Can stochastic geographical evolution re-create macroecological richness-environment correlations?

Aim Richness gradients are frequently correlated with environmental characteristics at broad geographic scales. In particular, richness is often associated with energy and climate, while environmental heterogeneity is rarely its best correlate. These correlations have been interpreted as evidence in favour of environmental determinants of diversity gradients, particularly energy and climate. This interpretation assumes that the expected-by-random correlation between richness and environment is zero, and that this is equally true for all environmental characteristics. However, these expectations might be unrealistic. We investigated to what degree basic evolutionary/biogeographical processes occurring independently of environment could lead to richness gradients that correlate with environmental characteristics by chance alone. Location Africa, Australia, Eurasia and the New World. Methods We produced artificial richness gradients based on a stochastic simulation model of geographic diversification of clades. In these simulations, species speciate, go extinct and expand or shift their distributions independently of any environmental characteristic. One thousand two hundred repetitions of this model were run, and the resulting stochastic richness gradients were regressed against real-world environmental variables. Stochastic species-environment relationships were then compared among continents and among three environmental characteristics: energy, environmental heterogeneity and climate seasonality. Results Simulations suggested that a significant degree of correlation between richness gradients and environment is expected even when clades diversify and species distribute stochastically. These correlations vary considerably in strength; but in the best cases, environment can spuriously account for almost 80% of variation in stochastic richness. Additionally, expected-by-chance relationships were different among continents and environmental characteristics, producing stronger spurious relationships with energy and climate than with heterogeneity. Main conclusions We conclude that some features of empirical species-environment relationships can be reproduced just by chance when taking into account evolutionary/biogeographical processes underlying the construction of species richness gradients. Future tests of environmental effects on richness should consider structure in richness-environment correlations that can be produced by simple evolutionary null models. Research should move away from the naive non-biological null hypotheses that are implicit in traditional statistical tests.