Forest site classification and grading using mixed-variables clustering and nonlinear mixed-effects modeling based on forest inventory data

Abstract Site classification is the basis for evaluating forest productivity and is essential for tree species selection, soil fertility maintenance, forest management, and securing forest carbon sinks. Despite extensive research on site classification and evaluation, it remains unclear how to incorporate mixed variables (discrete and continuous) from climate, soil, geographical, and topographic factors into site classification and how to rank the classification effectively. Based on a large dataset from 16 162 sample plots throughout Jilin Province in Northeast China, we identified environmental variables (geography, topography, climate, and soil factors) that affect site form, which is an indicator of site quality, and classified plots as 10 site types using mixed-variables clustering via the expectation–maximization algorithm. Subsequently, these site types were ranked as site classes based on growth performance. A mixed-effects site form model was developed with dummy variables accounting for differences among six forest types (coniferous forest, hardwood broadleaved forest, softwood broadleaved forest, coniferous mixed forest, broadleaved mixed forest, and coniferous broadleaved mixed forest) and random components describing site classes. The model was utilized to evaluate the reasonability of site classification. The final site classes were determined by combining the nonlinear mixed-effects model with hierarchical agglomeration. We conclude that multifactorial mixed-variables clustering had a good performance, and the mixed-effects site form model effectively describes the differences among site classes and forest types. The results demonstrate that site classification, which integrates both environmental factors and growth data, achieves good performance. This study presents a novel and practical framework for site classification and site quality assessment, with a focus on mixed forests, providing valuable tools for forest management and planning to support tree species (mixture) selection, site management, and silviculture.