease of the sea surface temperature than the western sides. The coastal SST during March 1993 at Assinie and Tabou (Fig. 4) show a global decrease from March 6 to March 11. The minimum SST is observed on March 11. The relative decrease of the SST was not otiserved on the corresponding TIR images because of the strong nebulosity dynamic during this period. The same figures show another decrease of the sea surface temperature during the fourth five-day period of March. The cooling trend sel-ms more persistent at Tabou than at Assinie. The analysis of the SST images corresponding to March shows that the second minor upwelling took place during the fourth five-day period of March. The daily synthesis image of Marc11 23 (Fig. 3cr) was selected to illustrate this phenomenon, located in the western coast of Côte-d'Ivoire, from Tabou to Grand Lahou ak-)ng 2i3Ci km approximately. The observed sea surface temperature was about 23°C. 29 T Tabou 1 6 II 16 21 26 Date in March 1993 3.2. Major upweiiing In contrast to the minor upwelling, the major upwelling is not difficult to observe, as it lasts approximately 3 months. Thus, it is straightforward to observe this phenomenon with Meteosat. However the occurrence of cloud is important di~ring this period; it is quite common to have four or five consecutive cloudy days. 'Rie available Meteosat SST images reveal that in 1993, the major upwelling effectively began on July, or even at the end of June. It was located al1 along the coast of Côte-d'Ivoire and Ghana and its offshore extension reached to 2"s. Figure 3e stiows the clearest five-day period synthesis image for July. 3.3. Cumulative negative thermal anomalies on the coastal stations: case of the minor upwelling A computation of the cumulative thermal negative anomalies (TJ was carried out from January to March in relation to tkie monthly mean SST value. A. AMAN AND 5. FOFANA 745 For a given site, Ta is obtained by: where Tm represents the monthly mean SST value and Ti the daily SST value, with Ti < Tm Four sites have been chosen: Assinie and Abidjan (eastern coastal stations) and San Pedro and Tabou (western coastal stations). Ta gives an indication on the intensity of the upweiling (Fig.5). Two patterns were observed: a) The cumulative index of negative anomalies observed dunng the study penod is relatively high for the eastern coastal stations (Fig. j). 'ihere is a weak decrease of Ta in February and March; b) In the western stations, the cumulative total index of negative anomalies is relatively important in January and Marc11 and very weak in February. Fig. 5 shows that Tabou is the only station affected by the cooling water in March. These patterns show that there is a spatial variability of the SST between Assinie and Tabou. In the West of Côte-d'Ivoire, there is a global trend of increase of the SST between the two periods of upwelling. In conclusion, the net cold ocean negative anomalies observed in January and March are in agreement with the temporal evolution of coastal SST measurements (Fig. 2 and 4) In this paper, coastal SST measurements and images derived from Meteosat TIR channel are used to study the spatio- temporal evolution of the minor upwelling off Côte-d'Ivoire. This study shows that: a) in spite of its local aspect, the use of the coastal SST measurements can be considered as a good parameter to detect the presence of the coastal upwelling. The TIR images derived from Meteosat provide a comprehensive view of the spatial dynamics of the upwelling off Côte-d'Ivoire. However, the cooling water has been detected by Meteosat sensors because there is generally a clear sky situation and a sufficient spatial extension of the upwelling each time it takes place. The use of the Meteosat data seems to be sufficient to localize the surface coolings off Côte-d'Ivoire; b) the minor upwelling is more persistent in the western sides than in the eastern sides off Côte-d'Ivoire; c) there is a need for collecting data along the two capes from January to March for a better knowledge of the decrease of sea surface temperature as these play an important role in determining the minor upwelling. For Future investigations, it should be required to analyze the impact of SST on rainfall dunng the dry season, when there is an important development of convective clouds in the study region; this may be very useful for local and global climate monitoring. 146 Coastal Upwelling off Côte-d'Ivoire Assinie Abidjan San Pedro Tabou Ttie present study was funded through ORSTOM, by the Centre de Recherches Océanologiques and the Département de Physi,.1ue de la Faculté des Sciences et Techniques de l'université d'Abidjan. We also wish to acknowledge the important contnbutions of H. Demarcq (UTISDakar), who provided most of the algorithms used for the restitution of the sen surface tempc:mture. Bakun A. 1990. Global climatic change and intensification of coasral ocean upwelling. Science, 247: 198-201. Binet J. and Servain J. 1993. Have the recent hydrological changes in the northem Gulf of Guinea induced the Sardinella auda outburst? Oceanologica Acta, 16 (3): 247-258. Bro~~n C.W. and H.E. Winn. 1989. Relationship between the distr:bution pattern of right whales, Eubalaena glacialis, and satellite-derived sea surface thermal structure in the Great South Channel. Cent.-Shelf-Res., 9(3): 247-260. Cautenet S. 1979. interaction d'une circulation méso-échelle de brise de mer avec un cisaillement de vent synoptique. Appi'ication au golfe de Guinée. Thèse de Doctorat, Univt:rsité Biaise Pascal (Clermont-Ferrand II). Ingham M.C. 1970. Coastal upwelling in the northwestern Gulf of Guinea. Bull. Mar. Sci, 20(1): 1-34. McCiain E., W.G. Pichel and C.C. Walton. 198j. Comparative performance of AVHRR-based multichannel sea surface temperatures.J Geoph. Res., 90 (C6): 11,587- 11,601. Mendelssohn R. and P. Cury. 1989. Fluctuations of a fortnightly abundance index of the Ivoirian coastal pelagic species and associated environmental conditions. Gan. J. Fish. Aquat. Sci., 44: 408-428. Pezennec 0. and F.X. Bard. 1992. Importance écologique de la petite saison d'upweiiing ivoireghanéenne et changemenrs dans la pêcherie de Sardinella aulita. Aquat. Liu. Res., j(4): 249-2 j9. A. AMAN AND S. FOFANA 147 Spatial and Temporal Dynamics of the Upwelling off Senegal and Mauritania: Local Change and Trend HERVÉ DEMARCQ OR'>TOM Laboratoire Halieutique et Ecosystèmes A~L atiques BP ~1045 34032 Montpellier cedex 1 FRA' JCE A specific processing cl-iain applied to the infrared data of the Météosat series satellites lias been elaborated. The higli repetitiveness of the observations allows to obtain j days synthesis SST maps over West Africa at a 6 km resolution. The resulting data set precisely describes the spatio-temporal dynamic of the coastal upwelling, from hlauritania to Guinea (21°N-9"N), since 1984. The study area is dominated by the seasonal coastal upwelling which displays important intennnual vartations. The study of the superficial thermal field structure enables to link the mean position of the upwelled water to the topography of the continental shelf. Continuous monitoring of SST along the shelf allows the spatial estimation of an upwelling index and to characterize the seasonal dynamic of the upwelling via parameters such as the intensity and the duration of the seasonal transition phase and its seasonal lags. The example of an anomalous migration of Snrriina pilchardus in Senegal leads to the hypothesis that neither the mean seasonal intensity nor the precocity of the upwelling is sufficient to initiate an abnormal southward migration and that the seasonal transition leads may be a key parameter in this process. L'élaboration d'une chaîne de traitement spécifique aux données infrarouge thermique des satellites de la série METEOSAT a été réalisée. L'abondance des données satellitales liées à la répétitivité des observations permet d'accéder à une résolution de 5 jours et 6 km, soit très en deçà des méthodes d'investigation utilisables en routine en océanographie côtière. La description spatio-temporelle précise et régulière de I'upwelling côtier de la Mauritanie à la Guinée est ainsi accessible depuis 1984. La zone étudiée est soumise à la très forte saisonnalité de l'upwelling côtier, lequel présente aussi de fortes anomalies interannuelles. La connaissance de la structure des champs thermiques superficiels permet de relier la position moyenne des zones de remontée à la topographie du plateau continental. I;i détermination de la TSM sur une bande continue centrée sur le maximum de résurgence permet le calcul d'indices d'upwelling spatialisés et de caractériser la dynamique saisonnière de I'upwelling à travers des paramètres tels que : intensité et durée mais aussi décalages saisonniers et progressivité des transitions saisonnières. A partir de l'exemple de Sardina pilchardus au Sénégal, on émet l'hypothèse que ni l'intensité saisonnière moyenne, ni la précocité de I'upwelling ne permettent d'expliquer à eux seuls certaines migrations exceptionnelles de cette espèce, mais que la dynamique des transitions saisonnières semble également déterminante. Remote sensing of sea surface provides synoptic and repetitive overviews, especially for large scale monitoring of climatic parameters. At a lower space and time scale, satellite infrared imagery allows satisfactory obseivation of coastal areas. Due to their low cloud coverage, coastal upwelling areas may be particularly well monitored via sea surface temperature (SST) mapping, at a time and a space scale adapted to their particularly high dynamic. A specific data processing chain has been developed from METEOSAT infrared data and ships of opportunity data (Citeau and Demarcq, 1990; Demarcq and Citeau, 1995) and tested in West African upwellings. The upwelling zone studied extends from North Mauritania to Guinea and corresponds to the seasonal zona1 displacement of the trade winds along the western African Coast. Directly depending on this dynamic, the seasonal variability of the SST reaches 14°C (Rébert, 1983), and is one of the largest in the world. The high pressure regime of the northern anticyclone which govems NE trades leads to weaker cloudiness (and permits better remote sensed 7 50 Dynamics of the Upwelling off Senegal and Mauritania otiservations of SST) during the cold season (October to June, depending on the latitude). The enrichment of these coastal areas depends on both intensity and variability of the corresponding upwellings. Irriportant fluctuations of pelagic fishes population abundance and particularly of Sardinella species, a major resource for Senegal, have been recorded, in spite of the ability of these species to tolerate some environmental fluctuations (Fréon, 1988; Cury and Fontana, 1988). The irregular presence of species depending of geographically neighboring stocks (as for Sardinapilcbardus) is noticeable and may also be related to environmental fluctuations. 1. THE SPEClFlC DATA PROCESSING FOR SST RETRIEVAL FROM METEOSAT INFRARED IMAGERY In terms of radiometric and spatial resolution, the accuracy permitted by geostationary satellites (O.j°C and 5 3 5 km subsatellite in the case of METEOSAT infra-red channel) is lower than the accuracy currently obtained from polar orbital satellites (0.12"C and 1 km for NOWHRR). Nevertheless, this lowest resolution is not really a coiistraint, even in coastal areas, if conipared with the size of the oceanic structures observed at sea level, on the one hajid, and with the strong thermal gradients encountered, on the other hand. On the contrary, the regularity of the earth scan provided by METEOSAT allows a simpler processing for geometric corrections, while its repetitiveness (el ery 30 minutes) allows improvements in discriminating the sea from clouds. Data pre-processing takes advantage of the half-hourly availability of earth scans by METEOSAT. Each satellite vie\v of the earth is classically calibrated (transformation of the energy emitted by the earth to temperature by irnr'ersion of Planck's Law). An extraction of the working area is then performed and the image is geometrically cor.rected to a linear latitude and longitude projection. In iropical areas, the infra-red radiance measured by the satellite sensor is systematically lower than the infra-red rariance emitted by the sea surface (except in the presence of suspended dusts), due to cold atmospheric water vapor. Consequently, by assuming that the SST is constant over 24 hours, the 48 images of the day are combined into a nt:w image synthesis, retaining for each pixel the 'warmest' one of the time series. Cloud cover in West Africa may strongly Vary during one day, especially when the trade winds are weak. The efficiency of the 'maximum temperature method' is shown for 27 days, from 5 to 31 May 1991 (Fig. 1) by comparing the insrantaneous cloud cover at 12h00 GMT (generally low cloud cover) and the daily synthesis index. The advantage of the repetitiveness of observations by a geostationaty orbit appears clearly. This important benefit in term of usable pixels for SST retrieval will also determine the performance of the sea-cloud discrimination, the major step of the processing in SST restitution. Fig. 1: Reduction of the cloud cover (%) on the daily thermal synthesis (solid line; crosses) cornpared to the instantaneous cloud cover (dotted line; open dots) at 12 ho0 CMT. . m 5 7 9 11 13 15 17 19 21 23 25 27 29 31 Day (May 1991) In remote sensing processing, sen-cloud discriminations are very often based on visible/infra-red comparisons. Nevertheless, this technique lias some constraints. The major one for acquisition and processing is the large amount of data required, five rimes more in the case of METEOSAT. Furthermore, some low level clouds are strongly absorbent in the IR channel and rernains transparent in the VIS one. The visiblehnfrared algorithm is then unusable. The method we developed for sea-cloud discrimination is based on a comparison of a daily synthesis with the 'most probable' real SST field. This field is provided either from a climatology of SST or, more often, frorn a previously processed SST field. For adequate masking, this reference situation is chosen as close as possible to the daily synthesis, in terms of upwelling spatial extent. A comparison of the radiative temperature synthesis (Fig. 2a) with this reference is then performed and the values with temperature deviation greater than a definite threshold (around 3°C according to the similarities of both fields) are considered to refer to clouds, and are masked (in black on Fig. 2b). 1.3. Atmospheric correction and SST restitution The above resulting temperature field remains affected by atmospheric absorption, mainly due to the atmospherical water vapor. In tropical area, the apparent thermal absorption generally lies between 2°C (trade wind region) and j°C or more (equatorial region). According to the previous pre-processing steps (maximum ternperature synthesis), and considering the difficulty to obtain direct information on atmosphere structure compatible with space and rime resolution of the SST fields (6 km and j days in Our case) the most adequate way to correct this temperature field from the atmospheric absorption is to use an exogenous source of SST data. The ships of opportunity data disseminated by the Global Transmission System (GTS) in the 'SHIP' meteorological messages (including SST, wind, air temperature, etc.) and synthesized in - 152 Dynamics of the Upwelling off Senegal and Mauritania Longitude ("W) Longitude ("W) Fig. 2: Raw daily infrared synthesis on October 20, 1994 (a) and after cloud rnasking (b) in the beginning of the cold season. The SST decrease frorn dark grey to white while the cloudy area is displayed in black. the COADS database (Roy and Mendelssohn,1994, this vol.) are rather convenient for this, by providing an adequate derisity of SST measurements, especially in Mauritanian and Senegalese areas (Fig. 3). Note that Figure ja corresponds to the satellite situation displayed in Figure 2 and that SHIP data would not allow to detect the presence of t'le coastal upwelling in South Mauritania and North Senegal. Because of their generally irregular spatial distribution, their poor sampling of the coastal area, (especially damaging during the beginning and the end of the upwelling season) and their relatively high instrumental noise, the usefulness of t!ie SHIP in-s'tu data for precisely describing the SST field in coastal upwelling areas is generally very low. The suspect SHIP data are first eliminated from the original data set, initially by comparison with a global SST clirriatology Le., the Reynolds monthly SST climatology (Reynolds, 1982) or with Our own climatology, preliminay computed from 1984 to 1989 (Demarcq and Citeau, 1795). Only values whose departure [rom the climatology is greater than 5°C are removed given the strong SST anomalies that are encountered in this upwelling area. Despite the above limitations, the SHIP data provide a very satisfying estimation of the residual atmospheric absorption field. The latter is obtained by coupling ship data with the uncorrected satellite data (Fig. 2b) in order to give corrected SST field: the field of 'atmospheric correction' is then computed as the statistical departures of the satellite synthesis from the in-situ SHIP SSTs. An example of atmospheric field and the resulting corrected satellite SST field is displayed on Figure 4. Standard SST processing was applied on a temporal basis of 5 days from 1784 to 1995. During the upwelling season off Senegal and Mauritania (from October to June), approximately 90% of the daily METEOSAT infrared synthesis can be iised. This percentage is in fact seasonally variable, and depends on the mean coastal nebulosity, which is invetsely proportional to trade wind intensity. SHlP SST data 16-2011 011 994 1 1 1 1 2; 23 21 1 b 17 15 Longitude ("W) SHlP SS 16-2010 1 - data 1994 I . ... I I I 1 1 25 23 21 19 17 15 Longitude ("W) Fig. 3: Typical examples of "SHIP" da ibution offshore Mauritania and Senegal during 5-days periods at the beginning (a) and in the middle (b) of the upwelling season. Longitude ("W) Longitude ("W) Fig. 4: Example of atmospheric abs Id (a) on 20 October 1994 (beginning of the upwelling season) and the resulting corrected satellite SST 754 Dynarnics of the Upwelling off Senegal and Mauritania For each daily synthesis, the cloudy area is masked and a radiative temperature field is calculated by a j day period. This field is then atmospherically corrected by adjusting the raw temperature values with the corrected SHIP SST measurements as described above. 2. COMPUTA~-ION OF COASTAL UPWELLINC INDEXES FROM SST FIELDS OFF MAURITANIA AND SENECAL The upwelling structures observed off Mauritania and Senegal from October to lune are representative of a complex spatial dynamic, chancterized by several local SST minima, mainly depending on wind direction and local bathymetry. The SST contrast with offshore waters depends mainly on the history of the upwelling in the preceding few weeks and tends to decrease during weak upwelling episodes. Superficial upwelling filaments moving offshore are frequently observed and reveal the concentrating effect of the shelf topography. Figure 5 displays some commonly observed features. The main differences in SST field are linked to the large scale wind field variation, in both intensity and direction. According to Ekrnan's theory (Ekman, 190j), upwelling is maximum along coast lines parallel to the wind. The localization of this maximum varies according to wind direction and is panicularly visible during the beginning of an upwelling event (see for example Figure ja, b). During more intense phase of the trade winds, the cooling extent is continuous along the coast line, from 21°N to 10°N approximately (Fig. 5c). The southern most extent occurs around March, according to the most southern latitudinal position of the ITCZ/trades system which occurs in Febniary and March (Citeau et al., 1989). Tht localization of the maximum flow of upwelled waters at the sea surface can be defined by a continuous area of minimum SST. This area is relatively f~ed and closely linked to the local bathymetry (Fig. 6). SST at these locations is related to the instantaneous response of the upwelling system to wind forcing. This local spatio-temporal signal does not reflect the dilution effects due to past upwelling events that would be reflected in the mean SST calculated on a larger space scale. An SST based upwelling index is calculated by differencing the local SST ('SSTsat') located at the minimum SST line (sec Fig. 6) and a reference offshore temperature at the same latitude, to avoid taking into account large scale SST anonlalies due to planetary clirnatic anomalies, not reflected in coastal areas. This reference temperature is chosen as the climatic SST temperature (and not the current offshore SST), calculated from 1984 to 1994 in the tropical Atlantic frorn a routinely elaborated product calculated from METEOSAT and SHIP data (Demarcq and Citeau, 199 5; Demarcq and Suisse de Sainte-Claire, 199 5). According to Jacques and Tréguer (1986)) the upwelled water off Mauritania and Senegal is essentially composed of SACW (South Atlantic Central Water). Regular coastal measurements in several oceanographic stations in Senegal (Roy et al., 1985) show that the extreme coldest events correspond to very stable values of SST between 14.0°C and 14.5 OC. In this case, the salinity of the upwelled water (between 3 j.4%0 and 3 j. j%o) confirms its SACW origin. Acc:)rding to the Ekman's theory and oceanographic coastal measurements off Senegal, the departure of the SST (mcasured as close as possible to its arriva1 location at the sea surface) from its minimum value (pure SACW) is Longitude ("W) Fig. 5: Commonly observed superficial SST fields during the upwelling season off Mauritania and Senegal. The SST decrease from black (27°C) to white (1 7°C) for al1 images. Numerous filaments of upwelled waters moving offshore are clearly visible. Fig. 6: Localization of the maximum flow of upwelled waters at the surface in relation to the local bathymetry and localization of coastal areas for 4 upwelling index computation. 156 Dynarnics of the Upwelling off Senegal and Mauritania Fig. 7: Offshore SST (SSTrnax [lat,monthl) at 2 3"W off Mauritania and Senegal, calculated for the upwelling season of the years 1984-94 frorn satellite clirnatology. Pa 8 0 ~+ 2 N.Maur. - ---- S Maur. 0 / .-- * N Sen. - - S.Sen. T7.-... . . . . SONDJFMAMJ Months dirclctly linked to the upwelling flow. On the other hand, the maximum SST recorded frorn oceanographic coastal medsurements in upwelling season during very weak upwelling phases varies seasonally, and converges towards the off:,hore SST at the same latitude, where the upwelling influence is negligible (because of the dilution of the up~velled water due to wind-generated turbulence). Tht: minimum value of SST expected at the upwelling centers, noted SSTmin, is the temperature of SACW as it reaches the surface. The maximum temperature, noted SSTmax, is chosen as the offshore clirnatic SST recorded at 23"W. Figure 7 displays this mean seasonal signal, calculated from the satellite clirnatology elaborated for the 1984- 1994 period, frorn North Mauritania to South Senegal. This offshore signal is representative of the rnean offshore upwelling influence. For a given year, it reflects the rnean 'seasonal past' of the upwelling in the coastal area, but not its current intensity. As reported in the time senes of coastal oceanographic measurernents of wind and SST, the seasonal variation of the observed value of SSTsat reflects the fact that, for a definite level of wind forcing, SST cooling is greater at the beginning or at the end of the upwelling season in a relatively warm environment than during the middle of the season in colder surrounding waters (Teisson, 1982). This makes it possible to compare the upwelling intensity during the whole season. The main difference with the Ekman index is the spatio-temporal integrating effect intrrnsically linked to an SST based index and clearly displayed in Figure 8. Important discrepancies remain between these two pararneters (Fig. 8) partly due to the spotty sampling of the ship data close to the Coast (especially in the south Mauritania region), because of the ship route locations (see Fig. 3). Thir; fact is clearly shown across the differences in rnean SST separately calculated from SHIP data (by objective ana ysis) and from satellite data over the sarne coastal area (Fig. 9). This difference leads to a severe under-estimation of upwelling extent and intensity calculated frorn the SHIP data. This under-estimation is high at the beginning of the upn-elling season (when the offshore extent of the upwelling is generally weak, see for example Fig. 4 and jd). In addition this under-estimation is different from one year to another, depending on the variability in the distribution of SHIP data. Frorn these observations, a relative SST-based upwelling index, ('SSTI') was calculated from the deviations of the loca Ily observed SST from their extreme theoretical values, respectively fured and seasonally varying. To take into account the effect of the spatial dilution of the upwelled waters at the surface rnixed layer, the SSTI may 1984 1986 1988 1990 1992 1994 Year Fig. 8: Direct compari index and satellite SST ["C). l5 1 satellite SST ------ I ship SST 1988 1990 Year Fig. 9: Simple SST-ba culated from SHlP data only (dotted line) and from the satellite product (solid line) in the south Maurit 7 58 Dynamics of the Upwelling off Senegal and Mauritania be expressed by the relation: SSTI = (SSTsat-SSTmaX[ht,month]) /(SSTminSSTmz[lat,monthl) Figui e 10 shows the upwelling dynamic calculated using this index for 4 areas (see also Fig. 6) from Nortli Mauritania to S 1 a O - 6 5. o. Yoff b I. _I.. / ...-. ----- 3t1 1 Thiaroye c JFMAMJJASOND V .. '---__ 0 '*.-- -5- -10- -15- Anchovy collapse Namibia h o m 10 20 30 40 50 60 32 40 48 56 64 72 Fig. 12: Plot of the ACE transformation of u Biornass (10%) Wind (mis) anchovy recruit (a) biomass and (b) al numbers and mean SE winds (m.s-l) from E October to March in the southern do Benguela. 0. s b 20 40 60 80 100 120 3.2 4.0 4.8 5.6 6.4 7.2 Nurnbers (log) Wind (mis) The South African anchovy fishery relies mainly upon catching the annual recruitment run, and TACS are based upon hydroacoustic estimates of recruitment undertaken in the middle of the fishing season. (Hampton, 1992). However, advance knowledge of recruitment would be of great value (Cochrane and Starfield, 1992). 206 Food, Transport and Anchovy Recruitment in South Africa In relating environmental factors to fish recmitment there is often the problem of which came first, the index which corr'rlates with recmitment or the rationale for choosing a suitable index. In this instance, we were fortunate, but also restricted, in having a single, long time-series of fish oil-to-meal ratios which could be used as a food index and an appropriate long term wind record which appears to tie in with transport. Rationale involved choice of the temporal and spatial averaging in producing the indices. The approach we took here with regard to the oil data differs fundamentally from previous studies (e.g., Cochrane and Hutiihings, 1995) which looked at the oil yields of the previous year as a potential predictor of the recmitrnent the next year. (Such studies argued that high oil yields operating through adults having sufficient lipid reserves would lead to successful spawning and thereafter to good feeding for juveniles). By choosing oil-yields frorn the pre-recmitment period (although frorn fish of commercial size), we delay any forecast but hopefully make the relationship more robust. Both appi.oaches do have problerns, however. In order to justify using oil-yields to indicate likely future spawning success, the fish~iry would have to target on these spawners to a far greater extent than the SA pelagic fishery does. On the other hand by using the oil data from largely before the beginning of the season to indicate recmitment (as done here) we implicitly suggest a common favourable environment for adults of various species and anchovy pre-recmits. Witli regard to the wind index, the transport rationale is generally given an equal or lesser rating than wind-driven turbulence, as described by Lasker (1975, 1978). We do not exclude the possibly important effect of turbulence on recruitment, but rather show that advective losses arising frorn the position of spawners in the flowfield can, and perturbations of the surface flowfield attributable to winds could, lead to a good proportion of the observed recmitrnent variability. The recmitment of anchovy, and possibly other pelagic species, could be investigated in the future b!r using the model to sirnulate transport under different flowfield scenarios, which could arise as a result of a changing global clirnate. Future work could build upon the use of oil-yield data as a proxy for anchovy recmitment and investigate the relationship of tkis variable, as dependent upon SE winds, using a longer rime-series. Lastly, it should be rnentioned that, although SE winds can be strongly implicated in the anchovy recmitrnent failure during 1994 (through transport and/or turbulence), additional factors have been proposed by sorne of our colleagues. These include the possibly poor condition of young spawners and the sharp divide between very cool upwelled and warm Agulhas water in the spawning area, leading to fish being relatively far offshore in a small region of favourable temperatures. This anomalous year merits a study in its own right, particularly because it was studied by means of eight consecutive monthly cruises throughout the spawning season as part of the South African Sardine and Anchovy Recruitment Prediction programme. This programme also addresses the sardine, the spawning biomass of which had, by late 1994, reached the sami: as that of anchovy for the first time in over 25 years. Whilst the increase in sardine, resulting frorn conservative management plus unknown environmental cues is to be welcomed, it rnay affect the relationship between oil-!.ieid data and anchovy recmitment, if anchovy are no longer the main contributor to the pelagic reduction plants. (;. Nelson is acknowledged and thanked for constmcting the anchovy transport model used in this study, and for his guid.ince in the project. We are grateful to L. Underhill for his statistical advice in the oil-yield project which forms part of this paper. The Benguela Ecology Programme is thanked for funding enabling A.J. Boyd to attend the First International CEOS meeting held in California in September 1994, and at which this paper was presented. -- A.]. BOY~ ET AL. 207 Crawford R. J. M. 1980. Seasonal patterns in South Africa's wes. terii cape purse-seine fishery.J Fisb Biol., 16(6): 649-664. Cntvford R. J. M., L.V. Shannon and D. E. Pollock. 1987. The Ben- gutla ecosystem. 4. The major fish and invertebrate resources. In: M. Barnes (ed.). Oceanography and marine biology. 4n amiual review. Aberdeen, University Press, 2 5: 3 53-505. Cnnlford R. J. M., L.1. Shannon and G. Nelson. 1995. Environ- merital change, regimes and middle-sized pelagic fish in the south-east Atlantic Ocean. Sci Mar., j9 (3-4): 417-426. Cuiy P. and C. Roy. 1989. Optimal environmental window and pekigic fish recruitment success in upwelling areas. Can.J Fisb. Aq~at. Sci, 46 (4): 670-680. Cuiy P., C. Roy, R. Mendelssohn, A Bakun, D.M. Husby and R. H. I1arrish. 1995. Moderate is better: exploring nonlinear clima- tic c'ffect on Californian anchovy (Engraulis nlordm). Climate change and fish population. Beamish, R.J. (ed.) Can. Spec. Publ. Fisi?. Aquat. Sc;. , 121: 417-424. Haiiipton 1. 1987. Acoustic study on the abundance and distri- bu tion of anchovy spawners and recruits in South African waters. In: .ILL. Payne,J.A Gulland and K H. Brink (eds.). The Benguela and comparable ecosystems. S. Ah:J Mar. Sci., 5: 901-917. Haiiipton 1.1992. The role ofacoustic surveys iri the assessnient of pelagic fish resources on the South African continental shelf. In: A.I.L. Payne, K.H. Brink, K.H. Mann and R. Hilborn (Eds.). Ber~guela trophic functioning. S. AfrJ Mar. Sci, 12: 1031-1050. Huichings L. and A. J. Boyd. 1992. Environmental influences on the purse seine fishery in South Africa.1nvestigacion Pesq., San- tiago', 37: 23-43. L~sl mm---- Senegal ,' ' z O- \- Fis. 9: Corr monthly values of alongshore wind stress and sea surface temperature. Negative correlations indicate Eh man upweliing. Li . • ~-0.6i~m~v~m~~m,m c - JFMAMJJASOND Months 3.2. Changes du ring the last three decades Then we consider the alongshore wind stress (ASWS) series as a broad scale proxy for upwelling intensities. From 1964 up [O 1993, ASWS (Fig. 10) and SST (Fig. 11) show roughly inverted patterns. Positive anomalies of ASWS (upwelling favclrable) occurred at approximately the same time in the three study areas : from 1971 to 197j, in 1986 and from 1991 onwards. Strong negative anomalies of SST were clearly associated to the first two events, while period of strong winds in the early 1970s corresponds only to a slight cooling. Vi a 1,5 - Sahara . - Ld -1,5 m r r rn 1964 1974 1984 1964 1974 1984 1964 1974 1984 Y ear Year Year Fig. 10: Norrnalized series of alongshore mean annual wind stress anomalies (1964-1993). The wind stress is parallel to 208" off Sahara, and is southwards off Mauritania and Senegal. 4. 1. Southwards spreading of Sardine 2 - z 1- .- - Ld O- s % rB -1 - -2 . The fishery for Sardina pilchardus in northern Morocco is relatively recent, as it began only during the 1720s (Belvèze, 1784). The fishing area progressively spread southwards, into what has been called zone A (32" to 30°N), (Belvèze, 1784). Occasional occurrence of small sardines was recorded farther south, in the Baie du Lévrier (near Sahara 1964 1974 1984 1964 1974 1984 1964 1974 1984 Year Year Year Fig. 11 : Norrnalized series of rnean annual sea surface temperatures anomalies, 1964-1 993. Noiladhibou) as early as 1923 by Monod. In June 1941, Spanish trawlers caught some small specimens (7-8 cm) off the Cape Blanc; in September 1952, a stomach of Orcynopsis unicolor provides another record. In 1953, the regular presence of sxdine in the Baie du Lévrier was established by beach seine sampling from April toJuly (Cadenat and Moal, 1955). All werï of small sizes (7 to 12 cm) but they reached sexual maturity and the catch of very small fish (3 cm) proved that a population has established itself. Furnestin (19%) attributed the small size of these fish to a low growth rate in tlie souihem limit of their province, and he thought that no commercial yield was foreseeable in this area. From 1965 a second fishery developed between 29" and 27"N (zone B), then a third, south of 26"N, after 1969 (zone C). Catches of several tons were obtained in the Baie du Lévrier and north of Cape Blanc in 1972 and 1973 (Maigret, 1974). The southern boundary of the sardine fishery was estimated at 28"N in 1966, 21°N in 1970 and 18"N in 1973 (Domanovsky and Barkova, 1976). In 1974 some sardines were caught off Senegal (Conand, 1975; Boëly and Fréon, 1979). Then the schools came back northwards and the species almost disappeared from Mauritanian waters in 1982-83. However, from 1984 the species was again fished off the Banc d'ilrguin (20°N) and a new southward displacement appl.:ared to begin (FAO, 1985). Indeed, the Senegalese small scale fisheries caught 77 t of Sardina pilchardus in 1991, and 1100 t in 1994, mainly by purse-seines. During a short period of the winter 1994, sardines were the main species caught in certain beaches south of Cape Vert, (Petitgas, pers. comm.). The present expansion of Sardina pilchardus in Senctgalese waters is by no means comparable with the preceding scarce records. The rare records of sardine concerned maiiily young fish, caught in waters between 16°C and 19"C, during the cold season, although, in 1954 and 1976, young fish were fished in the Bay de Gorée, in 25"-28°C waters (Fréon, 1988). On the contrary, the beach seine sampling, between January and March 1994, proved that the schools were made of 20-23 cm ripe sardines, weighing 100-130 g. Remote sensing thermographs indicate a strong cooling of superficial waters during the winters from 1986 onwards (excëpt 1990). In 1986 and 1994, surface cold waters spread southwards from the Mauritania to soutliern Senegal (Dernarcq, this vol.). This cooling may be responsible of the exceptional abundance of sardine south of Dakar. In the course of May 1994, the species was also caught in the bottom nets of the small scale fisheries, and it disappeared soon after, probably escaping in deeper waters as the warm season was advancing. On the contrary, the first months of 1995 were rather warm and no sardine was reported. Although the Senegalese catch was very limited until1994, it has clearly demonstrated, twice in twenty years, a southwards spre.dng of the geographical range of the sardine, following, one or two years later, the huge catch off Sahara in 1976- 197; and in 1989-1990. Moreover, each of these peaks was approximately in phase with a strengthening of the trade winds. 4.2. Seasonal pattern of catches and CPUE off Sahara and Mauritania E'easonal and spatial distribution of pelagic catches can be used to infer ecological preferences of different fish species. The iishing strategy, i.e. the distribution of fishing days north and south of Cape Timiris, was approximately the same for the Soviet and Romanian fleets (Fig. 12). Most of the effort was in the south from April to June, then the boats moved northwards until the end of the year. This shift of the fishing boats is related to the seasonal displacement of the strongest upwi:lling. Chavance et al. (1991) observed that the fleet mainly worked in the region of the steepest SST gradients. Indecd, Tracburus spp. are the main catch of pelagic trawlers and this fishing strategy enables a regular yield of either of the tao main Tracburus species. On the 1985-1991 Romanian catch avemges (Fig. 13), the CPUE of the temperate species T. trachurus highest during winter and almost nul1 in summer, while the tropical T. trecae was mainly fished in the southern area, al1 along the year. Chavance et al. (1991) noted that the two peaks of yield correspond respectively to arriva1 of spawning concentrations of each of these two species into Mauritanian waters. The seasonal patterns of CPUE for Decapterus rhonchus and Scomber japonicus differ from the preceding (Fig. 13). The higher catches occur after the maximum upwelling period. The seasonal changes in CPUE are related to temperature optima and feeding regimes. Phytoplankton predominates during period of maximum upwelling intensity, while heavier concentrations of zooplankton are delayed until wind stress and offshore transport relax (See Section 1.3). S. japonicus and D. rhonchus having a carnivorous diet avoid the newly upwelled waters, more than do Tracburus spp. The tropical species Sardinella aurita and S. maderensis are more abundant during the summer non-upwelling period and higher concentrations of these two species are encountered south of Cape Timiris (Fig. 13). The best catches of Sardina pilchardus clearly come from northern Mauritania, though the fishing effort of Romanians and Soviets were roughly equivalent in north and south areas (Fig. 12). The best fishing months were January-May and, secondly November-December. In other words, sardine was mainly fished dunng its two spawning seasons, and the catch were almost nul1 from June to September, during the warmer period. ---- North 4000 -( South JFMAMJIASOND J FMAMJ JASOND . W 2 10 10 U \ \ \ J FMAMJ 1 ASOND J FMAMJ JASOND ;; 600 Fig. 12: Seasonal features of the Romanian (left) and Sovietic (right) sardine fisheries off Mauritania, North and 300 South of 19"N: c 3 200 a: total catch; ~r b: catchleffort; 1 O0 c: fishing effort. JFMAMJ JASOND 1 FMAMJ J A SOND Months Months 226 Pelagic Fisheries Changes in the Southern Canary Current 20 1 Sardinella aurita 20 . Sardinella maderensis 20 1 Trachurus trachurus 20 Trachurus trecae 1 20 Decapterus rhonchus 1 20 , Scomber japonicus J FM A MJ J A SOND 1 FM A MI J A SON D Months Months Mauritania (1 979-1 992) , Decapterus rhonchus and Scornber japonicus, 4.3. Changes in species dominance A change in the relative abundance of pelagic species during the mid 1970s wind event was described by Fréon (1988) and Binet (1988). While the sardine landings were growing, the relative abundance of mackerels (Scomber japonicus), horse mackerels (Trachurus spp.), jack mackerels (Decapterus rhonchus) and Sardinella spp. were diminishing (Fig. 14). The new data set (1979-1992) shows a basically similar pattern. After the 1970s wind event, the proportion of sardine decreased, the catches were again dominated by Trachurus spp and Decaptem. Then, the 1986 ASWS peak was followed by ii sevenl year increase of Sardina pilchardus at the expense of the other species, up to 1989. These changes closely resemble those which happened 10 years earlier. W Sardinella O S. pichardus 1979 1984 1989 Trachurus O Scomber japonicus SPP. Fig. 14: Changes in the species composition of pel rea (left), from Fréon (1 988) and Binet (1988); and off Mauritania (right), from FA0 4.4. Wind - sardine recruitment relationships A linear correlation was sought between the time series of wind stress and sardine catch off Sahara and Mauritania (1969-1990), (Fig. 15). These were first calculated between annual values of ASWS and sardine catches. The best correlation is obtained between the annual catch and the wind of the previous years (Table 2). Assuming that catch variability reflects - to a certain extent - changes in the stock abundance, these correlations mean that if sardines are mriinl!r recmited at two years old, the recruitment is favoured by increase, in the wind, induced upwelling during the first yearof life of the fish, and the result appears in the fishery two years later. According to Barkova (in FAO, 1990) stock biomass began to increase slowly in 1982, and quickly from 1987 (Fig. 2), that is only one year after the 1986 wind strengthening. Year year n year n-1 year n-2 year n-3 r 0.126 0.242 0.3j4 0.237 Table 2: Correlations between the yearly catch of sardine (Sahara and Mauritania) and the alongshore wind stress between 22" and 26"N. Then, in order to identify the season whose climate determines recruitment levels, the regressions were calculated between the annual catches and the twelve monthly ASWS series, with time lags ranging from zero to two years (Fig. 1 j). The correlations are weak, but positive for the largest part of the year. Negative values are observed only in May, July, August and panicularly December. The coherence of these results has some significance. If the spawning periods are in October- December and April-May, strong winds when larvae are less than three months old, negatively affect their survival; on the other hand, strong winds during the rest of the year irnprove sunival. In other words, strengthening of Ekman upwellings are beneficial to the food web and to sardine feeduig, insofar they do not occur within the very f~st months after spawning of sardine. 228 Pelagic Fisheries Changes in the Southern Canary Current 5: Analysis of sardine recruitrnent. a) ardized anomalies of alongshore wind s; b) total sardine catch off Sahara and uritania (1969-90); c) correlations between annual sardine catch off Sahara and Mauritania and monthly alongshore wind stress, for time lags from O to 3 years. 2 -1,5 3 1964 1974 1984 Year -1,5 4 . , . . . 1964 1974 1984 Year -v,v - JFMAMJJASON D Months In the southern part of the Canary Current a large fishery has developed off the Western Sahara, based on a sardine population, (stock C), which was absent before 1965. Two southwards expansions of this sardine occurred at 23 years inteival. Sardina pilchardus was fished off Mauritania, where large industrial fleets were exploiting it and off Senegal, wht:re moderate numbers were caught by the small-scale fishery. Southwards extensions are correlated to multi-year periods of trade wind strengthening which occurred in 1972-72, in 1986 and from 1991. We described likely environmental changes associated to the new climate pattern: intensification of upwelling regime, southward transport, and decline of SST. Phytoplankton production was probably boosted as well, but not matched by zooplankton grazing, due to the brevitv of the residence time of waters over the shelf. Thus, sardines larvae were strongly advected towards the south and, as adult sardines are able to feed on phytoplankton, they were favoured instead of the zooplankton feeders, or the carnivorous sial1 pelagic fish. Off Northwest Africa, sardines spawn during the whole year, with a distinct maximum in winter. In the Cape Blanc area, larvae were abundant at SST between 16" and 17"C, scarce at temperature above 18°C and absent when the SST exceeded 21°C oohn et al., 1980). These authors described an absence or scarcity of lamae in 1968,1970 and 1972, during weak or absent upwelling, while they record high catches of lmae in 1974,1975 and 1977. These fluctuations are in good agreement with the changes of upweiiing intensity, and corroborate the correlations we found between recniitment and annual wind stress. However, spawning is limited by temperatures below 155°C (John et al., 1980), and during strong upwellings, eggs are only found in small number over the shelf break. The negative correlation we found between catch and MWS, during the spawning period, probably means that during early life, the advective losses are more deleterious for the larval survival than food limitations. This is in contrast to other parts of the year, when strong upwellings enhance the food content of coastal waters and the survival of young sardines. In the Canary Current, the drift of fish larvae is basically directed southwards. Lloris et al., (1979) propose a biological cycle of a demersal fish (Pagellus acame) based on a latitudinal separation of adults (northern, upstream group) and juveniles (southern, downstream group). Eggs and larvae released by the adults drift southwards and lead to the southern group. The reverse link between young and adult is assumed to be a countercurrent migration off the edge of the shelf. (This return migration might be helped by the deep northwards current). The iife span of Sardina pilchardus eggs is 2 to 4 days, while larvae may reach up to 9 weeks aohn et al., 1980). Along a transect parallel to the Coast, from Morocco to Mauritania, the larger larvae were found in the south, indicating the direction of their drift. However, John et al. noted that, in 1977, the southern drift of near-surface larvae may be an exception due to the strong upwelling rather than a regular feature. However, during the last decades of strong upwellings, a general trend to a southwards extension of the geograpliical ranges of pelagic species has occured (Ehrich et al., 1987). Thus, we may reasonably suppose that the southwards circulation was enhanced and northwards surface transport inhibited during these windy years. It became unlikely for juvenile tropical fish to settle north of Cape Blanc. On the contrary, during wind relaxation periods, the slowing of alongshore and cross shelf circulations lengthen the residence time of water over the shelf, improves food web transfers and favours carnivorous fish. The decrease of northerlies enables a northwards surface transport and a colonization of the Mauritanian shelf by tropical species. The first records of a species, out of its usual range, are generally from isolated individuals, which can be considered as vagrants, according to Sinclair (1988). If these vagrants are numerous enough, and if they encounter good environmental conditions, including circulation features enabling a complete life cycle on the shelf, their offsprings may initiate a new, self-sustaining population. The recent history of cod settlement off West Greenland, from larvae advected from the Iceland, is a similar example (Dickson and Brander, 1993). Thus, the southward extension of sardines from Morocco to Senegal, during the last decades, probably went off through the successive settlement of spawning areas, heading to self sustaining populations off Cape Bojador-Cape Barbas, Cape Blanc-Cape Timiris and possibly on the Senegalese Petite Côte, south of Cape Vert. This colonization was probably facilitated by the heavy exploitation of other fish stocks. It required 20 years at least, between the first spawning, observed in 19j0, in the Baie du Lévrier (Mauritania) and the first industrial fishing in 1970. Maigret (1974) noted that the three years preceding the first landings in Nouadliibou were especially cold. The frst observations of Sardina pilchardus in Senegalese waters, previously reported by Fréon (1988), concerned young individuals probably carried by the southward circulation. In 1977 some ripe specimens were caught and 230 Pelagic Fisheries Changes in the Southern Canary Current agiin in 1994. Thus, it seems that a new population is settling south of Cape Vert, but due to lack of observation in 1995, we cannot confirm it. Although a Senegalese sardine fishery looks very unlikely in the future, let us remind that to Furnestin (1955) a regular sardine fishery off Mauritania was quite improbable. It is satisbing to see that the same ecological relationships held during three decades and that chaotic dynamics did not prctvail against them. The reasons are probably because upwellings are young ecosystems where any strengthening of Ekman pumping increases offshore transport, stops the maturation, and resets al1 the ecosystem, preventing chaotic evolution. According to the 50 years wind stress time series compiled by Bakun (1990, 1992), it seerns that the present climatic chkinge was beginning at least 50 years ago. If, according to Bakun, this increase in the eastern boundaries trade winds is duc to the increased contrast in temperature between heated land masses and the oceans, in relation to the 'greenhouse effect', we may expect a continuation and a strengthening of these phenomena during the coming years. Probably, the strengthening will not be regular but fluctuating, and we can expect a continuation of the alternance of sardine and horse mackerel-sardinella periods, with more frequent sardine penods. Alc araz M. 1982. Zooplankton biomass and its relation to total Binet D. 1988. Rôle possible d'une intensification des alizés sur pai-ticulate carbon and nitlogen off northwest Africa. ~app. P.- le changement de répartition des sardines et sasdirielles le long v. Réun. Cons. int. E~plor. Mer, 180: 270-273. de la côte ouest africaine. Aquat. Living Resou,:, 1: 115-132. Bakun A. 1990. Global clirnate change and intensification of coas- Binet D. 1991. Dynamique du plancton dans les eaux côtières ouest africaines : écosystèmes équilibrés et déséquilibrés. 111: P. tal ocean upwelling. Science, 247: 198-201. Cury et C. Roy (eds.). Pêchen'es ouest-afi-icaines: unriabilité. trends, and potential impacts on coastalpelagic fish populations. ~i~~~ D, and E. suisse de sainte-claire. 1975, contribution à IC1SMar. Sci. Synzp., 195: 316-325. l'étude du copépode planctonique Calarzoidescaiinatus : répar- tition et cycle biologique au large de la Côte-d'Ivoire. Cab. ORS- Barkova N.A. and M.V. Domanevskaya. 1990. Utilisation des Ockanogr, 13 : j-30, injbnnations satellitaires pour l'étude des pa aicularités de ré,tjaaition de sardine dans ['Atlantique centre-est. ICES. Blackburn M. 1979. Zooplankton in an upnrellirig area off nor- c.A~. 1990/~9.12 p. thwest Africa : composition, distribution and ecologv. Deep-Sen Res., 26, 1 A: 41-j6. Belvèze H. 1984. Biologie et ~nanziquedespopulations desar- dine (Sardina pilchardus Wa1bautn)peuplant les côtesatlan- Boëly T. and P. Fréon. 1979. Les ressources pélagiques côtières. 1n:J.P. Troadec and S. Garcia (eds.). Les ressources halieutiques "9"esmamcai~e~e~P~0Po~"ionsPo~r~n a'zé'agmen'de' de ~~Atian~ique centre-ot, lere : les ressources du golfe de pêcheries. Thèse Univ. Bret. occident., j32 p. Guinée, de l'Angola à la Mauritanie. FA0 Doc. tecl?. , 186, 167 p. Biiiet D. 1977. Cycles biologiques et migrations ontogéniques cadenatj, and R. ~~~1, 1955. N~~~ sur la sardine (sa,.dinapil- chez quelques copépodes pélagiques des eaux ivokiennes. Cab. chardus Walbaum) de la région du cap Blanc. Rapp. P.-v. Cons. OR:)TOhJ, sér. Océanogr., 15 : 11 1-138. int. Explor. .Ver. 137: 21-23. Catewicz Z. and R. Siwecki. 1985. On the currents in the coas- ta1 zone of African shelf off Saint-Louis. Oceanologia, 21: 59- 75. Chavance P., Y. Loktionov and M. hlahfoud. 1991. Importance des saisons de transition hydrologique et impact des anomalies climatiques sur l'activité d'une flottille industrielle pélagiqiie en LEE mauritanienne. In: P. Cucy et C. Roy (eds.). Pêchwiesouest- africaines: ~~ariabilité, instabilité et changement. ORSTOM, Paris: 246-258. Conand F. 197 5. Distribution et abondance des larves de clu- péidésau largedes côtesdu Sén&al et de la Mauritanie en sep- tenzbre, octobre et no~wnbre 1977. ICES, CM 1975J 4,9p. Conand F. 1977. Oeufs et larves de la sardinelle ronde (Sardi- nella aurita) au Sénégal : distribution, croissance, mortalité, variations d'abondance de 1971 à 1976. Cab. ORSTOh~f, sér Océa- nogr., 15: 201-214. Cushing D.H. 1978. Upper trophic levels in upwelling areas. In: R. Boje and M. Tomczak (eds.). Up~~~elling ecoystenzs. Springer- Verlag: 101-110. Cushing D.H. 1982. ClinzateandFisheries. Academic Press, Lon- don, 373 p. Dickson R.R. and K.M. Brander. 1993. Effects ofa changing wind field on cod stocks of the North Atlantic. Fisheries Oceanogra- phv, 2 (3/4): 124-153. Domanovsky L.N. and N.A. Barkova. 1976. Sonzepeculari- ties of sardine (Sardina pilchardus Walb.) distribution and spaiuningalong the northwestAfnca. ICES. CM 1976/116,6, 15 p. Ehrich S., H.C. John and P. Westhaus-Ekau. 1987. Southward extension of the reproductive range of Macroranlphosus sco- lopm in the upwelling areaoff north-west Africa. In: A.I.L. Payne, J.A. Gulland and KH. Brink (eds). The Benguela and compa- rable ecosystem. S. Afv. J. Mar. Sci., 5: 95-105. F.A.O. 1985. Rapport de la troisiènie réunion du groupe de tra- vail ad hoc sur la sardine Sardina pilchardus (Walb.). COPA- CEIPACE Séries, 85/39,157 p. F.A.O. 1990. Rapport des groupes de travail ad hoc sur la sardi- ne et sur les chinchards et les maquereaux dans la région nord du COPACE. COPACEIPACE Series, 90/j0,372 p. F.kO. 1994. Rapport du groupe de travail ad hoc sur les sardi- nelles et autres espèces de petits pélagiques côtiers de la région nord du COPACE. Dakar 1993, COPACEIPACE Series, 91/j8, 295p. Fréon P. 1988. Réponses et adaptations des stocks de clupékles d'Afrique de l'ouest à la ua~iabilité du nlilieu et de l'exploita- tion. Analyse et rejlerion à partir de l'aenzple du Sénégal. Ed. de I'ORSTOM, Coll. Etudes et Thèses. Paris. 287 p. Fréon P. and B. Stéquert. 1979. Note sur la présence de Sardi- nupilchardus (Walb.) au Sénégal: étude de la biométrie et inter- prétation. Cybiunz (3ènie sér.), 6: 65-90, FurnestinJ. 19jj. Remarques sur quelques caractères de la sar- dine de Mauritanie (Sardina pilchardus Walbaum) . Rapp. P.-v. Cons. int. Exploc Mer, 137: 24-2 5. Grall J.R., P. Laborde, P. Le Corre , J. Neveux, P. Tréguer and A. Thiriot. 1974. Caractéristiques trophiques et production planc- tonique dans la région sud de l'Atlantique marocain. Résultats des campagnes CINECA-CHARCOT 1 et III. Tétbl~s, 6 (1-2): 11-28. Grall J.R., P. Le Corre, et P. Tréguer. 1982. Short-terin variabili- ty of priniaq production in coastal upwelling off Morocco. Rapp. P.-v. Réun. Cons. int. Explor. iller, 180: 221-227. Hargreaves P.M. 1978. Relative abundance ofzooplankton groups in the northwest African upwellingregion during 1968 and 1972. ln: R. Boje and M. Toniczak (eds.). Upzvellingecoylstei7z. Sprin- ger-Verlag: 62-72. lacques G. and P. Tréguer. 1986. Ecoystènzespélagiques nrarins. Masson, 243 p. John H.-C. 198 5. Horizontal and vertical distribution patterns of fish larvae off NWAfrica in relation to the environnient. Int. Syiilp. Upzu. W. Ab, Inst. Inu. Pesq. Barcelona, 1: 489-j12. John H.-C., U.J. Bohde and W. Nellen. 1980. Sardinapilchar- dus larvae in their southemniost range.Arch. Fisch Wisr., 31 (2): 67-8 5. Lemasson L. and J.P. Rébert. 1973. Les courants marins dans le golfe ivoirien. Cab. ORSTOIM, sér. Océanogr., 11: 67-96. Lewis R.C. 1929. The food habits of the California sardine in rela- tion to the seasonal distribution of microplankton. Bull. Scripps Instit. Oceanogr., 2 (3): 155-180. Lloris D., J. Rucabado, J. Lleonart and P. Rubies. 1979. Adapta- ciones de los organismos frente a variaciones ainbientales sos- tenidas: Pagellusacaine como modelo de ciclo vital en el aflo- ramiento del NW de Africa. Inu. Pesqu., 43 (2): 179-490. MaigretJ. 1974. La sardine sur les côtes de Mauritanie (Sardina pilchardus Walb.). Bull. 1F.A.N ,36 A (3): 714-721. Mittelstaedt E. 1976. On the currents along the northwest Afri- can Coast south of22" North. Dt. hvdrogi-. Z., 29 (3): 97-117. 232 Pelagic Fisheries Changes in the Southern Canary Current Mi ttelstaedt E. 1982. Large-scale circulation along the coast ofnor- thrvest Africa. Rapp. P.-v. Réun. Corn int. Explor. Mer, 180: 50-57. Mittelstaedt E. 1983. The upwelling area off northwest Africa. A description of phenomena related to coastal upwelling. Prog. Oceanog., 12: 307-331. Mittels~iedt E., D. Pillsbiiry and R.L. Smith. 1975. Flow patterns on thc northwest African upwelling area. Dt. lyd1~)gr. Z, 28: 145-167. Nieland H. 1980. DieNahmng von Sardinen, Sardinellen und M@schen vor der Westkiiste Aji-ikas. Berichte aus dem Inst. für Meereskunde. Univ. Kiel, 75, 137 p. Nieland H. 1982. The food of Sardinella aurita (Val.) and Sar- di,zella eba (Val.) off the coast of Senegal. Rapp. P.-v. Réun. Cons. int. Explor. Mer, 180: 369-373. Olivar M.P., P. Riibiès and J. Salat. 1985. Zooplankton biomass in upwelling regions off northwest and southwest Africa. Int. Si'inp. Upw. WAfr,, lnst. lnv. Pesq. Barcelona, 1: 471-477. Picaut J., J. Servain, P. Lecomte, M. Séva, S. Lukas and G. Rou- gicr. 1985. Climatic atlas of the tropical Atlantic luind stress antlsea suface temperature 1964-1979. Univ. Bret. Occident. Br 0 8- 8 C B 5 6 4 2 O 1950 1B56 1962 1968 1974 1980 1986 1992 1950 1956 1962 1968 1974 1980 1986 1992 Year Year ), and b) northern Chile (SAC, 1950-77; Varisbility is an inherent feature of these resources. This variability is generally associated with both the intensitv of exploitation and changes in environmental conditions (Csirke and Sharp, 1983; Canon, 1986; Yanez, 1989). This variabilitv, well analyzed may become a source of information for a better understanding of the dynamics of the above mentioned species (Bernal, 1990). Quirin et al. (1978) estimated that cenain features of the Southem Oscillation can be used as precursors of El Nino events. Midielchen (1985) suggested that the interannual variations of coastal upwellings in West Africa are related to the variability of the Southern Oscillation. Binet (1988) discussed the possible role of an intensification of the westerly winds in the 1,iistribution changes of pelagic fishes of West Africa. Bakun (1992) suggested that an intensification of the winds causing upwelling may be due to greenhouse effects. Parrish and MacCall (1978) analyzed the horse mackerel fishery off California, and incorporated oceanographic variables into the stock-recmitment models, thus explaining 75% of total variance. Mendelssohn and Cury (1987 analyzed the catch per unit effort (CPUE) of smaU pelagic fishes Côte-d'Ivoire (1966-82)) as a function of the sea surface temperature (SST) co1lei:ted by merchant ships, which explained 43% of the variance. Cury and Roy (1989) indicated that there exists an 'optimum environmental window' for the success of the pelagic resources recruitment in upwelling areas. Patterson et al. (1993) analyzed the collapse of the horse mackerel in the Eastern Central Pacific. They found that catchability varied with environmental conditions and stock size. Fréon (1988), analyzing the small pelagic fisheries of West Africa, proposed the incorporation of environmental variables in global production models. Later, an interactive software was developed for this purpose (Fréon et al., 1993). Mendelssohn (1989) fitted an additive non-linear mode1 using the parental biomass and Trujillo's transport in Peru, thus explaining 75% of the variance of recruitment of anchovy in Peru. Muck (1989) analyzed biomass changes, individual growth, dominance of species, feeding strategies and oceanographic parameters off Peru. He concluded that overfishing and high temperatures affected the anchovy which led to increases of sardine, jack mackerel and horse mackerel, among other species. Muck et al. (1989) showed that the anchovy's area of distribution is biomass and SST related. Yinez (1991) showed that the decrease of anchovy CPUE from 19j7 to 77 could be explained by fishing effort and SST; whereas the change in the CPUE of sardine from 1973 to 88 was explained by fishing effort and Bakun's upwelling index (1973). Yahez et al. (1994) showed that the distribution of anchovy and sardine in time and space in the north of Chile varied along with intra and interannual changes of SST as measured by NOAA satellites. This brief review establishes - if needed be - the need to consider environmental variables when assessing the pelagic fish stocks of upwelling systems. Thus, we move on to describe these environmental variables. Time series of environmental changes are analyzed, these include sea surface temperatures (SST) from 1950 to 1990 off Peru (4"-18'S), from the COADS dataset (Mendelssohn and Roy, this vol.); SSTof tidal gauges of Arica (18'28's) (19 j1- 93, and Antofagasta (23'40's) (1950-93); magnitude and direction of the wind from the meteorological station of Antofagasta (1950-93), used to obtain an upwelling index (Bakun, 1973) and a turbulence index (Elsberry and Gaimood, 1978); and atmospheric pressures of Darwin (12'26's-130°j2'E) and Tahiti (17'33'S-l49"20'\X1 (1950-93), used to estimate the Southern Oscillation index (SOI) (Ropelewski and Jones, 198T). Monthly anomalies were computed for each of the series and smoothed by 13 month centered moving average procedures. The monthly anomalies were also integrated to generate a series of accumulated values, taking into account their monthly signs. The relationship of the Chile-Peru environmental system with changes of the ocean-atmosphere system is of a global nature. In fact, the SOI shows aperiodical decreasing trends associated to the occurrence of the El Niho events (Fig. 2a); since 1976 negative anomalies prevailed, due to a long term weakening of the South-East Pacific antic)iclone (Fig. 2b). It should be noticed that after the 1987 El Nino event, the SOI recovered its positive values, then diminished again when the El Niiïo of 1992-93 developed. Associated with such SOI variations, the monthly mean SST off Peru showed positive anomalies during the El Nino events (Fig. 3a); after the 1950-75 period, a clear dominance of positive anomalies settled in at least until 1990 (Fig. 3b). Along the Coast of the north of Chile, monthly mean SST also shows the effects of El Nifio events: there is a predominance of negative anomalies from 1950 to 1975, followed by a warm period, and a cooling trend in the last period under study (Fig. 4, 5). 2 78 Pelagic Fish Stocks in the South-East Pacific 1950 to 1993: a) anomalies, and b) added a + El Nino events 90 - W) - ?-.. 2 70- D .- - 7 60- - I 50- .- - m 40- m $ 30- u 4 20- 10- O, a 20 - b I -BO - ............ 1950 1960 1970 1980 1990 .lW?,.! ........................ 1 1950 1980 1970 1980 1990 Year Year ........................................... 1 Fig. 3: Monthly mean sea surface temperature off Peru (1 950-90): a) anomalies, and b) added anomalies. 1950 1960 1970 1980 1990 1950 1960 1970 1980 1990 Year Year The monthly anomalies of the upwelling index at the meteorological station of Antofagasta showed a predominance of negdtive values from 1950 to 1975; later on, the anomalies became mainly positive, with a tendency to decrease from the mic 1980s (Fig. Ga). The anomalies of the turbulence index follows the same trend, as an effect of the S-SW predominant winrts (Fig. 6b). Bakun (1990) showed similar trends for wind stress along the coast of California, the Iberian Peninsula, Morocco and Peru, from 1950 to 86. This author suggests the existence of a mechanism through which the greenhouse effect would strengthen the upwellings by intensifying wind strength dong the coast. Thus, there appear to be a positive relationship between SST trends and the wind indexes. It is likely that the obseived warining trend starting in 1976, may have been caused by an invasion of subtropical waters from the north and coastward, assciciated with the long term weakening of the Pacific anticyclone. The intrusion of these waters would have caused a E. YANEZ, M. GARCIA AND M.A. BARB~ER~ 279 a f Ei Nino events - Moving average (13) ,,.,,,,. ,.3,,,,.,,,,,,~.,,,,,,,,,.,,,,,,r,,, - Anomalies 1950 1960 1970 1980 1990 Year Yeai Fig. 4: Monthly mean s I station (1951-93): a) anomalies, and b) added anomalies. - E 5 400 0' E 200 C - 3 O -200 -400 - Maving average (13) a 20 T b ....................................... ,.,+ - Anomalies 1950 1960 1970 1980 1990 10 - O- F - .10 1 -20 - 5 30. m $ -40. v 2 .50. -60. -70. .80 0- cf 400 - 8 200 m - z 0 z -200 -400 - Moving average (13) . Y , , , , , , , , . - , , , . , , , , . , , , , , , ~ , , , , , , . ,A . . .. /,,,.,..,.,,..._,,, ,,,,,,, ,,, ,,,,,,, - Anomalies 1950 1960 1970 1980 1990 1950 1960 1970 1980 1990 1950 1960 1970 1980 1990 Year Year Fig. 5: Monthly mean sea gasta coastal station (1 950-93): a) anomalies, and b) added anomalies. ~ear Year of: a) upwelling index; b) turbulence 280 Pelagic Fish Stocks in the South-East Pacific decpening of the thermocline; thus, the upwelling would not bring cold waters to the surface, but warm and nutrient-poor waiers (Guillén, 1983; Ramage, 1986). Ca!ion (1986) indicated that this effect become strongest during the extraordinary strong El Nino event of 1982-83: the sut~tropical layer of water reached a thickness of 150-200 m, over a large and extensive region. Th:: seasonal variations of the Pacific convergence zones are influenced by variations in intensity and position of the sul~tropical anticyclones (Rutliant, 1985). In general, during the Southern Hemisphere winter the Pacific anticyclone is well dewloped, and the south Pacific convergence reaches its most westerly position, crossing 20"s at 17jW. In summer, it crcsses the 20"s at 14jW, and is located to the equator. The interannual variations of the SOI show a behaviour similar to that of the seasonal variation. It i,i therefore deduced that the predominance of SOI negative anomalies since 1976 may be associated with a long term easiward displacement of the climatic action centers, in particular areas of pressures. This would explain the increase of wirds favourable to upwellings in the north of Chile and in Pem. On the contrary, in the area of Talcahuano (37"S), decrease of SOI is observed after 1975, which may be associated with a period of anticyclonal weakness (Yanez et al., 1992). 2. PELAGIC FlSHERlES AND ENVIRONMENTAL CHANGES 2.1 . The anchovy fishery IMARPE (1970) indicated that the anchovy fishery extends almost along the entire Peruvian Coast, and penetrates wati.rs of the northernmost extreme of Chile, without any clear discontinuity to suggest the presence of isolated and indears linear or at least monotone (Fig. 62). The model selected bv CLIMPROD combined a conventional exponential suilplus production model with an exponential effect of the environment: CPLE = a@ exp (c @ E) where V represents the upwelling index averaged over four years and E represents fishing effort weighted over three fisliing years. Fits were made both with CUEl and CUE2, providing slightly better results for the first index. This was mainly du(: to the lack of response of the stock to the 1971-1972 Ekman transport anomaly. Moreover, the CUE2 series is longer. Th(-refore, in the rest of the paper we onlv present results obtained 84th the latter index, especially since we determined tkit differences were negligible compared to those obtained with CUE1. The R2 coefficient was equal to 65% but was not validated by the jackknife method. The non-biased R2 estimate was equal to 43%, and al1 regression coefficients, except c, were significantly different from zero, which suggests that fishing effort has no major effect on CPUE as opposed to the upwelling index (Fig. 6b). The ACE algorithm applied to the variables Log CPLTEi (dependent variable year i), Ei and Vi (independent variables year i) shcws positive linear transformations for CPUEi and Vi and negative non-linear for Ei, which approaches a function of the typi: l/Ei (Fig. 7). The form l/Ei may imply a total independence between catch and effort (C/E = f(l/E)) or that the relation is not parabolic, but linear with an origin different from zero. To verify this, the algorithm was applied to the dependent variable Ci (catches year i) and Ei and V, as independent variables (Fig. 8). In this case, al1 transformations are positive and linear, with an R2 value of 95% and a strong contribution of Ei. Finally, an exploratory analysis of the relation between Ci and lagged values of E and V (results not presented) shows that the major effects of environmental variables on catch occurs without lags. ). ,MENDOZA ET AL. 299 ~ ~~ Catch (1.10~) Eiiorl (sets) Fig. 5: Univariate distributions of a) Catch; b) Effort; c) CPUE; and d) CUE2. CPUE Upwelling index (CUE2) , O 2 4 6 8 10 Upwelling index (CUEP) 60 - C O 2 4 8 10 Upwelling index (CUE2) iE and CUE2; c) Catch and effort; and d) Catch - - - - 81 - O-, 80 50 - - 70 g 80 300 Sardinelia aurita Population Dynarniu in the Scwthern Caribbean O 400 800 1200 1600 2000 2400 O 400 800 1200 1600 2000 2400 Eiioil (sels) Effort (sets) W 3 30 a U 20 10 O 6- - 0 50- - 81 82 83 85 C 40 30- - 20 - 10 -, Log (CPUE Eiiort i Upwelling index In iew of these results, instead of a surplus production model, we estimated a multiple linear regression of the form: Ci= a + bEi + cVi fit coefficient of determination of this regression was 90% and aü regression coefficients were significantly different from zero (P 2. F ig. 1 b: lvoirian and Chanaian continental shelves and coastal stations. B 6 4 2 O Longitude ("W) 2. OBSERVED CHANGES 2.1. Catch, abundance and dynamics 2.1.1 - Evolution of fishing effort The activity of the small-scale fishing fleets (canoes and beach seines) increased in bot11 countries from the mid 1970s to the mid 1980s (Fig. 2). From 1983 to 1989, the number of fishing trips by canoes fishing for small pelagics in Ghnna first declined, then increased. In Côte-d'Ivoire, the effort of the Ivoirian purse-seiners decreased between 1969 and 1975; there was a fifty percent reduction in the number of boats in 1973. Effort began to rise in 1976, reaching a maximum in 1979 and then declined again aftemards. There has been a steady growth in the effort of this fleet since 1985. 2.1.2- Trends in the landings and abundance of S. aurita Until 1972, total landings of the two sardinella species in the whole of the western Gulf of Guinea never exceeded 50.103 t. During the 1972 fishing season, over 70.103 t of S. aurita was caught off Ghana alone, leading to an apparent "overfishing". Between 1973 and 1975, catches dropped to an average of 4 400 t from the whole sub-region. In 1976, a recovery was noticed with the Ghanaian catch alone reaching about 14 000 t (Fig. 3). In subsequent years, catches fluctuated with an underlying increasing trend. During the same period, catches off Côte-d'Ivoire increased - 700 J . . . PS Cbte-d'ivoire - 6000 .- 9 . , - APW Ghana - 500 ,,,' :, - ES Ghana 6- - ; '... - 4000 a - 0 'C - a - - 300 - O C - 2000 n: 5- I - 100- U1 > 1 1970 1975 1980 1985 1990 -O 0" Year f the Ghanaian beach seine (BS) and Ali-Poli-Watsa nets (APW) and of lvoirian purse seiners (PS), 1972 (or 1966) to 1993. Fig. 3: Annuai catch of S. aurita off Côte-d'ivoire and Ghana, 1966 to 1993. 19-r0 1975 1980 1985 1990 Year 332 Pelagic Fisheries off Côte-d'Ivoire and Ghana ten fold. Since 1983 high catch and abundance have been observed in both countries, and catches recorded after 1985 have either been similar to, or exceeded those of 1972. The highest total catch, in excess of 154.103 t, was in 199;i. 2.7.3- Catch of the other small pelagic fish species lnndings of S. nladerensis remained relatively stable from 1972 to 1992 (20 - 40.103 t per year) in the whole sub- region, except in 1987 when 48 200 t were landed (Fig. 4). There was also a steady increase in the landings of anchovy (Enkraulis encrasicolus) during the last two decades, with relatively low landings in some years (1973, 1986). Over 90.11!3 t were caught in both countries in 1987. Large fluctuations in the landings of chub mackerel (Sco~~zbe~japonicus) wert observed. The combined effect is a global increase of catches of coastal pelagic species of this ecosystem in the period for which data are available. 2.1.4- Variation of catch and abundance with fishing effort - -Total For .Y, aurita, the trend of catch per unit effort (CPUE) of APW canoes or beach seines in Ghana and for purse seiners and ilPW canoes in Côte-d'Ivoire are presented on Figure 5. Catch and CPUE of S. aurita are distributed in accordance with :he state of the resource (Fig. 6): for the Ivoirian seiners, the inter-annual variability of total catch or CPUE with effon is very important, especially in view of the lower level of effort. Before 1981, catch and CPUE were low and increiised with fishing effort. After 1981 catch and CPUE, which were much higher than in previous years, decreased with ieffort. The analysis of the Ghanaian data showed that in the case of S. aul-ita a situation is quite similar to that in Côte.dtIvoire. Here, in the period 1973-1980, there were generally low values of CPUE with increasing fishing effort. With the exception of 1986, high values of CPUE have been recorded since 1985. For S. maderensis, in Côte-d'Ivoire, the y:ars are randomly distributed and both catch and CPUE are highly variable given an average level of effort (Pezennec, 1994). Similar analysis of the Ghanaian data for this species showed that catch and CPUE decreased with incre~ising canoe effort. 300 - G L 200 - - Z. O. PEZENNEC AND K.A. KORANTENC 333 - S. aurrta - S. maderensrs \ ----- E. encrasicolus 1 \ --- S. japonrcus I, P 4 1.7.- 1 .- J Fig 4: Total annual landi pelagic species in Ghana (Sardinelia aurita, S. maderensis, Engraulis ent rasicolus, Scomber japonicus). 1 970 1980 1990 Year - APWGhana --. . BS Ghana 1975 1980 1985 1990 Year b - r 2 16.00- - PS Côte-d'Ivoire ' m - 1200 , ln a 'L - r ; 12.00- O x m 3 - 800 a 05 8.00- 2 - 0 W 3 Fig. 5: Catch per unit of effort W 2 4.00 - - 400 % (CPUE) of Sardinella aunta for (a) O the Ghanaian and (b) lvoirian fieets, 1972 to 1993. O 1975 1980 1985 1990 Year aurita off (a) Côte-d'lvoire and (b) 16- w 12- 2 - O 8 - 4 - O 2501 Ghana 91 b Côte-d'lvoire 985 8. 92 87. ?3 "83. 90. .91 81. 88* d9 84 58 z2 75 7% .73 ?7 76.a8 l;Ois.. 80 69. 6: I Ghana and theoretical effort (catchICPU E). 500 1000 1500 2000 2500 3000 3500 Fig. 6: Annual abundance (catch Effort per unit of effort) of the Sardinella Effort 150- O 50 - 334 Pelagic Fisheries off Côte-d'lvoire and Ghana Côte-d'lvoire: CPUE: tonnes per 89, 9 ?3 day of search; effort: days of 78. *a5 search. 87. 90. Ghana: CPUE: kg per trip; effort: 8 b 8.3 t6 thousands of trips. For Ghana the i6 a$ 7.9 '84 year 1972 is not represented 74 . . 73. 75, (CPUE= 400). O ' I I 200 ' 400 ' 600 800 1000 2.1.5- Development of S. au rita abundance in Côte-d'/voire and changes in the species composition of catch In Côte-d'Ivoire, the average abundance (CPUE) of S. aurita observed during the eighties was ten times higher than duiing the previous years (Fig. 7). Indeed, this species became more important than S. maderensis and Bracbydeuterus au,-itus (Pezennec, 1994), which had dominated the catch of small pelagics in the 1960s and 1970s in this country. Year 2.1.6- Extension of the spatial distribution of S. aurita The increase of the abundance of S. aurita off Côte-d'Ivoire was accompanied by a spectacular extension of its distribution to the western part of the county. Since 1980, the CPUE of purse seine operations in western Côte-d'Ivoire have exceeded those from the east except for four years (Fig. 8). A similar increase was seen in the western part of Ghana whcre, although landings in the western areas always exceeded those from the est, the difference has widened since 1987. There were no sirnilar changes in the distribution of the other small pelagic species in either country. 4 Côte-d'Ivoire t ,' \, East /-. \ 1970 1975 1980 1985 1990 Year Fi);. 8: Abundance of Sardinella aurita off the wmtern and eastern parts of Côte-d'lvoire (1 066-1 993) and of Ghana (1 982-1 993). Cote-d'Ivoire: t per day of search for purse seiners. Ghana: kg per trip for APW nets. 01 1985 1990 Year 2.7.7- Importance of the S. aurita abundance during the minor upwelling season Figure 9 presents CPUE values recorded during two periods, the major upwelling season and the minor upwelling season in Côte-d'Ivoire and Ghana. Whereas, in Ghana, CPUE values in the major upweliing are always higher than those in the minor upwelling, the situation is different in Côte-d'Ivoire. In the latter country, CPUE values for the GSF and PSF were equally important. Values in the GSF increased drarnatically in 1981 and in subsequent years. The PSF has assumed greater importance since 1983-84, and in 1987, the CPLTE recorded during this season exceeded that of the GSF. Also in Côte-d'Ivoire, there is a difference between the western and eastem parts of the country. In western Côte-d'Ivoire, the minor upwelling is as important as the major upwelhg for the sustainability of the species (Pezennec, 1994). Côte-d'lvoire 30 Year 500 Ghana Fig. 9: Annual abundance (CPUE) of Sardinella aurita during the minor (PSF) and major (GSF) upwelling seasons off w 300 / Y:''! 3 Côte-d'Ivoire (1 966-1 903) and off Ghana 3; '! / '% O 200 : . . _-a V (1 982-1 993). Côte-d'Ivoire: t per day of search for purse seiners. Ghana: kg per 1 O0 PSF trip for APW nets. 1985 1990 Year In al1 these, the limitation on the use of CPUE in pelagic resource assessment, as discussed in many studies (e.g., Saville, 1980) need not to be overlooked. One such limitation is that of the spatial distribution of the stock and (or) the fleet is reduced, the CPUE could remain constant (or even increase) even though the biomass have decreased. In this case, the abundance of S. auv-ita has increased as well as his spatial distribution. 2.2. Biological changes 2.2.7- lncrease in sires of fish caught In the early 1960s and 1970s, the modal size (fork, length) of S. aurita caught off Côte-d'Ivoire was between 15 and 18 cm (ORSTOMPRU, 1976; Fig. 10). During the 1980s, this modal size was between 18 and 24 cm and a similar increase 336 Pelagic Fisheries off Côte-d'Ivoire and Ghana of the maximum size of the fishes caught was also observed. In Ghana, an increase of the modal size was also observed, from 14-17 in the early 1960s to 17-21 in the 1980s. Also 3bserved is an increase in the length at first maturity of S. aurita in Côte-d'Ivoire from 15-16 cm in 1969 to 19-20 cm in 19% (Fig. 11). Quaatey (1993) similarly noted an increase for female fsh caught in Ghanaian waters, from 14. j cm to 17.1 cm. Côte d'Ivoire OoO ,. O Maximum observed length 1 . Maximum modal length ximum modal length and erved length (fork length, in cm) of Sardinella aurita caught off Côte- d'Ivoire and Ghana from 1963 to 1990. Fig. 11: Sexual maturity of the Sanjinella aurita caught off dur ng the 1969 and 1990 upwelling seasons: percentage of mature fishes obser adjusted values (logistic function i . Maximum modai length 10J , 1965 1970 1975 1980 1985 1990 Year - Observed. minor upw. season -- Adjusted, minor upw. season - Observed, major upw. season - - Adjusted. major upw. season - Observed. minor upw. season - - Adjusted, minor upw. season - Observed. major upw. season - - Adjusted. major upw. season 14 16 18 20 22 24 Fork length (cm) O. PEZENNEC AND K.A. KOFANTENG 33 7 2.2.2- Changes in spawning activity S. aurita was known to spawn mainly during the upwelling seasons (ORSTOMPRU, 1776), especially during the major upwelling. In Côte-d'Ivoire, observed GSI were during the minor upwelling season and part of the warm season (March, April), as large as during the major upwelling season (Fig. 12). In recent years, it has become clear that both cold periods occurring off Côte-dlIvoire,are fully utilized by S. aurita for spawning. Quaatey (1793) also reported an increased gonadai development and spawning activity of the S. aurita outside the major upwelling season. No such changes were obseived for S. maderensis (Pezennec, 1774). -. IGS - Temperature Fig. 12: Gonado-somatic index of Sardinella aurita caught off Côte-d'lvoire and mean temperature (from warmer to colder) off Tabou and Abidjan. Monthly means, 1989-1 991. O . 29 1989 1990 1991 Year 2.3. influence of the biological changes on the dyna.mics of fish and fishery The general increase in size of fish has effect on the total weight of fish landed. In Côte-d'Ivoire, for example, similar numbers of S. aurita individuals were caught during the 1986 (120 millions) and 1988 (140 millions) fishing seasons, but the landed weight in 1986 was double that of 1788 (Pezennec, 1994). Al1 things being equal, the increase in size at first maturity and modal size should result in increase in the fecundity of the fisli. A fish of 23 cm long is expected to release a quantity of eggs twice as high as a 18 cm long fish (Fontana and Pianet, 1973; Boëly, 1982). This increase in fecundity may result in an increase of recmitment and hence, of the total biomass of the fisli. 3.1 . Hypotheses The observed changes in the dynamics and biologv of S. aurita surely constitute a puzzle, the solution of whicli possibly requires a deeper understanding of the fishery and the nature of changes in the physicochemical parameteis of the 338 Pelagic Fisheries off Côte-d'Ivoire and Ghana Ivoiro-Ghanaian ecosystem. Some of these changes in the past led to certain hypotheses being proposed by researchers. These hypotheses were based on observations of the fluctuations in the biotic and abiotic components of the ecosystem. One of the earliest suggestions attributed the decline of S. aurita, in part, to the increase of triggerfish Balistes capriscus, a semi-pelagic fish in coastal waters off Côte-d'Ivoire, and especially off Ghana (ORSTOMPRU, 1776). B. capriscus drastically declined since 1988 (FAO, 1992). It is obvious now that, although the rise and faIl ofB. capriscus was obsenred beween 1970 and 1988, and the decline of S. aurita occurred during the early part of this period, the recovely of the rourid sardinella began before the decline of the triggerfish. Furthermore, except perhaps for their juveniles, the two spec.ies do not have the sarne ecological requirements (Caverivière, 1991). Varb~us models developed to facilitate understanding of the dynamics of the pelagic resources of Côte-d'Ivoire and Ghana (Binet, 1782; Culy and Roy, 1987) failed to explain the increase in abundance of S. aurita in the 1980s (Pezennec, 1974). 1ncr::ase in wind speed has been suggested to lead to an increase of upwelling off Côte-d'Ivoire and Ghana (Roy, 1992). However, these upwellings are not entirely related to the wind and in fact, the annual values of mean temperature and wind speed or Ekman transport,are positively correlated during the GSF and showed no relationship during the PSF (Perennec and Bard, 1992). Thus, increase of wind speed has not led to an increase of upwelling. However, an increase in this environmental factor leads to increase of superficial mixing and turbulence, which have been hypothesized to increase protluctivity only if wind speed does not exceed 6 m.s-l (Cury and Roy, 1989). Binet et al., (1771) and Herbland and Marcha1 (1991) have attributed the increase of abundance of S. auiita off Côte- d'hi-)ire to changes in water currents and (or) intensity of the upwelling in the western and eastern regions off this couiitry. The 'current hypothesis' is based on the notion that an increase of the westward circulation may have increased the 1:lrift of S. aurita lanqe from the Ghanaian shelf and their retention off Côte-d'Ivoire, resulting in increased recruitment off the latter country. This hypothesis thus implies displacement of the Ghanaian stock of S. auî-ita towards Côte-d'Ivoire and a decrease of recruitment of the species off Ghana. Following this hypothesis, a decline in abundance of S. aurita shoiild occur off Ghana. However, catches and abundance off Ghana have increased just as in Côte-d'Ivoire. Another hypothesis postulated a displacement of the centre of the upwelling off Côte-d'Ivoire from West to east. This, however, is bascd on a short time series of coastal sea surface temperature and obviously contradicts the observed increase in catch and Libundance of S. aurita in the western part of Côte-d'Ivoire. The above hypotheses have failed to explain the dynamic changes in the Ivoiro-Ghanaian ecosystem and in the biology of sardinella, particularly on its Ghanaian side. These changes constitute a puzzle of observed facts which need to be expliined by a hypothesis that would take into consideration the dynamics of the fishery and of the ecosystem, and the biology of the species. 3.2. Ecoiogicai importance of the minor upwelling season .i new hypothesis, which deals with the part played by the second or minor upwelling season in the changes obseived in tlie Ivoiro-Ghanaian coastal pelagic ecosystem was proposed by Pezennec and Bard (1992). This ecosystem is char~cterized by two independent upwellings. However, the influence of the minor upwelling has never been taken into con5ideration as an important event for the productivity of the ecosystem and for the dynamics of the pelagic species. This hypothesis gives greater importance to the minor upwelling in the sustenance of the Ivoiro-Ghanaian pelagic ecosystem, and the role that it plays regarding the biology of the Sardinella aurita and dynamics of the small pelagic fislieries. The hypothesis in question assumes that the intensity of the minor upwelling may have been underestimated and that the difference between the intensities of the two upwellings exhibited a decreasing trend between 1970 and 1990, as slionn by Pezennec and Bard (1992) and Koranteng and Pezennec (this vol.). 3.2.1 - Favourable and unfavourable periods Outside of the major upwelling season, the pelagic species of the ecosystem studied here are faced with unfavourable condition. Several ecological theories insist on the necessity of a global approach to the population-environment system (Barbault, 1981). Taking qualitative approach to the problem of food limitation, one notes that it may be seen as sufficient on a global (annual) basis, but insufficient during a critical period or season. In this case, production of food during this period will be a limiting factor. 3.2.2- The minor upwelling season and productivity of the pelagic ecosystem Studies into the productivity of the upwelling ecosystem have shown the importance of cooling periods outside the main upwelling season. Zooplanctonic biomass is highly correlated with these coolings (Binet, 1983). The minor upwelling season and the other cooling events occuring outside the main upwelling season are low productivity periods for the pelagic ecosystem, and may function as a 'bottlenecks' in term of productivity. Sol an increase of the strength of the minor upwelling season in the ecosystem should be of great importance. 3.2.3- Importance of the minor upwelling season for S. aurita It has been shown that the spawning activity of S. aurita is similar during the minor and major upwelling seasons. This provides S. aurita with extended opportunities for exploitation of the ecosystem in terms of spawning and recruitment. The Guinea Current creates on the eastem side of Cape Palmas and Cape Three Points two areas of cyclonic circulation which favours larval retention (Marcha1 and Picaut, 1977). Thus, the western part of the Ivoirian continental shelf (where the minor upwelling season is most intense) is a favourable area for lama1 survival. The parailel changes of maximum length and length at first maturity are in conformity with the usual relationsl-iip between these two lengths (Beverton and Holt, 1959). According to Pauly (1984), the increase of these sizes should correspond to changes in key environmental factors (decrease of temperature or increase of the availability or density of food) which limit the growth of fish in an ecosystem. Further, the increase of the abundance of S. aurita in Côte-d'Ivoire, first during the major upwelling season, and later during the minor upwelling season, may be explained by MacCall's theory (1990) of density-dependent geographic distribution of biomass. 340 Pelagic Fisheries off Côte-d'Ivoire and Chana 3.2.4- Recovery from a depleted state Figure 6 shows that the stock of S. aurita recovered from its previous depleted state (see Peterman et al. (1979) and CII~ (1991)). liere have been significant changes in the dynamics of small pelagic fishenes in the Ivoiro-Ghanaian coastal mxine ecosystem in the last two decades. After the decline of the fisheries in the early 1970s, total catches of the principal small pelagic species in this ecosystem (especialiy of the round sardinelia, Sardinella aurita) increased iri the 1980s and early 1990s. The stock of S. airrita appears to have recovered from its depleted state, especially off Côte-d'Ivoire. Also observed are changes in some asnects of the biology of S. aurita. These changes are not in conformity with earlier hypotheses put forward to explain the d)riarnics of smaii pelagics in the western Gulf of Guinea or other ecosystcms. However, most of the observed changes in the bildogy and dynamics of S. aurita resources can be attributed to the increasing impact of the minor upwelling on the ec:)svstem. This minor upweüing, which is more intense off Côted'Ivoire than off Ghana, occurs during an environmentaliy ur Favourable pend of the year for the productivity of the pelagic ecosystem, and which acted as a 'bottleneck'. No comparable changes have been observed in the biology and dynamics of the other small pelagic species of this ecosystem. Cury and Fontana (1988) have shown, for example, that S. aurita and S. maderensis have different demographic and adaptive strategies. According to these authors S. aurita is more sensitive to environmental fluctuations and could take advantage of them. Therefore, utilization of the relative changes in the intensity of the two upwelling se~isons could be an illustration of this difference between the mo sardinella species. The stocks of the Indian oil sardine (Sardinella longiceps; Longhurst and Wooster, 1990) and sardinellas in the Benguela system (S. aurita and S. maderensis; Crawford et al., 1987, have experienced similar changes in abundance to those objerved in the Ivoinan and Ghanaian ecosystem. It would be very interesting and useful to do a comparative study of these ecosystems and their populations. We are grateful to Dr. F.X. Bard and Mc. S.N.K. Quaatey for their comments. Through the French Ministry of Cmiperation and Development, the DLTSRU (Dynarnics and Uses of Sardinclla Resources from Upwelling off Ghana and Côte-d'Ivoire) programme provided the means to perform the work presented here. ive thank the CEOS (Climate and Eastern Ocean Systems) project for sponsoring our participation at the Monterey meeting of September 1994. Barbault R. 1781. Ecologiedespopulationsetdespeuplen~ents. Masson, 200 p. Beverton*R.J.H. and S.J. Holt. 1757. Areview ofthe lifespans and mortality rates of fish in nature, and their relation to growth and other physiological characteristics. p. 142-177. In: G.E.W. Wol- stenholmen and M. O'Connor (eds.). Ciba Found Colloquium on Ageing: the lifespan of animals. Vol. 5, Churchill, London. Binet D. 1782. Influence des variations climatiques sur la pêche- rie desS. aurita ivoiro-ghanéennes: relation sécheresse-surpêche. Oceanol. Acta, 5 (4) : 443-452. Binet D. 1783. Zooplancton des régions côtières à upwellings sai- sonniers du golfe de Guinée. Océanop. trop., 18 (2): 357-380. Binet D., E. Marchal and 0. Pezennec. 1771. Sardinella aurita de Côte-d'Ivoire et du Ghana. Fluctuations halieutiques et chan- gements climatiques. In: P. Cury and C. Roy (eds.). Pêcheries ouest-africaines. Variabilitk instabilité et changenzent. ORS- TOM, Paris: 320-342. Boëly T. 1782. Etude du cycle sexuel de la sardinelle ronde (Sap dinella aurita Val. 1847) au Sénégal. Océanogr. trop., 17 (1): 3-13. Caverivière A. 1771. L'explosion démographique du baliste (Balistes carolinensis) en Afrique de l'ouest et son évolution en relation avec les tendances climatiques. In: P. Cury and C. Roy (eds.). Pêcherles ouest-africaines. Variabilité, instabilité et changement. ORYTOM, Pans: 354-367. Cu. P. 1771. Les contraintes biologiques liées à une gestion des ressources instables.in: P. Cury and C. Roy (eds.). Pêcheriesouest- afiïcaines: uariabilitk instabilité et changement. ORSTOM, Paris: 506- 518. Cury P. and C. Roy. 1787. Upwellinget pêche des espèces péla- giques côtières de Côte-d'Ivoire: une approche globale. Ocea- nol. Acta,lO (3): 347-357. Cury P. and A. Fontana. 1788. Compétition et stntégies démo- graphiques comparées de dei~x espèces de sardinelles (Sardi- nella aurita et Sardinella twaderensis) des côtes ouest-afri- caines. Aquat. Liuing. Ressour., 1: 165-180. Cury P. and C. Roy. 1787. Optimal environmental window and pelagic fish recruitment success in upwellingareas. Can. J. Fish. Aquat. Sci ,46: 670-680. Crawford R.J.M., L.V. Shannon and D. E. Pollock. 1787. The Beii- gueia ecosystem. Part IV. The major fish and invertebrate resources. Oceanog?: Mar. Biol. Ann. Rev., 25: 353- 50 5. FAO. 1772. Rapport dugroupe de travailad hocsur lesstocks depélagiques côtiers du goye de Guinée ouest (Côte-d'ivoire- Ghana-Togo-Bénin). 10-16 Décembre 1770, Abidjan. Fontana A. and R. Pianet. 1773. Biologie de Sardinella eba pal.) et Sardinella aurita pal.) des côtes du Congo et du Gabon. Doc. scient. Cent. ORSTOM Pointe-Noire, 430,21 p. Herbland A. and E. Marchal. 1771. Variations locales de l'up- welling, répartition et abondance des sardinelles en Côte-d'ivoi- ce. In: P. Cury and C. Roy (eds.). Pêcheries ouest-africaines. Variabilité, instabilité et changenzent. ORSTOM, Paris: 343-3 j3. Longhurst AR. and W.S. Wooster. 1770. Abundance of oil sar- dine (Sardinella longiceps) and upwelling on the southwest coast ofIndia. Can. J. Fish. Aquat. Sci, 47: 2407-2417. MacCall A, 1770. Dvnarrlic ge0graph.v of nlarinefish popula- tions. Washington Sea Grant, University of Washington, 153 p. Marchal E. and J. Picaut. 1777. Répartition et abondance éva- luées par échointégration des poissons du plateau continental ivoiro-ghanéen en relation avec les upwellings locaux. J. Rech. Océanogr., 2(4): 37- j7. Mensah M.A and K.A. Koranteng. 1788. A review of the ocea- nography and fisheries resources in the coastal waters of Ghana, 178 1-1786. ~MarlneFisher)~ Research Reports, 8. Fisheries Depan- ment, Research & Utilization Branch, Tema, Ghana. Ni I.H. and E.J. Sandeman. 1784. Size at maturity for northxest Atlantic redfishes (Sebastes). Can. J. Fish. Aquat. Sci. ! 41: 17 j3- 1762. ORSTOM/FRU 1776. Rapport du groupe de trauail sur la Sar- dinelle (Sardinella aurita) des c6tes iuoiro-ghanéennes, Abid- jan 28 juin3 juillet 1776. Pauly D. 1784. A mechanism for the juvenile-to-adult transitiori in fishes. J. Cons., Cons. int. Explor. Mer, 41: 280-284. Peterman M.R., W.C. Clark and C.S. Holling. 1777. The dynaniics ofresilience: shifting stability domains in fish and insect systeriis. In: R.M. Anderson, B.D. Turner and L.R. Taylor (eds.). Popula- tion dwamics. Blackwell Scientific Publications: 321-341. 342 Pelagic Fisheries off Côte-d'Ivoire and Ghana Rzennec 0. 1994. Instabilité et changenzents dans l'écoq~stè- nirpélagique côtiuiuoiro-ghanéen. Variabilité de la ressour- ce en sardinelles: faits, hypothèses et théorie. Thèse de Docto- rar Océanologie Biologique. Université de Bretagne Occidentale, Brest. Pezennec 0. and F.X. Bard. 1992. Importance écologique de la petite saison d'upwelling ivoiro-ghanéenne et changements dans la pêcherie de Sardinelin aurita. Aquat. Liuing Resour., 5::!49-2 59. Quaatey S.N.K 1993. Longteniz changes in the rep~oductice ci~cle andpopulationparanletus of Sardinella auiita rn Ghanninn zunters. M.Sc. Disseitation, University of Wales, U.K. Roy C. 1992. Réponses des stocks depoissonspélagiques 2 la dllnanzique des upiuellings en Afnque de l'ouest: anall~se et ~nodélisation. Tlièse dr. Univ. Bretagne Occid., Brest, 149 p. Sa\.ille k (ed.). 1980. The assessilient and nianagernent of pel;igic fish stocks. Ropp. P.+. Réun. Cons. int. hplo~: ~Zler, 177, j17 p. O. PEZENNEC AND K.A. KORANTENG 343 Stock Assessrnent of Sprat and Whiting in the Western Black Sea in Relation to Global and Local Anthropogenic Factors KAMEN B. PRODANOV* GEORGI M. DASKALOV* * KONSTANTIN R.MIKHAILOV* * KONSTANTIN MAXIM* * * EMIN OZDAMAR* * * * VIADISLAV SHLJAKHOV* * * ** ALI~XANDR CHASHCHIN* * * * * AI-IIXANDR ARKHIPOV* * * * * * lnstitute of Oceanology, Varna, 9000 PO Box 152, BULGARIA ** lnstitute of Fisheries, Varna, 9000 PO Box 72, BULGARIA ***lnstitute of Marine Research, Konstanta, 8700, ROUMANIA ****College of Fisheries, Ondokus Mayis University, Sinop 5700 TURKEY *****South Research lnstitute for Fisberies and Oceanolography, Kerch 334 500, UKRAINE Historical stock assessments of the Black Sea sprat (1957- 1992) and whiting (1976-1992) have been performed using Virtual Population Analysis. Relationships between fish stock parameters (recruitment, spawning biomass, mortality rates) and environmental variables (wind speed and duration, sea temperature, light, phyto- and zooplankton biomasses) have been analyzed using multiple regression mociels. Strong correlation has been found between sprat recruitment and western winds dunng November-December and January-March. The western winds force the upwelling of deep waters and their progress shorewards. As the upwelled waters are rich in nutrients and organic matter, they contribute to the intense productivity in the Black Sea. The role of the other variables appears to be less significant. The need for including more reliable data on plankton and ctenophore !Mnenziopsis maccradii in the analysis is pointed out. L'évaluation à long terme du stock du sprat Sprattus sprattus L pour la période 1957-1992 et du merlan Merlangius merlangus euxinus de 1976 à 1992 en mer Noire a été effectuée par l'analyse des cohortes (N'A, analyse virtuelle des populations). Les relations entre les paramètres des stocks de poisson (recrutement, biomasse féconde et taux de mortalité) et des variables environnementales (vitesse du vent et durée des événements de vent d'ouest, température de la mer, activité solaire, biomasses du phyto- et du zooplancton) ont été analysées avec des modèles de régression. Une corrélation importante a été trouvée entre le recrutement du sprat et les données du vent d'ouest pour novembre-décembre et janvier-mars. Le vent d'ouest est responsable d'un upwelling côtier qui apparaît à certains endroits en fonction de la topographie. La remontée d'eau enrichie en éléments nutritifs et en matière organique, qui contribue aux taux de production, augmente dans ces régions. La contribution des autres variables est moins significative. L'importance d'analyser des données de meilleure qualité sur le plancton et sur le cténophore Mnevziopsis maccradii est soulignée. During the last 30-35 years, the Black Sea ecosystem has been subjected to dramatic changes due to the increased pollution of the basin and the overexploitation of the main commercial fish species. The period of eutrophication dating back to the early 1970s is characterized by structural and functional alterations in the ecosystem as a result of the intensification and spreading of both local and regional phytoplankton blooms. These blooms over the last decade attain their maximum intensity in late spring-summer, a period abnormal for the Black sea, where peak production normally occurred in early spring and autumn. Changes have also been registered in the taxonomie composition of bloom- producing phytoplankton species with succession shifted towards the predominance of Dinophyta and since 1989 - towards an increasing importance of Cbrysopbyta species - Emiliania buslqi and Pbaeocystis poucbettii (Moncheva, 1991, 1992). Recently some new phyto- and zooplankton species for the Black Sea ecosystem have invaded the basin resulting in drarnatic alterations in the food web (Moncheva et al., 1995). During the period under consideration, the abundance of the most common carnivores has sharply decreased: bonito (Sarda sarda Bloch), bluefish (Ponzato~~zus saltator) and mackerel (Scomber scombrus) have almost become extinct in the Black Sea since 1986. This has been the period of rapid intensification of fishing particularly of sprat, horse mackerel (Tracburus mediterraneus ponticus) and - - - 346 Whiting and Sprat in the Black Sea anchovy (Engraulis encrasicolusponticus) catches of which have ken extended from 3.1,4.9 and 193.3 103 t (1970) up to 105.2 (1989), 147.7 (1985) and 502.6 (1984) thousand t, respectively. Shljakhov et al. (1990) claim that the rapid decline of sprat stock is related hth to the deteriorated environmental conditions and to overfishing. In 1982, the ctenophore Mzerniopsis leidyi invaded the Black Sea (Zaitsev, 1994; Konsulov and Konsulova, 1993) with a biomass attaining its mnimum in 1990-1991 and decreasing thereafter. In 1991 and 1992, the ctenophore biomass was 40.9 and 18.6 million t, reipectively. This has resulted in a reduction in the biomass of zooplankton consumed by fish, and of copepod species in pa~ticular (Vinogradov et al., 1989; Zaika and Sergeeva, 1991). Taking into account the fact that M. leidyi feeds on eggs and laivae of spawning fish, although at less significant rate (Eremeev and Chudinovsky, 1990) it is reasonable to assert that the shlirp reduction in sprat, anchovy and horse mackerel stocks is due mainly to the complex impact of the four above m-ntioned factors: pollution, eutrophication, structural alterations in the ecosystem and intensification of exploitation. Al1 fox factors are of local origin and should be distinguished from factors such as global climatic changes, and tlieir impact ori the hydrology and hydrochemistry of the basin (Brjantzev, 1989). An example of such global effects is that of sunlight ori phytoplankton (Petrova-Karadjova, 1993; Petrova-Karadjova and Apostolov, 1988). Tlie results of virtual population analyses (VPA) are reported; they aimed at the assessment of sprat abundance and bii:)mass dynamics over the period 1957-1992. Based on these results, an attempt is made to highlight the impact of some at iotic and biotic factors related to global and local anthropogenic causes. 1. MATERIAL AND METHODS The sprat data for VPA to be applied over the period 1957-1992 were collected and analyzed in conformity with the rejearch program of the international project Environmenral management of Black Sea fish resources and their rational exploitation, funded by the Central European University Foundation. Experts from Bulgaria, Romania, Ukraine and Turkey p rticipate in this project, which facilitated the integration of available data, including that of fishery statistics. Thus, it becme possible for the first time to perform a retrospective assessment of sprat stock in the western part of Black Sea, wliere most of the catch originates. Tlie age composition of sprat catches in the ahvementioned regions are presented in Table 1 and Figure 1. Until 1974, thc fishery was carried out mostly by coastal fishing gear (trap nets), while thereafter, it was carried out mainlv from ships. Thus, after 1974, the catch by coastal fishing gear does not exceed 5% of total catch. That is why the calculated fishing ef'ort for the 1975-1992 period is based on the catchleffort of the fishing fleet and those for the preceding period, on the number of trap nets in the Black Sea countries. Another peculiarity that was considered is the space distribution of the di'ferent age groups, as the 3, 4 and 5 years old fishes normally line further offshore, which is why trap net catches are dominated by 1- and 2-year old individuals. With the beginning of the fishing fleet operation deeper waters became eaploited, from 20-25 down to 80-110m which resulted in considerable changes in the age composition of the catch (Table 1, Fig. 1). Still, the 1- and 2-year old fishes dominate over the 3-year old individuals, and especially on the 4- and j- year old fishes (integrated in the age group 4+). The explanation lies in the fact that trawl sprat fishing is of highesr in:ensity in spring-summer months (April-July and in some years in August) at 2040m depth, when the young-age groups (1. and 2-years), move towards the coastal zone For feeding. During this period the thermocline is located at 1 5-30m. Tliat is why the older age groups, which prefer colder waters, may be found in deeper waters, although food availability there is Table 1 : Age composition ( of sprat catches in the w lower. The two periods (1957-1973 and 1974-1992) differ by their predator abundances, which dropped sharply after 1971- 1972, with favorable effect on their prey: sprat, anchow and horse mackerel. Stoyanov (1966) estimated a total mortality rate of 1.14 year-' in 1959 for 1- and 2-year old sprats. The same author, however noted that this figure may be an underestimate, due to the underestimation of 1- year old fish abundance. The analysis of sprat catch suggests that, in that particular year, total mortality was almost equal to natuml mortality, e.g. in the range 1.15-1.20 year-l. Our estimation of sprat natural mortality for the period 1957-1973, based on indirect estimates of the biomass of their predators (mackerel, bonito, bluefish), are presented in Table 2. 348 Whiting and Sprat in the Black Sea 1957 1862 1967 1972 1977 1982 1987 1992 Year Ailei. 1973, a constant M of 0.95 yearl value mras assumed (Domashenko and Ijurev, 1978). The relationship between spawning bioniass (BI*+) and recruitment abundance (R) is determined by the equations of Ricker (197 j) and Ivanov (1977: R = aB e-bB (9 R=~B-~B~ (2) The influence of environmental factors on sprat recruitment was evaluated by applying linear and nonlinear relationships as wcll as multiple linear regression. The c cor relation coefficients (r) for the corresponding equations were calculated from: r = [l-x(Rsi - R~~)~/(R~~)'] ln (3) where : Rsi = theoretical value of Ri; RFi = the observed value of R,; RF = mean value of 4. The :oefficient of determination (D%) as a measure of the extent of the variability in Ri value related to the variability of the esamined factors is defined by: D % = 100 12 (4) The 1:oefficient of indetermination (S%) is a measure of the extent of the variability in Ri due to random factors, and is defin-d by: s % = 100 (1-12) (5) Equz ion (3) is applicable both in linear and nonlinear correlations. - K.B. PRODANOV ET AL. 349 2. RESULTS AND DISCUSSION The changes sprat in spawning standing stock biomass (BI*+) from 1957-1992 in the western part of Black Sea are depicted on Figure 2. During the 1967-1992 period, sprat biomass was determined by tnwling, and hydroacoustics. The results of these surveys can be compared with the VPA estimatesi Table 3). Fig. 2: Sprat spawning biomass (by age groups) in the western part of the Black Sea during 1957-1 992. 1957 1962 1967 1972 1977 1982 1987 1992 Y ear ) Year Surveys 1 VPA Year 1 Surveys VPA 1 ass (1 O3 t) in the western part of ack Sea during 1967-1992, as estimated by nd by VPA. 350 Whiting and Sprat in the Black Sea Th':: biomass estimated by hydroacoustics refer to the entire Black Sea. As obvious in some specific years the differences bei:ween the two estimates are considerable, particularly in 1980 and 1986 (880 and 384 t.103, respectively). Similar differences occur beween the VPA estimates of recruitment and the fingerling abundances estimated by ichthyoplankton surveys conducted in May-June (Table 4). 1 year 1 surveys 1 VPA Y ear Survey s VP A uring 1967-1 992, from ichth surveys and VPA. Th,: sources for the differences in the survey and VPA estimates in Table 3 and 4 are not obvious, and, pending detailed stuiiy assumed to be due to the heterogeneity of the data we used. 331: estimates of the parameters of equation (1) and (2) and their cormpnding dues for Bop, and are presented in Table j. Equations / a b r 1 Il% 1 S'Y6 1 B,t 1 Rmax 1 e correspon pt (t.l 03) and R,, K.B. PRODANOV ET AL. 357 The multiple regression model used to investigate the impact of vanous factors on sprat recruitment (R) has the form Where: B is the parental biomass; Z the total mortality; X1 the phytoplankton standing stock; X2 the zooplankton standing stock; X3 the sea surface temperature; X4 the amount of sunlight; Xg the intensity of cosmic rays; and % an index of the Earth's geomagnetism These variables were introduced one at a time, starting with R=f(B) and ending with the full model in (6a). Table 6 summarises the results thus obtained. Sprat biomass along the Bulgarian Black Sea Coast from 19j2 to 1992 is illustrated in Figure 3. The overall trend has a maximum in the period 1974-1979, and decreases to about 35.10~ t after some fluctuations in 1990, after which biomass again increases. The variability in sprat biomass seems mainly dependent on the number of its predators. Their sharp decrease after 1968-1971 resulted in an increase in sprat standing stock, which coincided with a period (1975-1978) of sprat fishery intensification. In 1981 the Bulgarian catches reached the figure 19.109 t, where after a decline occurred, especially during 1990-1991. The latter penod coincides with the invasion of the new ctenophore whose biomass attained its peak in this particular period. The present trend of sprat biomass recovery may be related to the decrease in ctenophore abundance and a decrease of fishing effort. During 1984-1990 regular hydroacoustic assessment of sprat biomass have been conducted in Bulgarian waters, in June- September, excepting the last three years, when the surveys were conducted in June. The results of these investigations together with the VPA estimates are presented in Table 7. betwe bundance of sprat recruits ta tirn 352 Whiting and Sprat in the Black Sea Year long ack Sea coast from Fro1r.i 1957 to 1992, the sprat biomass along the Bulgarian Black Sea coast comprised about 23% of total sprat biomass in the ,NeStern Black Sea, but in 1989 and 1990, it represented only 8. 5 and 12.0% respectively. Mean 51.6 56.6 The values of the parameter of equation (1) and (2) concerning the relationship between sprat biomass and recruitment off t'le Bulgarian coast are presented on Table 8. Year Surveys VPA [ Equations 1 a b r 1 D% 1 S% 1 Bop, 1 Kmax 1 / 0.98834 0.00607 1 0.390 1 15.2 1 84.8 1 164.8 1 59.9 1 1987 16-28 57.6 The impact of fishing mortality rate during 1973-1989 is estimated by the equations: 1984 24.0 76.2 The estimated parameters of these equations are presented on Table 9. This shows that fishing mortality exens a strong impact on the stock-recruitment relationships. 1988 74.0 50.1 K.B. PRODANOV ET AL. 353 1985 69.0 71.3 1989 25.0 37.9 1986 77.0 67.8 1990 70.0 35.6 Table 9: Parameter equations (7) and (8). The second multiple regression model used to investigate the impact of various factors on sprat recruitment (R) has the form : R= aB + bF + cX1 + dX2 + ex3 + fXq + gX5 + 4 (6b) where B is the parental biomass; F the fishing mortality; X1 the duration of the western wind component in November-December; X2 the mean speed of the western wind component in November-December; X3 the duration of the western wind component in January-March; X4 the mean speed of the western wind component in January-March; X5 the intensity of cosmic rays; and X6 the amount of sunlight. r 0.771 0.796 D% 59.5 63.4 As for equation (6), these variables were introduced one at a time, starting with R=f(B) and ending with the full model in (6b). S % 40.5 36.6 Equations 7 8 It is known that the western winds generate an upwelling of deep waters and their shoreward progression. As these waters are rich in nutrients they contribute to phytoplankton bloom in summer. Because of their low temperature and low O2 content they may also be responsible, however, for regional zoobenthic mortality, including of fish species (Rozhdestvenski, 1767; Kolarov, 1770; Dimitrov and Yaneva, 1772). c 1.4732 0.9109 During the winter sirnilar upwelling of deep waters occurs, but unlike in summer, their temperature is higher than at tlie surface. For species that reproduce in winter, such as sprat, the upwelling of the deep waters towards the shelf is very dangerous such that this tends to lower the survival of the eggs and lanlae. d - 0.0092 a 1.1356 1.1082 That is why the wind factor is represented by four variables: XI (western wind duration in November-December with a speed above 5m/sec - in hours); X2 ( average wind speed in November-December) with X3 and X4 as the corresponding values for January-March. b 0.0041 0.0035 As is apparent from the results (Table IO), the influence of wind speed is higher than that of wind duration. Wind velocity has a favorable effect while wind duration has a negative one. The joint impact of these variables on recruitment survival is considerable with r = 0.843 (D% =73.0%; S% = 27.0%). The stock-recruitment relationship of sprat, depends both on the predator abundance and on the number of the days with water temperature below b"C during the reproduction period (November-March). The same author points to the fact that the influence of water temperature is indirect and occurs via the food web. Similar conclusions are drawn by Feldman (1786) who has reported a strong correlation between water temperature and the recruitment of sprat and cd (Gadus morhua) in the Baltic Sea. During severe winters sprat recruitment is less abundant while that of cod, increases. In warmer 354 Whiting and Sprat in the Black Sea winiers this relationship is inversed. The same author claims that low temperatures in February-March restrict zooplankton reproduction rate in the North and Baltic Sea. Gapshko and Malyshev (1990) documented that small sprat (2.5-3.0 cm of length) feeds principally on Calanus and Pseicdocalanus nauplii. According to these authors the number of individuals with empty stomach depends on the stability of the ratio berween nauplii, copepodites and adult forms, as well as on the availability of other nutritive and non- nutritive zooplankton. Protlanov and Konsulov (1987) established that sprat recruitment along the Bulgarian Coast is strongly influenced by status of the food web and by whiting biomass, whose reproduction occurs during the same months and at the same depths for sprat reproduction. During its reproductive period, whiting still actively feeds on sprat and large decapod crustaceans. Small whiting has almost the same food spectrum as small sprat, which implies a high food competition. Since 198j, the recriitment of both species declined resulting in a reduction in their standing stocks. Table 11 presents the numbers and total biomasses of sprat and whiting for the period 1976-1991, and estimates of whiting recr~its (O+) in the Western Black Sea. As might be seen, the sprat abundance decreased by a factor of two from the begirining to the end of the series, while that of whiting declined by 1.4. In this particular case the critical factor seems to be the expansion of ctenophores, which, after their introduction in 1982, sharply expanded in biomass. The estimates of sprat biomass from 1957 to 1992 show great fluctuations, as it has already been stated this may be due to a variety of reasons. Until 1971-1972, environmental conditions and predators play a decisive role. During 1978-1989 the key factors are ihe intensification of the fishery and environmental alterations, especially the invasion by ctenophores. After 1989 a lowtring of pollution of the basin occurred as a result of the economical recession in the former USSR, Romania and Bulgaria followed by a reduction of their industrial production: the chemical industry in particular is operating at present at oiily 30% of its total capacity. As a consequence during the last 1-2 years, a decline in the phytoplankton blooms occurred, along with a decline of their adverse effects: toxicity, O2 deficiency, asphyxia, etc. (Moncheva et al., 1995). Since 191, the ctenophores also exhibit a decreasing trend. As a consequence a recovery in sprat stock has occurred. The redu~tion in fishing effort after 1W is also thought to here contribute positively to this recovery. K.B. PRODANOV ET AL. 355 Year 1 Sprat 1 Sprat 1 Whiting 1 Whiting Whit ing 1 biomass recruits 1 biomass recruits 1976 186.1 52.7 27.3 1.55 40.0 Fish recruitment abundance depends on numerous natural and anthropogenic factors and its dynamics are difficult to predict. Among the natural variables, the strongest correlation is found between recruitment and the western wind average velocity dunng November-December and January-March when sprat is actively spawning, at 25-100m depth. The western winds force the upwelling of deep waters and their shorewards progress. As they are rich in nutrients and organic matter, they contribute to the intensive productivity in these regions. The role of the other variables is less important. The comparatively low correlations in equations (6) could be explained to a certain extent by the fact that there is no term in the equation accounting for role of the ctenophore (which is impossible at this stage as there is no long-term data set). The same is valid for phytoplankton bloom data except for the data reported by Moncheva et al. (1995). In the case of 1 Mean 1 90.6 1 29.85 356 Whiting and Sprat in the Black Sea * abundance of whiting recruits in the entire Black Sea, as estimated from ichthyoplankton surveys. 17.5 0.87 11.1 phl,toplankton, the great interannual variability of the species producing blooms restricts the possibility to collect staiistically significant long-term data reliable enough to be included in equation (6). Because they were unsure about the rel.itive importance of various factors they considered, Brjantzev (1989) considered the correlations tliey established as too risliy to rely upon for the determination of sprat recmitment. This is similar to the situation with equations (64 b) wliicli 1e:ives much variance unexplained. The task is now to reduce this unexplained variance and thus to obtain better prcdictions for the factor that determines biomasses. Br .irizev V.A. 1989. Long-term variability of the environment Moncheva S. 1991. Eutrophication, phj~toplankton blooi~is, an11 some problems of fishery prognoses. Coll. Rep. VNIRO, hypoxia. International workshop on the Black Sea. Variia, 30 M( iscow: 282-291 (in Russian). Sept. - 04 Oct. 1991. Diiiiitrov P. and A. Yaneva. 1992. On some negative impact of upwelling phenonienon upon the Black Sea coastal ecosystem. Rep. Conf Eco/ogicai Problerrzs of the Black Sea, Bourgas: 244-2 j4 ( in Bulgarian). Doinashenko G.P. and G.S. Ijurev. 1978. On the problem of spmt toril allowable catch. Coll. Rep. WRO, 128: j7-60 (in Russian). Erc,~meev V.N. and T.V. Chudinovsky. 1990. Hydrobiology of tlic Black Sea. Applied Ecology of Sea Regions. The Black Sea, Kiev: j7-106 (in Russian). Fel3man V.N. 1986. Interannual variations of sprat growth and recr,utment of commercial stock in relation to temperature regims of the North Sea. Dynamics of the commercial fish ablindance. Science, Moscow: 103-112 (in Russian). Moncheva S. 1992. Ecology of comriion Black Seu phytoplankton species utzder the influence ofunthi.opogenic eutrophication. Ph.D thesis (in Bulgarian). Moncheva S., V.1. Petrova-Karadjova aiid A. Palasov. 1995. Harmful algal blooms along the Bulgarian Black Sea Coast and possible patterns of fish and zoobenthic i~iortalities. In: P. Lassus, G. Arzul, E. Erard-Le-Denn, P. Gentien and C. MarcaiIlou-Le Baut (eds.). Proceedings of 6th Int. Conf Toxic Marine Phj~top[ankton, 18-22 Oct.1993, Nantes, France, Lavoisier, Paris: 193-198. Petrova-Karadjova V.J. and E.M. Apostolov. 1988. Influence of solar activity upon the diatoms of Black Sea plaiikton. Rapp,, P.-v. Réun. Cons. int. Exploi: ,Ver, 31 (2). Ga:iishko A.I. and V.I. Malyshev. 1990. Calculating daily Petr0v"Karadjova V.1. 1993. Solar control upoli the rat:«ns for sprat under natusal conditions at the time of ~h~to~lankton in the Black Sea. Rapp. P.-LI. Réun. Cons. int. spawning and feeding migration. Living resources of Black &plor 12/ler, 33. Seri Coli. Rep. VNIRO: 39-4j (in Russian). Prodanov K. alid A. Konsulov. 1987. 011 sonie factors Ivaiiov L.S. 1977. A symetric curve on the reproduction of fish determine the d~ndance of O+ !rear old fish (fingerlings) of pol~ulations. Proceedings of the lnstitute of Fisjeries-Varna, whiting off Bukarian Black Sea Coast. 4id~obiolo~i Sofa, 15: 7-1 5 (in Bulgarian). 30: 44-55 (in Bulgarian). Kolîrov P. 1970. On some biological peculiarities of the Black Ricker W.E. 1975. Computation and interpretarioii of biological Sea lichnus. Proc. NIORS-Vama, 10: 91-92 (in Bulgarian). statistics of fish populations. Bull.Fish.Res.Boud Canada! 191. Koiisulov A. and Tz. Konsulova. 1993. Biodiversity on the Rozhdestvenski A. 1969. Hj~drochemical and meteorological Bla:k Sea plankton and benthos. National Biological characteristics of the year 1968 regarding tlie Black Sea Diz,:mi8 years). Fisheries catches are generally highly variable. Caddy and Gulland (1983) 360 Variability of Fish Catches distinguished empincally four classes of fisheries in terms of pattern of their species catch variability: steady-state, cyclical, irregular and spasmodic fisheries and similar classification was provided by Kawasaki (1983). In addition, it is well known th,it the major upwelling areas of the world sustain some of the most productive and irregular fisheries (Bakun, 198 5). Yet, th'-se and related issues of fish catch vanability have not so Far been approached quantitatively. In the present study, the effect of the number of years over which catch variability is calculated as well as between-species and within-species variability were determined for a variety of species in different FA0 subareas (Atlantic, Pacific, Mtiditerranean). The following species/genus were considered: Engraulis spp., Dachurus spp., Sar-dinops spp., Sco~nbei, jai'onicus and Merluccius spp. (the main species found in the four major upwelling areas of the world: Bakun, 198 5) as wt,ll as Sardinella spp., Sardina pilchardus, Mallotus villosus, Scoilzber spp., Clupea spp., Micr-onzesistius poutassou, Pll?uronectes platessa, Hippoglossoides platessoides, Gadus spp., Ther-agra chalcogranznza, Thunrzus albacmes, k2zîsutuonuspelamis, Xiphias gladius and Elasmobranchii. The above mentioned species make up more than 40% of the anlual world marine catch. 1. MATERIAL AND METHODS Data was extracted using FISHSTAT-PC (FA0 Fisheries department, Release 19934 April 1993). Ovenll, 109 species pe- FAO-subarea catch records, referring to the 1970-1991 period, were analyzed. Catch records of species/subarea intluding many zero values were not taken into account. Scientific names of species are according to FA0 Bulletins. The mc i-hodology used has been described by Pimm and Redfearn (1988). The measure of variability (standard deviation of logged catch: SDL) and time scales over which SDL was calculated were both selected such that the results pi-esentecl here coiild be directly comparable with those of Pimm and Redfearn (1988). For al1 species/subarea catch records analyzed, SDL was calculated over 2,4, 8, 16 and 22 years (le., the maximum penod available) and calculations refer to nested data (i.~ , first two years, first four years, etc.). The results indicated that for the majority of species/subarea catches analvzed, variability increases with the length of timi: over which it is calculated (Tables 1 and 2; Fig. 1). Hence, the slopes of the regressions between SDL and time period ovc-r which SDL was calculated, were significantly (P<0.05) different from zero and positive for 63 (58%) species/subarea catches when only 2,4,8 and 16 years were considered (i.e., 4-point regressions) as well as when regressions also includecl SDI. calculated over the total period (Le., 22 years, j-point regressions; Table 1). Overall, slope values were negritive for five species/subarea catch senes (4-point regressions: Katsutuonuspelamis in SW and SE Pacific and Xiphiasglaclius in SE Pac fic; 5-point regressions: Katsutuonus pelamis pelamis in SW and SE Pacific and Xiphias gladius in SE Pacific; 5-point regïessions: Katsutvonus pelamis in SW Pacific and Thunnus albacares in NW Pacific: Table 1). The arithmetic values of the slopes (excluding the negative ones) ranged from 0.0024 to 0.0590, for the 4-point regressions, and from 0.0016 to 0.1>00, for the 5-point regressions (Table 1). Species FA0 Subarea 24,816) (2,q,6,8,16.22) Engrnulis anchoita Engraulis capensis Engraulis encvasiculus Engraulis encrasicirlus Engraulis encrnsiculus Engraulis japoninis Engraulis lliordm Engraulis nngens iV1erlziccius bilinearis :tl.capensis, M.paradox ~Merluccius gqi ,VIerluccius hubbsi :Ilerluccius nrerluccius ~V1erluccius ii~erluccius ~lerluccius ~~ierluccius :1.lerluccius productus ,Werluccius senegalensis Scirdina pilchardus Sardina pilchardus Sardina pilchardus Sardinella aurita Sardinella aurita Sardinella brasiliensis Scirdinella gibbosa Sardinella laiiuni Sardinella lnaderensis Sarn'inops caeruleus Sardinops elano an os tic tus Sardinops ocellatus Sardinops sagm Sconiber japonicus Scoinber jnponicus Scolilber japonicus Scon~ber japonicus Scoli~ber japonicus Scolnber japonicus Sconrber japoiiicus Scolnber japonicus Scoli~ber japonicus Scowrber spp. Trachunrs capensis Trachunrs japonicus Tracbutus nrurphj'i Trachurus picturatus Trachurus yilriizehicus rbr Atlantic, Soiitliwest 0.46 - 0.32 Atlantic, Soiitheast 0.73 - 0.89 Atlantic, Northeast 0.93 0.015 0.91 Atlantic, Eastern central 0.95 0.007 0.96 Mediterranean 0.94 0.0032 0.96 Pacific,Northwest 0.94 0.004 0.95 Pacific, Eastern central 0.92 0.008 0.95 Pacific,Soiitlieast 0.88 0.031 0.82 Atlantic, Northwest 0.95 0.016 0.88 Atlantic, Southwest 0.96 0.009 0.96 Pacific, Southeast 0.91 0.015 0.88 Atlantic, Soutliwest 0.93 0.013 0.89 Atlantic, Eastern Central 0.52 - 0.41 Atlantic, Northeast 0.49 - 0.62 Mediterranean 0.95 0.007 0.95 Pacific, Northeast 0.96 0.009 0.98 Atlantic, Eastern Central 0.71 - 0.69 Atlantic, Eastern Cenral 0.57 - 0.6 5 Atlantic, Northeast 0.47 - 0.42 Mediterranean 0.99 0.0044 0.98 Atlantic, Eastern Central 0.93 0.0198 0.95 Atlantic, Western Central 0.81 - 0.89 Atlantic, Southwest 0.61 - 0.86 Pacific, Westen Central 0.94 0.014 0.92 Pacific, Western Central 0.96 0.0175 0.94 Atlantic, Eastern Central 0.94 0.0341 0.96 Pacfic, Eastern Central 1 0.0197 0.98 Pacific, Northwest 0.93 0.0324 0.87 Atlantic, Soutlieast 0.93 0.022 0.92 Pacific, Southeast 0.92 0.0496 0.86 Pacific, Northa7est 0.98 0.004 0.97 Atlantic, Soutlieast 0.97 0.013 0.96 Atlantic, Northeast 0.98 0.017 0.9 Pacific, Eastern Central 0.99 0.059 0.95 Aitantic, Southwest 1 0.018 0.94 Pacific, Western Central 0.98 0.018 0.95 Atlantic, Western Central 0.23 . 0.1 Atlantic, Eastern Central 0.9 0.007 0.94 Pacific, Southeast 0.87 0.022 0.76 Mediterranean 0.98 0.0078 0.98 Atlantic, Southeast 0.82 - 0.7 Pacific,Northwest 0.89 0,008 0.89 Pacific, Southeast 0.98 0.03 0.97 Atlantic, Southwest 0.72 - 0.7 Pacific, Eastern Central 0.61 - 0.82 min max niean 10137 41100 20565 Table 1 (to be continued on next two pages): Minimum, maximum, mean and (maximum/minimum) ratio values of the 107 species/FAO-çubarea catches over the 1970-1 991 period examincd in the present study together with the correlation coefficient (r) and the slope (b) values of the regressions of SDL versus the length of period for 4 data points (i.e., 2, 4, 8, and 16 years) and 5 data points (Le., 2, 4, 8, 16 and 22 years). Slopes not shown were not significantly different from O (Pz0.05). 362 Variability of Fish Catches Species FA0 Siibarea (2,4,8,16) (2,4,6,8,16,22) r b r b min ma mean mau/rnin Trtrchums trachums Atlantic, Eastern Central 0.7 - 0.57 - O 3118 846 - Trachums trachums Trt~chums trecae Trtichums spp. Cli peu harengus Clupea harengus Cli!pea pallasii Clilpeu pallasii MG lloius vilIosis hf~:lloius villoms ML llotus villosus Mi:.rontesistius poutmou PIE lrronectes platessa Hij~poglossoides platessoides Grirlus lnorhua Gridus l~orhua Gridus ntacrocephalus Gt~~ius nlacrocephalus Thtrragra chalcograninia Thcragra chalcograntnta Ka.J-uwonus pelaniis Ka;mwonus pelantis Ka;~uuionus pelantis Kaïwtuonus pelantis Ka,::uiuonus pelaniis Ka, ::utuonus pelaniis Kaisuwonus pelamis Karsutuonus pelantis Kal::uulonus pelamis Kaizufuonus pelaniis Ka~::uwonus pelantis Kai:wtuonus pelaniis Thzrnnus albacares Thiinnus aibacares Thirnnus albacares Thiinnus albacares Thunnus albacares Tbirnnus albacares Thi i nnus albacares Thir nnus albacares Th~!rinus albacares Thi~nnus albacares Th1 !nnu albacares Xipbias gladius X@bias gladius Xipbias gladius X$ bias gladius XipSias giadius Atlantic, Northeast 0.93 0.008 0.1 Atlantic, Southeast 0.32 - 0.35 Mediterranean 0.93 0.0055 0.88 Atlantic, Northwest 0.95 0.0122 0.85 Atlantic, Northeast 0.91 0.0092 0.85 Pacific, Northwest 0.99 0.0177 0.96 Pacific, Northeast 0.74 - 0.76 Atlantic,Nortliwest 0.39 - 0.22 Atlantic, Northeast 0.78 - 0.89 Pacific, Northwest 0.97 0.02258 0.98 Atlantic, Northeast 0.98 0.032 0.93 Atlantic, Northeast 0.99 0.003 0.95 Atlantic, Northwest 0.91 0.0027 0.96 Atlantic, Northwest 0.73 - 0.65 Atlantic, Northeast 0.95 0.0024 0.96 Pacific, Northwest 0.96 0.0086 0.98 Pacific, Northeast 0.55 - 0.75 Pacific, Northwest 0.28 - 0.37 Pacific, Northeast 0.5 - 0.6 Atlantic, Northeast 0.92 -0.0133 0.87 Atlantic, Western Central 0.99 0.0233 0.92 Atlantic, Eastern Central 0.79 - 0.88 Atlantic, Soiithaest 1 0.0529 0.97 Atlantic, Soiitheast 0.97 0.0232 0.98 Indian Ocean, Western 0.97 0.012 0.97 Indian Ocean, Eastern 0.95 0.0147 0.95 Pacific, Northwest 0.2 0.38 Pacific, Western Central 0.95 0.0071 0.98 Pacific, Eastern Central 0.84 - 0.75 Pacific, Soiithwest 0.98 -0.0399 0.99 Pacific, Soiitheast 0.7 - 0.72 Atlantic, Western Central 0.99 0.0163 0.88 Atlan~ic, Eastern Central 0.82 - 0.63 Atlantic, Soiithwest 0.83 - 0.72 Atlantic, Soiitheast 0.86 - 0.78 Indian Ocean, Western 0.99 0.0134 0.99 Indian Ocean, Eastern 0.83 - 0.73 Pacific, Northwest 0.76 - 0.9 Pacific, Western Central 0.99 0.0095 0.95 Pacific, Eastern Central 0.2 - 0.83 Pacific, Southwest 0.43 - 0.45 Pacific, Soiitheast 0.69 - 0.9 Atlantic, Northwest 0.82 - 0.83 Atlantic, Northeast 0.97 0.0085 0.99 Atlantic, Western Central 0.94 0.0156 0.93 Atlantic, Eastern Central 0.99 0.0079 0.99 Atlantic, Soiithwest 0.82 - 0.52 Xiphias gladius Atlantic, Southeast 0.92 0.0106 0.96 0.139 317 9308 2145 29.4 Xiphias gladius PaciBc, Northwest 0.73 - 0.63 - 5574 10839 8483 1.9 Xiphias gladius PaciBc, Western Central 0.91 0.0057 0.94 0.0051 1931 4997 3281 2.6 Xphias gladius Pacilc, Eastern Central 0.77 - 0.54 - 2697 8463 5474 3.1 Xiphias gladius Pacific, Southwest 0.81 - 0.75 383 1865 1091 4.9 Siphias gladius Pacific, Southeast 0.96 -0.0131 0.34 - 500 8403 2177 16.8 Ela~iiobrnnchii Atlantic, Northeast 0.91 0.0286 0.93 0.023 895 14200 6183 15.9 Elas~irobranchii Atlantic, Western Central 0.99 0.0116 0.98 0.0097 3100 17642 9229 5.7 Elm~~obranchii Atlantic, Eastern Central 0.86 - 0.58 - 14544 32942 23876 2.3 Elasiitobranchii Atlantic, Southwest 0.71 - 0.71 - 2300 29864 16635 13 Elasi~robranchii Atlantic, Southeast 0.32 0.0149 0.88 0.0102 1230 9538 3683 7.8 Elasnlobranchii PaciBc, Northwest 0.87 - 0.92 0.0023 54834 96888 80602 1.8 Elasi~~obranchii Pacific, Western Central 1 0.0066 1 0.0065 19900 53317 33033 2.7 Elasl~robranchii Pacific, Eastern Central 0.92 0.0062 0.92 0.0048 9430 22218 15813 2.4 Elas~~robranchii Pacilc, Soiithwest 0.5 - 0.83 1800 7601 3860 4.2 Elasniobranchii Pacific, Southeast 0.93 0.0076 0.9 0.0088 300 3255 1652 10.9 Table 1 (concluded). For the different species/genus considered in the present study, the percentage of catch/subarea for which SDL increases over the different time periods examined ranged between 55% and 100%, with the following exceptions (a) il.le1.1uccius spp., Sardinapilchardus, Sardinops spp., Scomberjaponicus and Clupea spp. for the 16- versus 22-year comparison, for which the percentage ranged between O and 44%; and (b) Tracbu?-us spp. and Mallotus villosus for the 8- versus 16-year comparison, for which the percentage was 38 and 33%, respectively (Table 2). Overall, 78% of the total species/subarea catch records showed an increase in SDL from 2 versus 4 years, 75% from 4 versus 8 years, 76% from 8 versus 16 years and j7% from 16 versus 22 years (Table 2). The percentages for the successive changes in SDL in the first three time periods were generally higher for the relatively small-sized pelagic and demersal species (Engraulis spp., Sal-dinops spp., Sardinella spp., Sardina pilchardus, Mallotus villosus, Clupea spp., Scomber spp., Trachurus spp., Merluccius spp. and 'other') than for the relatively large-sized pelagic and demersal ones (Gadus spp., Thunnus albacares, Katsuu~onus pelamis, Xiphias gladius and Elasmobranchii). The opposite was true of the percentages of catch records showing changes in SDL from 16 to 22 years (Table 2). Al1 regressions between the mean SDL of the catches of each genus/species over the different subareas and time period over which SDL was calculated, had positive slopes, significantly (PO.Oj). 364 Variability of fish catches 2 4 8 16 Size Species N versus versus uersus versus Ca tego? - 4 8 16 22 - Number Sa idinops 4 3 4 4 1 siilal1 Sc: :~m ber, japonicus 9 9 7 9 4 siiiall Tr:rchurus 8 6 8 3 5 sinall Su idinella 6 5 5 4 6 small Sudina pilchardus 3 2 7 2 1 srii:ill Cl:~pea 4 3 3 4 O siiiall 8 6 6 8 Er,graulis > siiiall Mt ~rluccius 9 6 6 7 3 siiiall ,441 rllotus villosus 3 3 3 1 2 small 0iIiel"' 7 6 7 7 5 s111aIl Tl~unnus albacares 11 11 7 8 6 large Xij~hiasgladius 11 6 6 6 7 large Gl !,dus 4 4 3 3 3 large Kfi tsuzvonus pelanzis 12 9 9 8 8 large El:ismobranchii - 10 6 6 9 6 large Percentages Su idinops - 75 100 100 25 siiiall Scl)mb. japonicus - 100 78 100 44 s~iiall Tr,:fchurus - 75 100 38 63 sriiall Sa r.dinella - 83 83 67 100 siiinll Sa rdina pilchardus - 67 67 67 33 s~iiall C1,ipea - 75 75 100 O siiiall Engraulis - 75 75 100 63 siiiall Mi"rluccius - 67 67 78 33 siiiall Mrrllotus villosus - 100 100 33 67 siiiall Ot tier* - 86 100 100 7 1 siiiall Thi,rnnus albacares - 100 64 73 55 large Xiphias gladius - 55 55 55 64 large G~dus - 100 75 75 75 large Kalsuzvonus pelamis - 75 75 67 67 large El:.smobranchii - 60 60 90 60 large Total 109 85 82 83 62 botli % .:if Total 100 78 75 76 57 botli Total 1 61 49 5 1 49 32 siiiall % - 80 84 80 52 siiiall To ta1 2 48 36 3 1 34 30 large % - - 75 65 7 1 63 l;ir,ge a) ::onsisting of Micromesistius poutassou, Pleuronectes platessa, Hippoglossoides platessoides and Theragra ch,~Icogramma. er and percentages of species/FAO-subarea catches for which SDL increases were examined over diflerent tirne periods. Sard other upwelling areas > non up\velling Atlantic and Pacific areas > Mediterranean. This suggests that in the case of within-species/genus, spatial diffzrences in variability are a function of the community/ecosystem characteristics of the marine region in concern, such as the trophic potential and complexity of the food web (i.e., primary production and number of trophic levels) and the enrironmental dynamics affecting community characteriitics. Indeed, the four major upwelling ecosystems are generally cor trolled by similar environmental dynamics and inhabited by similar communities of exploited fisheries stocks that most prcbably have adapted to similar environmental/community characteristics (Bakun, 1985, 1990). Yet, the Peruvian upwelling region is characterized, when compared with the rernaining major upwelling areas, by: (a) a longer effective up~velling period, Le., more or less throughout the year; (b) more intense effect of ENS0 events (Mann and Lazier, 1991); and (c) the higzest fishery catch densities (Cury, 1995; Faure and Cuq?, this vol.). In contrast, there is a pronounced oligotrophy in the surice waters of the Mediterranean Sa, especially so in its eastern part (Aegean and Levantine Seas) because of the low nuiiient concentration in the trophogenic layer. 'The latter is attributed to: (a) the lack of significant upwelling areas, resillting to upwelling of 'new' nutrients from deep waters in the euphotic zone, the key to high biological productivit!r; (b) the relatively small amounts of discharge from land; and (c) the fact that in the Mediterranean Sea, where total evapoiation exc::eds precipitation and river runoff (Hopkins, 1978), the conservation of mass and salinity is maintained by the balance of a nto-layer flow through the Strait of Gibraltar: surface, nutrient-poor Atlantic waters inflow in the upper Iqer whereas Metliterranean deeper waters outflow in the lower layer. In addition, in the Mediterranean Sea, especially in the south- eas:ern part, the importance of picoplankton increases, a fact which presumably increases the number of tropllic levels and, her.i:e, may limit the potentiai production at higher trophic levels (Azov, 1991). As a result of this, the Mediterfanean Sea is chaiacterized by a low fish catch density (1.4 th2 of continental shelf), which is many times lower than that in upwelling as well as some other non-upwelling ones areas (Stergiou and Christou, this vol.). The low trophic potential of the Mediterranean sets an upper limit in the carrying capacity of the region and, hence, in the level of the catch variability of a givcn species at a given time (Le., lower maximum slope values: Fig. 3 and 4). Tht.re remain the questions of whether catch trends reflect abundance trends of similar scales and of what causes trends in fish catches. With respect to the first question, one rnay assume that fish abundance may exhibit trends similar to those of the a catches for the following reasons: (a) although zero catches do not imply zero abundance (e.g., closed fisheries, im~lossibility to completely fish out a population/species from a given geographic area), annual catches are usually smaller thaii biomasses; (b) al1 or most species examined here have been traditionally fished in most FA0 subareas and, since catc,hes refer to years following 1970, one may assume that most of the fisheries examined were not in the initial stage of dev::lopment, at which catch is not related to abundance (Hilborn and Walters, 1992); and finally, (c) many of the catch series used here possibly reflect, at least to a certain extent, management measures that were set based on forecasted abundance level. Wit!~ respect to what causes trends in fish catches, climate, predation and its special form of fishing, species' dynamics and lie-tiistory, managerial decisions, economic and social factors may ail, in a synergetic dynamic fashion, affect fisheries catches (Caidy and Sharp, 1986; Hilborn and Walters, 1992). The effect of climate cannot be distinguished from that of fishing inamuch as fisheries managers will tend to respond to catch declines by assuming that fishing is the main factor and, hence, borli effecrs will be reflected in catch records (Hilborn and Walters, 1992). However, it is worthy to mention that the fact that the majority of the catch records examined in the present study are characterized by 'reddened' spectm, as is also true of marine physical parameters (Steele, 1985), probably reflects the effect of the such parameters on fish catches. The author wishes to extend his gratitude to Drs P. Cury, D. Pauly and P. Smith, for their critical comments and to Drs P. Cury and C. Royfor providing access to FA0 data. AzovY. 1991. Eastern Mediterranean - a marine desen?Mal: Pol- lut. Bull., 23: 22 j-232. Bakun A. 1985. Comparative studies and the recruitment pro- blem: searching for generalizations. CalCOFl Rep., 26: 30-40. Bakun A. 1990. Global climate change and intensification ofcoas- ta1 ocean upwelling. Science, 247: 198-201. Caddy J.F. and J.A. Gulland. 1983. Historical patterns of fish stocks. Mal: Poliq, 7: 267-278. Caddy J.F. and G.D. Sharp. 1986. An ecologicai framework for marine fishery investigations. FA0 Fish. Tech. Pap., 283, 152 p. Cuv P. 199 j. Comparison of the Ivoiro-Gha~iari fisherywith other exploited upwelling systems of the world. ln: F.X Bard and KA. Konnteng (eds.). Dynamicsand use ojmdinella resourcesJi-onz upwellingoffGhnna andlvoy Coast. ORSTOM, Pans: 169-193. Hilborn R. and C.J. vl'alters. 1992. QuantitativeJsheries stock assessment: cboice, dvnanzicsand uncertninty. Chapman and Hall, New York, 570p. Hopkins T.S. 1978. Physical processes in the Mediterranean basin. ln: B. Kjerfve (ed.) Estuarine transpon processes. Bell W. Baruch Lib. in Mar. Sci.,7. Kawasaki T. 1983. Why do some pelagic fishes have wide fluctua- tions in tlieir numkrs? Biological basis of fluctuation hm the view- point ofevolutionary ecology. In: G.D. Sharp and J. Csirke (eds.). Proceedings of the Expen Consultation to Exariiine Changes in Abundance and Species Composition of Neritic Resources. Saii José, CostaRica, 18-29April1983. FA0 Fish. Rep., 291(2): 1065-1080 Mann K.H. andJ.R.N. Lazier. 1991. Dynanzics of~ila~.ine eco- ysteiiis: biological-plysical interactions in the oceans. Black- well Scientific Publications, Boston, 466p. Pimm S.L. 199 1. The balance ofnature?Ecological issues in the conservation ofspecies and coiiziizunities. The University of Chicago Press, Chicago, 434p. Pimm S.L. and A. Redfearn. 1988. The variability of populntioii densities. Nature, 334: 313-314. Steele J.H. 1985. A comparison of terrestrial and iiiarine ecolo- gical systerns. Nature, 313: 3 55-358. Stergiou K.I. 1984. Capelin and cliniaticcbnnge in t/~eBai.ents Sea. MSc Thesis, McGill University, Montreal, Canada, 33 jp. Taylor G.T. and F.J. Prochaska. 1984. Incorporating uiiobsen~ed cyclicalstock movements in fishery catchequations: an application to the Florida bliie crab fishery. N Ati2.J Fisb. 'Ilmage, 4: 67-74. Warner S., K.E. Limburg, A.H. Arino, M. Dodd, J. Diishoff, K.I. Stergiou and J. Potts. 1994. Coriiparisons of time series across the land-sea gradient. ln: J. Steele and T. Powell (eds.). Ecolo- gical filnesuies. Chapnian and Hall, New-York. Whitehead P.J.P., M.L. Bauchot, J.C. Hureau, J. Nielsen and E. Tortonese (eds.). 1984. Fishesofthe noi-th-eastel-n iitlnnticatid theMediterranean. LTNESCO, Paris, Vol. 1, 2 and 3. 3 70 Variability of Fish Catches Desperately Searching for Natu rai Eutrophication: the Case of the NE Mediterranean * A.istotle University of Thessaloniki School of Bioiogy Depirtrnent of Zoology Laboratory of Ichthyology, PO 3ox 134 5000 Thessaloniki GREI i3E ** National Centre for Marine Research Agius Kosmas, Hellinikon 16604 Athens GREICE In the present work, we review and analyse the major features of the Greek marine fisheries (catch species composition and densities) for the 1982-1989 period and the results are discussed in the context of the trophic potential of the Greek seas (phytoplankton and zooplankton productivity and abundance) and compared with those of other upwelling and non-upwelling areas of the world ocean. Multivariate analyses (classification and ordination) performed on the mean (1982- 1989) commercial catch weight of 66 species (or groups of species) from 16 fishing subareas indicated that the 16 subareas fall into three main groups, generally representing the S-SE Aegean (and NW Levantine Seas), the Ionian and central Aegean Seas, and the N-NW Aegean Sea, respectively. The species compositions of the mean catches of the groups identified by multivariate analyses differed considerably. The mean catch of S- SE Aegean was dominated by pickerel and bogue and, to a lesser extent, by horse mackerels, that of the central Aegean and Ionian Seas by sardine, horse mackerel, bogue and pickerel, whereas that of N-NW Aegean Sea by anchovy and sardine. The mean pelagic, demersal and total fisheries catch densities between 1982 to 1989 aii decrease from 1.3,0.83 and 2.13 t/km2 in N-NW Aegean Sea to 0.25, 0.37 and 1.23 t/km2 in S-SE Aegean Sea, respectively, with the latter densities being comparable to those in the Ionian Sea. In addition, such an increase goes along with a decrease in the relative importance of demersal species. The mean pelagic catch density in the Ionian and S-SE Aegean Seas is from 3 to 1 j0 times lower than those in other marine areas of the world ocean. In contrast, the mean pelagic catch density in the N-NW Aegean Sea is: (a) comparable to that in the California coast (22-38"N), (b) higher than those in the Gulf of California and Indian waters (7-2 jON) and (c) 2 to 30 times lower than those in the major upwelling areas of the world. Natural eutrophication processes are responsible for such a geographical differentiation in catch species composition and densities, namely the gradient in eutrophy, river runoff, temperature and salinity of the Greek marine waters along a NNW to SSE axis. In contrast to other areas where seasonal upwelling takes place (e.g., Ivoirian and Ghanaian coasts), seasonal upwelling along the E Aegean coast, dnven by the Etesian winds (dry and cool N-NE-E winds blowing over the Aegean Sea in summertime) probably has no significant impact on primary production and fisheries catch densities, most probably because upwelled waters reach the surface from layers immediately below the seasonal thermocline, its depth in the E Aegean Sea generally being <50 ml and, hence, are already poor in nutrients. Dans ce travail, nous analysons les principales caractéristiques des pêcheries marines grecques (composition des captures et densités) pour la période 1982-1989 ; les résultats sont discutés en termes de potentiel trophique des eaux maritimes grecques (abondance et productivité du phytoplancton et du zooplancton) et comparés à ceux obtenus dans d'autres zones (d'upwelling ou non). Des analyses multivariées (classification et hiérarchisation) sont faites sur les captures commerciales moyennes de 66 espèces (ou groupes d'espèces) de 16 zones de pêche appartenant à trois groupes : la zone S-SE de la mer Egée (et NW de la mer du Levantin), la mer Ionienne et la zone centrale de la mer Egée, et le N-NW de la mer Egée. La composition des captures diffère beaucoup d'une zone à l'autre. La capture moyenne de la zone S-SE de la mer Egée est dominée par les brochets et les bogues et, dans une moindre mesure, par les chinchards ; celle de la partie centrale de la mer Egée et de la mer Ionienne par les sardines, les chinchards, les bogues et les brochets ; et celle de la zone N-NW de la mer Egée par les anchois et les sardines. Les densités moyennes des captures de poissons pélagiques, démersaux et des captures totales diminuent entre 1982 et 1989 de 1.3,0.83 et 2.13 thm2 respectivement dans le N-NW de la mer Egée, à 0.2 5, 372 Natural Eutrophication ofthe NE Mediterranean 0.37 et 1.23 t/km2 dans le S-SE de la mer Egée, ces dernières valeurs se retrouvent aussi dans la mer Ionienne. De plus, un tel accroissement va de pair avec une décroissance de i'importance relative dans chacune des zones des espèces démersales. La densité des captures de poissons pélagiques dans la mer Ionienne et dans le S-SE de la mer Egée est de 3 à 150 fois plus faible que celle d'autres zones marines. Par exemple, la capture moyenne des pélagiques dans le N-NW de la mer Egée est (a) comparable à celle de la Californie (22-38 ON), (b) plus importante que celle du golfe de Californie et de l'Inde (7-2 5 ON) et (c) de 2 à 30 fois plus faible que celle des principales zones d'upwelling mondiales. Les processus d'eutrophisation naturelle que sont le gradient d'eutrophisation, le débit des rivières, la température, la salinité des eaux le long de l'axe N-NW à S-SE, sont responsables de ces différences de composition des captures et des densités. Par comparaison avec les autres upwellings saisonniers d'autres zones (par exemple Côte- d'Ivoire-Ghana), celui des côtes orientales de la mer Egée, soumis aux vents Etésiens secs et froids, N-NE-E, qui soufflent sur la mer Egée en été, n'a probablement pas d'impact sur la production primaire et sur les captures. Les eaux upwellées atteignent la surface depuis des couches situées juste en dessous de la thermocline dont la profondeur dans l'est de la mer Egée est généralement inférieure à 50 m. D e ce fait, elles sont généralement pauvres en sels nutritifs. The Mediterranean Sea has a maximal depth of 5 140 m, maximal width of 800 km, an area of 2.5 million kmL (0.8% of the total marine area of the world) and a mean depth of 1 470 m (Azov, 1991). There is a pronounced oligotrophy in the surl':ice waters of the Mediterranean Sea, especially so in its eastern part (Aegean and Levantine Seas). The lack of sigriificant upwelling areas resulting from upwelling of 'new' nutrients from deep waters in the euphotic zone, an important key to high biological productivity, and the relatively small amounts of discharge from land, results in low nutrient concentration in the trophogenic layer. In addition, the Mediterranean Sea is a concentration basin in which total evalioration exceeds precipitation and river runoff (Hopkins, 1978) and the conservation of mass and salinity is maintained by ille balance of the two-layer flow through the Strait of Gibraltar: surface, nutrient-poor Atlantic waters inflowing in the upper layer and Mediterranean deeper waters outflowing in the lower layer. The low concentration of phosphates in deei~er waters, reflect the restricted exchange of the landlocked water with the waters of the adjacent Atlantic Ocean. The oligotrophy of the Mediterranean Sea is also reflected in the level of fisheries catches (300 kgkm2 for al1 areas and 1,400 kgh2 over the continental shelf, Ben Tuvia, 1983). The marine waters of Greece include the Aegean Sea and a part of the Ionian and NW Levantine Seas. The Greek part of the Ionian Sea comprises a part of a larger area in which the existing stocks are fished by a number of other major fishing nations (especially Albania, Italy, Libya, Malta and Tunisia). The mean Greek catch from the Ionian Sea over the perioci 1982-1987 represented only 7% of the total Ionian Sea catch (FAO, 1989). The Greek part of the Levantine Sea also comprises a part of a larger area in which the existing stocks are fished by a number of other major fishing nations (Lebanon, Turkey, Israel, Syria, Cyprus, Egypt and Gaza Strip). The mean Greek catch from the 1W Levantine Sea over the peilod 1982-1987 represented less than 1% of the total Levantine catch. In contrast, the Aegean Sea is mainly fished by the Greek fleet. Although the Turkish fleet fishes along the eastern Aegean coast, the Turkish catch from this area is relatively insignificant when compared with that of the Greek Aegean Sea fisheries; it comprised about 20% of the mean total Aegean catch over the period 1982-1987 (FAO, 1989). In the present work, we analyse the major aspects of the Greek marine fisheries (catch species composition and densitv) for the 1982-1989 period. The results are discussed in the context of the trophic potential of the Greek Seas (phytoplankton and zooplankton productivity and abundance) and compared with those of other upwelling and non- upwelling areas of the world ocean. 1. MATERIAL AND METHODS Fisheries statistics for the waters of Greece have been recorded on a monthly basis since January 1964 by the National Statistical Service of Hellas (NSSH Bulletins, 1965-1990). For a better evaluation of the available data, the waters fished bv Greek vessels have been divided into 18 statistical fishing subareas (Fig. 1). Fishing subareas 1 and 2 (not shown in Fig. 1) refer to the Atlantic Ocean and the northern coast of Africa, respectively. Catch data are collected dijectly from a sample of fishing vessels (stratified random sampling) that are sunleyed by local customs authorities. For each vessel surveyed, a statistical questionaire is completed showing the quantities of each major fish species (or group of species) caught during the previous month (or that the vessel did not work during that period). Details on the validity of the NSSH data have been presented by Stergiou et al. (1994). In general, the Greek fishing fleet includes: (a) fishing vessels operating in distant waters (Atlantic Ocean and northern African coast, and thus of no concern to the present study); (b) traders operating in Greek open-sea waters; (c) purse seiners operating in Greek open-sea and coastal waters; (d) beach seiners operating along the Greek coasts; and (e) 'other coastal boats' (including smal ring netters, drifters, liners, etc.) operating along the Greek coasts. Since 1969 the catches of the smaller inshore ring netters, drifters and liners (Le., boats with engine horsepower of less than 20HP) have not been recorded by the local customs authorities. In addition, the catches from the marine sport fisheries, which locnllv may be relatively important (e.g., 11.8% and 4.5% of the total catches from subareas 9 and 5, respectively; Stergiou et al., 1989) are not included in the totals. For the period 1964-1981, separate catch stntistics are available for 23 species or groups of species only ivhereas for the years following 1981 separate catch statistics have been available for 66 species (or groups of species) of commerciallv important fishes, cephdopods and crustaceans (Stergiou et al., 1994). In the present study, the annual landed catches of the 66 species (or groups of species) in the 16 statistical fishing subareas (Fig. 1) for the years 1982-1989 inclusive, were analysed using two main categories of statistical techniques, as follows. 374 Natural Eutrophication of the NE Mediterranean Fig. 1: Map showing the 16 fishing subareas of Greece (Hellas). The dotted line represents the 200 m isobath; tiatchings show areas where anthropogenic eutrophication is locally important. The following univariate measures were computed: mean number of species, species diversity, richness and eveness for eai:h subarea over the 1982-1989 period. Diversity was calculated using the Shannon-Wiener diversity index H', species rictiness using Margalefs D, and eveness using Pielou's J (Maguran, 1988). Al1 computations were carried out using the PRIMER algorithms (Clarke and Warwick, 1989). A ri~atrix comprising the mean landed commercial catch weight of each species (or groups of species) from each subarea ov:r the 1982-1989 period was compiled. From that matrix, a triangular matrix of similarities between al1 pairs of subareas was computed using the Bray-Curtis coefficient (Bray and Curtis, 19j7). Prior to this computation, the data was standardized by the mean total landed catch per subarea in order to compensate for between-subarea differences in fishing effort, for which data per subarea is not provided in the NSSH Bulletins. Subsequently, the similrii-itv matrices were suhjected to both clustering (employing group-average linking) and ordination (employing non-metric multi-dimensional K./. STERCIOU AND E.D. CHRISTOU 3 75 scaling, MDS) analysis techniques. The adequacy of the ordination in two, rather than more than two, dimensions is expressed by a 'stress coefficient' (Field et al., 1782). In general, stress values <0.1 imply good representation (Clarke and Warwick, 1787). Discontinuities between subareas may be accepted as real when the results of the two methods agree (Field et al., 1782; Clarke and Green, 1788; Gray et al., 1788). Al1 multivariate analyses were carried out using the PRIMER algorithms (Clarke and Warwick, 1787). Overall, the two pelagic species anchovy and sardine were dominant, comprising 18.5% and 11.4% of the mean total Greek catch, respectively, over the 1782-1787 period (Fig. 2). In addition, bogue, horse mackerel, pickerel, hake and chub mackerel comprised 8.2%, 8.0%, 7.3%, 3.3% and 3.3% of the mean catch, respectively pig. 2). Aü remaining species each contributeci less than 3%. However, catch species composition diiers greatly between fishing subareas and component fishenes. The species composition of the mean catch per component fishery over the 1782-1787 penod is shown in Table 1. The mean trawl catch was dominated by hake, pickerel, horse mackerel and red mulet; the mean purse seine catch mainlv by anchovy and, to a lesser degree, by sardine, horse mackerel and bogue; the mean beach seine catch mainly by pickerel and, to a lesser degree, by sardine and bogue; and, finally, the mean 'otlier coastal boats' catch by bogue, pickerel and grey mullets. The mean annual commercial catch weights per fishing subarea over the 1782-1787 period are shown in Table 2. The mean annual Greek catch was 72 841 t, of which 86 584 t were caught in the Aegean Sea (less than 1% was derived from the hW Levantine Sea) and 6 257 t in the Ionian Sea. Fish made up 74.2% of the rnean total Aegean catch, cephalopods 3.2% and crustaceans 2.6%. In the Ionian Sea, fish made up 757% of the mean total catch, cephalopods 3.2% and crustaceans 1.1%. The mean annual catch weight, mean number of species comprising the catch, and the mean diversitv (Shannon-Wiener diversity, evenness and richness) over the 16 fishing subareas are shown in Table 3. Tlie total number of species ranged from 56 (in subarea 6) to 66 (in subareas 8, 10 and 12 to 16). Richness ranged between 6.37 (in subarea 13) and 12.0j (in subarea 6), diversity between 1.67 (in subarea 11) and 3.38 (in subarea 6) and evenness between 0.41 (in subarea 11) and 0.84 (in subarea 6). The following significant regression was found between the log-transformed mean catch weights (C) and number of species (N) comprising mean catches for each subarea over the 1782-1787 period: Ln(C) = -72.83 + 24.25Ln(N), SE(slope)=3.75, r2=0.75, n=16, P<0.01. Bogue (8.296) ,& Pickerel (ï.3%) Hake (3.3%) Chub rnackerel (3.3%y Fig. 2: Species composition of the mean total Greek marine catch, 1982-1 989, in % of total weight. 376 Natural Eutrophication of the NE Mediterranean -- - - Cornrnon name - Scien tific name T P S O FIqHES Arichovy Ariglerfish Arinular sea brearn BI ick sea bream BE ~tched pickerel Bl.ie whiting Bl.iefin tuna Bl.iefish Bcgue Bonito Br i Il Ctiub mackerel Ccinber Ccanrnon dentex C(liiimon sea bream Dogfishes D~isky grouper Eu ropean eel Eu:.opean sardine E~ropean sea bass G; rfish Gilt sardine Gi 1 thead sea bream Gicater amberjack Giciy mullet Giiitarfish Giirnard HF ke Horse mackerels Jat.k mackerel Joiin dory La geeye dentex Ma ckerel Pic kerel Ra !,s Red mullet Re:l pandora Ro: kfish Sa'idled bream Salema Engraulis encrasicoius (L.) Lopbius spp. Diplodus annulans (L.) Spondyliosortla cantharus (L.) Spicara flexuosa Rafinesque, 1810 Micmrnesistiuspoutassou (Risso, 1826) Thunnus thnnus (L.) Pmatonzus saltator (L.) Bmps bcqps (L.) Sarda sarda (Bloch, 1793) Scophthalnzus rhonzbus (L.) Sconzberjaponicus (Hou ttuyn, 1782) Serranus ca bfilla (L.) Dentex dentex (L.) Pagruspag~us (L.) Squ alidae Epinephelus guaza (L.) Anguilla anguilla (L.) Sardina pilchardus (Walbaum ,1792) Dicentrarchus la brax (L.) Belone belone gracilis (Lowe, 1839) Absa falax nilotica (Geofsoy Saint-Hilaire. 1808) and Sardinella au~ita (Valenciennes, 1847) Sparus aurata L. Seriola dunzenli (Risso, 1810) Mugilidae Rbinobatos spp. Triglidae Meduccius rire~Succius (L.) Trachu?us spp. Trachuruspicturatus (Bowdich, 182 j) Zeus fa ber (L.) Dentex nzacl~phthainzus (Bl~h, 1791) Sconzber sconzbrus (L.) Spicara stnarij (L.) Raja spp. Mullus ba ha tus (L.) Pagellus erythrinus (L.) Helicolenus dac@lopterus (Delaroche, 1809) Oblada melanura (L.) Saqa salpa (L.) 1.9 9j.4 0.9 1.8 94.9 0.4 0.6 1.0 8.8 13.3 0.8 77.1 13.7 6.8 6.0 73.6 49.5 7.6 11.6 31.2 93.8 O. j 3.3 2.4 0.3 39.7 0.6 j9.4 6.0 36. j 2.0 jj. j 11.0 jl.8 11.8 3j.i 1.4 64.j 2.1 32.0 83.4 3.8 0.2 13.6 3.3 82.7 3.8 10.3 32.6 5.8 3.1 j9.6 9.3 2.6 j.3 83.8 6.9 1.7 2.1 89.2 37.0 4.3 0.6 58.1 3.5 j.2 1. j 89.8 2.2 j. 1 0.7 92.0 1.7 77.9 9.0 11.4 24.6 14.4 3.0 j8.0 0.9 11.3 10.3 77. j 5.4 70.2 2.1 22.2 Table 1: Cornrnon narnes, scientific names and percentages of mean annual total catches of the 66 species (or groups of species), recorded by the Greek National Statistical Service, fished by trawlers (T), purse seiners (P), beach seiners (S) and 'other coastal boats' (= small ring netters, liners, drifters, etc.; 0) over the period 1982- 1087. Species are listed in alpha espect to common name. Sum across lines is 100% (from S ergiou and Pollard, in press). ---- Coni ~non riame Scientific name BONY FISH Scorpionfish Shi drum Skipjack Smoothhound Sole Sprat Stone bass Striped red mullet Swordfish Thick blotched pickerel Thornhck ray Tub fish White grouper White sea bream Whiting Other fishes Scolpaena spp. L'mbrina cinosa (L.) Katsuzuo~z us pelanz is (L.) Mustellus sp p. Soleidae (rnainly Solea tlulga~is Quensel, 1806) Sprattus sprattus (L.) Polvprion anlericatzus (Schneider, 1801) Muilus sunr~uletus (L.) Xiphins gladius (L.) Spicam nzaena (L.) Raja clat~ata (L.) Trigla luce~na (L.) Epinephelus aetzeus (Geofioy Saint-Hilaire, 1817) Dipldus salgus (L.) Merlangius nielhngus euinus (Nordmann) (numeroiis spp.) CEPHALOPODS Coninion squ id Loligo vulgalis (Lamarck, 1798) Cuttlefish Sepin oficinalis (L.) Flying squid Illex coindetii (Verany, 1839) Octopus Octopus vulgalis (Cuvier, 1797) Poulpes Eledone spp . CRUSTACEANS Crabs Brac hyura 30.5 0.0 0.8 68.7 Crayfish Nephlops no~uegicus 92.9 0.6 1. j 4.9 Lobsters Honzatus gant ntatus (L.) and j.1 0.6 0.9 93.4 Palinuvus elephas (Fabriciiis, 1787) Coninion prawn Penaeus kerathu~us (Forsskil, 177j) 658 1.0 3.1 30.1 Shrimps (no data on spp. composition) 69.8 0.4 O.j 29.3 Table 1 (continued). The classification of the rnean (1982-1989) commercial catch weight of al1 species for each subarea (Fig. 3) indicated that, at the j0% (Bray-Curtis) similarity level, the 16 subareas fa11 into three main groups: Group 1 (subanas 7 and 16 to 18), Group II (subareas 3 to 6,s to 10 and 15) and Group III (subareas 11 to 14), generally representing the southern Aegean and NW Levantine Seas, the Ionian and central Aegean Seas, and the N-NWI Aegean Sea, respectively. The results of the ordination (MDS) of these subareas (Fig. 3) agree with the above pattern. The resulting stress value for the two- dimensional plot (Fig. 3) was low (=0.08), implying the adequacy of the MDS representation in these two dimensions. The species compositions of the mean catches of the groups identified by multivariate analyses also diffend considerably (Table 4). The mean catch of Group 1 was dominated by pickerel and bogue and, to a lesser extent, by horse rnackerel, representing 21.0%, 17.6% and 8.5% of the mean total catch, respectively (Table 4). The mean catch of Group II was dominated by sardine, horse mackerel, bogue and pickerel, representing 12.1%, 12.0%, 11.3% and 10.2% of the mean 378 Natural Eutrophication of the NE Mediterranean Species Fishing subarea 34567 8 9 10 11 12 13 14 lj 16 17 18 Anchwy Blutfin tuna Blur6sh Bogiie Bon ro Chub mackerel European sardine Flyiiig squid. Gariish Gilt sardine Hor:je mackerel Jack mackerel Mac kerel Skip iack Spriit Swcr,dfish 320 5 O 1101 700 47 O40 j 202 11 136 1104 25 2 1 853 10 2 4 60 293 7 27 1369 19 O O 51 1 O 10 79 304 48 9j j 19 2008 17 O 1 164 O00 31 O00 18 10 O O 10 69 O 14 llj Tot:il pelagic 185 929 1072 36 218 7063 352 3679 361 4229 14606 1028j 37j9 1692 3087 1308 % -. 0.35 1.76 2.03 0.07 0.41 13.36 0.67 6.96 0.68 8.00 27.63 19.46 7.11 3.20 5.84 2.47 Anpierfish Anr ular sea bream Blac k sea bream Blot c hed pickerel Blu II whiting Bril Coiiiber Coiiimon dentex Coriiinon prawn Coiiirnon sea bream Coiiimon squid Cra a Cu t i lefish Dq fish Du:ky grouper Eus' ipean eel Euur ipean sea bass Gilrliead sea bream Gre'iter amberjack Grev mullet Guirarfish Gui nard Hake T,ible 2: Mean catch (in t) per speci Creek waters, 1982-1 989. Separation of species (or grou pelagic ones according to Stergiou and Pollard (in press) K./. STERCIOU AND E.0. CHRISTOU 3 79 Species Fishing subarea 34j67 8 9 10 11 12 13 14 lj 16 17 18 John dory Largeeye dentex Lobster Nomay ldxter Octopus Other fish Pickerel Octopi Rays Red niullet Red pandora Roc kfish Saddled bream Salema Scorpionîîsh Shi drum Shrimps Smmthhound Soles Stone bass Striped red mullet 'Thick blotched pic. Thornback ray Tub fish White grouper White sea breani Whiting Total demersal 420 j80 1251 80 788 3207 346 3j64 87 1221 11266 6314 248j 2096 4292 1983 % 1.05 l.4j 3.13 0.20 1.97 8.02 0.87 8.91 0.22 3.0j 28.18 1j.79 6.22 j.2410.73 4.96 Table 2 (continued). catch, respectively (Table 4). Finally, the mean catch of Group III was dominated by anchovy and sardine, representing 29.7% and 14.3% of the mean catch, respectively (Table 4). The mean catch density (t/km2) per group identified by multivariate analyses is shown in Table 5, together with mean catch density values from other ecosystenis of the world ocean. It is evident that the mean catch density of pelagic, demersal and al1 fishes combined, increase from South to riorth. In addition, such an increase goes along with a decrease in the relative importance of demersal species. Hence, in Group 1 the mean catch density of the demersal species is higher than that of the pelagic ones whereas in Group III the former is about two rimes less than that of the pelagic ones (Table j) . 380 Natural Eutrophication of the NE Mediterranean Fig. 3: Dendrogram (upper) for group-average clustering and MDS plot (lower), based on Bray-Curtis similarities between standardized mean catch weights, of al1 species from the 16 Greek fishing subareas, over the 1982-1 389 period. For MDS plot orientation and scale are arbitrary. Stergiou and Petrakis (1993) analysed the annual (standardized) commercial catch weights of the 66 species (or grc ups of species: Table 1) per component fishery for the years 1982-1987 (24 fisheriedyear combinations). The c1a:;sification and ordination of al1 species for each fishery/year combination indicated that the four fisheries faIl into t~vo ma n groups; Group A (trawlers, beach seiners and 'other coastal boats') and Group B (purse seiners), exploiting primarilv der-!iersal/inshore and primarily pelagic fisheries resources, respectively (Stergiou and Petrakis, 1993). Their results clearly rer8::al the multi-species and multi-gear nature of the Greek demersal/inshore and, to a lesser degree, pelagic/semi-pelagic fisheries. In addition, the results of the multiva'ate analyses (Fig. 3) suggested that the 16 Greek fishing subareas may be gecerally grouped into three geographical clusters that differ considerably from each other in terms of species coniposition (Table 4) and catch density (Table 5). Stergiou and Pollard (in press) did a similar analysis considering only K./. STERCIOU AND E.D. CHRISTOU 381 Fishing C N D H' .l Subrea 3 581 60 7.27 2.97 0.72 4 1480 62 8.36 2 .70 0.65 5 2272 65 8.27 3.01 0.72 6 26 56 12.0i 3.33 - 0.84 7 781 64 7.15 2.19 0.53 8 10237 - 66 7.04 2.77 0.66 9 667 64 7.68 2.70 0.67 10 7213 - 66 7.32 3.23 0.77 11 430 57 7.23 -Z - 0.41 12 j422 - 66 7.56 2.43 O. 58 13 25842 - 66 622 2.63 0.63 14 16568 - 66 6.67 2.71 0.67 15 621 5 66 7.44 3.01 0.72 16 3757 - 66 7.87 2.57 0.62 17 7349 65 7.17 2.62 0.63 18 3260 64 7.77 2.73 0.71 Table 3: Mean total annual catch (C, in t), number of species (N), richness (D), Shannon-Wiener diversity (H') and evenness (J) per fishing subarea, Greek waters, 1982-1 989. The highest and lowest values are underlined. Groups identified my multivariate analysis Species Group 1 Group II Groiip III Catch % Catch ?6 Catch % Pic kerel 3243 - 21.0 2745 10.1 - - Bogii e &?Ji 3284 11.3 1586 3.2 Horse mackerel 1312 8.5 - 12.0 2643 5.5 Striped red mullet 824 5.3 665 2.3 - - Swordfish 675 4.4 - - - - Chub rnackerel 574 3.7 1186 4.1 1284 2.7 Bonito 303 2.0 1207 4.2 - - Scorpionfish 278 1.9 - - - - Gurnard 260 1.7 - - - - Bluefin tuna 253 1.6 - - 1266 2.6 Sardine - - rn - 12.1 f&l Anchovy - - 2701 7.3 rn SpIZ Hake - - 1174 4.1 1676 3.5 Red mullet - 664 2.3 897 1.7 Grey mullet - - - - 1831 3.8 Sole - - - - 1040 2.2 Others 5007 32.3 8190 28.2 14847 30.6 Table 4: Mean annual catches (in t) and percentages of the main species (or groups of species) per group identified by multivariate analysis (Fig. 3), Greek waters, 1982-1 989. Catches and percentages for the two dominant species per group are underlined. 382 Natural Eutrophication of the NE Mediterranean the ten Aegean Sea subareas for each year of the 1982-1987 period; tliey found that the within-subarea similarities were much more intense than the between-subarea ones (Le., for each subarea and year of the 1982-1987 period subarea/year conlbinations were closely tied). The main factors which may contribute to such a geographical differentiation most probably are the gradient in eutropliy, rivc:r runoff, temperature and salinity of the marine waters of Greece along a NNW to SSE axis. The eastern Mediterranean Sea is known to be one of the most oligotrophic marine regions of the world (Friligos, 1980, 1987; Azov, 1991). In the Aegean Sea, annual gross primary production in the outer part of subarea 8 and in the southern part of subarea 10 (see Fig. 1) was found to be 64 and 30 &/rn2, respectively, the latter figure being among the lowest recorded for the Mecliterranean Sea (Becacos-Kontos, 1968, 1977), comparable to those of the oligotrophic oceanic areas (50-70 gc/m2: Azov, 1991) and one to two orders of magnitude lower than those in the major upwelling areas (e.g., Peru, 1100 g~/m2: Ch:ivez et al., 1989; Cape Blanc and California, 730 and 150-720 g~/m2, respectively: Mann and Lazier, 1991). Pavlova (19(;6), based on indirect estimations, suggested that secondary productivitv in the Aegean Sea is 12 to 18 times lower thin that in the Black Sea, four times lower than that in [lie Adriatic Sea, and similar to those of the Tyrrhenian and Libyan Seas. Dai:i pertinent to secondary production in the Greek seas are only available for the Saronikos Gulf (Christou, 1991). Secondary production in the coastal area of the Saronikos Gulf (0.12-0.19 g/m3/yr) is low when comparecl witli those rec:~rded in other marine areas of the world (Table 6). Hence, the generally low secondary productivity of the Greek seas is t\ident, especially if one takes into account that Saronikos Gulf resembles, in terms of zooplankton densitv and species cor:.iposition, many other gulfs of Greece, and its northern part is considered one of the most eutrophic areas of the coiintry (Christou, 1991). lit: oligotrophic nature of Greek waters is clearly depicted in the Coastal Zone Colour Scanner (CZCS) image of the ph!rtoplankton-like pigment (PLP) distribution (not shown here, but see Stergiou and Georgopoulos, 1993), and reflected alsi! to fisheries catch densities which, on the average, are lower than those of the Mediterranean Sea as a whole and in otl.~i:r areas of the world ocean (Table 5). However, within this generally oligotrophic environment relativelv eutrophic are.iis do exist. Indeed, fishing subareas 8 and 10 to 14 are al1 characterized by one order of magnitude higher PLP cor.r:entrations as compared with the S-SE Aegean, MV Levantine and Ionian Seas (with the exception of subareas 5 and 91, the latter areas being comparable to the most oligotrophic open ocean areas of the world. A similar distribution pattern of PU is also indicated for the wintertime, although concentrations, because of the winter mixing, are generallv higher (N- hW rim of the Aegean Sea >0.5 mg/m3; Ionian, S-SE Aegean and NW Levantine Seas: about 0.2 mg/m3; Georgopoulos, unpubl. data). This general pattern of PLP distribution is also true of al1 the 46 images examined for the period 1981-1982 (Gt~orgopoulos, unpubl. data), and coincides also with those of nutrient distributions in Greek waters, the latter being derived from historical data and from recent cruises in open sea (during 1986-1990) and coastal waters (during early 19P!)s) (Stergiou and Georgopoulos, 1993). Hence, the satellite image is representative of the general pattern of the spatial diff rrentiation in PLP distribution in the Greek seas (Stergiou and Georgopoulos, 1993). Thta spatial distributions of phytoplankton and zooplankton abundance in the surface waters as well as that of fisheries cat, h densities generally parallel that of PLP concentration. Hence, with respect to the offshore waters, pliytoplankton ab~rldance is higher in the N Aegean Sea, especially in the area bounded bv 40" and 41"N latitude (subarea 14: 9 120- 58c 160 cellP; Pagou and Gotsis-Skretas, 1989). Relatively high values (up to 72 090 cellsA) of ph~oplankton abundance ha\:: been also recorded in the northern part of the Greek Ionian Sea (Pagou and Gotsis-Skretas, 1989). In contnst, the 1obt:st values of phytoplankton abundance (<29 200 cellP) have been recorded in the central and S-SE Aegean, central Ionian and NW Levantine Seas (Pagou and Gotsis-Skretas, 1989). With respect to zooplankton, its abunciance is generally higlier in some enclosed or semi-enclosed bays and gulfs (enclosed gulf in subarea 4, northern part of subarea 10, part of K./. STERGIOU AND E.D. CHRISTOU 383 subarea 8 close to Pireas Port, subarea 11: as high as 5 000 ind./m3) and in the offshore waters of the North Aegean Sea (from j00 to 5 000 ind./m3), especially in the area bounded between 40" and 41°N latitude (Le., subareas 13 and 14), sometimes extending southward to 39.j0N (Siokou-Frangou and Pancucci-Papadopoulou, 1989; Siokou-Frangou et al., 1991). The S-SE Aegean and NW Levantine Seas (southern part of subarea 16: Fig. 1) are oligotrophic areas comparable to open ocean ones, with zooplankton abundance usually ranging between 100 and 300 ind./m3 (Siokou-Frangou et al., 1991). Moreover, zooplankton abundance attains relatively higher levels (up to 900 ind./m3: Siokou-Frangou et al., 1991) in the northern part of the Greek side of the Ionian Sea (subarea 3: Fig. l), when compared with the remaining part of the Ionian Sea (up to j00 ind./m3: Siokou-Frangou et al., 1991). In agreement with the above mentioned patterns, the mean (1982.1989) pelagic, demersal and total fisheries catch densities al1 decrease from 1.3, 0.83 and 2.13 t/km2 in N-NW Aegean Sea to 0.2 j,0.37 and 1.23 t/km2 in S-SE Aegean Sea, respectively, with the latter density being comparable to those in the Ionian Sea (Table j). The mean pelagic catch densitv in the Ionian and S-SE Aegean Seas is from 3 to 150 times lower than those in other marine areas of the worlcl ocean (Table 5). In contrast, the mean pelagic catch density in the N-NW Aegean Sea (Group III) is: (a) comparable to that in the California Coast (22-3a0N), (b) higher than those in the Gulf of California and Indian waters (7-2j0N), and (c) 2 to 30 times Catch (t) . . Gecgraphic area pelagic* ~emersal" Greek seas (total) Group 1 Group II Group III Ionian Sea Aegean Sea N-NW Aegean Sea S-SE Aegean Sea Mediterranean West Africa (6-36N) Côted 'Ivoire (8W-19 Sou th Africa (6-37N) California (22-38N) Gu If of California (24N) Sou th America (143N) India (7-25N) Spain (3644N) North Sea Cape Hateras to Nova Scotia Eastern Bering Sea Central Baltic Sea Area Density (t/km2) in (km2) Pelagic Demersal Total 75294 0.70 0.53 1.23 24998 0.25 0.37 0.62 27613 0.62 0.43 1.05 22683 1.30 0.83 2.13 9824 0.26 0.27 0.53 65470 0.77 0.57 1.31 22010 1.32 0.85 2.18 43460 0.49 0.43 0.91 - 1.40 - - - 3.14 - - - 2.17 - - - 3.75 - - - 1.75 - - - 0.87 - - - 38.40 - - - 0.97 - - - 3.35 - - - - - 4.70 - - - 1.54 - - - 2.10 - - - 2.75 *seParation of the 66 species (or groups ofspecies) into pelagic/semipelagic and demersal/ishore according to Stergiou and Pollard (1994). Table 5: Catches and densities of pelagic and dernersal fishes in Greek waters and other ecosysterns of the world ocean. Data for Mediterranean frorn Ben Tuvia (1 983), West Africa to Spain frorn Cury (1 9951, and for North Sea to Central Baltic Sea frorn Sparholt (1 990). 384 Natural Eutrophication of the NE Mediterranean lower rhan rhose in the major up~velling areas of the world (Table 5). The mean total catch density in S-SE Aegean Sen is frorii 2 to 7 times lower than those in the remaining areas of Greece, the Mediterranean as a whole and other areas of the world ocean (Table 5). In contrast, the mean total catch density in N-NWAegean Sea is comparable to tliat in the E Bering Sea and two times higher than that of the Mediterranean as a whole (Table 5). It rriust also be stressed that in the SE Mediterranean Sea the importance of picoplankton is high, a fact wliich presumably increases the number of trophic levels and, hence, may limit the potential production at higlier trophic levels (Azov, 1991), possibly resulting in a lower biomass of small- and medium-sized pelagic fishes. This is consistent with the north-to-south decline in the mean pelagic catch density (Table 5) and with the results of echo-sulveys undertaken during May 1989 - May 1992 in fishing subareas 12 to 1 j. The echo-surveys revealed that the echo-abundance (mm deflection per km2) of small- and medium-sized pelagic fishes in fishing subareas 13 and 14 are higher by an order of magnitude than tliose in fisliing sub.ireas 12 and 15 (Stergiou et al., 1993; Papaconstantinou et al., 1994). It is worthy to point out also that in the summertime (mainly fromJuly to September) upwelling, driven by the Etesian winds (dq and cool N-NE-E winds blowing over the Aegean Sea) which very often reach gale force, takes place along tlie eastern Aegran coast (e.g., Metaxas, 1973). In contnst to other areas where seasonal upwelling takes place (e.g., Ivoiian and Ghanaian co2i:s: Koranteng and Pezemec, this vol.; Venezuelan coast: Mendoza et al., this vol.) seasonal upwelling along the E Aegean coa.~ probably has no significant impact on the pnmary production of the xea (Georgopoulos, pers. comm.) as well as on fishc:ries catch densities ('Table 5). This could be attributed to the fact that upwelled waters reach the surface from layers imrrediately below the seasonal thermocline, its depth in the E Aegean Sea generally being O D O :O 8 !! 9 1' -0.4 s 5 4.11 J% -1 Fig. 5: Optimal empirical -0.8 O 2 4 6 8 10 1214 transformations from the ACE 0.2 O 6 1 1.4 Maximum total pelagic catch Upwelling index (m3/s/m) algorithm using maximum total productivity (103t) pelagic catch productivity as the O 4 O 2 15 '9 ependent variable and D pwelling index, turbulence 2 5.3 Li nd continental shelf surface as 9 .I I predictor variables. The $ E - 7. :4 transformation of the rerponse 6 O ô U> variable is forced to be linear. R2= 0.44, Numbers identify !! l- ecnsystems (see Tab. 2 and 3). - 6 -0.4 -0 4 O 200 400 600 800 O 04 018 1'2 1'4 2 Turbulence (m3/s3) Continental shelf surfacqlo3 km) Fig. 6: Optimal empirical transformations from the ACE algorithm using maximum sardine catch productivity as dependent variable and upwelling index, turbulence and sea surface temperature as ctor variables. R2= O. umbers identify ec s (see Tab. 2 and 3). productivity (1 03t) 0.2 0.6 1 1.4 Upwelling index (rn3/s/rn) Turbulence (rn3/s3) Sea surface temperature (OC) 2.3. Maximum anchovy catch productivity as response variable Optimal transformations (Tl, T2, T3, T4) for the multiple regressive model (8) were calculated using the maximum anchovy catch productivity index versus upwelling index, continental shelf surface and turbulence index, i.e., T1 (maximum anchovy catch productivity) = T2 (CUI) + T3 (CSs) + T4 013) R2 : 0.40 (8) The transformation of maximum anchovy catch productivity is forced to be linear (Fig. 7a). The upwelling index transformation is on the whole linear and positive, particularily beyond a value around 0.7 mj/s/m (Fig. 7b). Continental shelf surface is transformed to a nearly log shaped curve with a breaking point around 100 000 km2 (Fig. 7c). Turbulence is transformed to a linear and negative transformation (Fig. 7d). The model explains 40% of the observed variance of maximum anchovy catch productivity. Similar results are found when using continental shelf length instead of its surface. Also, using mean fish catch productivity instead of maximum fish catch productivity indices provide similar results (not shown). Results using the GAIM algorithm instead of the ACE algorithm are similar as well (Fig. 8). Both monovariate and multivariate analyses suggest similar patterns for the relations among fish catch productivity and environmental variables : -the transformation of the upwelling index is mostly linear and positive; -the transformation of the turbulence index is close to be linear and negative paniculary after a value around 200-2 50 m3/s3; productivity (103t) 0) a, 3 - - F 0 2- E" - 0' - E 1- m l= O Fig. 7: Optimal empirical transformations from the ACE algorithm using maximum anchovy catch productivity as dependent variable and upwelling index, continental shelf surface and turbulence as predictor variables. The transformation of the response variable is forced to be linear. R2 value is 0.4. Numbers identify ecosystems (see Tab. 2 and 3). *2 A -8.90~ .l1 I I Continental shelf surface(10~km) Turbulence (m3 /s3) 0 2 4 6 8 10 12 11 Maximum anchovy catch 402 Pelagic Fisheries in Upwelling Areas 02 0,6 I 194 Upwelling index (m3/s/m) 0 200 400 600 800 Turbulence (m3/s3) 0 500 1000 1500 2000 Continental shelf surface (1 C12km) - th~. transformation of sea surface temperature is linear and negative; and - th~: transformation of continental shelf surface is close to a log transformation with a breaking point around 100 000 km2. Thc comparative and exploratory analysis of the relationship among estimates of fish productivity and environmental feat,rres of upwelling systems reveals that a combination of several factors is necessary for high productivity: - a I! .gh upwelling intensity (near to 1.28 m3/s/m); - a ~xoderate turbulence (around 200-250 m3/s3); - a rrledium sea surface temperature (l j-1G"C); - a :i.:latively large continental shelf (approximately 100 000 km2). Thc results of our statistical analysis must be considered with caution, however, as important limitations do exist: - thlt:: comparisons are based on only eleven data points, and consequently, the statistical validity of the results is clt~estionable due to the low number of degrees of freedom; - onr: system with extreme values (Peru) plays an important role in all analyses. This represents important limitation in our comparative analysis. However, it is also true that: - the number of ecosystems with a documented pelagic fisheries and for which environmental data exist are limited; and - the number of environmental factors which have been hypothesized impact on productivity is large compared to the number of ecosystems that can be compared. Nevertheless, the present analysis gives some valuable information and cues. First, it appears that the size of the ecosistem is not the only parameter that influences its fish catches. Upwelling strength, turbulence, and sea surface temperature also play an important role. Only a combination of several environmental factors ensure a high fish productivity. The relationships between fish catch productivity and environmental variables appears to be in agreement with independent ecological knowledge on ecosystem functioning. High upwelling intensities as source of food availability (Wroblewski and Richman, 1987; Cushing, 1990) and small-scale turbulences that increase the encounter rate between food particles and larvae (Rothschild and Osborn, 1988; MacKenzie and Leggett, 1991) are thought to be beneficial to larval survival. The positive relationship between upwelling intensity and fish catch productivity could be related to these combined effects. In contrast, intense wind-driven turbulent mixing that mixes up patches of lamal food appears to be detrimental (Lasker, 1975; Peterman and ~radford, 1987; Cury and Roy, 1989). Bakun (this vol.) identified a 'fundamental triad' of three major processes that combine to yield favorable environmental conditions for fishes: an enrichment process (upwelling, mixing ...), a concentration process (water column stability, convergence ...) and processes favoring retention within appropriate habitat. In some degree, the environmental parameters we selected may be considered as proxy variables that account for some of the processes involved in such triad. For example, the size of the ecosystem combined with upwelling intensity determines global enrichment of the ecosvstem while turbulence is involved in processes that concentrate and retain food and larvae. A comparison of the environmental values of the upwelling areas with the 'optimal environmental values' is presented on Figure 9 and the limiting factor(s) to productivity are identified (Table 4). First, it is apparent that the Peruvian ecosystem is the only one which combines al1 the optimal environmental conditions (Fig. 9). In Chile and Narnibia, the upwelling index is favorable; however, it is associated with an excessively high tui*bulence index. The same high upweiiing index is found off Côte d'Ivoire-Ghana, but is associated with low turbulence. In South Africa, Spain and California, the turbulence index is high and associated with a low upweliing intensity, therefore limiting productivity. In every upwelling areas, except Peru, at least one environmental condition differs from the 'optimal conditions' and consequently tends to limit productivity. But what will happen under changes of one or several environmental factors? The effect of a gradua1 or a rapid climatic change on living marine resources is a challenge as numerous parameters are involved. There is no reliable computer-generated climate impact scenario about the next several decade, but generalizations derived from case-to-case assessments of past and present experiences can be used (Glantz, 1992). Such assessments can indeed provide first approximations on how fisheries might respond to environmental changes. Comparative analysis constirutes a good base of information to begin such assessment of possible impacts of environmental changes on fish productivity. Some scenarios for fish productivity under climatic chmges derived from previous studies (Fig. 9) can thus be considered for forecast by analogy (Glantz, 1992). How and how much will productivity evolve if one or several environmental parameters change? Let's assume for example two simple scenarios. First, a drastic increase of upwelling intensity that provides more nutrients would probably improve the fish productivity in major ecosystems. The consequences may be stronger in areas where low upwelling intensity is the main limiting factor: Morocco, Senegal and Venezuela (Table 4). Secondly, under a decrease of the intensity of turbulence, a higher fish productivity may be expected in areas where high turbulence is limiting factor: California, Chile, Spain-Portugal, Morocco, Namibia, South-Africa and Venezuela (Table 4). 404 Pelagic Fisheries in Upwelling Areas The reality is obviously more complex as environmental factors change sirnultaneously. A scenacio involving one environmental pxameter is thus only a very simplified view of what might occur under climatic changes. However, a qualitative approach aiiows to predict the increase or the decrease of the fish productivity and can give some preluninary answers. e environ- Tab. 4: The limiting factor(s) to productivity in upwelling areas. The signs (+ and -) indicate negative or positive deviation from the 'optimal erivironrnental values'. The limiting factors are noted in decreasing order of deviation. - - Upwelling areas Limiting factors 1 California V3+ CUI - SST- 2 Peru SST + 3 Chile V3+ CUI - 4 Spain-Portugal V3+ CUI- S Morocco CUI - V3+ SST+ 6 Senegal CUI- SST+ V3 - 7 CBte d'Ivoire - Ghana SST+ V3 - CUI - 8 Namibia V3+ CUI+ 9 South Africa V3 + CUI - SST+ IO Venezuela CUI- SST+ V3+ 11 India SST+ CUI - V3 + 3. OPTIMAL ENVIRONMENTAL CONDITIONS IN THE PERUVIAN ECOSYSTEM: REALITY OR TAUTOLOCY? t is possible to compare environmental variables in a given upwelling area to whar appears to be the optimal environmental conditions. These, however, were largely derived from the Pemvian ecosystem's values. Peters (1991) V. FAIIRE AND P. CURY 405 defines a tautology as "purely logical constructs rhat describe the implication of given premises and never reveal more than those premises contain". As Peru is known to be the most productive upwelling area, it is clear that using Our approach it will define the optimal environmental combination of factors. Thus, our results may be regarded as a tautology of poor scientific value. However, our comparative analysis did provide a framework for considering the relative impact of several environmental factors on fish productivity. It emphasized the importance of liniiting factors such as turbulence, upwelling intensity or size of the ecosystem. This should promote new insights of how to relate environmental variables to fis11 productivity in a multivariate context. Paleoecological studies reveal that pelagic fish populations experienced large natural fluctuations which were clearly unrelated to fishing pressure and that past abundances in California or in Peru were sometimes much higher than during the last century (Soutar and Isaacs, 1974; De Vries and Pearcy, 1982; Baumgartner et al., 1992). For California, we identified factors that limit fish productivity but whithout any reference point; this was not possible for Peru. This stresses the limit of Our approach as it may be that the Peruvian ecosystem is able to piotluce even more under other, but still undefined, environmental conditions. This work was financed by the Programme National du Déterminisme de Recrutement (PNDR). We thank Claude Roy for providing the environmental data used here. Bakun A. 1985. Coniparative studies and the recruitmerit Crawford R.J.M., L.V. Shannon and D.E. Pollock. 1987. Tlie problem: searching forgeneralizations. CalCOFI Rep., 26: 3040. Beriguela ecosysterii. Part IV. The major fish and invertebrate resources. Oceanogr. Mal: Biol. Ann. Reu, 2 5: 3 53- jO j, Bakun A. 1994. Clinfate change and narine populations: interactions of physical and biological dvnatizics. Paper Cury P., C. Roy, R. Mendelssohn, A. Bakuii, D.hl. Husby and presented at Greenpeace/University of Rhode Island R.H. Parrish. 1995. Moderate is better: exploring iionlinear Workshop on the Scope, Significance, and Policy Implications climatic effect on Californian anchovy (Engraulis 11101z(ns). of Global Change and the Marine Environment, 16 p. Climate change and fish population. R.J. Beamish (ed.). Can. Spec. Publ. Fish. Aquat. Sci.: 121: 417-424. Bakun A. et R.H. Parrish. 1982. Turbulence, transport, and pelagic fish in the California and Peru Current systems. Cury P. and C. Roy, 1989. Optiinal environmental window aiid CalCOFI Rep., 23: 99-1 12. pelagic fish recruitmeiit success in upwelliiig areas. Can. ,]. FisIl. Aquat. Sci., 46 (4): 670-680. Baumgartner T.R., A. Soutar et V. Ferreira-Bartrina. 1992. Reconstitution of the history of Pacific sardine and northern Cushing D.H. 1995. Population production and wgulation anchovy populations over the past two millennia from sediments in the sea. A fisheries peupectiue. Cambridge Uiiiversity of the Santa Bahm basin, California. CalCOFI Rep., 33: 2440. Press, 324 p. 406 Pelagic Fisheries in Upwelling Areas Cushing D.H. 1990. Plankton production and year-class s rength in fish populations: an update of the match/ niismatch hypothesis. Advances in Marine Biologl~, A-ademic Press, 26: 249-293. Ili: Vries T. J. and W.G. Pearcy. 1982. Fish debris in sediments O! the upwelling zone off central Peru: a late Quaternary record. Deep. Sea. Res., 29: 87-109. F.10. 1995. Production halieutique nzondiale. 1950-1993. Siipplément au volume 76 de l'annuaire statistique des pèches 1!)93. Captures et quantités débarquées de la FAO. Roiiie. C;iizman R., P. Fréon and J. Meiidc~za. In press. La pesqueria dt, sardina Sardinella auda en el oriente de Venezuela. su e\ olucion en el périoao 1973-1989. Bol. Inst. Oceanogi: C: iiv. Oïiente. G1antz M.H. 1992. Introduction. In: M.H. Glantz (ed.). Cil:nzate variab~lity, clinzate change and fisbenes. C,iinbridge University Press: 1-14. H,irvey 1). H. and M. D. Pagel. 1991. Vie comparative method in ievolutionar)~ biology. In: R.hl. Mny and P.H. Harvey (eds.). O.,ford series in ecologv and evolution. Oxford University Pr S. 239 p. Hiistie T.J et R.J. Tibshirani. 1990. Generalized additive ~wodels. Chapinan et Hall, London, 33 j p. Eiwasaki T. 1992. Clinlate-dependent fluctuations in the fat ea:itern sardine population and their impacts in the fisheries arid society. In: M.H. Glantz (ed.). Clilliate variability, clilrlate change andfisheries. Canibridge University Press: 32 ;-354. La: ker R. 197j. Field criteria for survival of anchovy lame: the relitionship between inshore chlorophyll mzimum layers and suc cessful first feeding. US. Fish. Bull., 73: 4 j3-462. LIiii.:h-Belda D., R.J.M. Crawford, T. Kawasaki, A.D. MacCall, R.11. Parrisli, R.H. Schwartzlose and P.E. Smith. 1989. W/orld- witle fluctuations of sardine and anchovy stocks: the regime prc~blem. S. Afr.J hilla. Sci, 8: 19j-205 Lluch-Belda D., D.B. Lluch-Cota, S. Hernadez-Vasquez and C.A. Salinas-Zavala. 1991. Sardine and anchoq spawning as related to temperature and upwelling in the California Current system. CalCOFI Rep., 32: 10j-111. Longhurst A.R. and W.S. Wooster. 1990. Abundniice of oil sardine (Sul-clinella longiceps) and upwelling on the southwest coast of Iridia. Cari. J Fish. Aquat. Sci . 47 (2): 2407-3419. MacKenzie B.R. and W.C. Leggett. 1991. Quailtifying [lie contribution of sriiall scale turbulence to the encouilter rares betïveen larval fish and their zooplankton prey: effecrs of wind and tide. Mai: Ecol. Prog. Sei:, 73: 149-160. Mainard Smitli J. and R. Holliday. 1979. Preface. Iri: J. Maynard and R. Holliday (eds.). The evoluiion of'ahptcrtion by natural selection. The Royal Society. Londori: 5-7. Parrish R.H. et A.D. MacCall. 1978. Ciinlatic vari:itioii aiid exploitation in the Pacific mackerel fisheq. Col(/: Dept. Fisb. Ga~rze Fish. Bull.. 167: 1-110. Peternian M.R. and M.J. Bradford. 1987. Wind speed and mortalitv rate of a niarine fish, the northern aiicliovy (Engrnulis tizo~dax). Scierice (Wdsh., DC), 235: 354-356. Peters R.11. 1991. Tautology. Ir/: R.H. Peters (ed.). A critique for ecologil. Canibridge University Press. 38-73. Rotliscliild B.J. and T.R. Osborn. 1988. The effects of turbulence on planktonic contact rates. ./. Plcrnkton Rej,. 10 (3): 465-474. Shai'p G.D. and J. Csirke. 1983. Proceedings of the Expert Consultation to Exaniine Changes in Abundance and Species Composition of Nerctic Fish Resources. FA0 Fish. Re) , 291 (3). Soutar A. and J.D. lsaacs. 1974. Abundance of pelagic fish during the 19th and 20th centuries as recorded III anaerobic sediinent of the Californias. Fish. Bull., 72: 257-27;. Wroblewski J.S. and J.G. Richnlan. 1987 The non-Iinear response of plankton to wind mixing eveilts. Iiiiplicatioiis for survival of larval northern anchovy. J. Plaiiktoti Res.. 9 (1): 103-123. Clu peoids Reproductive Strategies in U pwelling Areas: a Tentative Generalization * ORSTOM Laboratoire d'informatique Appliquée (LI/\) 32, avenue Henri Varagnat 93' 43 Bondy cedex FRANCE ** ,:entre ORSTOM de Brest BP 70 29:i80 Piouzané FRAIICE *** Sea Fisheries Research lnstitute Pri\ ate Bag X2 Rofge Bay 801 2 Cape Town Socr~ AFRICA Using a comparative approach, the reproductive strategies of sardines, anchovies and sardinellas in twelve upwelling areas over the Atlantic, Indian and Pacific Oceans are investigated in relation with major environmental processes affecting recruitment success. The main reproductive habits are identified, in each area and for each species, based on an extensive survey of the literature. From this, the main spawning grounds and the months corresponding to the peaks of the spawning season are identified. The monthly mean values of upwelling intensity and of wind speed are calculated in each spawning grounds using data from the Comprehensive Ocean Atmosphere Dataset (COADS). A comparison, between areas, of the value of these environmental parameters during the peaks of the spawning season is performed. Off Peru, spawning occurs when the intensity of upwelling is maximum (1.8 m3/s/m). Off Morocco, reproduction is out of phase with the upwelling process (upwelling intensity is in the range of 0.2-0.3 m3/s/m). Off Namibia, Chile and California, spawning occurs at intermediate values. Ir appears that the timing of the reproduction of sardines, sardinella or anchovies occurs over a wide range of upwelling intensity. There is no apparent link between the timing or the intensity of upwelling and the occurrence of the seasonal spawning peaks. A similar analysis is performed using the monthly values of wind speed during the seasonal spawning peaks. It turns out that for sardine and sardinella, reproduction occurs when the monthly values of wind speed are within a narrow window, bounded by 5.3 and 6.1 m/s for ten of the twelve upwelling ecosystems. The two outliers are the sardine popuhtions from the southern Benguela and the Iberian Peninsula. For anchovy, there is apparently little correspondence between the timing of reproduction and wind speed values. These results are discussed in the light of the 'optimal environmental window' concept of P. Cury and C. Roy, and a generalization of the spawning strategies of small pelagic fishes in upwelling areas is presented. A travers une approche comparative incluant douze régions d'upwelling des océans Atlantique, Indien et Pacifique, les stratégies de reproduction des sardines, anchois et sardinelles sont étudiées en relation avec les processus environnementaux majeurs affectant le succès du recrutement. Dans chaque région et pour chaque espèce considérée, une revue exhaustive de la littérature permet d'identifier les principales caractéristiques de la dynamique de reproduction. Les principales zones de ponte et les mois correspondant aux pics de la saison de reproduction sont ainsi identifiés. Les moyennes mensuelles de I'intensité de I'upwelling et de la vitesse du vent sont calculées dans chaque zone de ponte à partir de la base de données COADS (Comprehensive Ocean Atmosphere Dataset). Une comparaison des valeurs de ces paramètres environnementaux durant les pics de la saison de reproduction est effectuée entre les différentes zones d'upwelling étudiées. Au large du Pérou, la reproduction a lieu lorsque I'intensité de I'upwelling est maximale (1,8 m3/s/m). Au large du Maroc, la reproduction est décalée par rapport au processus d'upwelling (I'intensité de I'upwelling se situe dans l'intervalle 0,2-0,3 m3/s/m). Le long des côtes de la Namibie, du Chili et de la Californie, la ponte a lieu à des valeurs intermédiaires d'upwelling. II apparaît que le calendrier de la reproduction des sardines, anchois et sardinelles, s'étale sur une large gamme d'intensité d'upwelling. II n'y a pas de lien apparent entre l'occurrence du processus d'upwelling et celle des pics saisonniers de reproduction. Une analyse similaire est effectuée en utilisant les valeurs moyennes de la vitesse du vent durant les pics saisonniers de reproduction. Il en résulte que pour les sardines et sardinelles, la reproduction a lieu lorsque les moyennes mensuelles de la vitesse du vent sont comprises dans une étroite fourchette de valeurs de j,3 à 6,l m3/s/m pour dix des douze écosystèmes d'upwelling étudiés. Les deux exceptions sont les populations de sardine du sud du Benguela et de la péninsule ibérienne. 4 70 Clupeoids Reproductive Strategies Pour l'anchois, i1.y a apparemment peu de correspondance entre le calendrier de la reproduction et les valeurs de la vitesse du vent. Les résultats sont discutés à la lumière du concept de la u Fenêtre Environnementale Optimale )) de P. Cury et C. Roy, et une généralisation des stratégies de reproduction des petits poissons pélagiques dans les zones d'upwelling est présentée. Located in the tropical or subtropical zones, coastal upwelling ecosystems represent less than 0,1% of the entire ocriianic surface but are pan of the most productive oceanic regions and are able to produce between 20 and 30% of the wcrldwide annual fish catches (Cushing, 1969; Ryther, 1969; Pauly and Tsukayama, 1987). Coastal upwelling ecosystems artn mainly colonized by small pelagic fishes such as anchovies, sardines or sardinellas. These fish populations are chiracterized by imponant annual fluctuations of their abundance. For instance, after a peak of production in 1970, the Pei.uvian anchoveta stock collapsed in 1972-1973 (Valvidia, 1978; Pauly and Tsukayama, 1987); similarly, Pacific sardine sutidenly disappeared from the fishery in the 1950s (Lasker and MacCall, 1983). Although these populations are usuallv sulimitted to strong fishing pressure, these variations of abundance appear mainly due to recruitment failure related to chinges in the marine environment (Kawasaki, 1983; Shepherd et al., 1984). Upwelling ecosystems are characterized by a verv high rate of primary production. An upwelling ecosystem is also a dispersive environment where panicles tend to be swri:pt away from the coastal environment by the wind-induced offshore drift. Persistent equatonvard winds also induce a strong and continuous mixing of the surface water column. These are some of the major characteristic of coastal upwelling eccaystems; they can have strong ecological implications. In :L recent synthesis of the major environmental processes affecting fish reproduction, Bakun (1996) identified three maior classes of processes that combine to yield favorable reproductive fish habitat. They are: 1) enrichment processes (upwelling or mixing); 2) concentration processes (convergence, fronts, stratification) and 3) retention processes that maintain eggs and larvae in the suitable habitat. Despite the high rate of production, the triad indicates that the offshore flow and the intense wind mixing that characterize upwelling ecosystems can create adverse conditions for larval survival ancl subsequent recruitment success. The migrations that some of the major small pelagic fish population undertake in orcer to find suitable reproductive habitat confirm that an upwelling ecosystem can be an adverse habitat for fish to reproduce (Hutchings, 1992; Bakun, 1996). However, small pelagic fish populations are quite successful: they are well known for being able to develop very important biomass in eastern boundary ecosystems. Comparative studies, in several up~velling ecosystems, of the reproductive strategy of fishes helped identify common key environmental processes for sm;ill pelagic fish reproductive strategy: sardine, sardinella and anchovy tend to avoid spawning in areas dominated by strclng offshore transport and strong wind mixing (Parrish et al., 1983; Roy et al., 1989). Reproductive strategies of small pelagics appear to be tuned to minimize the detrimental effects of the environment on larval survival (Bakun, 1996). The Optimal Environmental Window concept (OEW; Cury and Roy, 1989) provides a simple model for relating the upwelling process to larval survival and recruitment success. Roy et al. (1992) used the OEW to account for the difference between reproductive strategies of small pelagic fishes observed in several areas within the Canary Current upwelling ecosystem. These authors showed that off West Africa, there is no apparent relationship between the upwelling process and reproduction, but rather a striking correspondence between the timing of reproduction and the occurrence of wind speed of about j-6 m/s. Tnis range of wind speed corresponds to the optimal wind conditions defined by the OEW. We present here an attempt to generalize the results of Roy et al. (1992) to other upwelling areas such as the Benguela Current system, the California Current system, the Humboldt Current system and the Malabar coastal upwelling ecosystem off India. These upwelling areas constitute a unique opportunity to develop a comparative approach. They share fundamental characteristics: wind is the driving force of the upwelling process in these areas; they are colonized by closely related species, such as anchovies, sardines and sardinellas (Table l), which are al1 small-sized, and have fast growth, a short life span, an early maturation and very high fecundity. Svstem Dominant ciu~eoids ~anarTcurren t - Sardina pilchardus ~ardinella aurita Sardinella nzaderensis Engraulis encrasicollrs Benguela Current Sa rdinops oceila tus Engraulis capensk California Cu rrent Sardinops caeruleus Engraulis nlmdax Humboldt Curren t Sardinops sagax Engraulis nngens India, Malabar Coast Sardinella longiceps Sardinelia Jinzblia~a Table 1 : Species of coastal pelagic fishes studied in each upwelling area. 1. BIOLOCICAL AND ENVIRONMENTAL DATA A review of the literature provides information on the reproductive seasons and locations for each spawning area and each species. Table 2 summarizes the information gained through this review. The identification of the spawning seasons results from a compromise between the information coliected. It can be considered as being, in average, valid for the period covering the 1950s to the 1990s. In some cases, data do not extend on a sufficiently long time interval. Then, some particular years are chosen to compare biological information and environmental data (Table 3). Environmental data are derived from the Comprehensive Ocean Atmosphere Dataset (COADS; Woodruff et al., 1987) using the software and CD-Rom produced for CEOS (Mendelssohn and Roy, 1996; Roy and Mendelssohn, this vol.). 4 12 Clupeoids Reproductive Strategies ECi3SYSTEMS Spiiwning grounds -- Spawning seasons Rderences SARDINES CALPORNlA CURRENT Sardùaops mlm Soiirhem Califoniia Bight MAMI (30 ,wN) Bail Califomia (îG30°N) A 5 CAIiIARï CURRENT Sardfna pilobardus Spain, Bay of Biscay - MAU mi .upl (37-41'~) 1EM N Q Mo r~cco (2830°N and 32-34"N) 1 F - D We lem Sahala (22-26"N) MAM QND HUMBOLDT CURRENT Saràinops saga Pen8 (614"s) Chi e, Arica (1824's) BENGUELA CURRENT Sarhops ocdlahrs Wal . is Bay (2&24'S) IEM Wesieni Agulhas Baiik 1 F (34 SOS, l82O0S) Ahlstroni (lm), Rosa aiid lamastu (lW), Alilstroni (1367), hmsli et al. (1981. 1383). Ahlsrmm (1367). \Vyatt aiid Ers-Gaiidam (1983). Sola (1987). lago de Laiizos el al. (1988), Sola et al (192). Ré (1981). Réel al. (1982), Figueiredo aiid hliguel Sanros (lm), Cuiilia aiid Figueiredo (lW), Ré el al (1%). Fumesriii and Fumes~iii (1959); hms li et al. (1983). Donuiievsky and Barkova (19?6), FA0 (1987). lAso Sliarp (1%0). Pairish et al. (1183), Muck et al. (1%7) Pamsii et al (1383). 18SO S O N Marrhws (1360). Pams h et al. (1983). Le Clus (19%). Hutching (192). s O N D De Jager (lm), Rosa aiid Lamastu (1360) Pairish el al. (1383). CAEIARY CURRENT Sur Wia aurlta hlauritania, Baiic d'Arguin (1822"N) Sou Iiem Salepl (12-lioN) INI)lA Sar.&ndia longiœps Mal. Inr Coast (816"N) 1AS Coiiand (1977), Boëly et al. (1982), Fréoii (1%). Li 1 O N Conaiid (1977j, Boëly el al. (198?), Fréoii (lm). ,IlAs Nair (1959, lm), Rosa and iaevastu (lm), Aiiioiiy Raja (1 W), Loiigliurs t aiid Woos rer (lm). AN( HOVIES - CALIFORNIA CURRENT Engraulis mordax Sou1 hem California Bight 1 HA Huiiter (1977), Lasker aiid Snu~li (19n Snurli aiid (30.: i"N) Ricliardson (1977), Snudi and Lasker (19?8), Pairish et al (1981, 1383, 1986) Bal.! ICalifomia (?G30°N) F MAM Sliarp (1980). Souihem Bala Califomia (22-26"N) J F M A Pams h et al (1383) CANARI' CURRENT Eng "aulis enuasiwlus hlonicco (2830"N and 32-34"N) 11A Fu~iiestiii alid Furiiesriii (1%')) HUMBOLDT CURRENT Engiaulis ringens Pem (614's) F M A 5 c! Valvidia (19?8), Sliarp (IW), Cusliiiig (1982). hirish et al (1!%3), Alha t et al (1984), Paul\! and Sonano (1983, hluck (198), Senocaket al (1989) Chilia Anca (1824"s) JAS0 Pams h et al (1983) BENGUELA CURRENT Engraulis capensis Walv \ Bq (î&?4"S) IEM Q Pdmsh el al (1%3), Le Clus (199), Hutchiiigç (1312) Ur. rc3m Agulhas Bank 1 F S Q N Q Sheltoii and Hutcliiiigs (1982), hirish et (il (1953) (3430'S, 182OoE) Hurcliing (1%2), Waldron et al (192) SARDINES ANCHOVIES Baja California 1952-1959 Southern Baja California 1970-1990 Spain 1980-1990 Walvis Bay 1970-1990 Portugal 1970-1990 Agu llias Bank 1970-1990 Chile, Arica 1970-1990 - - Table 3: Years considered in the study of biological and climatological data, by species. Monthly rime series of scalar wind speed and wind-stress, from 1950 to 1990, were constructed in each spawning areas (Table 2). A Coastal Upwelling Index (CUI) was calculated from the wind stress data following Bakun (1973). This index of the strength of the upwelling process is the offshore component of the wind induced Ekman transport. From the montlily rime-series of scalar wind speed and CUI, a mean monthly cycle was calculated. In most cases spawning grounds and nursery grounds have a similar location. This is not the case for the spawning areas located on the Agulhas Bank (Benguela) and in the Bay of Biscay (Spain). In these two areas, spawning occurs outside the upwelling area and eggs, once spawned, are removed from the spawning grounds and carried by coastal jets to the nursery grounds located in the upwelling (Shelton and Hutchings, 1982; Cabanas et al., 1989). Consequently, for these two examples, environmental data corresponding to the nursery grounds are considered: the Galician Coast (Spain: 42-44"N) and the area surrounding Sr Helena Bay (Benguela: 30-34"s). The duration of the upwelling season varies from one ecosFtem to the other (Fig. 1): it is a year-round process off Peru and South Africa but limited to spring and summer off California and Morocco. There is no apparent relationship between the timing of reproduction and the upwelling process (Fig. 2): for instance, off Morocco, sardine reproduces outside the upwelling season; on the contrary, in the California Bight or off Peru, spawning occurs when the upwelling is active. Following Bakun and Parrish (1982) and Parrish et al. (1983), we try to characterize the environmental conditions prevailing during the spawning season by using CUI and wind speed. These two environmental parameters are used as proxy-variables to estimate the strength of several environmental processes such as mixing by the wind, enrichment by the upwelling and offshore drift by the wind induced Ekman transport. These wind related processes are thought to be the key environmental processes to be considered when addressing the effect of the environment on fish population in upwelling areas (Iasker, 1975; Parrish et al., 1981; Cury and Roy, 1989; Bakun, 1996). Since the observed spawning habits reflect the net adaptive response to a history of annual successes or failure in reproduction, one may expect that spawning habits would be seasonally and geographically tuned in order to provide a compromise between the environmental processes affecting recruitment success (Bakun et al., 1991). 4 14 Clupeoids Reproductive Strategies SONDJFMAMJ JASOND CALIFORNIA CURRENT Southern California Bight Baja California HUMBOLDT CURRENT Peru Chile, Aiica CANARY CURRENT Western Sahara Morocco Spain Poitugal BENGUELA CURRENT Walvis Bay Agulhas Bank Fi;. 1: Temporal relationships between sardines reproduction seasons and upwelling seasons. Spawning periods arc represented by black points, upwelling seasons are in light grey, and upwelling peaks in dark grey (references for upwelling seasons : Cushing, 1971 ; Chesney and Alonso-Noval, 1989). The mean monthly values of the two parameters during the sardine peak spawning season in each ecosystem are selected and plotted against each other (Fig. 2). Each ecosystem is characterized by different CUI values, either high or low. Two groiips can be clearly distinguished. A first one corresponding to the sardine population off California, hloi.occo, Western Sahiira, Peru, Chile and Namibia; for this group the wind speed values reported during the spawning seasons are clusterecl witk in a narrow band of wind speed, between j and 6 m/s. The second group corresponds to the Iberian Peninsula (Spain, Poniigal) and the Agulhas Bank; in these ecosystems, the sardine populations do not follow the same pattern, wind values reported during the spawning season reach 7 to 9 m/s. We follow the same procedure for the anchow and sardinella populations. For anchow, there is no clear pattern of corrcspondence between spawning and wind speed. Reproduction occurs within a wide wind range: data points are scat ered between wind speed values of 5 and 8 m/s (Fig. 3). For the West African and Indian sardinella populations, spaxning appears to be restricted to a range ofwind speed between 5 and 6.8 m/s (Fig. 4). However, one should note that the wmber of data points for sardinella is rather limited. 'Yhrough the study of clupeoid reproductive strategies, mo categories of ecosystems can be identified: - Ecosystems of low latitudes: the main upwelling ecosystems of the world are part of this group: California, West Africa (hlorocco, Sahara, Mauritania, Senegal), Peru, northern Chile, Namibia and India (Malabar Coast). - Eccsystems of mid latitudes: the southern Benguela (South Africa) and the Iberian Peninsula (Spain and Portugal). -- Y.). SHIN, C. ROY AND P. CLIRY 4 1 5 4 6 8 1 O Wind speed (mls) Fig. 2: Plots of spawning peaks of sardines against rnonthly rneans of wind speed and of coastal upwelling index (CUI), by ecosystern, Fig. 3: Plots of spawning peaks of anchovies against m rneans of wind speed and of coastal upwelling index (CUI), by ecosystem. 4 16 Clupeoids Reproductive Strategies Wind speed (nils) ,. 4: Plot of spawni ella against monthly rneans of wind speed and of coastal upwelling index -.il), by ecosystern. 3 :1. Low latitude ecosystems 111 these ecosystems, a generalization of Roy et al's (1992) results is possible for the sardines and sardinella popiilation. The timing of reproduction for these two species coincides with the occurrence of j-6 m/s winds. This range of w.nd speed values is in accordance with the 'Optimal Environmental Window' defined by Cuv and Roy (1989). There is no apparent relationship between spawning and upwelling intensity. Spawning occurs sometimes during the upwelling season and sometimes outside the upwelling season. The ~:oincidence between spawning and the optimal wind range defined by the OEW suggests that reproductive strategies are strorgly influenced by the seasonal fluctuations of the wind regiie. Reproductive strategies appear to be seasonally tuned in orde- to coincide with the wind value that maximizes recruitment success. The wind value corresponding to the seasonal occurrence of the spawning peaks is constant over a wide range of latitude: from jON for Peru to 33ON for Morocco. This apparent constancy of the 'ideal' wind intensity around which the spawning activity of sardines and sardinella is maximum and is remarkable in view of the strong latitudiial dependence of some of the key environmental processes and scales that can Le expected to be involved (Bakun et al., 1991; Bakun, 1996). The magnitude of the enrichment by the upwelling and of the cffshore drift by the wind-induced Ekman transport are both related to the intensity of the wind, but are also Iatitude- dependent processes. Wind generated turbulent mixing is estimated to be proportional to the third power of the wind speed, and is a process independent of latitude. The apparent constancy of the optimal wind intensity over several ecosystems located at different latitudes, can therefore be interpreted as an indication of the dominance of wind mixing in the seasonal adjustment of smail pelagic fish reproductive strategy in upweliing areas. This may leave the choice of an adequate spawning location as an available means for dealing with limiting factors such as offshore transport and enncliment. The duration and the intensity of the upwelling process appears to have a limited effect on the timing of the reproduction. However, offshore Ekman transport can be detrimental for larval survival and the fate of a fisli population. Fish ma). have to select an adequate location for spawning in order to avoid the detrimental effect of wind-induced offshore drift. In these mid-latitude ecosystems, the spawning grounds are located in bays or in coastal indentations, downstream of intense upwelling centres (Parrish et al., 1983; Roy et al., 1989). Off Walvis Bay, the upwelled waters corning from the Lüderitz upwelling centre, are carried away by the main current, diffuse in the bal! and supply the nurseiy witli nutrients. The advection of cold waters in the bay induces at the same time the formation of convection cells, reducing larval offshore drift (Bakun, 1996). Furthermore, the larvae are sheltered from strong mixing by the wind. The width of the continental shelf is also an important characteristic. A wide continental shelf enables the formation of retention eddies (Brink, 1983; Nelson and Hutchings, 1987). As the retention process applies also to plankton, a wide continental shelf mai allow a berter coupling between primary and secondary productions. The inshore side of upwelling plumes also provide adequate locations for larvae retention (Graham and Largier, 1997; Roy, in press). 3.2. Mid-latitudes ecosystems There are two ecosystems for which the seasonal spawning is not related to the occurrence of the optimal wind value defined bv the OEW (Fig. 2). These ecosystems are the Iberian Peninsula and the southern Benguela. In these ecosystems, reproduction occurs during time period characterized by an intense wind regime. One also notes that the spawning grounds and nursery grounds are spatially distinct. In these mo areas, the configuration of the coastline is quite similar with a North-South coast where the upwelling develops and an East-West coast located polewarcl and up-wincl of the upwelling area. This configuration of the coast provides a unique opportunity for the fish populations to avoicl the reproductive difficulties inherent in an exposed upwelling coast (Bakun, 1996). In both cases, the spawning gsounds are located outside the upwelling coast and rather concentrate poleward along the East-West oriented coast (the Agulhas Bank off South Africa and the Bay of Biscay off Spain). Eggs laid in the Bay of Biscay are carried by a coastal jet towards the Galicilin Coast, in the North-West of Spain (Cabanris et al., 1989). Unlike most cases, eggs are thus laid upward the upwelling centres as regards to the main surface circulation. As spawning occurs outside the upwelling zone (Garcia et nl., 1991), early larval stages are not subjected to the detrimental effects of dispersion linked to Ekman transport. Furthermore, they probably take advantage of the spring bloom. This seasonal pnmary production peak indeed favors larval survival. After being transported along the Galician Coast, tliev can take advantage of the upper layers enrichment by the upwelling process. Moreover, at this stage of development, larvae are more mobile; therefore, their survival is supposed to depend less on concentration (pursuit and attack behavior) and retention processes (horizontal and vertical displacement). The reproductive strategies of sardine and anchovy in the southem Benguela follow a similar pattern. Sardine and ancliovy eggs are laid on the Agulhas Bank, upward the upwelling centre, and are then carried by a coastal jet toward the West coast upwelling area, North of Cape Columbine (Largier et al., 1992). Shelton and Hutchings (1982) have estimated tliat the time for the eggs to be transported to West coast upwelling is in order of days. Along the west coast, St Helena Bay is thought to be 4 18 Clupeoids Reprodudive Strategies an important nursery ground. It is a place where biological production can benefit from the input of nutrient b!~ the up~~elling. The upwelling plume that develops down-wind Cape Colombine creates a physical barrier allowing retention to occur within St Helena Bay (Graham and Largier, 1999. This area constitutes a place a priori favorable for a nurseiy grouncl. In these two ecosystems, the question of the evolutionaly advantage of developing such a strategy, i.e. to spatvn outside the up~ieliing area during time period characterized by intense wind induced mixing remains an open question. In both places, the spawning grounds seem to be charactenzed by a strong vertical stratification which may counteract the detrimental effect of wind mixing. Over the Agulhas Bank, warm waters advectecl from the Indian Ocean by the Agulhas Current overlie cocler and dense water from the Atlantic (Shannon, 1985). This allows to form a protected stable layer where fish can suc~:essfully reproduce under energetic wind conditions (Parrish et al., 1983). Egg development is also strongly affected by tempenture. The cold temperature encountered along the west coast upwelling may also be an important element in frivor of smwning in the warm waters off the Agulhas Bank. In the Bay of Biscay, spawning occurred in spring and is in phrise with the annual planktonic bloom, this might be an important element favoring lamae sumival within the Bay. The timing of sardine and sardinella spawning in low latitude upwelling ecosystems appears to be linked with the occiirrence of wind speed within a range of 5 to 6 m/s. There is no apparent relationship between spawning and upwelling inte-isity. Thus, it was possible to extend Roy et al.5 (1992) results to the major loui latitude upwelling ecosystems of the wor d. This optimal wind range is in accordance with the OEW (Cury and Roy, 1989) which defines 5-6 m/s wind as being the optimal condition for small pelagic fish recruitment success in uptvelling areas. The constancy over a wide range of latit~lde of the optimal wind range is an indication of the dominance of wind mixing in the adjustment of small pelagic fish reproductive strategy to seasonal upweliings. This may leave the choice of an adequate spawning location as an available means for dealing with limiting factors such as offshore transport and enrichment. The:e are two outliers for which the spawning is not related to the occurrence of the optimal wind value. These ecocystems are the iberian Peninsula and the southern Benguela. In these areas, spawning grounds and nursery grounds are ;ils0 spatially distinct. They both share similar topographical characteristics with a North-South oriented coast where the upwelling develops (the nursery grounds) and an East-West oriented coast located poleward and up-wincl of the upwdling area (the spawning grounds). This configuration of the coast provides a unique op port unit!^ for fish population to aiuid the reproductive difficulties inherent in an exposed upwelling coast. Anctiovy reproductive strategy appars to be quite distinct from sardine ancl sardinella strategies. There is no apparent relationship between the upwelling indices or the wind intensity and the timing of anchovy spawning. This remains an operi question. ACKNOWLEDGMENTS Support for this work was provided by CEOS and the Programme National sur le Déterminisme du Reciutement (PNDR). Ahlstrom E.H. 1960. Synopsis on the biology of the Pacific sar- dine (Sardinops caerulea). In: Proceedings ofthe WorldScien- tific Meeting on the BiologV of Sardines and Related Species, Ronte. FAO, 11, Species Svnopses: 415-451. Ahlstroni E.H. 1967. Co-occurences of sardine and anchovy lar- vae in the California Current region off California and Baja Cali- fornia. CalCOFl Rep., XI: 117-13 5. Alheit J., V.H. Alarcon and B.J. Macewicz. 1984. Spawning fre- quency and sex ratio in the Peruvian anchovy, Engraulis rin- gens. CalCOFl Rep., XXV: 43-52. Antony Raja B.T. 1964. Sonie aspects of spawning biologyof indian oil sardineSardinellalongicepsVal. IndianJ Fish. ,11(1): 45-120. Bakun A. 1973. Daily and weekly upwelling indices, west coast of North America 1946-71. U.S. Dep. Comm., NOAA Tech. Rep. NMFS SSRF, 671, 103p. Bakun A 1336. Patternsin theocean: oceanpmcessesand nzarine population dwmics. Uruversityof Cdiornia Sea Grant, San Diego, California, USA, in ccoperation with Centro de Investigaciones Bio- logicas de Nomste, ia Paz, Baja Califomia Sur, Mexico, 323 p. Bakun A and R.H. Parrish 1982. Turbulence, transport, and pela- gic fish in the California and Peru Current systenis. CalCOFl Rep., 23: 99-11?. Bakun A., C. Roy and P. Cury. 1991. The conzparntive approa- ch: latitude-dependence and effects of zuindforcing on repro- ductive success. ICES CM 1991/H45,12 p. Boëly T., J. Chabanne, P. Fréon and B. Stéquert. 1982. Cycle sexuel et migrations de Sardinella aurita sur le plateau conti- nental ouest-africain, des Iles Bissagos à la Mauritanie. Rapp. P.- u. Réun. Cons. int. Erplor. Mer, 180: 3 50-355. Brink K.H. 1983. The near-surface dynamics of coastal upwel- ling. Prog. Oceanogr., 12: 223-2 57. Cabanas J.M., C. Porteiro and M. Valera. 1989. Apossible rela- tion bet~veen sardinejsheries and oceanographic conditions inNWSpanish coastal zuaters. ICES CM 1989/H 18,12 p. Chesney E.J. and M. Alonso-Noval. 1989. Coastal upwelling and the early life history of sardines (Sardinapilchardus) along the Galician coast of Spain. Rapp. P.-v. Réun. Cons. int. Explor. Mer, 191: 63-69. Conand F. 1977. Oeufs et larves de la sardinelle ronde (Sardi- nella aurita) au Sénégal: distribution, croissance, riiortalité, variations d'abondance de 1971 à 1976. Cab. ORSTO&J sei: Océa- nogr., 15(3): 201-214. Cunha M.E. and 1. Figueiredo 1988. Reproductive qcle ofSn/.- dina pilchardus in the central region offthe Poituguese coast (197111987). ICES CM 1988/H61,9 p. Cury P. and C. Roy. 1989. Optimal environnielital window and pelagic fish recruitment success in upwelling areas. Can..l. Fish. aquat. Sn'., 46(4): 670-680. Cushing D.H. 1969. Upwelling and fish production. FA0 Fish. Tech. Pop., 84,40 p. Cushing D.H. 1971. The regularity of the spawniiig season of some fishes. J. Cons., Cons. int. Explor. Mer, 33(1): 81-92. Cushing D.H. 1982. Cli~ilate andfisheiles. London. Acadeiiiic Press. 373 p. De Jager B.D. 1960. Synopsis on the biology of the South Afri- can pilchard Sardinops ocellata (Pappé). In: Prozeedings ofthe World Scient& Meeting on the Biolog~ of Sardines and Rela- ted Species, Roine. FAO, 11, Species Svnopses: 97-1 14. Domanevsky L.N. and N.A. Barkova. 1976. Soiizepeculiaritiesof sardine (Sardina pilchardus Walb.) distribution and spaiil- ning along the noith~uest~lj?ica. ICES. CM 1976/56.6,1 jp. FAO. 1985. Rapport de la troisième réunion du groupe de tra- vail ad hoc sur la sardineSardinapilchardus (Walb.). COPACE! PACE Senes, 85/39,157 p. Figueiredo LM. and k Miguel Santos. 1988. On sexual nintuizr- tion, condition factorandgonadosonzatic index ofsardina pil- chardzrs Walb., offPortugal (198611987). ICES Ch1 1988/H70,4p. Fréon P. 1988. Réponses et adaptationsdes stocks de clupéidés d'Aft.ique de l'ouest à la variabilité du nzilieu et de l'exploita- tion. Analvse et réjlexion à partir de l'exerizple du Sénégal. Etudes et Thèses ORSTOM. 287 p. Furnestin J. and M.L. Furnestin. 19 59. La reproduction de la sar- dine et de l'anchois des côtes atlantiques du Maroc. Reu. Trau. Inst. Pêches marit., 23(1): 79-104. 420 Clupeoids Reproductive Strategies Ga~ia A., C. Franco, A. Sola and A. Lago de Lanzos. 1991. Sar- di7.e (Sardina pilcbardw Walb.) daib egg production offthe Gai'ician, Cantabrian and Bav of Biscav waters in April-May 19510. ICES C.M. 1991/H37. Gciham W.M. and!. L. Largier. 1997. Upwelling shadows as near- shore retention sites: the example of northern Monterey Bay. Co~ltinental SbeifResearch, 17 (5): 509-532. HutiterJ.R. 1977. Behavior and sumival of northern anchovy En;;raulis mordax larvae. CalCOFl Rep., XD(: 138-146. Hu chings L. 1992. Fish harvesting in avariable, productive envi- rorirnenc. Searching for rules or searching for exceptions ? In: A.I.I.. Payne, K.H. Brink, K.H. Mann, and R. Hilborn (eds). Ben- gut la trophic hnctioning. S. Afr. ]. mar. Sci., 12: 297-318. Kawasaki T. 1983. Why do some pelagic fishes have wide fluc- tuarions in their numbers? Biological basis of fluctuation from the viewpoint of evolutionq ecology. In: G.D. Sharp and J. Csir- ke i eds.). Proceedings of the Expert Consultation to Examine Chariges in Abundance and Species Composition of Neritic Fis11 Rest:iurces, SanJosé, Costa Rica, 18-29 April1983. FA0 Fish. Rep., 291 :3): 106 5-1080. Lag de Lanzos A., C. Franco and A. Sola. 1988. Conzpalutiue stw'p of tl~eannual variation in thespazuning intemihl orsar- dino Sardina pilchardus (Walb. 1792) of the Cantabnc and NVC'.$ain. ICES CM 1988/H73,5 p. Lar~ler J.L., P. Chapman, W.T. Peterson and V.P. Swart. 1992. The western Agulhas Bank: circulation, stratification and eco- log) In: A.1.L. Payne, KH. Brink, KM. Mann and R. Hilbom (4s.). Benqela trophic functioning. S. Afr. 1. niar. Sci., 12: 319-339. Lasker R. 1975. Field criteria for suwival of anchovy lamae : the celai ion between inshore chlorophyll maximum layers and suc- cessful first feeding. Fish. Bull., 73(3): 453-462. Lasker R. and P.E. Smith. 1977. Estimation of tlie effects of envi- roniiiental variations on the eggs and lamae of the northern and ovy. CalCOFl Rep., XD(: 128-137. Lask:r R and A MacCall. 1983. New ideas on the fluctuations of the clupeoid stocks off Califomia.In: ProceedingsoftheJoint Ocea- nog(zpl~icAssemblv1982, General Syniposia, Can. National Corn- mittt:e, Sc. Cominittee on Oceanic Research, Ottawa: 110-120. Le Clus F. 1990. Impact and implications of large-scale environ- men:al anomalies on the spatial distribution of spawning of the Namibian pilchard and anchovy populations. S. Afr.J nzar. Sci., 9: 14 1.-1 j9. Longhurst A.R. and W.S. Wooster. 1990. Abundance of oil sar- dine (Sardinella longiceps) and upwelling on the Soutliwest Coast of India. Can.J. Fish. aquat. sci., 47: 2107-2419. MatthewsJ.P. 1960. Synopsis on the biology of tlie south west African pilchard (Sardinops ocellata Pappé). In: Proceedings of the World Scientific ,Vieeting on the Biologv of Sardines aurd Related Species. Ronze FAO, II, Species Synopses: 11 3-135. Mendelssohn R. and C. Roy. 1996. Comprehensive Oceaii Data Extraction Users Guide. US. Dep. Conln~., NOA4 Tech. ~Llenlo. NOAA-Tech. Rep. NiVIFSSWFSC, La Jolla, CA., 228,67p. Muck P. 1989. Relationships between anclioveta spawning stra- tegies and the spatial variability of sea surface teinperature off Peru. In: D. Paul!; P. Muck, J. Mendo and 1. Tsukayania (eds.). The Peruvian upwelling ecosystem: dpnamics arid interactions. ICLARi21 Conj Proc., 18: 168-173. Muck P., O. Sandoval de Castillo and S. Carrasco. 1987. kuii- dance of sardine, mackerel and horse mackerel eggs and larvae and their relationship to temperature, turbulence and aiiclio- veta biornass off Peru.In: D. Pauly and 1. Tsukayiua (eds.).Tlie Peruvian anchoveta and its upwelling ecosysteiii : tliree decades of change.ICURbIStud. Reu., 15: 268-275. Nair R.V. 1959. Notes on the spawning habits and early life-liis- tory of the oil sardine Saidinella longiceps Cuv. et Val. Indiail J. Fish.: 6(2): 342-359. Nair R.V. 1960. Synopsis on tlie biology and fishery of the Indian sardines. ln: Proceedingsofthe \Vorld Scientific Meeting on thé Biologv oflardines and RelatedSpecies. Ro~ne. FA0 Vol. II, Spe- cies Synopses: 329-414. Nelson G. and L. Ilutchings. 1987. Passive transportation of pela- gic systeiii components in the southern Benguela area. ln: A.I.L. Payne, J.A. Gulland and K.H. Brink (eds.). The Benguela aiid comparative ecosystems. S. Aj?. 1. nza?: Sci., 5: 223-234. Parrish R.H., C.S. Nelson and A. Bakun. 1981. Transport iiiecha- nisms and reproductive success of fishes in the Califoriiia Cur- rent. Biological Oceanography. l(2): 175-203. Parrish R.H., A. Bakun, D.M. Husby and C.S. Nelson. 1983. Co~ii- parative clirnatology of selected environmental processes in rela- tion to eastern boundary current pelagic fisli repr,oduction. ln: G.D. Sharp and!. Csirke (eds.). Proceedings of the Ekpert Consul- tation to Examine Changes in Abundance and Species Compo- sition of Neritic Fish Resources,SanJosé, Costa Rica, April1983. FA0 Fish. Rep., 291(3): 731-777. Y.]. SHIN, C. ROY AND P. CURY 42 1 Parrish R.H., D.L. Mallicoate and R.A. Klingbeil. 1986. Age depen- dent fecundity, nuinber of spawning per year, sex ratio, and maturation stages in northern anchovy, Engraulis mordm. Fish. Bull., 84(3): j03-517. Pauly D. andM. Soriano. 1987. Monthly spawning stock and egg production of peruvian anchoveta (Engraulknngens), 1953 to 1982. In: D. Pauly and 1. Tsukqama (eds.). The Periivian ancho- veta and its upwelling ecosystem: three decades of change. ICURMStud. Rev., lj: 167-178. Pauly D. and 1. Tsukayama (eds.). 1987. The Peruoian ancho- veta and its upzuelling ecoystenz : three decades of change. ICL4RM Stud. Rev.,l5,3jlp. Ré P. 1981. Seasonal occurrence, mortality and dimensions of sardine eggs (Sa~dinapilchardus Walb.) off Portugal. Cybiunl, j(4): 41-48. Ré P., A. Farinha and 1. Meneses. 1982. Ichthyoplankton from the coast of Peniche (Portugal) (1979/1980). Arquiuosdo Museu Bocage, série A, l(16): 369-401. Ré P., R. Cabral, E. Silva, E. Cunha, A. Farinha, 1. Meneses and T. Moita. 19%. Sardine spawning off Portugal. Bol. Inst. Nac. Invest. Pescas, 1 j: 31-44. Rosa H. and T. Laevastii. 1960. Comparison of biological and ecological characteristics of sardines and related species -A pre- liminary study. In: Proceedingsof the World ScientiJ2c Meeting on the Biology of Sardines and Related Species. Ronze. FA0 Vol. II, Species Si~nopses: 523- 552. Roy C. In press. Upwelling-induced retention area: a mechanisrn to link iipwelling and retention processes. In: S. C. Pillar, C. Molo- iiey, A.I.L. Payne and F. A. Shillington (eds.). Benguela dynamics. S. Afr,J Mar. Sci, 19. Roy C., P. Cu., A. Fontana and H. Belvèze. 1989. Stratégies Spa- tio-temporelles de la reproduction des clupéidés des zones d'upwel- ling d'Afrique de l'ouest. Aquat. Living Resoui:, 2: 21-29. Roy C., P. Cuiy and S. Kifani. 1992. Pelagic fish recruitment suc- cess and reproductive stntegy in ii pwelling areas: environmental compromises.In: A.I.L. Payne, K.H. Brink, KH. Mannand R. Hil- born (eds). Benguela Trophic Functioning. S. Afr. J mar. Sci., 12: 13 5146. Ryther J.H. 1969. Photosynthesis and fish prodiiction in the sea. Science ,166: 72-76. Senocak T., D. Schnak and D. Pauly. 1989. Mean montlily distri- bution, abundance and production of anchoveta eggs off Peni (4-14"S), 1964-1971 and 1972-1986. In: D. Paiily, P. Muck., J. Mendo and 1. Tsukayama (eds.).The Peruvian upwelling ecosys- tern: dynamics and interactions.ICIARM Cotg Pm., 18: 143-1 j4. Shannon L.V. 198 j. The Benguela ecosystem. Part 1, Evolution of the Benguela physical features and processes. Oceanogt: Mai: Bioi. Ann. Rev, 23: 10j-182. Sharp G.D. 1980. Report of the workshop on effects of envi- ronmental variation on survival of laival pelagic fishes. In: Work- shop on the Effects of Environmentai Variation on the Suivival of iarval Pelagic Fislies, Lima, 20 April-5 May 1980. IOC Woik- shop Repoi-t, 28: 15-66, Shelton P.A. and L. Hutchings. 1982. Transport of anclioy, Engraulis capensis Gilchrist, eggs and early larvae by a frontal jet current. J. Cons., Cons. int. FAplor. Mer, 40: 18j-198. Shepherd J.G.,J.G. Pope and R.D. Cousens. 1984. Variations iri fish stocks and hypotheses concerning their links witli clinlate. Rapp. P.-v. Réun. Cons. int. Explor. Mer, 18 5: 2jj-267. Smith P.E. and S.L. Richardson (eds.). 1977. Standard tecliniques for pelagic eggand lama siirveys. FA0 Fish. Tech. Pap., 17 j. 100~. Smith P.E. and R. iasker. 1978. Position of larval fish in an eco- system. Rapp P.-v. Réun. int. Etplor. Mer, 173: 77-84. Sola A. 1987. i'ear1.y variation ofthe ichthyoplatikton O~COIII- tnercial interest species in the Santander- coast (no& of Spain). ICES CM 1987L32,bp. Sola A, C. Fnnco, A. Iago de Lanzos and L. Motos. 192. Ternponl evolution of Snrditiapilchardus (Walb.) spawning on the N-NW cmtof theIbenanPeninsula.Bo1. Insi. l3p Oceanog:, 8(1): 97-114. Valvidia J.E. 1978. The anchoveta and El Nifio. Rapp, P.-v. Réun. int. Explor. Mer, 173: 196-202. Waldron M.E., M.J. Amstroiig and B.A. Roel. 1992. Birtlidate dis- tribution of juvenile anchovy Engrnulis capensis caiight iii the southern Benguela ecosystem. In: A.I.L. Payne, K.H. Brink, K.H. Mann and R. Hilborn (eds.). Benguela tropliic functioniiig. S. Afi:.J nzar. Sci, 12: 865-871. Woodruff S.D., R.J. Sliitz, R.L. Jenne and P.M. Steurer. 1987. A Comprehensive Ocean-Atinosphere Data Set. Bull. Ait~t.,: ~Vleteoi: Soc., 68: 1239-12 j0. Wyatt T. and G. Pérez-Gandaias. 1989. Ekman transport aiid sar- dine yields in western Iberia. In: Int. S~~nip. Long-tel-iil Chai~ges Mar. Fish Pop., Vigo: 12 j-138. 422 Clupeoids Reproductive St rategies Comparative Modelling of Trophic Flows in four Large 1Jpweliing Ecosystems: Clobal versus Local Effects ASTRID JARRE-TEICHMANN* VI\-LY CHRISTENSEN* * * Iristitute of Marine Science &::)artment of Fisheries Biology Dui?sternbrooker Weg 20 24' 05 Kiel GEKMANY ** nternational Center for Living Aquatic Resources Management (ICLARM) P.Cl. Box 2631 Makati, Metro Manila PHI .IPPINES Trophic flou1 models of productive subsystems of the four large eastern boundaq current ecosystems, i.e., tlie Humboldt Current off northern and central Peru (4-14OS), the noithe1.n Benguela Current off Namibia (1 j-3j0S), the Canarv Current (12-2 jON) off Senegal and Mauritania, and the California Current (28-43"N) were constructed to represent different regimes in tliese systems during the mid-1960s to the eaily 1980s. The models were then analysed and compared by mcans of flow network analysis. The present contribution summarizes [lie results of a more detailed study with emphasis on tlie five dominant fish species: anchovy, sardine, horse-mackerel, mackerel and hake, and on the distinction between local and global effects. Whereas the general structure of the systems is very similar, characteristics pertaining to their size are system specific. Tlie systeiils are al1 rather inefficient in terms of energy transfer up the food web. Total catch is correlated witli primai? production, as well as with the trophic level of the fisliery Productivities of smail pelagics, as well as favorable conditions for them, appear of global nature, wl-iereas properties pertaining to tlie medium-sized fish, as well as tlie irihibition of al1 dominant fish groups, appear more similar within the systems. Properties related to the fishing regime, e.g., fisliing mortality, or the fraction of available primary production required ro sustain fishery catches, are highly variable among systems and regimes. A closer focus on rime-series of flows in the plankton compartments is required to obtain a more detailed understanding of regime-specific properties at the ecosystem level. Des modèles de flux trophiques de sous-systèmes productifs des quatre grands écosystèmes du bord est des océans, c'est-à-dire le courant de Humboldt devant le nord et le centre du Pérou (4-14 OS), le courant nord du Benguela devant la Namibie (15-35 OS), le courant des Canaries (12-25 ON) devant le Sénégal et la Mauritanie, et le courant de Californie (28-43 ON), sont construits afin de représenter les différents régimes de ces systèmes du milieu des années 60 au début des années 80. Les modèles sont analysés et comparés au moyen de l'analyse des réseaux de flux. La présente contribution synthétise les résultats d'une étude plus détaillée focalisée sur les cinq espèces dominantes : l'anchois, la sardine, le cliinchard, le maquereau et le merlu, et sur la distinction entre effets locaux et globaux. Tandis que la structure générale des systèmes est similaire, les caractéristiques propres à la taille des systèmes sont spécifiques. Les systèmes sont presque tous inefficaces en termes de transfert d'énergie au travers du réseau trophique. Les productivités en petits pélagiques, de même que les conditions favorables pour elles, apparaissent de nature globale, tandis que les propriétés relatives aux poissons de taille moyenne ainsi que les effets d'inhibition de tous les autres groupes dominants de poisson appanissent plus semblables à l'intérieur des systèmes. Les propriétés relatives au régime de pêche, c'est-i-dire à la mortalité par pêche, ou à la fraction de production primaire disponible et nécessaire pour maintenir les pêcheries, sont très fluctuantes d'un système ou d'un régime à l'autre. Une attention particulière est nécessaire pour analyser les séries temporelles de flux dans les compartiments planctoniques afin d'obtenir une compréhension plus détaillée des propriétés spécifiques de l'écosystème. 424 Comparative Modelling of Trophic Flows 1 NTRODUCTION Eastern boundary current systems, or upwelling ecosystems, are among the most productive marine areas of the wo-Id. The four largest of these are the Humboldt Current off South America, the Benguela Current off soutliern Africa, the Canary Current of northwest Africa and the California Current off North America. They support large fish stocks of a sirilac species composition (Table 1) and important fisheries, accounting for more than 25% of the world catch of marine fisk (FAO, 1990). Major scientific attention has been devoted towards the management of these fisheries, but to date no toc1 exists which would be capable of dealing with the considerable interannual fluctuations, notably of the anchovies and sar,.iines stocks. Recent approaches emphasize the necessity of managing fisheries a rnultispecies context (see, e.g., cor,tributions in Daan and Sissenwine, 1991), giving expression to the need of understanding the ecology of tlie systems corîponents and their interactions. Thii productivity of the upwelling systerns depends strongly on the oceanographic conditions (Parrish et al., 1983; Cul? antl Roy, 1989), which are likely to change along with intensijing coastal winds due to global warming (Bakun, 1990, 1953). Additionally, upwelling systems are carbon sinks (Walsh, 1989; Siegenthaler and Sarmiento, 1993), which is relevant to ciur understanding of the global carbon cycle. In :]ne with the CEOS concept (Bakun et al., 1993; Cury et al., this vol.), and based on published results of research relivant to models of upwelling ecosystems and earlier modelling studies on the Peruvian upwelling system Uarre et al., 195 1; Jarre-Teichmann, 1992), trophic flow models of subsysterns of the four upwelling regions (Fig. 1) were constructed anc. analyzed, aiming at a cornparison of energy flows and systems characteristics under different climatic ancilor fishing reg:mes Uarre-Teichmann and Christensen, in press). This contribution contrasts the results of this study relevant to global verius those relevant to local properties of the four upwelling ecosystems, with emphasis on the five comrnercially most im~~ortant fish anchovy, sardine, horse mackerel, mackerel and hake. Common narne Genus Soecies Pem California knguela Northwest &ica Anc hovy Engraulis ringens lrlordax cap ensis encrasicohrs Sardine Sardinops caeruieus ocellntus Sardina pilcl~ardus Horse rnackerel Tracburus murphjli s)ln met ricus capensis t rachurus, t recae Mackerel Sconlber 1 aponicus Hake hlerluccius gqi productus cap ensis, polis, Table 1 : Dominant fish species in the four large upwelling systems (modified from Bakun and Parrish, 1980). A. JARRE-TEICHMANN AND V. CHRISTENSEN 425 Fig. 1: The four large eastern boundary currents of the world (shaded) and the subsystems modelled for the present contribution (solid black). The subsystems were chosen such that they comprise the areas occupied by the dominant stocks of small pelagics. 1. MATERIAL AND METHODS 1 .l . Model construction A set of seIren models was constructed of the four large upwelling ecosysrems, averaging two periods each foi the systems off Peru, Namibia and California, and one period for the upwelling system off Northwest Africa (Table 2). As the models are described in detail in Jarre-Teichmann and Christensen (in press), we confine ourselves here to a biief overview of their approach. We used the Ecopath software (Christensen and Pauly, 1992), which based on the work of Polovina and associates (Polovina and Ow, 1983; Polovina, 1984, 1985). Assuming mass-balance over an appropriate period of time, trophic interactions between the components of an ecosystem (species or species groups) are described by a set of linear 426 Comparative Modelling of Trophic Flows eqi~ations, wherein the production of each component equals its withdrawals by other components in the system (predation mortality), its export from the system (fishing mortality and other exports), and the baseline mortality, Le.: Prcduction by (i) = ,411 predation on (i) + nonpredatory biomass losses of (i) + fishery catches of (i) + other exports of (i) The terms in this equation may be replaced by: I'roduction by (i) = Bi (PDi) I'redatory losses of (i) = M2 = C, (Bi ( QDj) DCjti) Other losses of (i) = (1 - EEi) Bi (PDi) anil this leads, for any component in the system, to: Lii (PDi) (EEi - Sj) (Bi (QD,) DCj,j) - Exi = O wh xe: i indicates a component (stock, species, species group) of the model; i any of its predators; Hi its biomass; l',/Bi the production of a component per unit biomass (= total mortality under steady-state conditions); ?Di the consumption of a component per unit biomass; !)Cj,i the average fraction of i in the diet of j (in terms of weight); 5Ei its ecotrophic efficiency (the fraction of the total production consumed by predators or exported from the system); ?xi its export from the system (e.g., by emigration or advection, or fishery catch). Th(, energy balance of each component is given by: ~:.:onsumption = Production + Respiration + Non-assimilated food whcrein consumption is composed of consumption within the system and consumption of imports (i.e., consumption 'ou side the system'), and production may be consumed by predators, exported from the system, or be a contribution to dei -itus. Thi:; structure defines the necessary inputs to the model. These are, for each component, an estimate of its . biomass; - production per unit biomass; total food consumption per unit biomass; - assimilation efficiency; - diet composition; - exports from the system; . ecotrophic efficiency. For each component, one of above parameters B, PD, QD, or EE may be unknown and is estimated when solving the systcim, along with the respiration of that component. If an acceptable result for each of the unknowns is achieved from the inputs, the model is mass-balanced and may be analysed Further. In tiiis study, the ecotrophic eficiencies were computed for most of the components (Table 3) and used to balance the models, where, for obvious reasons O 5 EE < 1 served as a constraint. For components for which the biomasses were not aval .ible or proved erroneous dunng the modelling process, EEs were set and the biomasses estimated. Sys teni Period Doininant"' pelagic Regime characteris tics, reniarks G~titude) fis h s pecies Peru, 1'96"+1971 Ancliovy Ovemlielming ancliovy biomass with world's larges t s irigle- (414 OS) species fisliery; prior to collapse of nncliovy 1973-1981 Anchovy Slow increase ofanchovy stock Northern Benguela, 1971-1977 Sardine Sardine biornass lower than in the late 1963, but still Iiigli. (15-35 Os) Strong fishery on sardine and liake. 19781983 Horse mackerel Sardineand hake biomasses strongly decraseci, nevertlieless Iiavily fislied. Nortlilvest Ahica. 197@1979 Sardine Svstem witli last publislied information. Sardinella liold (12-2 j ON) ecological niclie occupied by ancliovies in [lie other upmjelling systems. Strorig fisliery for sardine and horse ninckerel. Sasonal ~ipwelling, system comprises several tropical fish species. California, 1965-1972 Ancliovy Highly seasonal ~ipwelling. Very low 6sh bioninsses after (2843 ON) brakdown of sardine and mackerel stocks. Fisliirig nioratoria effective for both species. 19781385 Ancliovy Fisli biomasses recovering, end of Eisliing nioratoiia. " in terms ofproduction Table 2: Major characteristics of the four large upwelling systerns and the periods rnodeled 1.2. Anaiysis of the models After a model has been balanced, it is assured that the various estimates of biomass and turnover rates rire mutually compatible, and hence represent a possible and consistent picture of energy flows in the system. Only then is it meaningful to perform further analyses OF the model, e.g., for interactions between its components, or towards a holistic assessment of the system's structure based on the theories OF Odum (1969) or Ulanowicz (1986). If shifts in biomass or catches in an ecosystem reflect transitions between alternative States of that system (Steele and Henderson, 1984; Lluch- Belda et al., 1989), these changes should be reflected in such ecological properties. Various types of Favorable or inhibitive interactions are commonly described in ecology (see, e.g., Odum, 1971). Direct trophic interactions can be assessed by analysing partial rnortality coefficients of the prey groups. In order to also consider indirect interactions between the components in an ecosystem, such as cornpetition, we used the mixed trophic impact routine suggested by Ulanowicz and Puccia (1990). This approach assesses the relative impact tliat infinitissimally small changes in the biomass of a given group would have on the biomass of other groups, provided that the trophic structure in the system does not change. The latter is the reason why it cannot be used for predictive purposes; however, it can well be used ris a sensitivity analysis for interspecific interactions. 428 Comparative Modelling of Trophic Flows Estimated paramater Coinponent Peru Nortliern Northwes t California Benguela Afnca Phytoplankton EE EE B B Benthic producers Zooplankton Anchovy Sardinella Sardine Mackerel Horse mackerel Large scombnds O ther pelagics Meiobentlios Macrobenthos Hake Demersals Marine birds Marine mammals . - Detritus EE EE EE EE " Period 1971-1977 b' Period 19781383 eters in the models of the four large upwelling ecosystems constructed. 6: biomass, EE: trophic efficiency; -: component not included in model. List refers to al1 time periods modeled if not stated ot herwise. Frai:tional trophic levels may also be reexpressed into discrete trophic levels sensu Lindemann (1942) (Ulanowicz, 1995). Thu, a given consumer is not placed on a single (fractional) trophic level according to its diet composition, but is pen:eived as fnding on various (discrete) trophic levels simultaneously. The ratio between trophic flows consumecl or expxted from one trophic level and the flows entering it is defined as transfer efficiency. FisFeries in different areas may have catches of similar size, but their species composition can be rather different, based, of course, on the availability of fish and on the fishing regime. The exploitation of fish on higher trophic levels in tlie food web is niore costly in ecological terms than the exploitation of groups on lower trophic levels, because the energy transfer efficiency up the food web is far below unity. Hence, the maintenance cost of a fishery (or, generally, any system cou ponent) can only be compared across systems by using a common currency, e.g., pnmary production equivalents as irnplemented by Chnstensen & Pauly (1993). Following their approach, cycles in the system are removed first. The end flow of elch path in the system (e.g., fishery catch) is then traced backwards to the pnrnary producers, using, for each step, tlie ratio between consumption and production as a raising factor. Consequently, the sum of the pnmary production required at the :)asis of each path is the total primary production needed to sustain the system component in question, or the fisheiy. 2. RESULTS AND DISCUSSION 2.1. Trophic flow diagrams Exampies of trophic flow diagrams are given for the Peruvian upwelling ecosystem for the periods 1964-1971 and 1973- 1981 (Fig. 2). The general structure of the trophic flow diagrams looks similar for al1 four upwelling systems. With the primary producers and detritus situated on trophic level 1 (by definition), the planktonic and benthic invertebrate groups are located at trophic levels 2.0 - 2. j. Small pelagics and other pelagics ranged next, with trophic levels between 2. j and 3.0, while the predatory fish, as well as marine birds and mammals, were operate at trophic levels 3 and 4. Large scombrids and birds are the top predators in the system. The major flows in al1 systems occurred in the plankton. Other important flows in the Peruvian ecosystem comprised anchovy and the benthic invertebrates, as well as sardine during the later period. Flows towards anchovy were reduced by a factor of more than four between the two periods, while those to sardine increased by a factor of seven, towards values similar to those for anchovy. Flows toward macrobenthos and hake also increased with the higher abundance of these groups. Due to the overall shortage of small pelagics in the system, the trophic level of predatory fish decreased (as they switched to a larger fraction of zooplankton) while that of hake, mammals and birds increased. Feeding of marine mammals 'outside' the system (i.e., on oceanic squid and mesopelagics) was important during both periods. In spite of the considerable changes in the ecosystem, the general structure of the pathways in the system was not altered. 2.2. Systems characteristics The four upwelling systems ranked rather distinctly after the 'size' of their primary production, total biomass sustained in the system, catches and, consequently, total system throughput (Fig. 3). The Peruvian upwelling ecosystem was the largest of these four systems. It was also the system in which the most pronounced changes of system size occurred during the periods analysed. After the collapse of the anchovy stock, it became more similar to the northern Benguela system. The latter decreased in size from the mid-70s to the early SOS, due to the strong decrease of small pelagics (notablv sardine), not compensated by the increased abundance of horse mackerel. The upwelling system off northwest Africa was similar in size to the northern Benguela system, despite the seasonality of its upwelling. The California system, whose upwelling is also highly seasonal, was the smallest of these four systems. 2.3. Productivity of small pelagics The productivity (or PD ratio, equivalent to total mortality) of small pelagics ranged between 1.1 year-' and 2.7 year-l for anchovy, and between 0.4 year-l and 1.2 year-l for sardine in the balanced models (Fig. 4). The productivity of anchovy was highest off Peru, followed by Namibia, northwest Africa and California. Their natural mortality (1.1 - 2.1 yearl) was 430 Comparative Modelling of Trophic Flows - Primary production - 300 1~ Biomass Total system throughpul O Peru Peru Namibia Namibia NWAfr. California Calilorn 1964-71 1972-81 1971-77 1978-83 1972-79 1965-72 1977-8 Naturai moriality Fishing moriality Peru Peru Namibia Namibia NW Alr. Caiifornia Caiifornia 1964-71 1973-81 1971-77 1978-83 1972-79 1965-72 1977-85 Fig. 3: Surnrnary statistics of the seven balanced rnodels constructed, referring to systern size. Systerns are arranged after decreasing prirnary ' production. Note that systerns are set apart in geographic rather than in regirne-specific order. 1 Also note sirnilar trend, in al1 four parameters, of prirnary production, total biornass (excl. detritus), total catches and total systern throughput. Peru Peru Namibia Namibia NW Afr. Calilornia California 1964-71 1973-81 197177 1978-83 1972-79 1965-72 1977-85 Nalural morialfty Fig. 4: Breakdown of total mortality rates for anchovy (a) and sardine (b). Note consistent scale of vertical axes. Systems are arranged after decreasing prirnary production. ?... 1 2- 2. - m - m 433 Comparative Modelling of Trophic Flows Fishing moiiaiity - - alw lys considerably higher than their fishing mortality (0.1 to 0.7 yearl), indicating their importance as a food resource for oth::r system components. Fishing mortality of anchovy was highest off Pem during the late 1960s, followed by Namibia in the early 1980s. The latter is remarkable as anchovy were not dominant off Namibia, neither in the system, nor in the lanclings. Anchovy fishing mortality was still high in the Pemvian and Namibian systems during the other periods, while it waa lower in the northwest African and Californian systems, due to a lack of directed fishery in the former, and 3 more restrictive fishery management in the latter. Sardine were subjected to the highest fishing mortality off Northwest Africa, followed by the northern Benguela system and Pemvian system during the 1970s. Due to the closure of the sardine fishery off California, fishing mortality of sardine off California was negligible. The natural mortality of sardine was highest off Namibia during both periods, followed by the Penivian and the Californian systems. 2.4. Productivity of mackerel, horse mackerel and hake hlackerel tended to be more productive than horse mackerel, with productivities ranging from 0.5 year-l ro 0.9 and 0.3 year-l to 1.1 year-l, respectively (Fig. 5). Total mortality of mackerel was lowest off California due to the closure of its f shery, and similar in the other systems. Total productivity of horse mackerel was rather low in the Pemvian upwelling syst-m, where they grew relatively large. It was more than three times as high off Northwest Africa, were also the highest ftsh,ng mortalities were obsemed. Apart from the Namibian system in the early 1980s, fishing rnortalities were generally ION reflecting the lack of major directed fisheries. The productivity of hake nnged from 0.4 year-l to 0.9 yearl (Fig. 5). It was lowest off Pem in the 1960s, reflecting the foci s of the fishery on small pelagics. During the 1970s, hake were more strongly exploited off Pem. The natunl mortality of horse mackerel was similar in al1 systems except off Northwest Africa, where it was twice as high as in the other systems. For mackerel and hake, both components were approximately equal off Pem during the 1970s, as well as in the Namibian and Northwest African systems, reflecting their strong exploitation. In general, the results from our balanced ecosystem models confirm that the rates of natural mortality are not systern- spec:ific (see also Beverton and Holt, 1959; Pauly, 1980). The fishing mortalities, however, showed marked differences amr ng systems and regimes. In consequence, a fishing regime should be regarded as a local property of a system, whereas the ~roductivities of each of the major fish components are probably more similar on the global scale. 2.5. interactions between ecosystem components 12ecmitment success is largely determined by the dynamics of primary production (Cushing, 1982; Parrish et al., 1983), and inoderate upweiiing conditions are most favorable for small pelagics in upwelling regions (Cury and Roy, 1989; Cui? et al., 1995). Cushing (1982) also linked recmitment success to cornpetition and stressed the need for information on how recniitment is affected by predation, although he considered the latter a minor process. A. JARRE-TEICHMANN AND V. CHRISTENSEN 433 Natural mortalily Fishing moitality Naturai mortallty a Fishing mortality Fig. 5: Breakdown of total mortality rates for mackerel (a); horse mackerel (b); and hake (cl. Note consistent scale of vertical axes. Systems are arranged after decreasing primary production. - Natural moriality Peru Peru Namibia Namibia NW Afr. California California 1964-71 1973-81 1971 -77 1978-83 1972-79 1965-72 1977-85 - The dominant food items in the diet of the five comrnercially rnost important fish are given in Table 4. All of them were essentially pianktivores. Anchovy fed largely on phytoplankton, except off California, where they ingested a larger fraction of zooplankton. Sardine were predominantly phytoplanktivores in the Atlantic, and zooplanktivores in the Pacific. However, due to the well-known ambiguities in assessing the diet composition (see, e.g., James, 1988 for a revie~v), this geographic division should be viewed with caution. Mackerel and hake fed predorninantly on zooplankton throughout, and horse mackerel fed mostly on zooplankton except off Peru, where they fed mainly on anchovy. Fishlng morialiiy Although the dominance of planktivores suggests a food web of rather simple structure, mixed trophic impact analysis suggested some indirect effects, as some of the groups most strongly favoring those five fish species differed from the rnost important food items. For anchovy, sardine md mackerel off California, primary production was the most enhancing factor. This also held true for the Peruvian systern after the collapse of anchovy, where phytoplankon became the dominant factor for al1 fish groups. Hake off Namibia were also favored most strongly by primary production. 434 Comparative Modelling of Trophic Flows System/Group Anclio 1 Horse mackerel Hake Pem 1'364-1971 Pliytoplankton Zooplankton Zooplankton Ancliovy Zooplarikton Pem 1973-1931 Phytoplankton Zooplankton Zooplankton Ancliovy Zooplnnkton Namibia 1971-1977 Phytoplankton Phytoplankton Zooplankton Zooplankton Zooplanktoii Namibia 19781983 Phytoplankton Pliytoplankton Zooplankton Zooplankton Zooplarikton hWAfÏica 1972-1979 Phytoplankton Phytoplankton Zooplankton Zooplankton Zooplanktoii Cal'imia 1965-1972 Zooplankton Zooplanktoii Zooplankton Zooplaiikton Zooplanktoii Calibmia 19781985 Zooplankton Zooplankton Zooplankton Zooplankton Zooplankton Table 4a: Dominant fo (by weight) in the diet of the five cornmercially most important fish species by systcm and regime. System /Group Anchovy Sardine Mac kerel Horsemackerel Hake 196-1-197 1 Phytoplaiikron Zooplankroii Zooplnnktoii Xiiclio\y Zoop.nnk ton I1eru 1973-1931 Phytoplankron Phytoplankton Phytoplankton Phytoplxnkton Phyophnkton Aiicho\v Namibia 1971-1977 Phytoplankton Phytoplankton Zooplankton Zooplankton Phytaplankton Uamibia 19781983 Phytoplankton Phytoplankton Zoopbiikton Zooplankton ~hy&~lmkron ?NliAfrica 1972-1979 Phvto~lankton Phvto~iankton Zoo~lankton Zoo~lankton Zooulanktoii ~p ~ ; 1 \:aliforiiia 19651972 P ton ~h~o~iankwn ~liy~o~lniiktoii ~oo;,lankton ~oo;>l3iikton (:alifurnia 1970:935 ~hytoptuikton Phytopiankton I)li!tophiiktoii Zoopl3nkioii %oopl:iiirms and regimes. Anchovy were heavily exploited by the fishery in the late 1960s off Peru, while horse mackerel caused most of its mortality during the 1970s. Hake was the dominant predator on anchovy off Namibia, other pelagics off Nonhwest Africa, and horse mackerel and mackerel off California. Sardine were preyed upon intensively by mammals ancl hea\ ily fished off Pem; they were preyed upon most strongly by hake off Namibia, and by other pelagics off Northwest Africa. Mackerel were more strongly exploited by the fisheries off Peru, Namibia, and (after its reopening) off California than subjected to predation by any single group. Horse mackerel were preyed upon by mammals and other pelagics, except off Namibia, where predation by hake and exploitation by the fishery were the most important causes of mortality in the first and second periods, respectively. Hake was either influenced by cannibalism or by the fishery, except off California, where cannibalism was less important because the hake population consisted mainly of juveniles, preyed upon by the abundant marine mammals. Whereas anchovy was inhibited rather directly, indirect effects of trophic intenctions were more pronounced for the other groups. It is worth noting, though, that intraspecific competition apparently had a larger effect on anchovy duiing the 1960s than food limitation. Competition with anchovy was more important for sardine off Peru than predation by mammals. Inhibition by mackerel, exclusively based on indirect effects, was more important for sardine off Namibia in [lie 1970s than predation by hake. Competition for food could also have been limiting for sardine off northwest Africa. Ail in all, our results are thus supported by the competition model of Silvert and Crawford (1988). Mackerel were subjected to more direct effects rather than indirect ones, except off California during the early 1980s) where intraspecific competition for food inhibited its population growth more than any other group. Horse mackerel competed with the Fishery during the period of high anchovy abundance off Peru, and with each other after the collapse of its major food resource. Off Namibia, horse mackerel were inhibited by hake to the same extent as by the fishery. Inhibition by other pelagics was more important to horse mackerel than direct predation by mammals off California during the late 1960s. Hake were generally predator-controlled, either by each other, by the fishery, or (off California) by marine mammals. Indirect interaction with anchovy turned out to be s-ongly inhibiting for hake off California during the early 1980s. These results supplement earlier work by Korrubel (1992), who suggested that fisheries may induce species dominance shifts while emphasizing the need for further assessrnent of the role of other ecosystem components, based on improved knowledge of their interactions. In general, the components the Peruvian system were indeed most strongly inhibited by the fishery, but those in the northern Benguela system most strongly by hake. The cross-impacts included more groups in the two systems with seasonal upwelling, off California and Northwest Africa. However, al1 in all, the inhibition of these five groups appeared to be a highly local property. 2.6. Transfer efficiency Restructuring of the fractional trophic levels (as used for the trophic flow diagrams) yielded six discrete trophic levels, i.e., producers, herbivores, and first- to fourth-order carnivores, for al1 of the models except Northwest Africa which had seven trophic levels. As the absolute flows on the topmost level were negligible, we computed the average transfer efficiency of the consumer levels II to V (Le., herbivores to third-order carnivores) (Fig. 6). The grand mean of al1 models yielded a transfer efficiency of slightly above 5% (range 3.6 - 7.4%), much lower than the general mean of about 10% computed for a cross-section of aquatic ecosystems (Christensen and Pauly, 1993; Pauly and Christensen, 1995). Hence, upwelling systems are al1 relatively inefficient systems regardless of the prevailing fishing regime. Moreover, despite relatively large changes of the transfer efficiencies between different regimes in a given system, there is some suggestion that the systems might be slightly different from each other, the California system being the least, and the two Atlantic systems the most efficient ones. 436 Comparative Modelling of Trophic Flows 6: Mean transfer efficiencies n discrete trophic levels II to V, i.e., herbivores to third-order carnivores, in seven models of the four major i~pwelling ecosysterns. Thin lines indicate ranges. Systems are arranged ,ifter decreasing prirnary production. 2.7. Primary production required to sustain the fishery Figure 7 gives the primary production required to sustain the fishery in the seven models constructed both in absolute tei.ms, and relative to the primary production available in the corresponding system. In absolute terms, the flows from tlie pr rnary producers required to sustain the fishery reflected the size of the catches, with high catches implying high primary pr1:iduction requirements. The systems could thus be ranked in the same way as after the 'size' parameters (Fig. 3). Hciwever, changes in the fishing regirne must be considered as well: although the magnitude of the catches off Peru was retluced by a factor of more than three between the two periods analysed, the primary production required to sustain tlie fisliery in the 1970s decreased only by about IO%, as sardine and hake, both situated on higher trophic levels, accounted for a considerable fraction of total catches. Th: fraction of the available primary production used to sustain the catches, however, showed a rather different picture. Alt îough the catches decreased significantly bemreen the mro periods analysed in the Peruvian system, a larger fraction of the a~ilable prirnary production was used in the later period. The fishery off Namibia during the early 1970s required the largest sh.re of the available pnmaq production, whereas the fishery off Califomia took only a minor share. Comparing the 1970s as a peiiod modeiied for al1 four upweiiing areas, we conclude that the primary production required to sustain the fishev is a local characteristic of the system, depending on the intensity of fishing and the trophic levels of its target species. It cannot str:iightforwardly be linked to species dominance, as the differences between sardine-dominated systems, such as the No-thwest African and the norrhem Benguela systems dunng the 1970s, were larger than the differences between an anchovy- doininated system (off Peni duriig the iate 1960s) and a horse-mackerel dominated one (off Namibia, dunng 19?8-83). Thc grand rnean of the primary production used to sustain the fishery amounted to 9.5% of the available primary prc duction in our models. This figure is higher than the global average of 8% estimated by Pauly and Christensen (1995), bu1 seems strikingly low if compared with their results for upwelling regions of 25%. This apparent discrepancv is A. IARRE-TEICHMANN AND . CHRISTENSEN 43 7 - - 3 14 X D % z .- 12 g . - z 2 1500 10 $ L P P .- E 8 .: a P 1000 - 2 6 c + Fig. 7: Flows from the producers's level to "7 5 500 4 U. the fishery, both in absolute (light shading) as well as in relative (dark shading) terms. 2 Systems are arranged after decreasing O O primary production. 1964-71 1973-81 1971-77 1978-83 1972-79 1965-72 1977-85 explained as follows: Pauly and Christensen (1995) used primary production estimates similar to the 1970s conditions in Our models, but (i) catches of small pelagics were considerably higher in 1990 than they were during the 1970s; (ii) we used, except for Peruvian anchovy, nominal catches in our models, whereas Pauly and Christensen (1995) accounted for additional 15% of discards for each group; and (iii) catches of horse mackerel were disproportionally large in the southern part of the Humboldt Current, not included in the present comparisons. We have, thus, been looking at very productive subsystems in the four upwelling regions, but not necessarily at those parts of the systems andlor periods subjected to the highest rate of exploitation by the fishery. Using the respective raising factors for the catches, we would, too, arrive at relative requirements of 64% to sustain the fishery of anchovy and sardine, 1517% for the fishery of horse mackerel and mackerel, and <1% for other pelagic groups, indicating the considerable degree of present exploitation of these systems. Additionally, it should be kept in mind that another 3-4% of the primary production are required to sustain the fisheiy of hake, not explicitly attributed to upwelling regions in the above study. 2.8. Primary production required to sustain the five major fish groups About 20% - 33% of the available flows from the producers' level were generally used to sustain the five dominant fish groups in the system (Fig. 8 and Table 5), with the exception of the Namibian system during the 1970s, where almost half of the available flow was required. The latter seemed particularly high at Brst glance, but went along with a slightly elevated mean trophic efficiency in the system, and was also consistent with the high fraction of primary production required to sustain the fishery. The shift in species dominance in the Peruvian system between the late 1960s and the 1970s was clearly reflected, and it is worth noting that -in spite of the considerably smaller total biomass in the system -a similar fraction of primary production was required to sustain the dominant fish species. It was also striking that the fraction of primary production used to sustain the fish in the Californian system wris not much lower than in the other systems, indicating that the structure of the Californian system was indeed very similar to the other 438 Comparative Modelling of Trophic Flows sy:tems, despite their large differences in size. We should thus moderate the statement of \Vare (1992) on the pa~~iculrir intfficiency of the Californian ecosystem with respect to fish production. Mithin the systems where anchow were the dominant fish by production (Le., off Peru and California, see Table 2), it \vas 31s:) anchow which required the largest single fraction of primanr production among the fish groups. The sarne heltl true for horse mackerel off Namibia in the early 1980s. Stnkingly, within the two systems dominated by sardine (the Namibian an(:\ northwest African systems during the 1970s), hake and horse mackerel required the single largest share of primai? production, respectively, but not sardine. It would be premature to draw any conclusion from this inconsistency ~vith re>pect to the persistence of species dominance or the diversity of flows in the systems (see also Shannon et al. (1988), anil LeClus (1991)). Sy; tem/Group - Anchow Sardine Mackerel Horse niackerel Hake Peru 1964-1971 Fis hery Mammals Fisliery Mammals Hake Peru 1973-1981 Horse mackerel Fishery Fishery Mammals Fis heql N!iiubia 1971-1977 Hake Hake Hake Hake Fishery Niinibia 1978-1983 Hake Hake Fishery Fishery Hake M';"Afnca 1972-1979 Other pelagics Other pelagics Lrg. scombrids Other pelagics Fishe~y Callfomia 1965-1972 Horse mackerel 2" Marine birds Mamnials Maiiiriials Ca!ifomia 19781985 Mackerel - Fishery O ther pelagics Mariunais 1) " '"he biomass ofsardine was so low that it could only be a marginal component in the diet ofits predators. It \vas hence im;~ossible to determine the main predator on sardine. Table 5a: Strongest predator group of the five cornmercially rnost important fish species by system and regirne. Nar ibia 1971-1977 Nai-ubia 19781983 NW.&ca 1972-1979 Other pelagics Zooplankton Calibmia 1965-1972 Horse iriackerel -" imp.)ssible to account for its proper role in the mixed trophic impacts routine. Tab e 5b: Strongest inhibition of the five commercially most important fish species by systern and reginie, bascd on inixed trophic impact analysis. Cornponents that differ from the strongest predator, and thus point at food limitation or cornpetitive inhibition of the respective fish group, are shaded. A. IARRE-TEICHMANN AND V. CHRISTENSEN 43 9 Anchovy Sardine Mackerel 40 Horçe Mackerel . Hake flows from the producers' level required to sustain the five dominant groups in the system. For the Northwest Africa system, the requirements of anchovy and sardinella have been combined, as sardinella occupies, in part, the ecological niche of anchovy. Systems are arranged after decreasing primary production. Peru Peru Namibia Namibia NW Afr. California California 196471 1973-81 1971-77 1978-83 1972-79 1965-72 1977-85 The different aspects of our analysis with respect to global versus local properties are summarized in T~ble 6. 'Ille consistent set of species dominating the flow on the higher trophic levels is one of the obvious global characteristics of upwelling ecosystems. Also, the distribution of the major flows in the models is also quite similar among systems. However, it is necessary to keep in mind that dl flows among the fish groups and toward the warm-blooded top predators are small in comparison to those in the plankton. Hence, for improved companson of the systems with respect to system- level properties (maturity, ascendency, etc.) under different environmental regimes, rime-series of flows in the plankton compartments, as well as the microbial food web, will need to be assembled and analped in more detail than it has been possible for the present contribution. The productivity of the small pelagics, as well as the natural mortality of al1 dominant fish stocks, was also similar between systems, and can thus be regarded global properties. This should not be regarded as a pure artifact of mode1 constiuction, as the models were balanced prior to any comparisons. It also emerged clearly that small pelagics increase with primary production regardless of sptem and regime. This does not appear, at first glance, to go along with the findings of Cury and Roy (1989) and Cury et al. (1995) that moderate conditions are most beneficial to successful recruitment, but it should be kept in mind that the present approach dealt only with trophic interactions, and not with the transport processes that determine the survival of eggs and larvae. The fraction of primary production used to sustain the five most important fish groups in the ecosystems was also quite similar between the systems (with exception of the Namibian ecosystem during the 1970s) and may thus be regardecl as a global property. Furthermore, our results suggest that flows from the primary producers' level required to sustain the dominant fish species (in terms of production) may be a regime-specific property, anchovy using the largest fraction in anchovy-dominated systems, but hake or horse-mackerel using the largest fraction in sardine-dominated systems. 440 Comparative Modelling of Trophic Flows C haractenstic /Property i~10bal'3' cal'^) Genenl sp tem stmcturc Species compositioii System size -. Major flow patterns Productivity of major Natural mortality ofall dominant 6sli Fisliingmortality 6sh species stocks, -. total mortality of small pelagics Species interactions Favonng of sinall pelagics Favoringof medium-sized 6sli; Inhibition of al1 major 6s1i groiips; -. Strongest predators ofall niajor 6sli groups Siistenance of tlie major 6sh Total baction of primary production Fraction of primary production required by the grou ps required (in general) dominant species Sustenance of the 6slie1-j~ Fisliing regime. - .. primary production required to sustain the 6sliery Relation between production Total catch vs. trophic level ofbliery - ~.nd fis lie^ - - System transfer eficiency Inw mean efficiency ofenergy transfer Eficiency on 'medium scale' 0) -. up the food web :bal functions Overall low maturitv; - similar iiiforination content offlow~~ low - relativeascendency Similar between sys rems tlirougii time, i.e., (i) regimedependent, or (ii) independent ofs ystem and regiiiie ' Moresimilar witliin systems throiigli time (Le., regime-independent), or systcm- and regime-specific Table 6: ersus local properties of the four upwelling ecosystcrns analysed. Most system-level properties of these ecosystems, such as their generally low transfer efficiency, were also of global nature. Furtlier attributes are discussed in detail in Jarre-Teichmann and Christensen (in press), notably those pertaining to the thecries of Odum (1969) and Ulanowicz (1986). The generally low maturity, and low relative ascendency of these systems are !:lobal properties as well, and corroborate the grouping of upwelling systems in Christensen's (1992) maturitv ranking of al luatic ecosystems. It should further be pointed out that the total catch was correlated with the trophic level of the fishtsry, and also with the pcimary production of the systems. The most obvious local properties of the systems were related to system size, as primary production, total system biomass, or total catches. Furthermore, the factors most strongly favoring the medium-sized fish were more similar within systems thro.igh time. The strongest predators of the five most important fish species were rather variable, but their inhibition apptbared to be a system-specific property, independent of the prevailing regime. Iastly, al1 properties relateci to the fishery, as fishing mortality of the groups, or the primary production required to sustain the fishery, were Iiighlv variable, not only among systems, but also among different regimes. There are a number of inconsistencies in the properties that we have categorized as 'global'. In terms of multispecies man:igement models, care should be taken before these estimates are transferred from one system to anotlier. Additionally, al1 aspects related to the fishing regime wiii need to be modeUed specifically for the system and period in question. The complexity and interaction of the various factors influencing fish populations in upwelling ecosystems have been highlighted earlier (Crawford, 1991), implying blurred borderlines that have also made it difficult for us to categorize a given aspect as global or local. Nevertheless, time series, long required for improved understanding of any kind of variability in upwelling systems (see, cg., Bakun and Parrish, 1980; Pauly, 1987; Sharp, 1991) are becoming increasinglv suited for ecosystem approaches such as the one presented here, and may justify optimism that they will subscquentl!r allow for refined future assessment of climatic effects on these four eastern boundary currents. Bakun A. 1990. Global climate change and the intensification of coastal ocean upwelling. Science, 247: 198-201. Bakun A. 1993. The California Current, Benguela Current and soutliwestern Atlantic shelfecosystems: a comparative approa- ch to identibiiig factors regulating biomass yields. In: K. Sl-ier- mari, L.M. Alexander and B. Gold (eds.). Stress, nzitigation and preseruation oflarge nzarine ecoqstenzs. American Association for the Advancenient of Science, Washington D.C., USA. Bakun A. and R.H. Parrish. 1980. Enviroriniental inputs to fishe- ry population models for eastern boundary current regions. Report on the workshop on the effects of environmental varia- tion on the survival of lanialpelagic fishes, Lima, 20April- j May 1980. Organized by FA0 as a contribution to the IOC Program- me of Ocean Sciences and Living Resources. UNESCO, Paris, IOC W'orkshop Rep, ,28: 67-104. Bakun A., V. Christerisen, C. Curtis, P. Cury, M.H. Durand, D. Husby, R. Mendelsson, J. Mendo, R. Parrish, D. Pauly and C. Roy. 1993. The Climateand Eastern Oceans Systems Project. ICLARM, Naga, 15(4): 26-30. Beverton RJ.H and S.J. Holt. 1959. A review of the lifespans and mortality rates offish in nature, and their relation to growth and other physiological characteristics. In: G.E.W. Wostenholnien and M. O'Connor (eds.). The lifepan of anit~ials. Ciba Found. Colloq. Ageing. 5, Churchill, London: 142-180. Christensen V. 1992. Netzuork analysis of trophic interactions in aquatic ecog!stems. PhD Thesis, Royal Danish School of Phar- macy, Copenhagen. j5 p + 13 Appendices. Christensen V. and D. Pauly. 1992. Aguide to the ECOPATH II soffiuares~sten~ (Ver. 2.1). ICLARM Software 6, Manila, Interna- tional Center for Living Aquatic Resources Management, 72 p. Christensen V. and D. Pauly. 1993. Flon7characteristics of aquatic ecosystems. In: V. Chnstensen and D. Pauly (eds.). T~ophic iilo~lels of aquatic ecoqsteius. ICLAR21 Conf. Proc., 26: 338-352. Crawford R.J.M. 1991. Factors influencirig population trends of some abundant vertebrates in sardine-rich coastal ecosysteiiis. S. Afi: J. mar. Sci, 10: 36 5-381. Cury P. and C. Roy. 1989. Optiiiial enviroriinental niiidow and peiagic fish recruitinent success in upwelling areas. Can.1. Fish. Aquat. Sci., 46: 670-680. Cury P., C. Roy, R. Mendelssohn,A. Bakun, D.M. Husbyand R.H. Parrish, 1995. Moderate is better: exploring non1ine;ir cliriiatic effect or1 Californian anchoy (E~2graulis1110da~). In: R.J. Bea- niish (ed.). Climate change and fish population. Carr.Spec Pub/, Fish. Aquat. Sci., 121: 417424. Cushing D.H. 1982. Clinzateandfsheries. Acadeniic Press. Lon- don, 373 p. Daan N. and M. Sissenwine (eds.). 1991. Multispecies iiiodels relevant to management of living resources. ICESIPI~~: Sci .$!n~p., 193,358 p. FAO. 1990. Yearbook offihery statistics: catches and l~indings 70,647 p. James A. 1988. Are clupeid microphagists herbivorous or oiniii- vorous? A review of the diets of some conimercially iniportaiit clupeids. S. AJi. J. Mar. Sci, 7: 161-177. Jarre-Teichmann A 1992. Steabstate modelling olthe Peiul'inn upwellingecosvsteiii. Doctonl thesis, University of Breiiien, F.K. Germaiiy, 153 p. 442 Comparative Modelling of Trophic Flows ],irre-Teichmann A, D. Pauly and P. Muck. 1791. Two approaches Ior modelling fish stock interactions in the Peruvian upwelling (,cosystem. ICEShfar. Sci. Svnlp., 193: 171-184. Jiirre-Teichmann A and V. Christensen. In press. Modelling tro- phic flows in large upwelling ecosystems: temporal and spatial c.omparisons. ICLARMStzidies and Reuiezus, 24. 1;ormbel J.L. 1992. Anage-stmctured simulation model to inves- iigate species replacement between pilchard and anchovy popu- Iitions in the southern Benguela. S. Afr. J. Mar. Sci, 12: 375-391. 1.1-Clus F. 1991. Hydrographic features related to pilchard and ; iichovy spawning in the northern Benguela system, comparing iliree environmental regimes. S. Afr. J. Mar. Sci., 10: 103-124. I indemann R. 1942. The tropho-dynamic aspect of ecology. Eco- l,)gv, 23: 399418. 1 luchBelda D., R.J.M Craford, T. Kawasaki, AD. MacCall, R.B. 1 '.irrish, R.A. Schwartzlose and P.E. Smith. 1989. World-wide flue- t .iations of sardine and anchovy stocks: the regime problem. S. P fr. J. Mar. Sci., 8: 19 5-20 5. Oduni E.P. 1969. The strategy of ecosystem development. Scien- L tJ. 164: 262-270. Odum E.P. 1971. Fundanlentals of ecology. Saunders, Phila- celphia, USA j74 p. E;irrish R.H., A. Bakun, D.M. Husby and C.S. Nelson. 1983. Coni- 1. ~irative climatology of selected environmental processes in cela- iion to eastern boundary current pelagic fish reproduction. In: C-.D. Sharp andJ. Csirke (eds.). Proceedings of the Fxpen Consul- Lition to Examine Changes in Abundance and Species Campo- s lion of Neritic Fish Resources, SanJosé, Costa Rica, 18-29April 11:183. FA0 Fisl~. Rep., 291 (3): 731-777. Piuly D. 1980. On the interrelationships between natural morta- li:y, growth parameters, and mean environmental temperature ii i 175 fish stocks. J. Cons., Cons. int. Fxplor. Mer, 39(2): 175-192. Piuly D. 1987. Managing the Peruvian upwelling ecosystem: a s!nthesis. In: D. Pauly and 1. Tsukayama (eds.). The Pemvian aiichoveta and its upwelling ecosystem: three decades of chan- gtr. ICWLZIStudiesand Reuiezus, 15: 32 5342. P:iuly D. and V. Christensen. 1995. Primary production required tc sustain global fisheries. Nature, 374: 2 j j-2 j7. Polovina J.J. 1984. Model of a coral reef ecosystern. 1. The ECO- PATH rnodel and its application to French Frigate Shoals. Colal Re& 3: 1-11. Po1ovinaJ.J. 1985. An approach to estimating an ecosystem box model. Fish. Bull., U.S., 83(3): 4 57-460. Polovina J.J. and M.D. Ow. 1983. ECOPATH: A user's lllat~ual andprogmni listings. Nat. Mar. Fish. Sem., NOAA, Horiol~~lu, Admin. Rep., H83-26,46 p. Shannon L.V., R.J.M. Crawford, G.B. Bundrit and L.G. Underhill. 1988. Responses of fish populations in the Benguela system to envi- ronmental change.J. Cons., J. Cons. int. E~ploi: Merer., 4j(1): 5-12, Sharp G. 1991. Climate and fisheries: cause and effect. In: T. Kawasaki, S. Tanaka, Y. Toba and A. Taniguchi (eds.). Long-tet-111 uariability ofpelagicfish populations and theirenuirou111 ent. Pergariion Press, Oxford, UK. 239-258. Siegenthaler U. and L. Sarmiento. 1993. Atniospheric carbon dioxide and the ocean. Nature, 365: 119-12 5. Silvert W. and R.J.M. Crawford. 1988. The periodic replaceillent of one fish stock by another. In: T. Wyatt and M.G. Larraneta (eds.). Long-temz changesin 17zatYnefishpopulations. Procee- dings of a symposium held in Vigo, Noveiiiber 1986: 161-180. Stee1eJ.H. and E.W. Henderson. 1984. Modeling long-term fluc- tuations in fish stocks. Science, 224: 985-987. Ulanowicz R.E. 1986. Givzutl~ anddevelopnzent: ecos~ateinsphe- nonzenologv. Springer, New York. 203 p. Ulanowicz R.E. 1995 Ecosystem trophic foundations: Lindeiiian exonerata. In: B.C. Patten and S.E. Jorgensen (eds.) Colilplex ecologv: The Part-Whole Relation in Ecosysteiils. Prentice Hall, Eaglewood Cliffs, Newlersey, USA. Ulanowicz R.E. and C.J. Puccia. 1990. Mixed trophic iiiipacts iii ecosystems. Coenoses, j(1): 7-16. Ware D.M. 1992. Production characteristics of iipwelliilg systeiiis and the trophodynamic role of hake. S. gfi J. ~i~ai: Sci : 12: 501- j13. Walsh J.J. 1989. How much shelf production reaclies the deep sea? In: W.H. Berger, V.S. Sinetacek and G. Wefer (eds.). Pro- ductioip of the ocean:present andpast. Dahlem Workshop Reports, Life Sciences Research Report. John Wiley & Sons. CIii- chester, UK. 44: 175-191. PART 3 Human facing Short and Long Term Changes Comparative Study of the Dynamics of Srnail-Scaie Marine Fisheries in Senegal and Ghana JOCELYNE FERRARIS* KWAME A. KORANTENG** AIASSANE SAMBA* * * * .irntenne ORSTOM au Centre IFREMER de Nantes BF 1135 4431 1 Nantes Cedex 03 FRINCE ** Fisheries Research & Utilization Branch (FFUB) P.(:). BOX B-62 Te ,na Gt ANA **+ Centre de Recherches O( Panographiques de Dakar-Thiaroye (CKODT) BP 2241 Dakar SEPIECAL The small-scale fisheries in Senegal and Ghana are described and compared. The structure of the canoe fleets, the trends in landing and the composition of catches are anal~lzed. Located in upwelling areas of the eastern Atlantic, these tropical multispecies fisheries are influenced by a strong environmental signal. Variability and instability of marine resources raise the problem of interactions arnong the different dvnamics of the environment, of fish stocks and of fisliing communities. The changes observed during a decade are interpreted from the perspective of the dynamics of the exploitation. Local and global changes were assessed in terms of fishing tactics and strategies that facilitate description of the adaptability of small- scale fishers to the variability of their environment. Les pêcheries artisanales du Sénégal et du Ghana sont décrites et comparées. La structure des flottilles, les tendances des débarquements et la composition spécifique des captures sont analysées. Situées dans les zones d'upwelling d'Atlantique est, ces pêcheries tropicales multispécifiques, multiengins sont influencées par une forte composante environnementale. La variabilité et l'instabilité des ressources marines soulèvent le problème de l'interaction entre les dynamiques de l'environnement, des ressources renouvelables et des sociétés de pêcheurs. Les changements observés sur une décade sont interprétés sous l'angle de la dynamique du système d'exploitation. Les changements locaux et globaux sont abordés en termes de tactique et de stratégie de pêche qui permettent de décrire l'adaptabilité et la flexibilité des pêcheurs artisans aux fluctuations de leur environnement. This paper presents a comparative studv of the small-scale fisheries of Senegal and Ghana, from the point of view of the dynamics of exploitation. With respect to fisheries, Senegal and Ghana have at least two things in common: a coastal upwelling system and an important small-scale fishery. The small-scale fisheries of Senegal and Ghana are the largest in West Africa and are among the most important economic activities in these countries. In terms of landings, these small- scale fisheries are the most important in the West African region (Everett and Sheves, 1991) and contribute over 70% of the total catch of fish in each of the two countries. The per-capita consumption of seafood per year is similar in the hvo countries, being about 2 5 kg per year (Horemans, 1993). The small-scale fisheries of Senegal antl Ghana bear n number of other resemblance, notably in the way thev changed over the years. The main objective of this study is to establish a synoptic description of the fisheries with regard to the structure of their fleet and catch compositions. The evolution of the two fisheries in the last decade is compared, to analyze the responses to changes in their environment and to distinguish global and local changes. Data on marine small-scale fisheries have been collected through sampling for over twenty years in Senegal and Ghana. The statistics used in this study were obtained from catch assessment surveys and canoe censuses that have been undertaken simultaneously since the beginning of 1980s. The statistics were gatheretl and preprocessed bv the Researcl~ & Utilization Branch of the Fisheries Department (FRUB) at Tema in Ghana, and by the Centre de Recherches Océanographiques de Dakar-Thiaroye (CRODT) in Senegal. I . CHARAC~~ERISTICS OF SMALL-SCALE FlSHERlES IN SENEGAL AND GHANA Situated respectively in the northern and central part of the eastern central Atlantic, Senegal antl Ghana are two West Afi-ican countries separated by thousands of kilometers (Fig. 1); yet they have similar important small-scale fisheries. These fisheries have a long history and have undergone great changes during the last thiity years, especially the introduction of outboard motors and the development of the purse seine nets (Koranteng, 1992; Kébé, 1995). 448 Srnall-scale Fisheries in Senegal and Ghana Many types of fishing gears are used by small scale fishers. In both countries, small-scale gears are classified into five groups: purse nets, dr'iting gill nets, set nets, hooks-and-lines and beach seines. The group named 'others' includes diverse gears, suc11 as traps or jigs for cuttlefish in Senegal, or no gear (with canoes used only for trans-shipment of catches). F g. 1 : Location of Senegal and Ghana ir Africa. TaMe 1 gives the percentage of canoes in each type of fishing gear as enumerated in the most recent census in each coiintty. The following features may be noticed: i.) a large proportion of nets for small pelagic fishes in Ghana; ii.) a large proportion of canoes with line gear in Senegal; iii.) an almost equal proportion of canoes with set nets in Ghana and Senegal; iv.) a much smaller proportion of canoes with drifting gill net and beach seines in Senegal compared to Ghana. In Ghana, the common fishing craft is a dugout canoe carved out of a single tmnk of wood. These canoes are symmetrical in :hape, double-ended and range in sizes benveen 3 and 18 m long and 0.5 to 1.8 m wide. About 49% of the over 8 600 Gh,inaian canoes are motorized, using outboard engines of between 25 and 40 hp (Koranteng et al., 1993). The level of mororization depends on the size of the canoe and on the gex operated. For example the canoes operating purse nets are uscally large and almost al1 of them are motorized. The engine is put on the side (normally on the right), at the rear of the canoe. Senegalese canoes are almost completely planked except for the area close to the 'keel' which is carved out of a small log. Thtt evolution of the canoe was dictated by geographic constraints, especially by the presence of a sti-ong surf along the northern Coast (Chauveau, 1984). The outboard engine is mounted at the rear or at the Stern of the çanoe through an inc:Yon made to the hull. About 90% of the 5 600 canoes used for marine fishing in Senegal are motorized (Kébé, 1995; Saniba, 1995). The canoes operating pursing gears and long-range line with insulated ice boxes have outboard engines frorn 25 to 40 hp; canoes with line, set net or beach seine generally use outboard motors of 8 up to 15 hp (Ferraris, 1993). In Cenegal, the total number of canoes recordeci for each gear is usually larger than the total nurnber of opentional canoes. Thi.; is because one fishing unit may have severai gears and undertake different fishing activities depending upon the target ~ - Feature Ghana Senegal Lengtli of the coastline j jCi km 700 kin Niimber of fishing villages (N) 187 incliiding 306 landing beaches 1 j j iricliiding 96 villages with iiiariiie canoes Niimber of fishers 96,400 = 35,000 Number of marine canoes 8,688 5,661 (Q = C qi, i=l to N) (north of Sine Salouni in warni seasoii) Index of canoes distribution 0.902 0.802 Shannon index (1987) (maximum value in wariii season 1788) (- 2 ((qi/Q) log2(qi/Q))hog2N Gears (as % of canoes) Purse Seine (PS) 37.8 8.6 Drift Gill Net (DGN) 2.7 O. 1 Set Net (SN) 27.7 25.7 Hand-Line (LN) 11.7 57.5 Beach Seine (BS) 8.7 1.3 Miscellaneous (DiV) 6.7 10.0 Catches (t.103) Pelagic 261 27 1 Demersal 46 46 Table 1 : Main characteristics of the small-scale fishery sectors in Ghana and Senegal (1 992). species. This phenomenon of rapid gear switching is one characteristic of smal-scale fisheries. In Ghana, only the main gear is taken into account during the census, but several cases of gear switching between set net and line or benveen the small-pelagic nets ('ail, 'poli' and 'watsa' nets) have been observed (Koranteng, 1990). Manv fishing units may also have a set net. Tlie joint utilization of different gears during the same fishing trip is best observed in the strategy of Senegalese fishers. Migration is an important characteristic feature of small-scale fisheries in West Africa. According to Chauveau (1990), migration of fishers from the Gold Coast (now Ghana), for example, had been recorded by the beginning of tlie 20th century. Todav, Ghanaian fishers mai be found in manv countries in West and central Africa, especially in Togo, Benin, Cameroon, Côte-d'Ivoire and Guinea. Senegal is also a net 'exporter' of migrant fishers (Diaw, 1991). Senegalese fisliers mav be found in Mauritania, Guinea, the Gambia, Guinea-Bissau and in other countries in West and central Africa. Distribution, abundance and movement of fish and fish schools are some of the factors that induce migration of fishers. Others are social and economic factors (Odotei, 1791; Haakonsen and Diaw, 1991; Koranteng et al., 1993). These migrations have, in part, contributed to the diffusion of small-scale fishing technology, ski11 ancl expertise. For example, it appears that the storage of fish in insulated, removable fish-holds on canoes was spread in the region bv migrant fishers. Ghanaian fishers mav have learnt this technique from their Senegalese counterparts in conjunction witli long-range and long-duration line fishing. Another example of technologv transfer through migration of fishers is the 'chorkor' smoking oven, which originated from a fishing vikiige in Ghana but is now known throughout tlie West African region (Odotei, 1791; Koranteng, 1995). The major technical innovations have an impact on tlie relations between producers and tradesmen and consequently on the small-scale fisheries development (hwson and Klvei, 1974). 450 Small-scale Fisheries in Senegal and Ghana Moj~ement of fishers with their canoes also has serious effect on the assessment of the quantity of fish caught by the small- scal:: fishing fleets (Koranteng, 1992). In Ghana, small-scale fishers land their catches at a certain spot at the end of a daily trip, then return home with emptv canoes. Assessment of catches is made by sampling and such an act introduces sampling and coverage errors in the estimations. Therefore, it is important to classify fishing villages between landing and refuge sites and to know and understand the temporal migratory flows between regions. As migration is influenced by vari:ition of fish availability, considering a stable fleet during a long time might underestimate total catches because the migrant canoes are then associated with less productive regions and the most productive regions with an underestimated flee,: (Ferraris, 1994). The oceanographic regime of the Ghanaian and Senegalese coastal waters is characterized by seasonal upwellings. In Ghiiria, a major upwelling occurs for approximately three months each year, beginning late June or early July and ending in late September or early October, and a minor upwelling occurs three weeks either in January or Febniary (Mensah and Koriinteng, 1988). The Senegalese Coast is under the influence of the trade winds which create local upwelling conditions and give rise to two main seasons: a cold season from December to May and a warm season from Julv to October (Roy, 199.!). During the upwelling season, biological activity is high and most fislies, both demersal and pelagic, spawn and migrate. Stock availability and catchability are thus variable in the course of the year. These events result in variability in lant ings. 2. FLEET STRUCTURE AND DYNAMICS 'I'he number of canoes at each landing area or village is obtained from the census of canoes which are conducted once in 2 4 years in Ghana and twice a year in Senegal. The census, also referred as 'frame surveys', form the basis of catch assesment surveys involving the small-scale fishing crafts. The stmcture of the Ghanaian fleet (Fig. 2) shows marked differences in gear composition on regional basis. This is due to diffeyences in specialization of the dierent ethnic groups. There are four regions for smail-scale marine fisheries in Ghana (Voli;i, Greater Accra, Central, and Western; Fig. 4) and each region presents the characteristic gears of the dominant etl-inic grouos (Ewe, Ga, Fanti). The occurrence of a paaicular type of gear in a particular location is determined by the txget species (Beciaseck, 1986). The same phenomenon is observed in the Senegalese small-scale fishery, pursued by three principal etlinic com nunities with long fishing tradition. These are the Wolof, Lebou, and Sérère-Nyominka, originating from Saint-Louis (Flei ve), Cap-Vert or Thies-Sud region and the Sine-Saloum Islands respectively (Fig. j). The riame surveys undertriken in Ghana show a relative stability in the stmcture of each region's canoe fleet (Fig. 2). For the cour try as a whole, the number of canoes increased by 2 j% between 1981 and 1992. The number of canoes di-opped signiïcantly in 1981, due to migration of fishers to neighbouring countries, possibly as a result of unfavorable economic conditions in Ghana at the time (Odoi-Akersie, 1982). Since 1986, the number of canoes operating drift gill nets has geneixlly decrc:ased. This decline has been attributed to the excessive and rising cost of operating this gear (Korinteng et al., 1993). Anot~ier possible reason is that with the gl-ddual depletion of the 'wawa' tree from which the dug-out canoes are produced, worn out large canoes are not being fuUy replaced (Wayo Seini, l9j). On the other hand, set net canoes have increased in numher by about 50%. It is possible that the canoes used for techniques showing a deciine mav now be used for set net fishing. /. FERRARIS, K. KORANTENC AND A. SAMBA 45 1 -m O PS BS LlNE DGN bN UIV lm b cc 8 1- c 2 5n .- O Il Li PS DS LINI: UUN SN DIV 0 PS BS I.INE UUN SN »IV Fig. 2: Structure of the small scale fleet in Ghana: number of canoes by gear and by region from 1981 to 1992 (PS: Purse Seine, BS: Beach Seine, LINE: Hand Line, DGN: Drift Gill Net, SN: Set Net, DIV: Miscellaneous). a) Volta; b) Greater Accra; c) Central; d) Western. , , , , , , PS BS LlNE DGN SN DIV In Senegal (Fig. 3), there nras an increase of 40% in the total number of canoes from 1982 to 1992. The increase was particularly important between 1989 and 1992 with the return of migrant fishers as a result of political conflict between Senegal and neighbouring Mauritania. The increase in line fishing in Senegal is due to the development of the activity of large canoes equipped with insulated ice boxes. Fishing is done far away from home bnses. The increase of these canoes may be the sign that fishers prospect for new fishing places. There is the same phenomenon in Ghana, where 157 canoes with insulated removable fish-holds were counted during the frame survey in 1992 (Koranteng et al., 1993). The fleet structure established through the frame surveys exhibits seasonal variation. For example, in Ghana, the number of canoes with net for catching small pelagic species and located at specific landing beaches, reflects the seasonal pattern of the upwelling (Fig. ha). This temporal variability in canoe numbers is well marked in the regions on the western part of Ghana where the sardinella season usuallp starts from (Ferraris and Koranteng, 1995). Fishers from other parts of the countly migrate to the west at the onset of the sardinella season. The large fluctuations that occur annually correspond to migration of canoes within and between regions. This movement used to be very important during the major upwelling period ('JulySeptember) but in the last decade migration during the minor upwelling season (January-February) has also been remarkable. This could be linked to the increasing importance of the minor upwelling and the resulting increase in the production of sardinellas in the western Gulf of Guinea (Pezennec, 1995). -- 452 Srnall-scale Fisheries in Senegal and Ghana cture of the small scale fleet in S~megal: number of canoes by gear and by rt%gion from 1982 to 1992 - in warm season (F'S: Purse Seine, BS: Beach Seine, LINE: Hand Line, DGN: Drift Cill Net, SN: Set Net, DIV: Miscellaneous. a) Fleuve; b) Thies Nord; c) Cap Vert; d) Thies Siid. un - 2m - 1, I m I PS BE LlNE DGN SN DIV a BS LlNE LZN SN L)Ik 800 - hl* - im - 21x1 - O 1 PS BF LlNE DGN SN !>IV Fip. 4: Map of the Chanaian Coast, showing the four 4 coastal regions considered he-e. 3 2 1 O 1 Longitude ("W or E) Fleuve / Thies Nord Layar - - - - - - - - - - - - - - Cap vert Senegal Petite Cdte y' Casamance Ziguinchor w Fig. 5: Map of the Senegalese coastline showing the coastal regions considered here. The same phenomenon of seasonal variation in the fleet is observable in Senegal (Fig. 6b). Here the profile of number of canoes in the Thies Nord region is complementary to that of the Fleuve region: during each upwelling season, some fishers migrate from the Fleuve region to the Thies Nord region and they return home after the upwelling period. This is related to seasonal migration of Saharian fish species such as grouper (Epinephelus aeneus), which migrate south and are then distributed al along the Senegalese shelf until the cold upwelled waters are replaced by warm tropical waters (Cuq ancl Roy, 1988). Since 191, a movement of fishers from the Thies Sud region to the Fleuve region is noticeable. These fishers, who have been in Senegal since the warm season 1989 (after the conflict with Mauntania), went back to Saint-Louis because of the reopening of the border, and also because of good fishing conditions in the north of Senegal during that time. with a smaller coastline Ghana has two times the number of fishing villages as in Senegal and a larger number of fishers and canoes (Table 1). The distribution of canoes along the coast was measured by the Shannon index wliich depends on the relative number of canoes per landing beach (Table 1). This index is usually used in ecology to measure the diversitv of species (Pielou, 197j). In this case, it was calculated during the period of low migration. The index was found to bc higher in Ghana than in Senegal indicating that canoes are distributed more evenly along the Ghanaian coast than along the Senegalese coast. In Senegal, the canoes are concentrated in about ten main landing centers which are close to consumption centers with road infrastructure. As the distribution of canoes varies between the seasons, the index is low during the cold season when the fishers are concentrated in a few strategic centers. For example, in Kayac located in the Thies Nord region of Senegal, the number of canoes is two times as high during the upwelling period than during the warm season. In Ghana, large fluctuations of canoe numbers are observed at important centers like Elmina (Central 454 Small-scale Fisheries in Senegal and Chana - Central a - - - Greater Accra - - - Volta -Western ---- - - - - Cap Vert 2000 - - - Thies nord - Thies sud ...------___ 2 1200 O D z 400 Fig. 6: Seasonal variation in the structure of the artisanal fleet by region in Ghana and Senegal. a Average number of canoes using pursing nets in Ghana; b Total number of operational canoes in Senegal. Reg ion) and Sekondi (Western Region). Large canoe populations at these centers correspond to the sardinella season whm many canoes migrate towards centers that are close to the bulk of the migrating fisli and where they are Iikely to fincl market and better price for their catch (Ferraris and Koranteng, 1995). In t lie last decade or so, the small-scale fisheries of the two countries have also undergone changes in nature as well as in the population of fishing craft. After the decline of the sardinella fishery in Ghana in the early 1970s, a classical case of tecl!nological innovation took place: the purse seine-type 'poli' net was developed from the 'watsa' ring net to catch anciiovies (Koranteng, 1992). Today, poli net is the major fishing gear used in the sardinella fishery in Ghana. Long-range line fishing has also increased considenbly. Some large-sized canoes that were used for purse seining mriy have been converted into long-range line canoes. These canoes carry ice and stay at sea for about three days (Koranteng, 1990). Set net fishers have also adopted the habit of attaching hooks on lines to the leadline of their nets (Koranteng, 199j). In !#enegal, the number of gears in al1 categories but 'line' and 'diverse' decreased between 1985 and 1989, due to a trarsfer of fishing effort to the exploitation of cephalopods as the biomass of octopus increased (Samba, 199j). New fish ng methods for cuttlefish, traps and jigs were introduced in Senegal from Japan in 1975. With the increase of octopus stoik size,the fishers adapted cuttlefish jigging for this species. Theie changes reflect the nature of small-scale fisheries, in which there is the propensity for innovation by the fisliers theiiiselves, in response to changing circumstances. Figure 7 shows the trend of fish landings by small-scale fishers in Senegal and Ghana from 1981 and 1992. The catch of small-scale fishers are composed mainly of pelagic fish species, although some demersal species are also caught. For example in 1992, pelagic species accounted for over 85 percent of the landing in each country (over 260 000 t). About 46 000 t of demersal fishes were landed by the small-scale fishers in each OF the two countries in the same year (Table 1). 351 Senegal Fig. 7: Total catches of the srnall-scale fisheries in Senegal and Ghana frorn 1981 to 1992. Year The changes in the composition of catches may be analyzed to detect changes in target species (Fig. 8, 9). In Ghana, the catches of pelagics are dominated by Engraulidae and Clupeidae (Fig. 8). The quantity of sardinellas and anchovies, therefore, greatly influence the percentage composition of the year's landings (Fig. 10a). Sardinella aurita is more abundant than Sardinella maderensis. Other important pelagic groups in the Ghanaian small-scale Fishery are the Scombridae, Carangidae and Istiophoridae. The pelagic catches in Senegal are dominated by Clupeidae (Fig. 8). The increase in landings since 1985 is due to Sardinella aurita (Fig. 11). At the beginning of the decade, catches of purse seine nets were relatively more diversified but less important than presently. Quite noticeable are the catches of bluefish (Pomatomus saltator) by purse seine and Iine gears, a species which disappeared after 1983. Samba and Laloë (1991) found a relation between bluefish catches, upwelling intensity and sea surface temperatures, supporting the idea that the upwelling is a favourable Factor determining the migration pattern of migrant Saharian species in Senegal. In the demersal sector, two families dominate the landings in Ghana (Fig. 9), the Pomadasyidae and the Sparidae. From 1973, the triggerfish (Balistes capriscus) became important in the landings of trawlers in Ghana (Mensah and Koranteng, 456 Small-scale Fisheries in Senegal and Ghana Senegal Carangidae igc 1- Ephippidae Clupeidae F g. 8: Pelagic catches, by farnily a~id country frorn 1981 to 1992. 18 1981 1983 1985 1987 1989 1991 Year Ghana 300 T Undetermined pelagic Scombridae Tetradontidae Year 198;;). However, the proliferation of triggerfish in the Gulf of Guinea did not show up early in the landings of small-scale fishm Later increases of triggerfish in the landings of the canoes followed an increase in demand of the fisli as a result of imp:ovement in its utilization. The increase of triggerfish in the landings of smdl-scale fishers resulted essentially from trade betvieen smali-scale and industnal fishers on the high seas, with the former buying the catch from the latter. Triggerfisli resources in the whole of the Gulf of Guinea have declined hence the reduction in landings of the species by the Ghanaian sma 1-scale fleet since 1988 (Fig. 9). The increase in the landings unidentified demersal species by canoes in Ghana is also noticeable since 1987 (Fig. 9, 10). Brachydeuterus auritus (Pomadasyidae) and Pagellus bellottii (Spar'idae) dominate demersal fish landings by small-scale crafts. In ttie last decade, the demersal catches of Senegalese smaii-scale fishers were dominated by Serranidae, Sparidae and molliiscs (Fig. 9). The increase in octopus catches since 1989 reflects the interest of fishers For a new target species ). FERDIRIS, K. KOR~N~~NG AND A. SAMBA 45 7 Senegal Rays Undeterrnined Year Fig. 9: Dernersai catches, by farnily and country from 1 981 to 1992 Ghana Rays Cephalopodal Mollusc Cynoglossidae Scianidae Lutjanidae Polynemidae \ Muranidae 1981 1983 1985 1987 1989 1991 \Ariidae Year resulting from the increase in biomass of this species since 1986 (Caverivière, 199ja) and from the development of a market for octopus. There was a simultaneous decrease in the landings of groupers (Serranidae). At the end of the tlecatle, catches of Sparidae (mainly of the genus Pagellus and Spnrus) have increased with expansion of the export market and improvement in the organization of trading (Kébé, 1995). The increase in the landings of says was a result of the introduction of long line fishing with multiple hooks and the development of the export of processed small-scale catches (Samba, 1995). The changes in catch composition of tlie line fishery shows a dccrease in pelagic species following the disappearance of bluefish (Fig. 11). There has never been a lucrative market for triggerfish in Senegal and the landings by the small-scale fleet were not important even though this species was abundant in the south of Senegal at the beginning of the 1980s (Caverivière, 1995b), tlien in the north. However, a trade between small-scale and industrial fishers at sea, similas to what has been observed in Ghana for the triggerfish also occurs in Senegal for cephalopods. 458 Srnall-scale Fisheries in Senegal and Ghana (a) PELAGIC SPECIES E. Iimbriata Other , Ethmalosa fimbriata l S :omber japonicus Harengula roux; Caranx rhonchus Sepia sppc Rays Sh rks- ~ynog~ossiaae- Lpeneus prayensis - Balistes capriscu Sparus caeruleostir 1 Harengula roux; (b) DEMERSAL SPEClES Polynemidae renalus 1992 Polynemidae Serranidae \ I , Priacanthidae Pseudotolithus entex . " P. prayensis F ig. 1 O: sity of arti h landings of Ghana, 1981and 1992; a: Pelagics; b: Demersals. /. FERRARIS, K. KORANTENC AND A. SAMBA 459 1982 (a) PLIRSE SEINE 1992 4z1 -65 l E. fimbriata - - - - - - - - 1 Euthynnus alletteratus - - - - - - - - - - - - - - - - - - - - - - - - - . 2 Scomber japonicus - - - - - - - - - - - - - - - - - - ---------- - - - - - - - - - . 3 Pomatomus saltator ---------- ---------- ---------- 4 Caranx rhonchus 1 5 Chloroscombrus chrysurus 6 Trachurus ----------- ---------- ----------. 7 Decapturus rhonchus ----.----- 8 Brachydeuterus 9 Pomadasys ----------. . - - - - - - - - - . - - - - - -, Ethmalosa (b) LlhlE rhonchus 1 Euthynnus alletterarr 3 Pomatomus saltator 9 Pomadasys caeruleostictus (c) SET NET sarda Dentex Sparus caerusleostictus 9 Pomadasys 10 Sphyraena 11 Galeoides decadactylus 12 Pseudotobthus 13 Argyrosomus 14 Scomberomorus Rays Sharks Fig. 11 : Changes of species diversity, by gear, of artis h landings of Senegal, 1981 and 1992; a: Purse seines; b: lines; c: set nets. 460 Srnall-scale Fisheries in Senegal and Ghana The species composition in the landings of set net and line fishing in Senegal (Fig. 11) shows a great diversity that reflects diffuences in target species which depend on the following factors: i.) kind of set net (for soles, gastropods, sharks or rays, surface or bottom setting); ii.) ~ishing season (iine fishers target lstiophorus in the warm season and Epinephelus in the cold or upwelling season); iii.) fishing practices in accordance with ecological characteristics, fishing villages, ethnic groups, etc. (e.g., in Kayar, Wolof fiihers prefer to catch species such as Povzatomus or Pagellus with line and sole with set net whereas Lebou fishers p refer Epinephelus aeneus and do not use futed gear such as set net). 'fie study of changes in the small-scale fisheries in Ghana and Senegal shows in both cases an expanding sector. The twci upwelling systems in Senegal and Ghana induce similar characteristics in the small-scale fisheries; particularly a doniinant fishery for small pelagic fish species. However, there are differences in species diversity. Environmental variability, a cclnsequence of the upwelling fluctuations, results in uncertainty and instability in marine resources and raises the prot~lem of interactions between the dynamics of the environment, fish stocks and fisher communities (Cury and Roy, 199.!). In die face of variability, small-scale tropical fishenes are characterized by their dynamics and adaptability. Small-scale fishing unit ; have flexibiiity and ability to switch between various target species in response to trends in relative abundance of fish and to chai,ges in market preference or technical innovation. Tne situation where a wide range of target species is exploited by bats, shifiing seasonally or from one trip to another, is often descnbed as the likely ultirnate stage of development of industrial muliispecies fisheries. Paradoxicaiiy, smaii-scale fisheries fa1 into th catego. (Guland and Garcia, 1984). Stuc ies of long-term fishery development often show a 'fishing-up' sequence with an expansion of the fisheiy as fisheis become more mobile and shift their effort to other species in response to a decline in landings of preferred species (Deimling and Liss, 1994). The fact that fishers in Senegal and Ghana developed canoes with insulated ice-boxes to exploit remote fishi.ig grounds or adapted to joint utilization of fishing gears is, perhaps, a sign of the 'fishing-up' sequence in response to yielc s decrease. But migration and the use of a multiple fishing gears are two intrinsic characteristics of tropical multispecies smail-scale fisheries. These characteristics, given alternatives, give them flexibility and provide stability. There are similarities in nature, extent and evolution of the small-scale fisheries in Senegal and Ghana. In the comparative stud!l of the evolution of these two fisheries over a decade, three principal factors were identified as common responses to char ges in the environment. These are: i.) Pigrations: motivated by various causes (socio-economic or biological); they are either short- or long-term, between cc untries in the sub-region, or between regions within one country; ii.) lechnological mutations: introduction of new technology, innovation, knowledge transfer; iii.) !;witching and joint utilization of fishing gears. For Senegalese fishers, Laloë and Samba (1990) identified two types of reactions: the strategic (migration and tech~~ological choice) and tactical (switching of effort depending on resource availability or market opportunities). The dynuinics of exploitation is thus approached in terms of strategy and tactics. A mode1 was developeci to describe the dynamics of the small-scale fisheries in Senegal on the basis of a stock production model adapted for the multispecies and multigear fisheries (Laloë and Samba, 1991). In this model, the terms 'tactics' and 'strategy' are used to describe the decision making process of fishers in response to accessibility of resource, biomass changes, variations in market and/oi socio-economic factors. A second model is being developed to simulate decision-making in the Senegalese small-scale fishery on the basis of an expert system and object-oriented formalism, where the whole fishery is viewed from a systemic point of view (Le Fur, this vol.). The development of the fishery must be studied in a natural-cultural context where each fisheq system is composed of interacting factors of physico-chemical, biological and cultural nature (Deimling and Liss, 1994). Catch is reflective of natural factors which create varying levels of abundance of species, and reflective also of cultural factors such as fishing technology, fisheries economics and market preferences. Catch is thus a product of the entire natural-cultural system. From the point of view of fishing operation, catch is the result of the choice of one gear, one fishing place and one target species. The choice of these three factors may be presented as a tactical decision that needs to be taken by the fislier before or during a fishing trip. Ferraris (1995) defines 'tactics' as a combination of fishing grounds, target species and gears. One can study changes of fishery dynamics in the short term. 'Strategy' is defined as a set of tactics. This concept integrates fishing activity in a given penod of time and allows the study of fishing dynamics on a longer term. Local and global changes, from a temporal point of view, may then be interpreted in terms of tactics and strategy. Tactics permit the analysis of the dynamics on the basis of daily fishing activity; a change in tactics is interpreted as a response to some local change in the fishers' environment. Strategy permits the analysis of the dynamics of a fishery on a seasonal or annual basis. A change in strategy reflects changes in global fishing conditions and it impacts on available tactics. However, a local change, for example the introduction of new tactics, may have an impact at the global level. The responses to biological, ecological or socio-economic changes observed in the Senegalese and Glianaian fisheries may also be described on a spatial scale. Local changes have reference to spatial peculiarities due to the natural-cultunl system and the history of each fishery. Global changes, on the other hand, generated similar changes observed in the two countries. The local dimension refers to the specific tactics and strategies of each fishery, while global dimension led to common responses. Despite observed differences in the small-scale fisheries in the two countries, the changes observed on a decadal basis underline some generic fishing behaviors. These global changes may be due to similar changes in the natural system (e.g., increase in Sardinella aurita abundance, proliferation of triggerfish, development of cephalopods, etc.) or in the human system (e.g., the opening of export market and increasing domestic demand related to human demographic growth). The study of the small-scale fisheries in Ghana and Senegal, through the structure and evolution of fleets and catches, underlines the importance of a good understanding of the dynamics of exploitation. The fishers' ability to adiust their activities and to react to perturbations in their environment confer on multispecies and multigear small-scale fislieries great flexibility and stability. The opportunistic behavior of the fishers may give some signals about the condition of the system and the wealth of the resource. Therefore, from a fisheries management point of view, it is important to better understand the reaction of fisher facing changes (Hilborn and IValters, 1992). 462 Small-scale Fisheries in Senegal and Ghana Ccrnparative studies of the dynamics of the small-scale fisheries in Senegal and Ghana underlines tliree important factors in ihe fishing decision-making: fishing location, fishing gear and target species. The dynamics of the fislieries mai be st~died by these three factors, expressed in terms of tactics and strategy. Changes in species composition of catches and vo urne of landings were observed. However, changes in fishing strategy and in landings may be confused with 'real' ch.inges in species composition or abundance. Similar migratory behaviors by fisher, resort to the use of a multiplicity of ge.irs and technologicd mutation were identified. Thus we found, the specificity and comrnon cliancteristics of tlie t\vo fisiieries facilitated the study of the dynamics of exploitation in the context of local and global changes. Bc rnaseck G.M. 1986. Profile of the marine resources of Ghana. Cl 'CAF/TECH, 86/71,10 j p. Overivière A. 199 ja. Le poulpe (Octopus oulgans) au Sénégal: ur t: nouvelle ressource. In: El~aluation des ressources trrpioi- ta~~lespourlapêcheal-tisanalesénégalaie Coll. et Séminaires, OItSTOM: 24 5-257. Caverivière A. 1995b. Les fluctuations d'abondance du baliste (6:tlistes carolinensis). In: Evaluation des ressources exploi- tadespour lapêcheal-tisanalesénégalaise. Coll. et Séminaires, OIBTOM: 257-264, Ck .luveau J.-P. 1984. Lapêchepiroguièreau Sénégal. Les leçons de /'histoire. Rewe Mer, no spécial, automne 1984,19 p. Cb.iuveau J.-P. 1990. The histoncalgeognphyof fisheries migra- ti~iis in the CECAF region (endXIXcentury to 1980's). In: J. Haa- koiisen and M. Diaw (eds.). Fishei7nen's nzigrations in iuest Afi-i- eu, IDAF/wp/36, IDAF Project (FAO), Cotonou, Benin: 12-35. Cuiy P. and C. Roy. 1988. Migration saisonnière du thiof (Epi- nq!~helus aeneu) au Sénégal: influence des upn~ellings séné- ga:;iis et mauritanien. Oceanologica Acta, ll(1): 25-36. Cuiv P. and C. Roy. 1992 (ed.). Pêchetiesouest-africaines: aria- Dil,;té, instabilité et changement. ORSTOM, Paris, 525 p. Deimling E. and W. Liss. 1994. Fishery development in the eas- triil North Pacific: a natunl-cultunl system perspective, 1888- 19 -6. Fish. Oceanogr., 3 (1): 6û-77. Diaw M. C. 1991. Mignnt fishermen froni Casamance and sou- them river areas. ln: J.M. Haakonsen and M.C. Diaw (eds.). Fisher- iizen's inigmtions in luest Afi.ica. IDAF/WP/36. IDAF Project (FAO), Cotonou. Benin: 77-93. Everett G.V. and G.T. Sheves. 1991. Recent tivnds in altisana1 fisheiies and report on alternatives to canoes. IDAFNP/40, IDAF Project (FAO), Cotonou, Bénin, 33 p. Ferraris J. 1993. Lapêchea~-tisanaleau Sénégal: ntlnsde ie~viiia- bilitéspatio-teniporelle. Doc. int. CRODT. 79 p. Ferraris J. 1994. A critical look at the sui.vqi q)steir? on aifisa- nalfisheiies in Ghana. Document de travail. Graiid Prograiii- rne Sardinelle, ORSTOM/CROA/FRUB, 29 p. Ferraris J. 1995 Déinarche niéthodologiqiie pour I'aiialyse des comportements tactiques et stratégiqiies des pêcheurs artisaris sénégalais. In: F. Laloë, H. Rey and1.L. Durand (eds.). Questions sur la d)jnainique de l'exploitation halieutique. Colloques et Séminaires, ORSTOM: 263-293. FerrarisJ. and K. A. Koranteng. 1995. Statistical analysis ofcanoe fisherydata iii Ghanawith particular reference tosarclinellas. In: F.X. Bard aiid K. Konnteng (eds.). Qlnainics and uses of snr. dinella i~esourcesfiorr~ upzvelling off Ghana and Côte-d'luoi- re. Colloqiies et Séminaires, ORSTOM: 20 j-22. Gulland J.A. and S. Garcia. 1984. Observed patterns iii iiiiiltis- pecies fisheries. In: R.M. May (ed.).E~ploitation ofirrarinc corrl- nzunities. Dahlern Konferenzen 1984. Berlin, Heidelberg, New York, Tokvo, Springer-Verlag: 1 j 5-190. Haakonsen J.M. and M.C. Diaw. 1991. Fishemzen's migrations in tuest9frca. IDAF/WP/36, IDAF Project (FAO), Cotonou, Beniti. 307 p. Horemans B. 1993. La situation de la pêche artisanale en Afrique de l'ouest en 1992. Cotonou, DIPA/WP/47,36 p. Hilborn R. et C.J. Walters. 1992. Quantitativefibenes stock assessnzent. Choice, dynanîics and uncet7ainh~. Chapman and Hall, N.Y., Lond., 570p. KébéM. 1995. Principales mutations de la pêche artisanale mari- time sénégalaise. In: Evaluation des ressources exploitables pourlapêche altisanalesénégalaise. Colloques et Séminaires, ORSTOM: 43- j8. Koratiteng K.A. 1990. Ghana canoe frame survq, 1989. Inf. Rep., Fish. Res. & Util. Branch, Tema, Ghana, 25, 57p. Koranteng K. A. 1992. The irzarine artisanaljshety in Ghana, recent developnlents and iiitplications for resource evalua- tion. Paper presented at the \Vorld Fisheries Congress, Athens, Greece. May 3-8,1992. Koranteng K A 1995. Marineartisanalf2sheries in Ghana and Senegal- sinîilarities and differmces. Paper presented at the Workshop on 'Evaluation des ressources exploitables pour la pèche artisanale sénégalaise'. CRODT, 8-13 February 1993. Konnteng K A, 0.0. Nmashie and A N. A Baddoo. 1993. Ghana canoe franze survq, 1992. Inf. Rep. Fish. Res. & Util. Braiich, Tema, Ghana, 28,48 p. L~loë F. and A Samba. 1990. Lapêche artisanale au Sénégal: res- sources et aatégies depêche. ORSTOM, Etudes et Thèses, 393 p. Laloë F. and A. Samba. 1991. A simulation niodel of artisanal fisheries of Seiiegal. ICES Mar: Sci. Symp., 193: 281-286. Lawson R.M.and E.A. Kwei. 1974. African entrepreneur sliip and ecotiomicgrowth: acase study of the fishing industry ofGhatia. Ghana Universities Press, Accra, 262 p. Mensah M. A and K A Konnteng. 1988. A reviem of theoceario- graphy andfisheries resources in the coastal zuateis oj'Ghana. Mar. Res. Rep. Fish. Res. & Util. Branch, Tema, Ghana, 8,35 p. Odoi-Akersie W. 1982. Canoefimne suruev Ghana - October 1981. Inf. Rep. Fisheries Dept, Accra. Odotei 1. 1991. Migrations of fatite fisherrnen. In: J.M. Haakon- sen and M.C. Diaw (eds.). Fisheniien's inigrations in zuest Ah- ca. IDAF/WP/36, IDAFProject (FAO), Cotonou, Bénin: 168-179. Pezeiinec 0.1995. Ecological importance of the Ivorian and Glia- iiaian minor upwelling season. In: F.X. Bard and K. Koraiiteng (eds.). D~~nanzics and uses of sardinella resourvesfi-oni upiuel- lingoff Ghana and Côte-d'ivoire. Colloques et Séminaires. ORS- TOM. 324-34 5. Pielou E.C. 1975. Ecological diversitl!. Wiley, New York. 16j p. \ Roy C. 1992. Réponses des stocks depbissonspél~7giques à la dynamique des upzvellings en Afrique de l'ouest: analilse et nzodélisation. Etudes et Thèses, ORSTOM, 146 p. Samba A. 1995. Présentation sommaire des différentes pèclie- ries artisanales. In: Evaluation des ressources exploitablespour la pêche artisanale sénégalaise. CRODT, 8-13 Februan 1993. Colloques et Séniinaires, ORSTOM: 1-9. Saniba A. and F. Laloë. 1991. Upwelling sénégalo-niauritanien et pêche au tassergal (Poiiratoiizussaltato~ sur la côte nord du Sénégal. In: Pêcheries ouest-afi-icaines: variabilité, instabilité et changement. ORSTOM, Paris: 307-310. Wayo Seini A. 1995. Economics of marine canoe fisheries in Ghana. In: F.X. Bard and K. Koranteng (eds.). Dvnairiics and uses of sardinella resources froin upiuelling of Ghana and Côte-d'ivoire. Colloques et Séminaires. ORSTOM: 426-436. 464 Srnall-scale Fisheries in Senegal and Ghana Fishmeal Price Behaviour: Global Dynamics and Short-Term Changes MARIE-HELENE DURAND ORSTOM Laboratoire Halieutique et Ecosystèmes Aqiiatiques B.F. 5045 34(132 Montpellier Cedex 1 FRA VCE ABSTRACT About 80% of the world's pelagic fish resources are processed into fishmeal. The price of fishmeal is set on the world market and imposed to local producers. The liigh variability of fishmeal prices on the world market is not wholly connected with the fluctuations of aggregate supply and demand; interdependencies with other markets and speculative activities determine a large amount of this variability. This study considers the relationsliip benveen the fishmeal and soyabean meal markets. The liypothesis tested here concerns the existence of a long-term relationship directing the behaviour of the prices of tliese two commodities. Tests for cointegration are performed, and an equilibrium relationsliip is estimated. The results show that soyabean meal market induces short- term fluctuations into the fishmeal market because of speculative effects, while fishmeal price changes influence soyabean meal prices through a modification of the demand for soyabean meal. Près de 80 % des espèces pélagiques capturées dans le monde sont transformées en farine de poisson. Le prix de la farine de poisson fixé sur le marché mondial s'impose à tous les producteurs quelles que soient les conditions locales de la pêcherie. Ce prix présente une forte variabilité qui n'est pas toujours en rapport avec l'évolution de l'offre et de la demande mondiale; les interactions avec d'autres marchés et les activités spéculatives déterminent en grande partie les variations de prix. Des tests de cointégration et l'estimation d'un modèle à correction d'erreur montrent l'existence d'une relation à long terme entre le marché de la farine de poisson et le marché du tourteau de soja qui dirige en partie l'évolution des prix sur ces deux marchés. C'est par un effet spéculatif que le marché du tourteau de soja induit des fluctuations à court terme du prix de la farine de poisson. L'évolution du prix de la farine de poisson provoque des modifications de demande sur le marché du tourteau de soja et entraîne des changements de prix. 1 NTRODUCTION Fishmeal is usually prepared from pelagic species (anchovy, sardine, jack mackerel or capelin), the most important fish resource available, but also the most unstable: sudden pelagic stock 'outbursts' or, on the contrary, sharp resource declines are frequent. Above and beyond the amount of study that goes into the reasons behind them, such variations in the availability of fish do have an impact on the overall market. While the available data are not very precise, it can be estimated that the pelagic catches used worldwide by the nduction industry represent roughly one third of world marine catches (Le., about 30.10~ t out of the 70-75.10~ t), and about 80% of world pelagic catches. Thus, fishmeal production is the main outlet of pelagic fisheries. World fishmeal production totals about 6.5.10~ t, more than half of which (around 3.5.10~ t) moves into international trade channels. About 70% of tliis international trade originates from five countries (Peru, Chile, Denmark, Iceland and Nonvay). The main areas of consumption are Europe (a traditional market centered mainly on Germany, a leading importer), East and Southest East Asia (China, now the biggest importer, Taiwan, Japan and, more recently, Indonesia, the Philippines and Thailancl), and North America (essentially the United States, although Mexico has recently developed a broader demand base). These regional destinations represent roughly 3 5,40 and 20%, respectively of total exports. Fishmeal is a commodity whose sole end-market is the feed indust~y, itself located upstream from the animal and me:it production sectors. Beyond the protective measures implemented in various regions as part of agricultur~l policy agricultural markets are al1 very competitive and under great pressure as far as pricing is concerned. The fishmeal mrket is a suppljr-limited market, and, due to the rapid development of aquaculture, an increasing demand helps to maintain a higli price level. However, as for major commodities traded in an increasingly globalized and competitive world, the worlcl fishmeal price shows a high variability. Apart from demand and supply factors, interdependence with other coinmodities and financial markets determine the evolution of prices. Local producers, facing instability in the input, are price-takers for their output. 466 Fishmeal Price Behaviour It is common knowledge that prices of commodities such as raw materials and agricultural products follow similar patterns (Deaton and Laroque, 1992). Such similarities in the behaviour of commodity prices can often be explained by the broader underlying macroeconomic factors affecting al1 prices in general, e.g., world inflation, interest rates or evolving demand and industrial production (Pindyck and Rotemberg, 1990). Some commodities' prices, however, can b.e seen to be more closely interrelated and their common trends do show additional links. Certain factors more specific to these commodity markets - e.g., substitution possibilities, complementarity, or orientation towards a same demand - have to be taken into acco.Int when explaining their CO-movements (Lord, 1991). It is generally admitted, for example, that substitutability amoiig several commodities has the effect of decreasing prices. The id en ce of a link between the fishmeal and soyabean meal markets is well known, particularly to the animal feed mil1 operitors and traders. However, this is only an empirical observation and, although generally postulated, it has never been tested as a forma1 hypothesis. The purpose of this study is to verify whether there really is a specific relationship between the p-ices of the fishmeal and soyabean meal markets. We test the existence of and quantify the common long-term trend to whic t i both prices may be related; also we investigate the causality links explaining the behaviour of prices. Interpretations are ( ffered of the long-run equilibrium between prices, as well as price-forming mechanisms on the two markets. In the first section, we shall briefly describe the markets' specif