West African rivers as biogeographic islands: species richness of fish communities.

Some factors influencing the species richness of West African fish communities were studied in a sample of 26 rivers using four habitat and hydrologic variables. Analysis of a larger sample of 39 rivers showed that species richness was positively related to area. A power function with an exponent of 0.32 gave the best fit. As the surface area used was that of the catchment area and not that (unknown) of the river, the biological significance of this relationship and the possibilities of comparison were limited. Ridge regression analysis and forward stepwise selection indicated that a model that explained ln(species richness) as a function of ln(mean annual discharge) and ln-(catchment surface area) was best, accounting for 90% of the variance of the dependent variable. The combination of surface area and discharge was presumed to act through the volume of water available for the fishes and habitat productivity. Habitat diversity, measured by the diversity of the terrestrial vegetation covering the catchment area, had no significant positive effect when surface area was used in the regression. Rivers ("islands") should have fewer species than tributaries of similar size since, for fishes within a river system ("continent"), there is free circulation between all its branches. The model derived from the river data underestimated the species richness of a sample of 11 tributaries. This was compatible with the hypothesis of higher population extinction rates in insular biotopes. The residuals of the linear model did not show random geographical distribution; the rivers in some areas had more species than expected. The possibility that historical factors, especially Quaternary climatic variations, might cause this distribution is discussed.