应用贝叶斯模型平均法建立东北云冷杉针阔混交林蓄积生长模型  

Stand Volume Growth Model of Spruce-fir Mixed Stands in Northeast China Using Bayesian Model Averaging Approach

作　　者：贾勃 王新杰[1] Jia Bo;Wang Xinjie(Beijing Forestry University,Beijing 100083,P.R.China)

机构地区：[1]北京林业大学,北京100083

出　　处：《东北林业大学学报》2023年第1期1-5,共5页Journal of Northeast Forestry University

基　　金：国家重点研发项目(2017YFC050410101)。

摘　　要：根据金沟岭林场198块固定样地数据,分别使用逐步回归方法和贝叶斯模型平均法建立东北云冷杉针阔混交林蓄积量生长模型,分析林分蓄积生长与林分因子、气候因子的关系,并比较两种方法的预测效果。结果表明:贝叶斯模型平均法建模时将多个模型组合进行加权建模,减少了冗余变量,建立的模型较逐步回归更为准确;海拔、林分断面积、林分密度、林分平均胸径、低于0℃积温、低于18℃积温、参考蒸发量是影响林分蓄积生长的主要因素;逐步回归的决定系数(R^(2))为0.95,均方根误差(R_(MSE))为17.53;贝叶斯模型平均法的决定系数(R^(2))也为0.95,均方根误差(R_(MSE))为37.51。With the data of 198 fixed plots of Jingouling Forest Farm,the volume growth model of spruce-fir coniferous and broad-leaved mixed forests in northeast China were established by using stepwise regression method and Bayesian model average method,respectively.We explored the relationship between stand volume and stand factors,climate factors,and the compared the prediction effect of the two methods.The results showed that Bayesian model average method combined multiple models for weighted modeling,which reduced redundant variables,the model established by Bayesian model average method was more accurate than stepwise regression.The altitude,stand basal area,stand density,stand average diameter,degrees-days below 0℃,degrees-days below 18℃,and reference evaporation were the main factors which affect the growth of stand volume.The R^(2) of stepwise regression was 0.95 and R_(MSE) was 17.53.The R^(2) of Bayesian average model was also 0.95,and the R_(MSE) was 37.51.

关 键 词：贝叶斯模型平均法 针阔混交林 林分蓄积量 先验信息 

分 类 号：S758.51[农业科学—森林经理学]