近天然落叶松云冷杉林单木枯损模型研究

为了预测近天然落叶松云冷杉林的生长和发展，该文建立了近天然落叶松云冷杉林中长白落叶松、红松、云冷杉、中阔和慢阔5个树种（组）的单木枯损模型。数据来自于20个长期固定观测样地，共得到12 853个观测数据，其中80%的数据用于建模，20%的数据用于模型验证。采用Logistic回归方法，考虑多重共线性逐步筛选出对枯损有显著影响的自变量；根据预测结果敏感度和特异度的最大和原则(MST原则)确定模型的概率阈值。结果表明:①起初胸径、竞争指数、林分密度、多样性指数、立地因子都是显著影响树木枯损的因子，但不同树种的影响因子各异。②起初胸径在所有模型中均极显著；竞争指数除在红松和慢阔模型中未进入模型外，在其他模型中均显著；林分密度和多样性指数在长白落叶松和慢阔模型中显著；立地因子在长白落叶松模型中极显著。③除红松、中阔模型由于参与建模数据过少而无法通过χ2检验外，其余分树种模型都有良好的解释性和预测性。④模型诊断的AUC值在0.618～0.815之间。因此，建立的不同树种的单木枯损模型可为近天然落叶松云冷杉林的生长预测提供依据。 

Individual tree mortality models for semi-natural larch-spruce-fir forests in Jilin Province,northeastern China

Logistic regression models were developed to predict the probabilities of individual tree mortality for larch,spruce and fir,Korean pine and two deciduous groups in semi-natural larch-spruce-fir forests in Jilin Province,northeastern China.The data were originated from 20 permanent sample plots and 12 853 observation data were obtained,of which 80% were used for model development and 20% for model validation.Logistic stepwise regression models were preformed to select significant explicatory variables,with ...