• 论文及研究报告 •

### 相容性生物量模型的建立及其估计方法研究

1. 中国林业科学研究院资源信息研究所,北京100091
• 收稿日期:1998-10-21 修回日期:1900-01-01 出版日期:2001-01-25 发布日期:2001-01-25

### STUDY ON ESTABLISH AND ESTIMATE METHOD OF COMPATIBLE BIOMASS MODEL

Tang Shouzheng,Zhang Huiru,Xu Hui

1. The Research Institute of Forest Resource Information Techniques, CAF Beijing100091
• Received:1998-10-21 Revised:1900-01-01 Online:2001-01-25 Published:2001-01-25

Abstract:

Forest biomass is a basic quantity character of the forest ecological system. Biomass data are foundation of researching many forestry and ecology problems, thus accurate measurement of biomass is very important. Establishing biomass models is a major way to biomass estimation. There were a serious shortcomings in the models established previously, i.e.the results were incompatible for models of each component, in other words, the sum of estimated biomass of wood, bark, branches and foliage was unequal to estimated biomass of total aboveground, the sum of estimated biomass of wood and bark was unequal to estimated biomass of stem, the sum of estimated biomass of branches and foliage was unequal to estimated biomass of crown. There fore how to obtain the compatibility is stile a difficult problem for biomass estimate. A new method, nonlinear joint estimate, was proposed in this paper, and compared with method of adjustment in proportion. To the different methods of establishing models, five alternative methods were designed, then one of them was determined as a optimum method through the analysis and comparison. The optimum method took stem as a basis component and adopted two steps joint estimate, structure of models was as follows; the first step, total aboveground\;W1=f2(x)+f5(x), stem W2=f2(x), crown W5=f5(x); the second step, wood W3=f3(x), bark W4=f2(x)-f3(x); branch W6=f6(x), foliage W7=f5(x)-f6(x)In this paper, the progressive variable selection method was used to select models structure, and weighted least squares method was used to estimate parameters for reducing errors of non-homogeneous variance. At meantime, the paper used five indices to evaluate models, they were coefficient of variation for parameters C%, total relative error RS%,average relative error EE%,average absolute value of relative error RMA% and prediction precision P%. All of researches shown above in this paper took Larix olgensis as an example.