• 研究简报 •

### 分段削度方程2种估计方法比较

1. 1. 中国林业科学研究院资源信息研究所, 北京 100091;
2. 中国林业科学研究院热带林业实验中心, 凭祥 532600
• 收稿日期:2015-02-09 修回日期:2015-10-08 出版日期:2015-12-25 发布日期:2015-12-29
• 基金资助:
中央级公益性科研院所基本科研业务费专项资金课题"目标树单株木管理和收获预估研究"(IFRIT2013);公益性科研院所基本科研专项"中德合作多功能森林抚育经营创新技术研究"(CAFYBB2012013&Lin2Value);河南省省院合作项目"河南省主要人工用材林木材造材最优方案"(122106000051)。

### Comparison of Two Parameters Estimation Methods for Segmented Taper Equations

Pang Lifeng1, Jia Hongyan2, Lu Yuanchang1, Niu Changhai2, Fu Liyong1

1. 1. Research Institute of Forest Resources Information Techniques, CAF Beijing 100091;
2. Experimental Center of Tropical Forestry, CAF Pingxiang 532600
• Received:2015-02-09 Revised:2015-10-08 Online:2015-12-25 Published:2015-12-29

Abstract: [Objective] Taper equation is a key tool to describe stem form variations. By far, different taper equations have been proposed in the world. Among them, the segmented taper equation is one of the most commonly used equation.The method of ordinary least squares regression (OLS) is commonly used to estimate the parameters in the model. However, in practical terms, the application of the segmented taper equation was restricted because the estimated parameters of inflection points a1 and a2 (the relative tree height with numerical region of 0 to 1) obtained by OLS regression do not ensure to fall into the region of 0 to 1. Based on the above issue, the aim of this research is to find out the optimal fitting method by using TS(the two-factor automatic selection algorithm)and OLS, which provides technical supports for the construction of tree curve model and fine material.[Method]The three rare species such as Erythrophleum fordii, Castanopsis hystrix and Tectona grandis were developed using 120 individual data (40 for each species), the segmented taper equation was constructed for each species using OLS and TS, respectively. They were evaluated and compared based on the indexes of the coefficient of determination, the residual sum of squares, the mean prediction error, the variance of prediction errors and the root mean square error.[Result]The results show that both the OLS regression and TS fitting accuracy are preferred more than 95% and their coefficient of determination, residual sum of squares are the same. For the variance of prediction errors and the root mean square error TS is the best for E.fordii, C.hystrix and T.grandis; for the mean prediction error OLS method is the best for E.fordii,but TS is the best for C.hystrix and T.grandis, however, the difference between the two methods corresponding the mean prediction error is small; stem profiles prediction results are also very similar; TS can ensure that the inflection point of the estimated parameters is in the (0,1),and it is simple, fitting result is stable.[Conclusion] TS could explained theoretically the optimal value of segmented taper equation parameters a1 and a2. Therefore, it was recommended to fit the segmented taper equations with TS.