Explaining gradients in plant community composition with a general multivariate model

Analysis and interpretation of community composition is difficult due to the high dimensionality of community datasets. Data reduction methods such as ordination can reduce a community matrix down to at least two orthogonal axes, suggesting that at least two major ecological processes generate gradients in species space. Based on assembly and response rule theory, we hypothesized that plant community composition is a function of the interactions among three general constructs: abiotic and biotic factors and disturbance history. We sampled vegetation and soils on seventy-five randomly located plots across a broad soil gradient within a 110,000 ha ponderosa pine forest in northern Arizona to evaluate this hypothesis. General multivariate models have appeal when analyzing ecological systems because they have the potential to shed light on the relative importance of multifaceted factors (e.g., abiotic gradients). We evaluated the relationships between ordination results and environmental conditions using structural equation modeling. The model employed composite variables, which specify the combined effects of multiple factors on a response, in order to simplify the complex specific model to address the general hypothesis, and to evaluate the relative importance of the three theoretical constructs. We identified two independent compositional gradients in ponderosa pine forest understory plant communities, suggesting that there were at least two underlying ecological processes that generated these gradients. Interestingly, the general model suggested that abiotic factors explained one compositional gradient (R²=0.71) and biotic factors and disturbance history explained the second compositional gradient (R²=0.61) in this system.