Can fish species co-occurrence patterns be predicted by their trait dissimilarities?

Trait-based analyses have been successful in determining and predicting species association outcomes in diverse communities. Most studies have limited the scope of this approach to the biotic responses of a small number of species or geographical regions. We focused on determining whether three biologically relevant traits (body size, temperature preference and trophic level) influence the patterns of co-occurrence between multiple species. We used fish species presence/absence from 9204 lakes in Ontario, Canada, to obtain effect sizes of 2001 species-pair co-occurrence values, using a null model approach. Euclidean distances between each species-pair were calculated for each of the three traits selected. Multiple regression models and randomization tests were used to determine the direction and significance of the relationship of each trait with the observed co-occurrence values. The results show that species temperature preference was significantly related to co-occurrence patterns, indicating the effect of environmental filtering. Trophic level was significantly related to co-occurrence values for both linear and quadratic terms, suggesting that segregation between species is driven by large differences in this trait (predation effects). Unexpectedly, body size was not significantly related to the observed co-occurrence patterns. We provide a new approach to test relationships between species assemblages and trait conditions.