云冷杉阔叶混交林不同树种组单木枝下高非参数模拟

[Objective]To compare the simulation differences and fitting accuracy of different non parametric modeling methods and variable types on the height of single branches of different tree species groups in a mixed forest of spruce and fir broad-leaved trees,and to provide theoretical basis for scientific management decision-making in this forest.[Method]Based on data from 12 large-scale plots of 1 hectare,the generalized additive model,BRT model,Cubist model,and MARS model were used to simulate the HCB of different tree species groups in a broad-leaved mixed forest of spruce and fir in Northeast China.The importance of different types of variables and their impact trends on HCB were evaluated,and the importance of various factors affecting branch height was assessed.[Results]In terms of variable importance,single tree variables(coniferous,spruce,hard broad-leaved,and soft broad-leaved groups accounting for 77.54％,31.02％,31.12％,and 73.69％,respectively)and forest partition variables(coniferous,spruce,hard broad-leaved,and soft broad-leaved groups accounting for 5.00％,20.34％,11.03％,and 8.71％,respectively)have relatively high relative importance,while forest spatial structure variables(coniferous,spruce,hard broad-leaved,and soft broad-leaved groups accounting for 4.57％,12.14％,21.91％,and 5.89％,respectively)and density index variables(coniferous,spruce,hard broad-leaved,and soft broad-leaved groups accounting for 10.79％,35.20％,31.90％,and 8.84％,respectively)have relatively low relative importance.[Conclusion]BRT and Cubist are the best methods for simulating single tree HCB of four tree species groups in a mixed forest of spruce and broad-leaved trees.The single tree and stand variables have the greatest impact on the HCB of different tree species groups.Although the relative importance of spatial structure variables is relatively small,further research is needed.