PLS-regressions highlight litter quality as the major predictor of humus form shift along forest maturation

Abstract

Using Partial Least Squares regression, we ranked the ability of leaf litter and topsoil properties to predict humus form shift along a 130-yr-old pure beech forest chronosequence. Three models were tested, including only litter properties (model 1), only topsoil properties (model 2) and both litter and topsoil properties (model 3). The first model was highly significant and explained more than 91% of the humus form variability with N-based variables, Mn, Mg and K as the best predictors. The second model showed lower goodness of fit (75%) with Ca and Mg contents, pHKCl and ΔpH as good predictors. The last model showed that litter traits were better predictors compared to topsoil variables, suggesting that beech trees may impact humus form along forest development mainly through aboveground pathways.