Simulating random species-environment relationships at broad geographic extents

Background/Question/Methods

Much research in macroecology and biogeography has been related to how patterns of variation in species richness at broad spatial extents have arisen. Two hundred years of research on the topic has generated a large number of possible mechanisms. Among them, environmental characteristics, such as energy or heterogeneity, have received much attention. From this research, the idea of a ubiquitous and strong species-environment relationship has developed; this has been used as evidence for the role of environment on determining species richness gradients. In particular, recent research has highlighted the potential importance of energetic determinants. Most studies investigating species-environment relationships have used traditional statistical null hypotheses of no correlation. However, species richness gradients at broad scales result from spatially explicit processes of clade diversification and distribution. Such processes have potential to create spatially structured variation in species richness independent of environmental variables. Thus, correlations between independent environmental and richness gradients are possible by simple coincidence. In this study, we quantify the expected relationship between species richness and environment produced by stochastically diversifying clades. We use computer simulations to produce artificial species richness gradients based on a purely stochastic process of clade diversification and species distribution. For each simulation run, we evaluate strength of the relationship between species richness and three environmental hypotheses: energy, heterogeneity and seasonality. Additionally, potential influence of extinction rate, continent, species range size and position of clade origin were investigated.  Finally, diversification models were compared to simpler mid-domain effect models.
Results/Conclusions

Results demonstrate that random diversification of clades can produce species-environment relationships that are different from mid-domain effects, and very different from naïve null hypotheses used in traditional statistical tests. Inclusion of high rates of extinction or variation in average range size does not strongly modify these results. This suggests that clades that have diversified stochastically can frequently lead to patterns of species richness that coincide with environmental gradients. Frequent spatial coincidence of independent species richness and environmental gradients can create spurious and strong relationships. Additionally, different spatial variation among variables representing separate environmental characteristics can lead to a false predominance of a particular hypothesis. For example in Africa, simulated species richness correlates more strongly and more often with energy than with heterogeneity. This suggests that the proposed primacy of energetic determinants over other environmental hypotheses could also be spurious. Overall, our results suggest that the notion of a ubiquitous species-environment relationship in macroecology might need to be revisited.