Abiotic factors influence species co-occurrence patterns of lake fishes

Abiotic factors are recognized for their strong influence on community structure. Habitat diversity is related to resource availability that influences species richness and abundance. In lakes, surface area and depth have been used as measures of the size and diversity of habitat, and have strong effects on the structure of entire communities. We tested whether abiotic variables, related to habitat size, influence co-occurrence patterns of species pairs of fishes by analysing groups of lakes within a specific area and depth categories in two regions in Ontario, Canada. We used null models to obtain co-occurrence patterns and standard effect sizes for each species pair within each area and depth category. We estimated standard effect sizes relative to lake area or depth and determined whether species co-occurrence patterns change systematically as these measures of habitat increase. We evaluated groups of species where factors such as predation and habitat filtering have been shown to structure those assemblages, and we tested whether area and depth alter the species associations and our interpretation of these relationships. We found significant differences between the observed and expected distributions of regression slopes relating co-occurrences to area and depth in both regions across all species, which indicated the strong influence of both variables on the overall co-occurrence patterns. We observed a significant negative trend of the co-occurrence patterns across lake area categories for the predator-prey species, indicating that the effect of predation was stronger in smaller lakes, but it was reduced in larger lakes, possibly due to increased habitat and resource availability. We show that pooling results as done in standard community null models can lead to Type II errors due to the 'cancellation' of opposing ecological signals. Our results demonstrate the effect of environmental variables on species co-occurrence patterns, but the divergent results obtained between geographical regions suggest that such patterns are context-dependent. This study emphasizes the importance of considering abiotic factors in null models of species co-occurrence to obtain reliable and detailed information about the association patterns between species.