Predicting the spatial distribution of leaf litterfall in a mixed deciduous forest

An accurate prediction of the spatial distribution of litterfall can improve insight in the interaction between the canopy layer and forest floor characteristics, which is a key feature in forest nutrient cycling. Attempts to model the spatial variability of litterfall have been made across forest types, but the reported models have not yet been compared. We predicted the spatial distribution of leaf litterfall for the same mixed hardwood stand using inverse distance interpolation, ordinary kriging, single and multiple regressions based on plot basal area, and three individual-tree models. Models were calibrated using litterfall data (n = 67) of white birch (Betula pendula Roth), pedunculate oak (Quercus robur L.), and northern red oak (Quercus rubra L.). Model performance was compared using an independent validation data set (n = 37). Interpolation techniques did not reliably estimate spatial patterns of leaf litterfall (r < 0.60, n = 37). However, models incorporating tree data, such as linear regressions and individual-tree models, successfully reproduced the observed spatial litterfall heterogeneity of each species (r > 0.80). No model was able to predict the variability of the total leaf litterfall of the three species. We conclude that, for an intimately mixed forest stand, a model that simulates leaf dispersal of individual trees is likely to be the best choice for predicting the spatial distribution of leaf litterfall.