基于3种常用的理论生长模型,利用代数差分法和广义代数差分法推导8个差分型地位指数模型。模型拟合采用加拿大哥伦比亚省美国黄松林分79个临时样地的纵向剖面解析木数据和分类变量回归的参数估计方法,模型筛选采用统计指标和图形分析相结合的方法,并着重突出了图形分析的重要性;同时,本文对于最佳模型的应用也进行了较为全面的分析研究。结果表明: 1) 广义差分型地位指数模型具有更高的预测性能; 2) 基于Hossfeld IV生长模型的广义差分型地位指数模型表现出了最佳的性能,推荐采用该模型对加拿大哥伦比亚省30~200年的美国黄松林分进行树高生长预测和立地质量分类; 3) 依据预测的相对误差,确定100年为最佳指数年龄。
Based upon 3 widely used base models, a total of 8 ADA/GADA site index models were derived. The data for these models in this study were obtained from 79 pith-split stem analysis plots and the estimation method was "indicator variable approach". We used both fit statistics and visual analysis to select the best-fit model, and attached more importance to the visual analysis. A comprehensive application analysis was also given to the selected model. The results showed: 1) GADA outperformed ADA with respect to predictions. 2) A GADA model derived from Hossfeld IV presented the best prediction ability. It was suggested that the model be used to predict dominant height and to estimate site index for ponderosa pine stands ranging 30-200 years in British Columbia, Canada. 3) The best site index age was age of 100 years, based upon relative errors of predictions.