We estimate biomass trends for demersal and large pelagic fishes, (i.e., excluding small pelagic and mesopelagic fishes) based on 17 Ecopath models from Mauritania, Cape Verde, Senegal, the Gambia, Guinea, and Guinea-Bissau, Sierra Leone and the open waters of the central east Atlantic, made to represent various time periods during the second half of the 20th Century. We use a published method developed for estimating fish biomass in the North Atlantic, and modify it to account for the specificity of West African fisheries and ecosystems. We show that, overall, fish biomass as defined here (i.e., excluding low-trophic level and small fishes) has declined over the forty year period from 1960 by a factor of 13. An implication of our results is that further increase in fishing mortality in the region will not lead to increased catches but will only drive biomasses further down. The economic and political consequences of our findings are briefly discussed. RESUME Nous avons evalues les tendances de la biomasse des poissons demersaux et des grands pelagiques (c’est-a-dire apres l’exclusion des petits pelagiques et des poissons mesopelagiques) sur la base de 17 modeles des ecosystemiques construits avec le logiciel Ecopath, et representant, pour differentes periodes de la deuxieme partie du 20ieme siecle, les ecosystemes au large de la Mauritanie, Le Cape Vert, le Senegal, la Gambie, la Guinee, la Guinee-Bissau, le Sierra Leone, et le systeme oceanique du l’Atlantique central, a differentes periodes de la deuxieme partie du 20ieme siecle. L’analyse de ces modeles a ete effectuee a l’aide d’ une methode de spatialisation appliquee auparavant a l’Atlantique nord, adaptee aux circonstances du Nord Ouest africain. Notons que la biomasse de poissons definie ci-dessus n’a en l’an 2000 qu’un treizieme de sa valeur en 1960. Ceci implique qu’une augmentation de la mortalite par peche dans la sous-region ne pourra pas augmenter les prises et ne fera qu’accentuer le declin de la biomasse. Nous discutons les consequences economiques et politiques liees a ces resultats. INTRODUCTION This contribution synthesizes the results of the ‘Ecopath module’ of the FIAS/SIAP project, devoted to the construction of ecosystem models for each of the major fishing areas in the countries of the sub-region 1 Cite as: Christensen, V., Amorim, P., Diallo, I. Diouf, T., Guenette, S. Heymans, J.J., Mendy, A.N., Mahfoudh Sidi, T., Palomares, M.L.D., Samb, B., Stobberup, K., Vakily, J.M., Vasconcellos, M., Watson, R., Pauly, D. 2004. Trends in fish biomass off Northwest Africa, 1960-2000, p. 215-220. In: Palomares, M.L.D., Pauly, D. (eds.) West African marine ecosystems: models and fisheries impacts. Fisheries Centre Research Reports 12(7). Fisheries Centre, UBC, Vancouver. Also available in Proceedings of the International Symposium on Marine fisheries, ecosystems, and societies in West Africa: half a century of change, Dakar, Senegal, 24 – 28 June 2002. 2 An earlier version of this contribution is also included in: P. Chavance, M. Ba, D. Gascuel, M. Vakily and D. Pauly (eds.). 2004. Pecheries maritimes, ecosystemes et societes en Afrique de l’Ouest: un demi-siecle de changement. Actes du symposium international, Dakar (Senegal), 24-28 juin 2002, Bruxelles, Office des publications officielles des Communautes europeenes, XXXVI532-XIV p., 6 pl. h.-t. coul., ISBN 92-894-7480-7 (coll. Rapports de recherche halieutique ACP-UE n°15). Trends in fish biomass off Northwest Africa, 1960-2000, V. Christensen et al. 216 covered by the CRSP, plus adjacent waters in Sierra Leone and off-shore. In this synthesis, we concentrate on the change in biomass of demersal and large pelagic fishes, i.e., we exclude small pelagic fishes (mainly sardinella and anchovies), whose environmentally-driven biomass fluctuations would tend to mask systematic, fishery-induced trends, and mesopelagic fishes, which occur only offshore in deeper waters and are not exploitable by current fisheries (Gjosaeter and Kawaguchi 1980). The methodological approach used here is similar to that developed for studying temporal trends in the biomass of high-trophic level fishes in the North Atlantic (Christensen et al., 2003), and hence it will be presented only in summary form, with some emphasis on the modifications required to adapt it to the conditions of Northwest Africa. We concentrate on biomass because this is usually proportional to the catch per effort of fishing vessels, and thus directly impacts on their profitability. Thus, we shall abstain from discussing here the implications on the biodiversity of West African fish that a massive reduction in their biomass is likely to have. MATERIALS AND METHODS Materials Table 1 summarizes the major characteristics of the mass-balance food web (Ecopath) models, used here as starting point for this analysis. Most of these models were constructed during the course of the FIAS/SIAP project, by members of that project (see Pauly et al., 2002). However, additional models were contributed through the Sea Around Us Project to ensure a wide and consistent coverage both spatially and temporally (Palomares et al., in press). We refer to Christensen and Pauly (1992), Christensen et al. (2000) and Pauly et al. (2000) for details on construction and interpretation of Ecopath models in general (see also www.ecopath.org). Some changes (most minor) were made to the models in Table 1 to make them mutually compatible, and in one case, to accommodate a publication with important information on biomass changes, not available when the model in question was constructed and balanced. This refers to the contribution of Myers and Worm (2002) on biomass change in tuna, which documents biomass declines stronger than estimated by Vasconcellos (this vol.). The Myers and Worm estimates have been contested by various tuna biologists, but the area of disagreement in the Central Pacific. There is, on the other hand, a broad consensus that tuna biomass declines have been extremely strong in the Central Atlantic. Thus, to align the abundance trends used here with those in Myers and Worm (2002), the 1950 tuna biomass estimates in Vasconcellos (this vol.) were increased, for the analysis presented here, by a factor of 5 for the North Atlantic, and 8 for the Central Atlantic. The spatially explicit primary production data used here originated as SeaWiFS data, as processed by the European Union’s Joint Research Centre, in Ispra, Italy (Hoepffner et al. unpublished data), based on a model that incorporates estimated chlorophyll, photosynthetically active radiation, and sea surface temperature patterns (Behrenfeld and Falkowski, 1997). The data used here are average values for 1998. An upwelling index was derived based on latitude and basin-specific temperature anomalies by 1⁄2 by 1⁄2 degrees of latitude and longitude. Depth information by 1⁄2 by 1⁄2 degrees of latitude/longitude was obtained from the ETOPO5 dataset available on the U.S. National Geophysical Data Center’s Global Relief Data CD (www.ngdc.noaa.gov/products/ngdc_products.html). Spatialized fisheries catches by half-degree squares were obtained for the years 1950-1999 from the Sea Around Us Project database, and are based on the rule-based method developed by Watson et al. (2001, 2004). Methods The methodology we have used to predict the biomass of fish draws on a combination of ecosystem modeling, information from hydrographic databases, statistical analysis, and GIS modeling (Christensen et al., 2003). The mapping of biomass changes were performed using a series of steps, as follows: West African marine ecosystems, M.L.D. Palomares and D. Pauly 217 • The 17 models of Table 1 were re-expressed on a spatial basis using 1⁄2 by 1⁄2 degree cells (corresponding to 30 by 30 miles at the Equator) using the Ecospace program (Walters et al., 1999). For each of the spatial models, the cells were distributed between habitats based on their mean depth. The following depth strata were used for all models: (1) 1000 m. These yielded estimates of biomass by Ecopath functional groups for each of the spatial cells covered by each model (see Table 1); • The biomass of different functional fish groups were re-expressed as a single value representing all fish with a trophic level of 3.0 or higher, excluding, however, the unexploited mesopelagics as well as the highly-variable small pelagics (see above); • Multiple linear regression analyses were performed using S-Plus 6 (Anon., 2001). We used additive and variance stabilizing transformation, (AVAS) to decide how individual variables are best transformed to obtain linearity (Figure 1); 00 0 5 10 00 0 5 10 00 0 5 10 00 0 5 10 • A multiple regression was identified which predicted fish biomass based on the year for which biomass was estimated, (log transformed) primary production in each half-degree cell, (log transformed) mean depth of each cell, distance from the coast (quadratic transformation), and cell-specific average temperature at 10 m depth in degrees Celsius, and catches of (i) medium demersals and (ii) large demersals. To prevent the records from models covering large areas from overwhelming those from other models, each of the records was weighted, in the regression analyses, by the inverse of the square root of the number of non-land cells in the model to which it belonged. From this, we extracted 5488 records based on the 1⁄2 by 1⁄2 degree spatial cells of the 17 ecosystem models in Table 1. Each of the records included estimates of fish biomass (trophic level ≥3.0), depth, primary production, and year of the model; effect r