吉林省森工林区形高表编制  

Compilation of Stand Form Height Table at Jilin Forest Industry General Administration Area

作　　者：王鹤智 刘奇峰 朱万才 Wang Hezhi;Liu Qifeng;Zhu Wancai(Forest and Grassland Survey and Planning Institute,National Forestry and Grassland Administration,Beijing 100714;Heilongjiang Institute of Forestry Science,Harbin 150080)

机构地区：[1]国家林业和草原局林草调查规划院,北京100714 [2]黑龙江省林业科学研究所,哈尔滨150080

出　　处：《中国林副特产》2023年第6期1-6,共6页Forest By-product and Speciality in China

基　　金：国家重点研发项目子课题(2022YFD2201004-02)。

摘　　要：在林业中,林分形高模型被用于预测和控制林分的生长和产量,对于森林资源的规划和管理具有重要意义。研究基于27884个样地的资料,归类形成了18种林分类型。然后,选择了8种函数进行模型拟合,以调整后相关指数(Ra^(2))最大化和均方根误差(RMSE)最小化的原则,优选出不同林分类型的最适模型。通过交叉验证,采用了总相对误差(TRE)、平均系统误差(MSE)、平均预估误差(MPE)和平均百分标准误(MPSE)4个指标对最适模型进行了评估。最终建立了18个林分类型的最优林分形高模型,模型的TRE和MSE均在±0.1%内,MPE均小于3%,MPSE均小于12%。这项研究形成了吉林省重点国有林区具有代表性的形高模型系统,弥补了该地区林分形高建模方面的空缺。这些模型与各相应的样本点之间均有较好的契合度,具有较好的预估性,可在森林资源规划设计调查及连续清查中推广使用。此外,研究还进一步验证了以全部数据拟合模型、以“刀切法”验证模型的方法是可行的。这种方法能够在一定程度上避免模型过拟合和欠拟合的问题,提高模型的预测精度。In forestry,stand height models are used to predict and control the growth and yield of stands,playing a significant role in the planning and management of forest resources.Based on data from 27,884 sample plots,this study classified and formed 18 stand types.Then,eight functions were selected for model fitting,following the principle of maximizing the adjusted coefficient of determination(Ra 2)and minimizing the root mean square error(RMSE),to identify the optimal models for different stand types.Through cross-validation,four indicators-total relative error(TRE),mean systematic error(MSE),mean prediction error(MPE),and mean percentage standard error(MPSE)-were used to evaluate the optimal models.Ultimately,optimal stand height models were established for the 18 stand types,with TRE and MSE within±0.1%,and MPE and MPSE below 3%and 12%,respectively.This study developed a representative stand height model system for key state-owned forest areas in Jilin Province,filling the gap in stand height modeling in this region.These models exhibited good fit with their respective sample points,demonstrating good predictive ability,and can be widely applied in forest resource planning,design,surveying,and continuous inventory.Furthermore,the study further validated the feasibility of fitting models using all the data and verifying models using the"knife-edge"method.This approach helps to avoid overfitting and underfitting issues to some extent,improving the predictive accuracy of the models.

关 键 词：林分形高 重点国有林区 吉林 形高模型 

分 类 号：S758.51[农业科学—森林经理学]