长白落叶松人工林单木冠幅模型

以黑龙江省伊春市朗乡林业局的49块样地共2 499棵长白落叶松为研究对象,基于区组和样地嵌套二水平构建非线性混合效应冠幅模型。从10种常用的冠幅—胸径模型中选出一种最好的作为建模的基础模型,从其他林分因子中选取与冠幅相关性较高的作为协变量参与建模。结果表明:与基础模型相比,混合模型的赤池信息准则(AIC)、贝叶斯准则(BIC)、负二倍的最大似然值(-2LL)和剩余均方根误差(RMSE)值均降低,调整决定系数(Radj2)提高,冠幅预测值的残差分布和密度分布相对更集中,说明模型的预测精度和拟合能力均有所提高。研究发现,以逻辑斯蒂模型作为基础模型,将树高、冠长率和高径比作为协变量,考虑区组和样地嵌套二水平的混合模型有利于提高冠幅模型的预测精度,从而有助于更好地模拟长白落叶松的冠幅胸径关系。

Individual Tree Crown Width Prediction Models for Larix olgensis Plantation

Taking 2 499 Larix olgensis trees from 49 sample plots in Langxiang Forestry Bureau, Yichun City, Heilongjiang Province as the research objects, nonlinear mixed effect crown-width model was constructed based on two levels of block formation and sample plots. One model had been chosen as the best basic model from 10 common crown width-diameter models. A few other stand factors that were relevant with crown width had been put into basic model as covariates. Compared with the basic model, the AIC, BIC,-2LL and RMSE values of the mixed model all decreased, the Radj2 increased, and the residual distribution and density distribution of predictions were relatively more concentrated, indicating that the prediction accuracy and fitting ability of the model had been improved. It was found that taking logistic model as the basic model, taking tree height, crown length ratio and height to diameter ratio as covariates, and considering the mixed model of two levels of block information and sample plots, it is beneficial to improve the prediction accuracy of crown width model, therefore, it is helpful to better simulate the relationship between crown-width and DBH of L. olgensis.