Complex trait–environment relationships underlie the structure of Wisconsin forest plant communities

Abstract Plant species shift in abundance as environmental conditions change because traits adapt species to particular conditions. As a result, trait values shift along environmental gradients—the so-called trait–environment relationships. These relationships are often assessed by regressing community-weighted mean (CWM) traits on environmental gradients. Such regressions (CWMr) assume that local communities exhibit centered optimum trait–abundance relationships and that traits are not independent from one another. However, the shape of trait–abundance relationships can vary widely along environmental gradients—reflecting the interaction between traits and gradients—and traits are usually interrelated. Accounting for these complexities should improve our ability to accurately describe trait–environment relationships. We tested these ideas by analyzing how abundances of 185 herbaceous understory species distributed among 189 forested sites in Wisconsin, USA, varied in response to four functional traits (vegetative height-VH, leaf size-LS, leaf mass per area-LMA, and leaf carbon content) and six soil and climate variables. A generalized linear mixed model (GLMM) allowed us to assess how the shape of trait–abundance relationships changed along environmental gradients for the 24 trait–environment combinations simultaneously. We then compared the resulting trait–environment relationships to those estimated via CWMr. The GLMM identified five significant trait–environment relationships that together explained ∼40% of variation in species abundances across sites. Temperature played important roles with warmer and more seasonal sites favoring taller plants. Soil texture and temperature seasonality affected LS and LMA more modestly; these seasonality effects declined at more seasonal sites. Only some traits under certain conditions showed centered optimum trait– abundance relationships. Concomitantly, CWMr identified 17 significant trait–environment relationships including effects of temperature, precipitation, and soil on LMA as often reported in other studies. Despite this overidentification, CWMr failed to detect significant temperature-seasonality effects found in the GLMM. Modeling the complexity of how traits and environments interact to affect plant abundance allows us to identify and rank key trait– environment relationships. Although the GLMM model was more complex compared to single CWM regressions, it identified a simpler hierarchy of trait–environment relationships that accurately and reliably predicted responses of forest understory species to gradients in environmental conditions.