Modelling site index of Chinese fir plantations using a random effects model across regional site types in Hunan province, China

Abstract

Site index is a popular method used for assessing site productivity. Based on sampling of 418 plots of Chinese fir (Cunninghamia lanceolata) across Hunan province, China, a site index model of this species in relation to site type as a random effect was developed. First, site factors (topography) influencing stand dominant height were screened as main factors through one-way ANOVA analysis. The results showed that elevation was the most significant factor, followed by soil type, aspect, and slope. Second, ten widely base models were developed, and found that these models performed poorly (R2 ranged from 0.3850 to 0.4977). Although the models performed poorly, the best base model (M4) was selected (R2 = 0.4977). Considering the influence of site factors on the site index curve, the different site factors and its combination as a random effect were simulated by nonlinear mixed-effect approach. The R2 values had improved from 0.3850 ∼ 0.4977 to 0.5132 ∼ 0.9018, and the model with combined site type (110 levels) performed best (R2 = 0.9018). K-means cluster method was used to cluster the 110 site types into 8 site type groups, and then a mixed-effects model with random effect of site type groups was developed and resulted in improved model performance (R2 = 0.9268). The results have practical utility and guidelines for regional level forest productivity estimation.