林分径级分布预测模型参数的贝叶斯法估计--以金沟岭林场样地数据为例  

Bayesian Method in Estimating Model Parameters to Predict Diameter Class Distribution of Stand-Taking Sample Plot Data in Jingouling Forest Farm as Example

作　　者：于汝川 张青[1] 岳朝方 亢新刚[1] 

机构地区：[1]北京林业大学,北京100083 [2]德国巴登-符腾堡州林业科学研究院

出　　处：《东北林业大学学报》2017年第6期30-35,共6页Journal of Northeast Forestry University

基　　金：“948”国家林业局引进项目(2013-4-66)

摘　　要：以吉林省金沟岭林场云冷杉针阔混交异龄林26块检查法样地的5次观测数据为基础,建立转移矩阵模型对一定周期的林分径级分布进行预测。分别利用经典统计学方法和贝叶斯方法对转移矩阵模型的参数进行估计,建立了固定参数的矩阵模型和贝叶斯矩阵模型,并对两种模型的预测效果进行对比。结果显示,固定参数的转移矩阵模型对林分径级分布的预测值比实际值偏高,贝叶斯模型的预测结果更接近林分的实际径级分布,证明了贝叶斯参数估计方法比固定参数平均的方法所建模型具有更高的预测精度。With the observed data in five different years from 26 sample plots in Jingouling Forest Farm of Wangqing Forestry Bureau,Jilin Province,transition matrix models were established to predict the stand diameter class distribution in a given period.Parameters of the models were estimated by classical statistical method and Bayesian method,respectively.After the fixed parameter model and the Bayesian model were built,their prediction results could be compared to test the prediction accuracy of the two models.The predicted values of fixed parameter model tend to be higher than the true value.The prediction of Bayesian model is relatively more accurate than that of fixed parameter model.

关 键 词：转移矩阵模型 林分径级分布 贝叶斯统计 MCMC方法 Gibbs抽样算法 

分 类 号：S758.5[农业科学—森林经理学]