欢迎访问林业科学,今天是

林业科学 ›› 2014, Vol. 50 ›› Issue (6): 42-54.doi: 10.11707/j.1001-7488.20140606

• 论文与研究报告 • 上一篇    下一篇

几种相容性生物量模型及估计方法的比较

符利勇1, 雷渊才1, 曾伟生2   

  1. 1. 中国林业科学研究院资源信息研究所 北京 100091;
    2. 国家林业局调查规划设计院 北京 100714
  • 收稿日期:2013-06-19 修回日期:2013-08-29 出版日期:2014-06-25 发布日期:2014-07-07
  • 基金资助:

    国家自然科学基金项目(31300534,31170588);国家863重点项目(2012AA12A306)。

Comparison of Several Compatible Biomass Models and Estimation Approaches

Fu Liyong1, Lei Yuancai1, Zeng Weisheng2   

  1. 1. Research Institute of Forest Resources Information Techniques, CAF Beijing 100091;
    2. Academy of Forest Inventory and Planning, State Forestry Administration Beijing 100714
  • Received:2013-06-19 Revised:2013-08-29 Online:2014-06-25 Published:2014-07-07
  • Contact: 雷渊才

摘要:

以南方150株马尾松地上生物量数据为例,在考虑林分起源和未考虑林分起源2种情形下,对非线性似然无关回归法、比例平差法和非线性联立方程组法3种方法进行综合比较研究。根据分配层次不同,比例平差法和非线性联立方程组法将进一步考虑总量直接控制和分级联合控制2种方案。从直径、树高、地径、年龄、枝下高和冠幅6个林分变量中选取不同的变量构建一元、二元和三元生物量模型,利用加权最小二乘回归法消除生物量模型中存在的异方差性。基于独立形式的一元、二元和三元模型,利用非线性似然无关回归法、比例平差法和非线性联立方程组法构建相应的相容性生物量模型。结果表明:在考虑和未考虑林分起源情形下,3种估计方法都能有效保证各分量生物量总和等于总生物量,预测精度较高。总体而言,非线性联立方程组法预测精度更高、稳定性更强,其次是非线性似然无关回归法,最差的是非线性比例平差法;根据建模数据和检验数据综合比较得知,在考虑和未考虑林分起源情形下,总量控制联立方程组法对应的二元相容性生物量模型预测精度最高。

关键词: 非线性似然无关回归法, 比例平差法, 非线性联立方程组法, 相容性, 生物量模型

Abstract:

So far, the approaches of nonlinear seemingly unrelated regression (NSUR), nonlinear adjustment in proportion (NAP) and nonlinear simultaneous equations (NSE) have been proposed to establish compatible biomass models. However, to our knowledge, systemic comparison of these methods is not studied. Therefore, the three methods NSUR, NAP and NSE were compared based on predictive accuracy using 150 masson pine (Pinus massoniana) biomass data in this study. Two alternative approaches, controlling jointly from level to level by ratio functions and controlling directly under total biomass by proportion functions were considered for the two approaches of NAP and NSE. Six candidate tree variables of diameter at breast height, tree height, ground diameter, age, under branch height and crown width were evaluated for their contribution to biomass models improvement. Single variable, bivariate and multivariate (three variables) biomass models were established based on the first three of the most significant tree characteristics. Heteroskedasticity in the biomass models was removed by weighted least square regression. Compatible biomass models were established and estimated based on single variable, bivariate and multivariate using NSUR, NAP and NSE. The results showed that the three analyzed methods could ensure efficiently that components of biomass added up to the total biomass with high prediction accuracy. However, overall, NSE had the highest prediction accuracy and most stable, following by NSUR, and NAP was the worst. For balancing the model prediction accuracy and survey cost, the NSE of controlling directly under total biomass with diameter at breast height and height as stand variables was proposed to construct compatible biomass model at considering or no considering stand origin situation based on the systemic comparison of modelling and validation data sets.

Key words: nonlinear seemingly unrelated regression, adjustment in proportion, nonlinear simultaneous equations, compatible, biomass model

中图分类号: