高山松单木地上生物量模型不确定性研究

以高山松为研究对象，基于113株单木地上生物量测量数据，在前期构建高山松一元和二元幂函数模型基础上，采用六步法和泰勒级数对模型的残差变异和参数误差进行不确定性分析。结果表明：基于113株高山松构建的单木生物量模型，一元回归模型的残差变异及模型参数引起的不确定性分别为6.20%、30.30%，综合不确定性约为30.83%；二元回归模型残差变异和模型引起的不确定性分别为5.17%、3.12%，综合不确定性约为6.04%。当建模样本量从38增加至71时，模型的不确定性变化显著。一元模型的参数不确定性和残差变异不确定分别减少14.50%和1.25%，综合不确定性减少14.52%；二元模型参数不确定性和残差变异不确定分别减少35.65%和5.80%，综合不确定性减少34.96%。随着分组样本数的增加，一元地上生物量模型和二元地上生物量模型，其模型的残差变异不确定性也逐渐减小。模型残差变异不确定性和参数不确定性是单木地上生物量模型构建中不确定性的主要来源，增加模型参量能有效降低单木地上生物量模型中参数引起的不确定性；建模样本数量的增加，能有效降低模型参数不确定性及残差变异不确定性。

Model Uncertainty Analysis of Aboveground Biomass Estimation of Pinus densata

In the study, the Pinus densata was taken as the research object, based on the measurement data of 113 individual-tree aboveground biomass. By constructing univariate and binary aboveground biomass models of P. densata, six-step method and Taylor series were used to analyze the uncertainty of the model's residual variation and parameter error. The data analysis shows that the uncertainty caused by univariate aboveground biomass model residual variation based on the biomass of 113 P. densata trees was about 6.20%, the parameter uncertainty was 30.30%, the comprehensive uncertainty was about 30.83%; the uncertainty caused by binary aboveground biomass model residual variation was about 5.17%, the parameter uncertainty was 3.12%, the comprehensive uncertainty was about 6.04%. When the modeling sample size increases from 38 to 71, the uncertainty of the model changes significantly. The parameters uncertainty and residual variation uncertainty of the unary model were respectively reduced 14.50% and 1.25%, the comprehensive uncertainty was reduced 14.52%; the parameters uncertainty and residual variation uncertainty of the binary model were reduced 35.65% and 5.80%, the comprehensive uncertainty was reduced 34.96%. With the increase in the number of grouped samples, the uncertainty of the residual variation of the model was gradually reduced. Model residual variation uncertainty and parameter uncertainty were the main sources of uncertainty in the construction of individual tree above-ground biomass models. Increasing model parameters was effectively reduced the uncertainty caused by parameters in individual above-ground biomass models; the increase in the number of samples were reduced the uncertainty of model parameters and the uncertainty of residual variation effectively.