Inferring nitrogen deposition from plant community composition

Chronically elevated reactive nitrogen deposition has a severe impact on many ecosystems, and there is widespread interest in the possibility of using plant community composition to estimate the level of nitrogen deposition and consequent impacts. Existing approaches use a variety of simple measures including functional type ratios, Ellenberg numbers, and diversity indices. We propose an alternative approach in which species–environment models are constructed using national datasets designed to capture broad-scale deposition patterns. We construct models using partial least squares, weighted average, and maximum likelihood Gaussian logit regression for two British semi-natural habitats, and test how well they predict N deposition by cross-validation. We find that performance is good with R2 values up to 0.7, and suggest that such models could be a useful addition to the bioindication toolbox.

► Bioindicators for N deposition are urgently needed for conservation management. ► Models trained on national datasets can be used to infer N deposition from vegetation. ► Such models offer considerable advantages and require further development and testing.