• 论文与研究报告 •

### 江西省不同立地等级的马尾松林生物量估计和不确定性度量

1. 中国林业科学研究院资源信息研究所 北京 100091
• 收稿日期:2016-03-04 修回日期:2016-04-22 出版日期:2017-08-25 发布日期:2017-09-27
• 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重点项目"基于轻小型无人机平台的多尺度森林生物量估计"（CAFYBB2016SZ003）；国家自然科学基金项目（31170588，31570628，31300534）。

### Biomass and Uncertainty Estimates of Pinus massoniana Forest for Different Site Classes in Jiangxi Province

Zhao Han, Lei Yuancai, Fu Liyong

1. Research Institute of Forest Resource Information Techniques, CAF Beijing 100091
• Received:2016-03-04 Revised:2016-04-22 Online:2017-08-25 Published:2017-09-27

Abstract: [Objective] To obtain the regional tree aboveground biomass and its uncertainty estimate on different site quality and choose the optimizational model for biomass estimation,this study presented a novel method to obtain more accurate estimates of forest biomass in the forest productivity estimation.[Method] The regional site quality classification in Pinus massoniana forests of Jiangxi Province was determined using the dominant tree height (H)-diameter at breast height (D) model. The aboveground biomass density and its root mean square error (RMSE) in each site class were estimated by the Monte Carolmethod based on the three allometric biomass models including (1) gi=aDib+ε,(2) gi=a(Di2Hi)b+ε,and (3) gi=aDibHic+ε,where gi is the individual biomass of the ith sample tree, Di and Hi are the diameter at breast height (DBH) and tree height for the ith sample tree, respectively; a,b and c are model parameters; ε is the error term.[Result] 1) The coefficient of determination (R2) obtained from the three biomass equations are more than 0.95, which indicated that the three equations have good fitting abilities. Among the candidate models, Model (3) showed the best performance. 2) The dominant H-D model showed a good fitting ability with R2=0.907, mean error (ME)=0.001, mean absolute error (MAE)=0.559, and RMSE=0.027. Plots classified by site quality were distributed to all the regions of Jiangxi Province and the sample plots in the same site level were relatively concentrated. 3) The simulation studies using Monte Carlo method were achieved stability by 10 000 times repeats. Aboveground biomass estimates calculating by the same individual tree biomass equation increased with increasing level of site. The middle site class level (the third level) represents the mean level of the regional site conditions and has similar biomass estimate with the whole region. Under the same site class, the order of mean aboveground biomass estimate values of the three models was the following:equation (1) > equation (3) > equation (2) and the order of both RMSE and relative RMSE estimates values was the following:equation (2) < equation (3) < equation (1).[Conclusion] 1) The equation (2) is better than equation (3) and then the equation (1) by comparing the relative RMSEs of the mean biomass density estimate. 2) The more similar the site quality is to the mean site quality level, the smaller the relative RMSE of the aboveground biomass density will be. 3) This study put forward a method to estimate the regional tree biomass and uncertainty in different site quality by combining the H-D model and the Monte Carlo simulation, and provides a probability and reference to accurately estimate the site productivity and biomass under different site quality.