Using a random forest model to predict volume growth of larch, birch, and their mixed forests in northern China

Accurately quantifying forest volume and identifying its driving mechanisms are critical for achieving carbon neutrality objectives. Using data from the National Forest Inventory (NFI), plot-level measurements, and environmental variables from pure larch (LP), birch (BP), and mixed larch-birch (LB) forests in the mountainous region of northern Hebei, China, this study employed random forest (RF) algorithms to evaluate the relative importance and partial dependence of biotic and abiotic factors on stand volume growth. A total of 33 predictors related to climate, topography, and soil were analyzed, and model hyperparameters were optimized through grid search combined with blocked cross-validation to mitigate spatial autocorrelation. The RF models exhibited strong predictive performance, with the BP model achieving the highest R² (0.92). The minimum temperature of the coldest month (Bio12) was identified as the most influential predictor across all stand types, while stand age also exerted a substantial effect on growth dynamics. Young and middle-aged forests demonstrated higher productivity compared with near-mature and mature stands, suggesting that the latter require improved management interventions to sustain growth. The LB stands exhibited higher productivity than pure stands, likely due to species complementarity and interspecific facilitation. In LP, growth was primarily driven by the interaction between stand age and canopy density, whereas in BP, slope position was more decisive. The management of LB stands offers potential to maintain or enhance forest productivity. The findings emphasize the importance of adaptive forest management strategies that optimize forest structure and mitigate climate change impacts. These insights contribute to advancing carbon sequestration efforts and supporting the development of carbon neutrality policies by enhancing forest productivity and resilience to climate variability.