基于多元线性回归的西昌市云南松蓄积量模型研究  

Study on Forest Volume Model of Pinus yunnanensis Based on Multiple Linear Regression in Xichang City

作　　者：闵志强[1] 胡云云[1] 顾丽[2] 

机构地区：[1]国家林业局西北林业调查规划设计院,陕西西安710048 [2]西北农林科技大学,陕西杨陵712100

出　　处：《西北林学院学报》2017年第3期186-190,共5页Journal of Northwest Forestry University

基　　金：国家自然科学基金(31300538)

摘　　要：利用2015年西昌市二类调查数据和2014年遥感影像数据,选择最小二乘法、基于AIC准则的逐步回归法、主成分法和偏最小二乘法分别建立云南松蓄积量线性回归模型并进行对比分析。结果表明:1)4种模型均为显著回归关系,除最小二乘法外,3种模型中的自变量T检验与因变量均显著相关。2)最小二乘法易导致模型自变量产生共线性,其他3种方法可有效消除共线性影响。3)模型综合评价由高到低为:偏最小二乘法>逐步回归法>主成分法>线性最小二乘法,通过预留样本进行检验,精度由高到低为:偏最小二乘法>逐步回归法为>线性最小二乘法>主成分法。综合分析认为,偏最小二乘法和逐步回归法综合效果最优,结果可为今后准确、高效地估测森林蓄积量提供参考。In this paper,Using 2015 Inventory Data in Xichang City and 2014 remote sensing data, forest volume linear regression modelsof Pinus yunnanensis were established respectivelyby the least-squares method,based on the AIC criterion ofstepwise regression method,principal component analysis and partial least squares, and alsocompared and analyzed. The results showed that：1） four models F tests were Signifi- cant regression relationship,except the least squares method,the other three models T test ofindependent variables and the dependent variablewere significantly correlated. 2） The least squares method was easy to produce independent variables collinearity, the other three methods can effectively eliminate the influence of collinearity. 3） Comprehensive assessment model descending order.partial least squares〉stepwise regres- sion〉 principal component analysis〉linear least squares,through the reserved sample test,precision from high to low： partial least squares〉stepwise regression as〉 linear least squares〉 principal component anal- ysis. Through comprehensive analysis,the combined effect of the partial least squares and stepwiseregres- sion method was the best in this study,and the resultprovided reference basisfor accurately and efficiently estimating forest volumein future.

关 键 词：森林蓄积量 线性回归 最小二乘法 逐步回归法 偏最小二乘法 主成分法 

分 类 号：S791.257[农业科学—林木遗传育种]