干旱区地理2017，Vol.40Issue(6)：1248-1255,8.
基于改进植被指数土壤水分遥感反演
Remote sensing inversion of soil moisture based on modified vegetation index
摘要
Abstract
Fast acquisition of the soil moisture content,characteristics,and spatial distributing are the objective needs of agricultural drought management and control.This paper focused on the soil moisture content on arid and semi-arid bytaking Delta Oasis of the Weigan and Kuqa Rivers in Xinjiang as an example.Based on the multi-spectral remote sensing image of Landsat 8 OLI which were collectedon July 15,2015,the traditional vegetation index (VI) was extended by adding the short-wave infrared band,then Grey Relational Analysis (GRA) was used to sort the vegetation index before and after the improvement,and the first three vegetationindiceswhich hada larger gray relational degree were screened out,at the same time,the Partial least squares regression (PLSR) inversion models of soil moisture content were built and validated.Finally,the spatial distribution of soil moisture content wasanalyzed by using the PLSR model in the study area.The study results show as follows:(1) On the basis of the traditional vegetation index,the shortwave infrared with large amount of information was introduced,by comparing the variance inflation factor (VIF) vegetation index before and after the improvement,the multi-collinearity between vegetation indices was greatly reduced by extending the traditional vegetation index;(2) Grey Relational Analysis was used to sort the vegetation index before and after the improvement,the analysis results were as follows:EEVI＞ ERVI＞ ENDVI＞ RVI＞ EDVI＞ DVI＞ NDVI＞ EVI,the gray relational degree between EEVI,ERVI,ENDVI and soil moisture were all over 0.8,and it also showed that the gray relational degreebetween the extended vegetation indicesand the soil moisture content was higher than the corresponding traditional vegetation index;(3)The first three vegetation indices which had larger gray relational degree were screened out to build the Partial least squares regressioninversion models of soil moisture content,in the PLSR model,coefficient of determination (R2) of the calibration and validation were 0.82 and 0.80 respectively,the root mean squares error (RMSE) of the calibration and validation were 0.030 and 0.034 respectively,and the validation relative prediction deviation RPD equals to 2.07,greater than 2.00.Themain reasons of improving the model precision were that the shortwave infrared (SWIR) on Landsat 8 OLI had more informationand could protrude the difference in vegetation coverage and production status.Therefore,the model had very high accuracy andreliability,and the spatial distribution of the soil moisture content in the studyarea was analyzed based on the PLSR model.The statistical information of the inversed soil moisture content showed that the soil moisture content was gradually reduced from the west to east,and north to south.However,the minimum soil moisture content was mainly distributed in the oasis-desert ecotone,which made the ecotone become "ecological rift".Therefore,the experiment indicated that the vegetationindex was modified by the introduction of the shortwave infrared based on Landsat 8 OLI,and the PLSR model of soil moisture content had been built,by which better inversion result of soil salinity spatial distribution could be obtained.But in order to achieve accurate inversion,in the future,to enhance the experimental dataprocessing abilityand improve the simulation accuracy,more samples and accurate modelswill be laid and builtas much as possible.关键词
土壤水分/植被指数/灰色关联分析/偏最小二乘回归Key words
soil moisture/vegetation index/grey relational analysis/partial least squares regression分类
农业科技引用本文复制引用
蔡亮红,丁建丽..基于改进植被指数土壤水分遥感反演[J].干旱区地理,2017,40(6):1248-1255,8.基金项目
国家自然科学基金(41771470、U1303381) （41771470、U1303381）
自治区重点实验室专项基金(2016D03001) （2016D03001）
