欢迎访问林业科学,今天是

林业科学 ›› 2014, Vol. 50 ›› Issue (3): 69-75.doi: 10.11707/j.1001-7488.20140310

• 论文与研究报告 • 上一篇    下一篇

基于贝叶斯法估计杉木人工林树高生长模型

张雄清, 张建国, 段爱国   

  1. 中国林业科学研究院林业研究所 国家林业局林木培育重点实验室 北京 100091
  • 收稿日期:2013-05-21 修回日期:2013-08-29 出版日期:2014-03-25 发布日期:2014-04-16
  • 基金资助:

    中央级公益性科研院所中国林业科学研究院林业研究所所长基金(RIF2013-09);国家自然科学基金项目(31300537,31100476);江苏高校协同创新计划资助项目。

Tree-Height Growth Model for Chinese Fir Plantation Based on Bayesian Method

Zhang Xiongqing, Zhang Jianguo, Duan Aiguo   

  1. Key Laboratory of Tree Breeding and Cultivation of State Forestry Administration Research Institute of Forestry, CAF Beijing 100091
  • Received:2013-05-21 Revised:2013-08-29 Online:2014-03-25 Published:2014-04-16
  • Contact: 张建国

摘要:

以江西杉木密度试验林为例,分别基于贝叶斯法和传统法(非线性最小二乘法)估计杉木人工林树高生长模型,并在贝叶斯法中考虑无信息先验分布和有信息先验分布。结果表明:利用贝叶斯法估计杉木人工林树高生长模型,预测值的可靠性比传统法好,而且基于有信息先验分布估计杉木人工林树高生长模型要略好于无信息先验分布。这是因为利用生长模型预测杉木人工林树高生长存在着一定的不确定性,使得利用传统的估计方法分析杉木人工林生长模型稳定性比较低,可靠性也相对较差。贝叶斯法综合利用了先验信息和样本信息,而传统法仅利用了样本信息,而且贝叶斯法把模型参数看作是随机变量,更能反映杉木人工林树高生长的本质,预测杉木人工林树高的可靠性更好,而传统法把模型参数看作固定值。研究结果为杉木人工林生长模型的估计提供一种新的思路。

关键词: 贝叶斯法, 有信息先验分布, 无信息先验分布, 树高生长, 杉木人工林

Abstract:

Chinese fir (Cunninghamia lanceolata), a special tree species in China, is one of the most important fast-growing tree species for timber production in southern China. Tree height is an important variable, not only reflecting the site index, but also estimating tree volume, and biomass. It is critical for exploring height-growth law of Chinese fir to develop tree-height growth model. Based on the periodic data of the Chinese fir in Jiangxi Province, tree-height growth model was developed. Bayesian method and classical method (nonlinear least squares method) were used to estimate the height growth mode. In the Bayesian framework, non-informative prior and informative were also introduced. Results showed that the model reliability using Bayesian method was better than classical method, and the informative prior was slightly better than non-informative prior. That is because that the uncertainty of tree-height growth results in low reliability using classical method. In contrast, relevant prior knowledge about the data can be incorporated into Bayesian analyses whereas classical methods ignore the relevant prior knowledge, and the parameters using Bayesian method are treated as random variables, which is a very different assumption from that of classical method, which treats parameters as fixed values. It provides a new method for estimating forest growth model of Chinese fir plantation.

Key words: Bayesian method, informative prior, non-informative prior, tree-height growth, Chinese fir plantation

中图分类号: