基于混合效应的湖南省栎林生长收获模型  

Growth and yield model of oak forest based on mixed effect

作　　者：刘洵 曾思齐[1] 贺东北 龙时胜 向博文 LIU Xun;ZENG Siqi;HE Dongbei;LONG Shisheng;XIANG Bowen(College of Forestry,Central South University of Forestry and Technology,Changsha,Hunan 410004,China;Central SouthForest Inventory and Planning Institute,National Forestry and Grassland Administration,Changsha,Hunan 410014,China)

机构地区：[1]中南林业科技大学林学院,湖南长沙410004 [2]国家林业和草原局中南林业调查规划设计院,湖南长沙410014

出　　处：《森林与环境学报》2019年第5期475-482,共8页Journal of Forest and Environment

基　　金：国家林业公益性行业科研专项“闽楠、青冈栎次生林提质增量关键技术研究与示范”(201504301);中南林业科技大学研究生创新基金项目(20181001)

摘　　要：为准确预测栎类林分的生长动态,建立预估精度较高的林分生长收获混合效应模型,以湖南省国家森林资源连续清查固定样地中的96块栎类次生林为研究对象,选用相容性林分生长收获模型作为基础模型,加入样地层次的混合效应,构建基于混合效应的相容性林分生长收获模型,并将其与加权二步最小二乘法进行比较。结果表明:(1)选择b 3作为期初林分蓄积量模型的混合效应参数,a 2作为期末林分断面积模型的混合效应参数,幂函数作为异方差结构和AR(1)结构矩阵作为自相关结构时的模拟效果最好;(2)利用平均误差、平均绝对相对误差、均方根误差(RMSE)以及预测精度4个指标评价模型的拟合效果,结果均显示混合效应模型方法优于加权二步最小二乘法的值,且混合效应模型方法的残差分布图分布均匀,分布范围也大大减小,说明基于样地水平的混合效应模型方法在预测精度方面要明显高于加权二步最小二乘法,能够更加准确地预测栎类林分未来的生长动态,为栎类林分的抚育经营提供理论依据。In order to accurately predict the growth dynamics of oak stand,a mixed effect model of stand growth and yield with high prediction accuracy was established.Ninety six secondary oak forests in the fixed sample plots of Human National Forest Resources Continuous Inventory were taken into account as research objects.The compatibility stand growth and yield model was selected as the basic model and combined with the mixed effect of sample plot level.The model growth and yield model based on the mixed effect was,afterwards,compared with the weighted two-step least squares method.The results show:(1)by selecting model variable of b 3 as the mixed effect parameter of the initial stand stock volume model,model variable of a 2 as the mixed effect parameter of future basal area model,power function as the heteroscedastic structure and AR(1)structure matrix as the autocorrelation structure,the best simulation results could be obtained;(2)four indexes(average error,mean absolute relative error,root mean square error and prediction accuracy),were used to evaluate the predictability of the model,the results show that the model combined with the mixed effect is more accurate than the weighted two-step least squares method,the residual distribution obtained from this model is uniform and the distribution range of its predictions is greatly reduced.Furthermore,the new model′s prediction accuracy is significantly higher than the weighted two-step least square method,and,therefore,can more accurately predict the future growth dynamics of the oak stand.Moreover,it provides a theoretical basis for the tending and management of oak stands.

关 键 词：栎类次生林 生长收获模型 联立方程组 混合效应模型 湖南省 

分 类 号：S757.2[农业科学—森林经理学]