Estimating dynamics of central hardwood forests using random forests

Abstract

Estimation of forest population dynamics is critical for forest management decisions making. In this study, we developed an innovative climate-sensitive matrix model using random forests (RF) algorithm to estimate tree diameter growth, tree mortality, and stand recruitment and consequently predict population dynamics of the central hardwood forests under four climate scenarios (i.e. Representative Concentration Pathway [RCP]2.6, 4.5, 6.0, and 8.5). Based on post-sample validation, this RF matrix (RFMatrix) model was more accurate than the traditional climate-sensitive matrix model and Landis pro 7.0. According to the importance values of all predicted variables, the variability in tree diameter growth, tree mortality, and stand recruitment was mainly explained by local tree and stand-level factors, followed by climatic and anthropogenic factors, and soil factors were the least important for all the species. Additionally, our model predictedthat climate change could substantially reduce total stand basal area. The RFMatrix model and its prediction results could assist future forest population dynamics studies on the central hardwood region under changing climate.