CAUSES OF LATITUDINAL GRADIENTS IN SPECIES RICHNESS: A TEST WITH FISHES OF THE TROPICAL EASTERN PACIFIC

Recent advances in analytical methods and exploration of regional diversity patterns in greater depth than simple whole‐fauna patterns may change our understanding of the determinants of latitudinal gradients in species richness. Using a comprehensive database on the geographical distributions of the large fauna of endemic shorefishes from the Tropical Eastern Pacific (TEP), we delineated latitudinal diversity gradients (LDGs) of species with different range sizes and assessed how the mid‐domain effect, energy supply, environmental variability, and habitat availability predicted the various LDGs. We used statistical methods that account for spatial autocorrelation within each variable and consider collinearity among them. We found marked variation in LDGs among species with different range sizes: species with large ranges contributed the most to the convex form of the entire‐fauna LDG while species with small ranges generated most deviations from that form. The mid‐domain effect was the strongest determinant for the entire‐fauna LDG, largely due to the strength of its effect on widespread species. Habitat variables were the best predictors for LDGs of small‐range species but had no significant effects on the entire‐fauna pattern. We found that any relationship between energy supply and LDGs likely is spurious and arises from statistical effects of (1) the marked spatial autocorrelation of the residuals in those relationships and (2) strong covariation between energy supply and the mid‐domain effect, two factors that have never been analyzed together in marine studies (and seldom in terrestrial studies). There was no indication that environmental stability was an important LDG determinant at any level within the fauna. We found that latitudinal (Rapoport) trends in range size in this fauna arise as corollaries of the mid‐domain effect. The disparate contributions of species with different range sizes to diversity patterns demonstrate the failure of traditional, whole‐fauna LDGs to adequately represent all faunal components and their determinants, particularly those of small‐range (and more threatened) species. We argue that, just as analyses of whole‐fauna patterns within a region obscure variation in determinants among faunal components, analyses of patterns at transoceanic (multiregional) scales are likely to conceal important regional variation in determinants of diversity gradients.