福建木荷多元生物量模型及2种权函数的比较

Multi-Biomass Model of Schima superba and Comparison of Two Weight Functions

为准确计量与监测木荷生物量以及准确评估其碳汇能力、生态效益等生态功能,基于160株木荷样木实测数据,以胸径(D)、树高(H)、冠幅(Cw)和冠长(Cl)作为模型的自变量,运用非线性最小二乘法,采用15种模型结构建立木荷各组分生物量模型,并以1/f(x)^2与1/f(x)^1.6分别作为权函数对模型进行异方差的消除,对比分析各模型拟合结果并选取各组分最优生物量模型。结果表明,木荷各部分的生物量模型采用同一模型结构所拟合的效果大致相同;各自变量对生物量模型的拟合效果与贡献程度从大到小顺序为D>H>Cl>Cw(其中自变量H与Cl的作用效果相近);随着函数模型的多元化,从一元到二元模型的提升效果明显,而后二元到多元模型提升效果不大,建议实际应用中采用二元模型W=a D b1 H b2即可;采用1/f(x)^1.6作为权函数消除异方差后模型整体的拟合效果与估计精度,优于未消除异方差与以1/f(x)^2作为权函数消除异方差的模型,证明以1/f(x)n作为通用权函数将更适用,但其具体n值需进一步研究。

In order to accurately measure and monitor the biomass of Schima superba and accurately assess its ecological functions such as carbon-sinking capacity and ecological benefits,by the measuring data of 160 sample trees,the biomass model Schima superba was established based on nonlinear least squares method with 15 kinds of model structure,and the DBH(D),tree height(H),crown width(Cw)and crown length(Cl)were used as the variables.The heteroscedasticity of model was eliminated by using 1/f(x)^2 and 1/f(x)^1.6 as weight functions respectively,and the optimal biomass model was selected after comparing and analyzing each fitting.The results showed that the biomass model fitting effect of each component is approximately the equivalence by using the same model structure.The order of the model fitting effect and contribution degree of each variable to each component biomass model is from large to small,D>H>Cl>Cw(where the variable H and Cl have similar effects).With the more variables of the functional mode,the model fitting effect is gradually improve,the improvement effect from the unary function to the binary function model is the obvious,and the effect of the binary function to the function of several variables model is not significant.Therefore,the binary model W=a D b1 H b2 in practical applications was recommanded.The overall fitting effect and estimation accuracy of the model,which is used 1/f(x)^1.6 as the weight function to eliminate the heteroscedasticity,are better than the model unresolved heteroscedasticity or using 1/f(x)^2 as the weight function.This means that 1/f(x)n,as a general weight function,is more applicable,but further research need be done for the specific value of n.