湖南省杉木二元立木材积模型的检验与调整

Test and adjustment of the binary standing volume model of Cunninghamia lanceolata in Hunan Province

【目的】对湖南省现有的杉木二元立木材积模型进行误差分析,建立新杉木二元立木材积模型,为湖南省杉木材积的精准预测提供理论依据。【方法】以湖南省杉木人工林为研究对象,采用配对t检验的方法对原二元立木材积模型进行检验。以山本材积式为基础模型构建固定参数模型、可变参数模型,以区域作为随机变量、哑变量构建混合效应模型和哑变量模型,对拟合结果进行对比分析。【结果】1)对原二元立木材积模型进行配对t检验,结果表明湖南现有的杉木材积模型与实际材积计算值存在显著差异;2)固定参数模型和可变参数模型的确定系数R2都在0.95以上。对检验数据进行检验,模型的总相对偏差(TRB)和平均系统偏差(MSB)均在±3%范围内,利用分径阶的方法进行检验,其固定参数模型在10、22和24 cm径阶的总相对偏差与平均系统偏差均超出±3%范围,其在24 cm径阶的总相对偏差与平均系统偏差均超出±7%,而可变参数模型在各径阶的偏差较小;3)混合效应模型与哑变量模型的确定系数均在0.95以上,从整体上看,哑变量模型的总相对偏差和平均系统偏差趋于0,而混合效应模型在10、12和24 cm径阶的总相对偏差均超出±3%范围。【结论】可变参数模型与固定参数模型相比较,可变参数模型拟合精度更高,总相对偏差和平均系统偏差较小,明显优于固定参数模型。哑变量模型的总相对偏差和平均系统偏差较小,赤池信息准则和贝叶斯信息准则较小,明显优于混合效应模型。

【Objective】Based on the error analysis of the existing binary volume model of Cunninghamia lanceolata in Hunan province,the binary standing volume model of Cunninghamia lanceolata was established to provide a theoretical basis for the accurate prediction of Cunninghamia lanceolata volume in Hunan Province.【Method】Taking the Cunninghamia lanceolata plantation in Hunan Province as the research object,the paired t-test method was used to test the original binary standing volume model,the fixed parameter model and variable parameter model were constructed based on the volume formula of Yamamoto,using the region as the random variable and dummy variable,the mixed effect model and dummy variable model were constructed and the fitting results were compared and analyzed.【Result】1)The paired t-test was carried out on the original binary standing volume model.The results showed that there was a significant difference between the existing volume model and the actual volume calculation value of Cunninghamia lanceolata in Hunan;2)The determination coefficients(R2)of the fixed parameter model and variable parameter model were above 0.95.The total relative deviation(TRB)and mean systematic deviation(MSB)of the model were both within±3%by testing the test data.The total relative deviation and average systematic deviation of the fixed parameter model at the diameter of 10,22 and 24 cm exceeded±3%,the total relative deviation and average systematic deviation at the diameter of 24 cm exceeded±7%,and the deviation of the variable parameter model in each diameter order was small;3)The determination coefficients of the mixed effect model and the dummy variable model were both above 0.95.On the whole,the total relative deviation and the average systematic deviation of the dummy variable model tended to be zero,while the total relative deviations of the mixed effect model at the diameter of 10,12 and 24 cm were all beyond the range of±3%.【Conclusion】Compared with the fixed parameter model,the variable parameter model had higher fitting accuracy,smaller total relative deviation and average systematic deviation,which is significantly better than the fixed parameter model.The total relative deviation and average system deviation of the dummy variable model are smaller,and the Akchi information criterion and Bayesian information criterion are smaller,which is obviously superior to the mixed effect model.