2种云南松天然次生林林分生长模型构建方法的比较  

Comparison of Two Stand Growth Modeling Methods of Pinus yunnanensis Secondary Forest

作　　者：陈建珍[1] 何超[2] 许彦红[1] 

机构地区：[1]西南林业大学林业调查规划设计研究院,云南昆明650224 [2]西南林业大学林业3S技术工程研究中心,云南昆明650224

出　　处：《西部林业科学》2015年第3期112-116,共5页Journal of West China Forestry Science

基　　金：西南林业大学校级科研专项(111131);云南省教育厅科研专项(2010Y300)

摘　　要：为更有效地经营管理云南松天然次生林,在云南省云南松主要分布区的昆明市、楚雄州设置55块云南松天然次生林标准地,从中随机抽取33块作为建模数据,剩余22块作为模型校验数据,采用逐步回归剔除法和BP神经网络建模法,建立林分生长模型。结果表明,使用逐步回归剔除法建立的平均胸径生长模型、平均树高生长模型、蓄积量生长模型总体拟合精度在89.22%-95.52%之间,校验精度在81.09%-94.15%之间;使用BP神经网络建立的各林分生长模型总体拟合精度在90.01%-98.62%之间,校验精度在92.63%-95.68%之间。可见BP神经网络所建立的林分生长模型精度较高,可为同类森林的经营管理提供参考。In order to more effectively manage Pinus yunnanensis secondary forest, 55 sample plots in secondary forest were set up in main distribution area of P. yunnanensis in Kunming and Chuxiong, of which 33 sample plots were randomly selected as the modeling data and the rest acted as the adaptability testing data. The growth models were established by using stepwise regression removal method and BP neural network. The result showed that the fit- ting precision of growth models established by stepwise regression removal method ranged from 89. 22 % to 95. 52 %, while test accuracy from 81.09 % to 94. 15 %. The fitting precision of growth models established by BP neural network was between 90. 01% and 98.62 % , and the test accuracy between 92. 63 % and 95.68 %. It was clear that the growth model established by BP neural network had higher fitting precision which could provide reference for management of similar P. yunnanensis forest.

关 键 词：森林经理 云南松 逐步回归剔除法 BP神经网络 林分生长模型 

分 类 号：S758.5[农业科学—森林经理学] S791.257[农业科学—林学]