Based on data from 56 plots of natural Larix gmelinii forest in Cuigang Forest Farm of Daxing'anling Mountains, we constructed the height to diameter ratio (HDR) model of L. gmelinii with exponential decay function as the base model. We used the tree classification as dummy variables and the method of reparameterization. The aim was to provide scientific evidence for evaluating the stability of different grades of L. gmelinii trees and stands in Daxing'anling Mountains. The results showed that except for diameter at breast height, the dominant height, dominant diameter, individual tree competition index all had significant correlations with the HDR. The involvement of these variables significantly improved the fitted accuracy of the generalized HDR model, with the adjustment coefficients, root mean square error and mean absolute error being 0.5130, 0.1703 m·cm-1 and 0.1281 m·cm-1, respectively. When the tree classification as a dummy variable was added to parameters 0 and 2 of the generalized model, the fitting effect of the model was further improved. The three above-mentioned statistics were 0.5171, 0.1696 m·cm-1 and 0.1277 m·cm-1, respectively. Through comparative analysis, the generalized HDR model with tree classification as dummy variable had the best fitting effect, which was superior to the basic model, and had higher prediction precision and adaptability.
基于大兴安岭地区翠岗林场兴安落叶松天然林的56块样地数据,以指数衰减函数为基础模型,利用再参数化的方法以林木分级为哑变量构建大兴安岭地区兴安落叶松的高径比模型,为大兴安岭地区兴安落叶松不同等级的树木和林分稳定性的评估提供科学依据。结果表明: 除胸径以外,优势树高、优势胸径、单木竞争指数与高径比有显著相关性,加入后明显提升模型的拟合精度,而且兴安落叶松高径比广义模型的调整系数、均方根误差和平均绝对误差分别为0.5130、0.1703 m·cm-1和0.1281 m·cm-1;将林木分级哑变量添加到广义模型的参数0和2上时,模型的拟合效果进一步提高,其高径比模型的3个统计量依次为0.5171、0.1696 m·cm-1和0.1277 m·cm-1。经过比较分析,以林木分级为哑变量的广义高径比模型拟合效果最好,不仅优于基础模型,且具有较好的预测精度和适应性。.
Keywords: dummy variable; height to diameter ratio; reparameterization; tree classification.