基于GF-1与Landsat-8的康保县叶面积指数遥感反演研究

摘要

Leaf area index (LAI) is an important indicator of forest structural parameter. In this study, a novel method that combined PCA with a linear stepwise regression, a logistic-model and GWR regression was developed to derive an integrated regression model of LAI.A total of 134 sample plots were systematically selected in the study area-Kangbao County, Hebei province and LAI data were collected. Landsat-8 and GF-1 image were acquired. The results were validated using the observations of sample plots and showed that:(1) In the desertification area, the vegetation index and LAI extracted by GF-1 and Landsat-8 had a high correlation. The PCA method can be used to eliminate the collinearity of the vegetation index factors. (2) The estimation accuracy of GWR regression was the highest for both GF-1 and Landsat-8 data with the greatest determination coefficient and smallest root mean square error (RMSE). (3) Inversion of LAI by domestically produced GF-1 data in the study area is better than that of Landsat-8, and can be used as a substitute for Landsat-8 data for estimation of LAI.%以GF-1和Landsat8遥感影像为数据源,采用逐步回归、非线性Logistic回归和基于空间位置的地理加权回归3种方法,结合134个野外样地调查数据,在河北省康保县开展叶面积指数反演研究,并对结果进行精度检验.结果表明:(1)在荒漠化地区,GF-1和Landsat-8遥感影像提取的植被指数因子与LAI均有较高的相关性.运用主成分分析方法对植被指数因子进行处理,可以有效消除各影响因子间的共线性.(2)基于GF-1和Landsat-8影像分别建立的3种模型,均以地理加权回归决定系数最大,均方根误差最小,反演精度最高.(3)国产GF-1数据反演LAI效果优于Landsat-8,可以代替Landsat-8数据进行叶面积指数的估测.