基于植被光谱和FY-3A/VIRRL1B数据的藏北地区土壤水分估算  

Estimation of Soil Moisture Content in North Tibet Based on Vegetation Spectra and FY-3A/VIRR L1B Data

作　　者：王凤杰[1] 冯文兰[1] 扎西央宗 牛晓俊[2] 刘志红[1] 王永前[1] WANG Feng-jie;FENG Wen-lan;Zhaxiyangzong;NIU Xiao-jun;LIU Zhi-hong;WANG Yong-qian(College of Resources and Environment Sciences, Chengdu University of Information Technology, Chengdu 610225, Sichuan, China;Tibet Institute of Plateau Atmospheric and Environmental Sciences, Lhasa 850000, Tibet, China)

机构地区：[1]成都信息工程大学资源环境学院,四川成都610225 [2]西藏高原大气环境科学研究所,西藏拉萨850000

出　　处：《干旱区研究》2018年第3期532-539,共8页Arid Zone Research

基　　金：国家自然科学基金项目(41465006,41301653,41631180);四川省教育厅项目(16TD0024,18ZA0110)资助

摘　　要：为了实现对藏北地区土壤水分和干旱情况的动态监测,基于藏北植被光谱、实测20 cm土壤水分以及FY-3A/VIRR数据,利用相关性筛选出对土壤水分敏感的植被光谱波段构建植被指数,并以此建立土壤水分估算模型,再结合FY-3A/VIRR L1B数据将建立的模型应用于藏北地区的土壤水分估算,通过比较决定系数和RMSE,确定精度较高的藏北地区土壤水分遥感估算模型。研究表明：NDVI（620,850）、EVI（450,620,850）、NDWI（850,1 330）和RVI（850,1 330）与实测20 cm土壤水分的决定系数分别为0.232、0.256、0.537和0.554,都能较好地表征土壤水分,分别利用每个指数建立的二次模型所获得的土壤水分估算结果与实测数据的RMSE均较小;以FY-3A/VIRR数据为基础,模型M（NDVI）和M（EVI）能够有效的估算藏北土壤水分,模拟值与实测值的相关系数r分别为0.50和0.51,RMSE分别为0.13和0.11,模型都可实现对藏北地区土壤水分的估算。研究可为掌握藏北地区土壤水分状况和制定农牧业发展决策提供依据。In order to dynamically monitor soil moisture content and drought in north Tibet,this study tried to build a model to estimate soil moisture content. The model was based on the measured data of soil moisture content and vegetation indexes sensitive to soil moisture content. Firstly,the sensitive bands of soil moisture content were screened out through correlation analysis between the hyper-spectral reflectance and the first-order differential reflectivity with soil moisture content. The sensitive spectral parameters of soil moisture content were identified through establishing vegetation indices,including the ratio vegetation index RVI（ 850,1 330）,normalized difference water index NDWI（ 850,1 330）,enhanced vegetation index EVI（ 450,620,850） and normalized difference vegetation index NDVI（ 620,850）. Secondly,the estimation model was established based on the data of soil moisture content and the vegetation indices from the measured hyperspectral values. Thirdly,the FY-3 A/VIRR L1 B data were preprocessed,including the atmospheric correction and geometric correction,and then the vegetation indexes were calculated. Finally,the vegetation indices of FY-3 A/VIRR data were replaced by the vegetation indices from the measured hyperspectral values in the estimation model. In order to find the best estimation model,the model accuracy was judged by the determination coefficient（ R～2） and the root mean square error（ RMSE）. The results showed that all the indexes could be used to reflect the soil moisture content. The correlation coefficients between the NDVI（ 620,850）,EVI（ 450,620,850）,NDWI（ 850,1 330）,RVI（ 850,1 330） and the soil water content were 0. 232,0. 256,0. 537 and 0. 554 respectively. Then,the corresponding models of estimating soil moisture content were established based on these indexes. Among them the best estimated results were obtained using the quadratic model because of the higher R^2 and lower RMSE. The vegetation indices of FY-3 A/VIRR data were brought into the mo

关 键 词：土壤水分 植被光谱 植被指数 FY-3A/VIRR 藏北地区 

分 类 号：S152.7[农业科学—土壤学]