关键词: 广义可加模型, 土壤有机质, 定性因子, 定量因子, 交互作用

Abstract: Understanding the influencing factors of forest soil organic matter can provide reference for forest management and quality improvement. In this study, we used Kriging interpolation method to analyze the spatial distribution of soil organic matter in forests of Guangdong Province based on 1182 sampling plots. With 52 factors of geography, topography, stand, climate, soil physics, and forest management as explanatory variables, and soil organic matter as the response variable, the factors that significantly influenced soil organic matter were selected by using the generalized additive model (GAM). The qualitative and quantitative factors with significant effects were interacted respectively to obtain the interaction terms. The results showed that: (1) The distribution of forest soil organic matter showed spatial heterogeneity, characterized by high content in the northwest areas, moderate in the west-central areas, and low in the southern coastal areas. (2) GAM could better fit the nonlinear relationship between the response variable and explanatory variables. The factors with extremely significant effects on soil organic matter included slope level, dominant tree species, soil water content, non-capillary water storage capacity, and non-capillary porosity capacity, while the factors with significant effects were longitude, slope direction, altitude, number of species, humus layer thickness, isothermality, annual minimum temperature and soil thickness. The modified determination coefficient and the explanation rate of variance of the single-factor GAM were 0.502 and 59.7%, respectively. The interactions between three qualitative factors (dominant tree species, slope level, and slope direction) and eight quantitative factors (longitude, humus layer thickness, isothermality, annual minimum temperature, soil water content, non-capillary water storage capacity, non-capillary porosity capacity, and soil thickness) had significant effects on soil organic matter. The modified determination coefficient and the explanation rate of variance were improved to 0.515 and 59.8% by adding the interactive items. Our results indicated that forest soil organic matter in Guangdong Province was mainly affected by geography, topography, stand, climate and soil physical factors, among which topography and soil physical factors had great effects. Our results can provide reference for forest quality improvement and rational forest management in Guangdong Province.

Key words: generalized additive model, soil organic matter, qualitative factor, quantitative factor, interaction.