Separating neighborhood structural information from morphological effects in DBH variation in even-aged plantation forests

This study examines whether competition indices capture neighborhood-scale structural information not fully represented by conventional morphological variables in diameter at breast height (DBH) variation of cedar (Cryptomeria japonica) and cypress (Chamaecyparis obtusa). Using terrestrial laser scanning (TLS) data from even-aged planted conifer stands, we compared Conventional models based on tree height, crown diameter, and species with Augmented models incorporating competition indices. Across statistical models, the inclusion of competition indices consistently provided additional explanatory information on DBH variation beyond morphological variables, increasing R² from 0.36–0.38 to 0.40–0.46, although improvements in predictive performance were modest. In contrast, tree-based machine learning models showed minimal differentiation between model variants and lower overall explanatory performance, likely because they can capture nonlinear relationships among predictors. Under such conditions, part of the structural information represented by competition indices may be internally integrated with that captured by morphological variables. We suggest that competition indices function as descriptors of neighborhood-scale structure rather than as variables primarily improving predictive performance. Our results provide consistent evidence for the structural role of competition indices in tree-level modeling and support integrating spatial structure into forest mensuration frameworks.