天然林保护的神经网络分级规划  

Neural network grade program of natural forest protection.

作　　者：罗传文[1] 陈焱[1] 胡海清[1] 沈海龙[1] 范少辉[2] 

机构地区：[1]东北林业大学,哈尔滨150040 [2]中国林业科学院林业研究所,北京100091

出　　处：《应用生态学报》2005年第6期1002-1006,共5页Chinese Journal of Applied Ecology

基　　金：国家"十五"科技攻关资助项目(2001BA510B08).

摘　　要：总结了用神经网络进行天然林保护规划的实施步骤,阐述了规划举例、小斑标识归一、小斑标识还原、倾斜因子等概念.应用Arc/Info地理信息系统对帽儿山国家森林公园的树种组成多样性、树种稀有性、受干扰性、沟系保护特性和分类经营等因子进行了描述,并分析了天然林保护分级与这些因子之间的关系,并将这些因子作为神经网络的输入,经过人为确定训练样本,最终建立了规划神经网络.用研究区内的全部小斑检验了网络的泛化性能,结果较为满意.规划结果表明,神经网络兼顾了分类经营规划的成果,并保护了各种森林群落类型,对生态环境保护有所体现;网络中激励函数没有严重饱和,体现了倾斜因子对神经网络优化的引导作用.In this paper,the implement steps of natural forest protection program grading (NFPPG) with neural network (NN) were summarized,and the concepts of program illustration,patch sign unification and regress,and inclining factor were set forth.Employing Arc/Info GIS,the tree species diversity and rarity,disturbance degree,protection of channel system,and classification management in Moershan National Forest Park were described,and,used as the input factors of NN,the relationships between NFPPG and above factors were analyzed.Through artificially determining training samples,the NFFPG of Moershan National Forest Park was built.Tested with all patches in the park,the generalization of NFFPG was satisfied.NFPPG took both the classification management and the protection of forest community types into account,as well as the ecological environments.The excitation function of NFPPG was not seriously saturated,indicating the leading effect of inclining factor on the network optimization.

关 键 词：神经网络分级规划 天然林保护 规划举例 倾斜因子 

分 类 号：S757.4[农业科学—森林经理学]