我国森林生物量和碳储量多层次联立模型系统
Multi-level simultaneous models of forest biomass and carbon storage in China
[目的]森林生物量和碳储量模型，是森林资源调查监测评价的重要计量工具。研建我国各种森林类型的材积源生物量和碳储量模型，既为估计国家尺度的森林碳储量变化提供科学依据，也为估计区域和全球尺度的森林碳储量及其碳汇潜力提供重要参考。[方法]基于我国第九次森林资源清查的 52700个固定样地实测数据，依据 35 个树种组的立木生物量模型和含碳系数计算出每公顷生物量和碳储量，再综合利用加权回归、哑变量建模和误差变量联立方程组方法，按 1 个总体、3 个大类(第 1 层)、20 个类型(第 2 层)、74 个亚类(第 3 层)建立了基于材积的森林生物量碳储量多层次联立模型系统，并对 3 个层次模型的评价指标进行了分析，还与已发表的相关模型进行了对比。[结果]结果表明森林生物量和碳储量模型的评价指标相差很小，总体平均模型及第 1、2、3 层哑变量模型的确定系数R2 分别在 0.78、0.85、0.92、0.95以上，平均预估误差MPE分别为0.42%、0.34%、0.24%和0.19%，平均百分标准误差MPSE分别在22%、21%、15%和12%左右，总体相对误差TRE均接近于0.所建 20 个类型的森林生物量和碳储量模型，其R2 都在0.87以上，MPE均在3%以内;所建 74个亚类的模型，其R2 都在 0.85 以上，MPE基本在5%以内。已发表的3组生物量模型的估计结果存在较大的不确定性，TRE的变动范围达到-20%～74%，但会随着建模样本量的增加而逐渐改进。[结论]对森林生物量和碳储量总量的估计，总体平均模型的R2 最低，第 1、2、3 层哑变量模型的R2依次逐渐提高，MPSE依次逐渐降低;但TRE和MPE差异不大。所建我国森林生物量和碳储量多层次联立模型系统，为准确估计国家和区域尺度的森林生物量和碳储量现状及变化提供了计量依据。为了提高森林生物量和碳储量估计结果的准确性，应该尽可能应用第3层次的模型。
[Objective]Forest biomass and carbon storage models are important tools for forest resource inventory,monitoring and assessment.Developing forest biomass and carbon storage models for the forest types in China not only provide a scientific basis for estimating forest carbon storage changes at the national level,but also provide an important reference for estimating forest carbon storage and carbon sequestration potential at regional and global scales.[Method]Based on the measured data of 52700 permanent plots from the 9th national forest inventory of China,the biomass and carbon storage per hectare were calculated according to the tree biomass models and carbon factors of 35 tree species groups.Then,the volume-based forest biomass and carbon storage models were established for one population,3 categories(level I),20 forest types(level II),74 sub-types(level III)by using weighted regression,dummy variable modeling and error-in-variable simultaneous equations approach.Additionally,the evaluation statistics of three-level models were analyzed,and comparison with previously published models was conducted.[Results]The results showed that evaluation statistics of forest biomass and carbon storage models are only slightly different.The determination coefficients(R2)of population average model,and dummy models of level I,II,and III are more than 0.78,0.85,0.92,and 0.95 respectively,the mean percent standard errors(MPSEs)are about 22%,21%,15%and 12%respectively,the mean prediction errors(MPEs)are also decreased gradually and all less than 0.5%,and the total relative errors(TREs)are all close to 0.The R2 of biomass and carbon models for 20 forest types are more than 0.87,and the MPEs are fewer than 3%;and those for 74 sub-types are 0.85 and 5%,respectively.The estimated results from three sets of biomass models published before had great uncertainty,the TREs ranging from-20%to 74%.But the models would gradually improve with the increase of modeling sample size.[Conclusion]For the estimation of total forest biomass and carbon storage,the R2 of population average model is lowest,and the R2 of dummy models of level I,II,and III are increased gradually while the MPSEs are decreased,and the MPEs and TREs are not significantly different.The multi-level simultaneous models of forest biomass and carbon storage for all forest types in China developed in this study would provide a quantitative basis for accurately estimating status and changes of forest carbon storage at national and regional scales.In order to improve the accuracy of forest biomass and carbon stock estimation results,the models of level III should be applied whenever possible.
曾伟生;邹文涛;陈新云;杨学云;
国家林业和草原局林草调查规划院,北京 100714;中国林科院科信所,北京 100091;国家林业和草原局林草调查规划院,北京 100714;国家林业和草原局林草调查规划院,北京 100714;
biomass conversion factor simultaneous equations error-in-variable dummy variable carbon storage
1314-1327 / 14
评论