基于哑变量的湿地松树皮厚度模型研究

拟合立地质量等级的湿地松树皮厚度模型,为测算生长于不同立地湿地松出材量、出材率及树皮材积提供依据。根据宁德市湿地松样木和样地的调查数据,运用R软件拟合4类树皮厚度模型,选用赤池信息准则(AIC)、贝叶斯准则(BIC)、偏差(B)、绝对偏差(AB)与决定系数(R2)为模型评价指标,使用含熵权值的TOPSIS法筛选最优模型,于得到的4类树皮厚度最优模型中加入立地质量等级哑变量,得到胸高处树皮厚度模型lnTBB=(-3.075S1-3.154S2-3.157S3-3.325S4)+1.125lnD1.3,任意高度处树皮厚度模型TB=(0.005S1-0.026S2-0.101S3-0.141S4)-0.616RC+0.761DR+0.040dob,相对树皮厚度模型TRB=0.137+0.903DR2.009S1+1.806S2+2.126S3+1.855S4,去皮直径模型dib=(-0.087S1+0.138S2+0.223S3+0.293S4)+0.881dob,这4个哑变量模型R2相对于原模型分别提升了8.9%、41.6%、20.4%和1.2%,各项低优指标均降低,经配对t检验,4个哑变量模型的预估值与实测值不存在明显差异,比原模型拟合效果更佳。

Estimation of bark thickness of Pinus elliottii plantations based on dummy variables

The model to estimate the bark thickness ofPinus elliottiiplantations based on dummy variables and site quality gradeshas been developed in this study to provide a basis for measuring the volume,rate of change in volume,and bark volume ofPinus elliottiicorresponding to varying levels of site quality.According to the survey data onPinus elliottiiwood and sample plots in NingdeCity,the R software was used to calculate the four bark skin thickness models.Akaike information criterion(AIC),Bayesianinformation criterion(BIC),deviation(B),absolute deviation(AB),and decision coefficient(R2)were selected to be the modelevaluation indicators.A TOPSIS method based on entropy weights was introduced to select the optimal model and dummy variableswere introduced into the optimal model.Bark thickness models were evaluated based on the dummy variables and the results obtainedwere as follows.lnTBB=(-3.075 S1-3.154 S2-3.157 S3-3.325 S4)+1.125 lnD1.3,TB=(0.005 S1-0.026 S2-0.101 S3-0.141 S4)-0.616 RC+0.761 DR+0.040 dob,TRB=0.137+0.903 DR2.009 S1+1.806 S2+2.126 S3+1.855 S4,dib=(-0.087 S1+0.138 S2+0.223 S3+0.293 S4)+0.881 dob.Further,R2 increased by 8.9%,41.6%,20.4%,and 1.2%for the four dummy variable models,respectively,and eachlow-quality indicator decreased corresponding to each model.After performing pairedt-tests,no significant differences between thevalues estimated by the four dummy variable models were observed,and the fitting effect was detected to be better than that of theoriginal model.