Modelling factors affecting litter mass components of pine stands

Litter mass represents a key factor in the process of carbon sequestration. Pine plantations are known to accumulate high amounts of litter, which may act as real carbon sink only if it persists for long time. Thus, predicting litter mass by means of robust and straightforward models which convey information from several ecological predictors become crucial in this framework. The aim of this paper was to test for relationships between environmental predictors and pine litter mass (total branch, needle and cone) by Generalized Linear Models, exploring the contribution of different environmental variables in describing patterns of pine litter mass. Different predictors accounting for seasonality, spatial and geomorphological variability, pine stand properties, remotely sensed derived biomass were taken into account. Considering total litter mass, observed vs. predicted values showed a statistically highly significant relation (p < 0.001) by retaining four variables: elevation, latitude, stand age and season. Similar results were achieved for the needle litter mass, which represented anyway the largest fraction of the litter. Regarding branch litter mass, only stand age appeared to be a significant variable. For cone litter mass, no variables were statistically significant in explaining its variance. Potential ecological background processes responsible for the correlations between variables are discussed. Abbreviations: GLM – Generalized Linear Model, NDVI – Normalized Difference Vegetation Index, NIR – Near Infrared, RMSE – Root Mean Squared Error, VIF – Variance Inflation Factor. established (Cseresnyés et al. 2006). Flammability of these pine stands were analysed by McArthur’s model (Noble et al. 1980), and considerable fire risks were predicted (Cseresnyés and Csontos 2004, 2006). All the facts mentioned above inspired us to increase our understanding on relationships between litter quantities in Pinus nigra stands and environmental factors. Least square regression analysis has long been used to investigate relationships between variables. In its simplest form, regression is based on a linear univariate relationship between one predictor and a response variable, such that