Soil chemical variables improve models of understorey plant species distributions

Aim Species distribution models (SDMs) assume that all ecologically relevant predictor variables are included. This assumption is frequently violated in SDMs of plant species, as soil rarely Here, we used in-situ, geochemical along with more commonly topo-climatic and remotely sensed to create understorey species. We evaluated whether the potential importance greater than generally described SDM literature, which predictors most important, a standard subset can be effectively model Location Northern Appalachian Mountains. Taxon Vascular, forest plants. Methods fit for presence 41 species across 158 plots using soil, topographic spectral predictors. Results Models containing three types performed best. Soil had comparable importance; were lesser importance. The best variable was B horizon carbon nitrogen ratio (B C:N), followed by position index, elevation exchangeable calcium Ca). A did not Main conclusions Our results those other include in-situ data suggest increasingly important detailed descriptions soils. fertility data, such C:N Ca, particularly acidic, soils where pH poor indicator fertility. Commonly provide meaningful predictions but limited their use indirect variables, inhibiting transferability interpretability. performance highlights uniqueness each species' niche need have variety variables.