Sample unevenness distorts species spatial turnover interpretation: The problem and a solution

Background/Question/Methods

The shape of beta diversity pattern, i.e., species spatial turnover, has been commonly used to infer ecological processes. A monotonic decay pattern is typically considered a benchmark, deviations from which are attributed to environmental filtering. Many beta diversity indices are notoriously sensitive to sample size, but surprisingly, no studies have systematically examined how spatial configuration of sample sizes (hereinafter sample unevenness) influence species spatial turnover. This paper provides such a systematic examination. A neutral metacommunity model was developed, ensuring the absence of environmental filtering. By excluding the effects of environmental filtering, we examined how sample unevenness alone influences the shape of species spatial turnover when only dispersal limitation is present. We also proposed a null-model corrective approach and evaluated its effectiveness.  The approach was applied to reexamine two published empirical studies. 

Results/Conclusions

Our results show that sample unevenness can yield a non-monotonic decay of species turnover pattern even when only dispersal limitation is present, calling into question the conventional practice of using a non-monotonic decay as an indicator of environmental filtering. This sample unevenness effect is more pronounced in systems with widely dispersing organisms and low diversity. Accounting for the sample unevenness effect altered the ecological understanding reached in the two published studies we examined. The proposed corrective approach is generally effective in removing irregular trends induced by sample unevenness, but some caveats apply under certain circumstances. We conclude that the monotonic decay is not always a valid benchmark for species spatial turnover. It is necessary to consider the statistical effect of sample evenness before invoking possible ecological mechanisms to explain the spatial turnover patterns. Failure to do so can result in mistaking a statistical artifact for ecological processes. Consequently, because different systems are likely of different spatial configurations of sample size, extra caution must be exercised for cross-system comparison.