Prediction of forest nutrient and moisture regimes from understory vegetation with random forest classification models

The proper choice of the tree species to be grown in a specific forest site requires a good knowledge of the tree species autecology and a comprehensive description of the local environmental conditions. In Belgium (Western Europe), ecological forest site are classified according to three major gradients: climate, soil nutrient (fertility) and soil moisture regimes. Understory indicator species are used by practitioners to determine nutrient and moisture regimes, but requires a significant expertise of forest ecosystems. The present work aims in a first instance at modelling the nutrient and moisture regimes based on species composition. Secondly, a practical decision support tool is developped and made available in order to predict forest nutrient and moisture regime starting from a floristic releve '. To do so, we collected floristic releve ' s representing understory vegetation di-versity in Belgium and covering all the nutrient and moisture gradient. The combination of soil and topographic measurements with the indicator plants presence/absence support forest scientists in inferring a nutrient and moisture regime to each releve '. The resulting dataset was balanced along the different nutrient or moisture regimes and Random Forest classification models were trained in order to predict the forest site characteristic from indicator species presence (or absence). One model was fitted for the prediction of the nutrient regime, exclusively based on the floristic information. A second one was trained to classify the moisture regime. Accurate predictions confirms the appropriate use of indicator species for the Belgian forest site classification. The two models are intregrated in a web application dedicated to forest practionners. This website enables the automatic determination of nutrient and moisture regimes from the species list of a floristic releve '.