樟子松人工固沙林冠幅——胸径模型

Canopy-DBH Models for Sand-fixing Plantation of Pinus sylvestris var. mongolica

基于章古台地区22块樟子松（Pinus sylvestris var.mongolica）人工纯林标准地的702棵樟子松立木数据,构建了樟子松固沙林冠幅—胸径关系的基础模型、广义模型及基于混合效应的基础模型和广义模型;比较了随机选择样本木、选择平均胸径树、选胸径较小树和选胸径较大树4种方案,计算混合模型随机参数时的混合模型预测精度;最后分析了不同林木因子和林分变量对冠幅—胸径关系的影响。模型评价指标包括决定系数（R2）、平均绝对误差（MAE）以及均方根误差（RMSE）。结果表明：枝下高（HCB）、相对植距（RS）和林龄（A）对冠幅—胸径关系影响最为显著;混合模型拟合精度（基础混合模型R2、MAE和RMSE分别是0.703 0、0.386 6和0.515 4;广义混合模型R2、MAE和RMSE为0.705 1、0.382 2和0.513 6）高于最小二乘法回归（OLS）模型（基础模型R2、MAE和RMSE分别为0.587 5、0.469 6、0.607 5;广义模型R2、MAE和RMSE分别为0.661 8、0.415 5和0.550 0）。基础混合模型和广义混合模型差异较小（2模型R2、MAE和RMSE均相差1%左右）。冠幅随HCB和A的增大而减小,随RS的增大而增大。进行冠幅预测时,推荐使用基础混合模型并从每块标准地选择2棵平均木冠幅计算其随机参数,或使用方法较为简单的OLS广义模型预测单木冠幅大小。

In this study,the data consisted of 702 individuals of Pinus sylvestris var.mongolica in 22 temporary sample plots of pure plantation in Zhanggutai of Liaoning Province,northeast China,were used to develop the canopy-DBH models,i.e.the basic model（fitted by ordinary least squares,OLS）,generalized model（fitted by OLS）,nonlinear mixed-effect basic model and nonlinear mixed-effect generalized model.The goodness-of-fits and prediction accuracy of the four models were compared.For the mixed-effect model,four sampling strategies,i.e.,random sampling,large-DBH tree sampling,small-DBH tree sampling and medium-DBH tree sampling,were designed to calculate the random parameters.The effects of individual factor and stand level variables on the canopy-DBH relationship were simulated.Model evaluation indices included the determination coefficient（R2）,mean absolute error（MAE）and root mean square error（RMSE）.Results showed that the height to live crown base（HCB）,relative spacing index（RS）and stand age（A）were the dominant factors in CW-DBH models.The goodness-of-fits of mixed-effect CW-DBH models（R2,MAE and RMSE）were 0.703 0,0.386 6 and 0.515 4,and those of mixedeffect generalized model were 0.705 1,0.382 2 and 0.513 6,respectively,which were better than those of the OLS models.However,the difference in goodness-of-fit between the mixed-effect basic and generalized models was not significant（below 1%）.Canopy decreased with the increase of A and height of HCB,but increased with the increase of RS.