摘要
为了探讨迎春5号杨树(Populus nigra×P.simonii)胸径和树高的关系,以黑龙江省尚志市林业局18块样地90株迎春5号杨树为研究对象,采用简单、可变指数、分段3类削度方程,利用R软件拟合得出最优基础模型,且在最优模型基础上,建立嵌套样地和样木效应的两水平非线性混合效应模型。在选定的20个基础模型拟合中,Kozak(2004)的拟合精度最高,它的参数估计在95%置信区间上均显著;基于样地、样木两水平构建的非线性混合效应模型显著提升了模型拟合效果。最终确定的最佳混合效应模型调整后决定系数(R_(a)^(2))为0.994 2,绝对误差(Bias)为0.003 8,均方根误差(RMSE)为0.604 5。预测时采用了两种不同的抽样方案,方案Ⅰ-5预测精度最好,平均绝对误差为0.510 4,平均绝对误差百分比为4.64%,而方案Ⅱ抽样数量从1增加到5之间的预测效果差别不是很明显,但整体上仍优于基础模型。
With the stem shape data of 90 Populus nigra×P.simonii trees in 18 plots set by the Forestry Bureau of Shangzhi City,Heilongjiang Province,considering the three types of cutting equation,the nls package of R software was used to fit and obtain the optimal basic model,and based on the optimal model,a two-level nonlinear mixed-effects model with nested plot and wood effects was established.Among the 20 selected basic model fits,Kozak(2004)had the highest fitting accuracy.After deleting an insignificant parameter,its parameter estimates were all significant at 95%confidence intervals.The nonlinear mixed effect model based on the two levels of the sample plot effect and the sample tree effect significantly improved the fitting effect of the model.The final best mixed effects model R^(2)_(a)was 0.9942,Bias was 0.0038,and RMSE was 0.6045.Inpredicting the hybrid model,when adopted two different sampling plan,planⅠ-5 prediction accuracy is best,an average absolute error MAE is 0.5104,the average absolute error percentage mape was 4.64%,and schemeⅡsampling number difference between the predicted effect of increased from 1 to 5 is not very obvious,but the overall is still better than that of the base model.
作者
常锦锦
苗铮
郝元朔
董利虎
Chang Jinjin;Miao zheng;Hao Yuanshuo;Dong Lihu(Northeast Forestry University,Harbin 150040,P.R.China)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2023年第9期59-66,共8页
Journal of Northeast Forestry University
基金
中央高校基本科研业务费专项资金项目(2572020DR03)
黑龙江省森林可持续经营试验示范区建设项目(201522)。
关键词
迎春五号杨树
非线性混合效应模型
削度方程
Populus nigra×P.simonii
Nonlinear mixed effects model
Pruning equation