基于多时相高分一号影像的土壤湿度反演  

Soil Moisture Retrieval Based on Multi-temporal GF-1 Images

作　　者：杨丹阳 严颂华 杨永立[1] 田茂[2,3] YANG Dan-yang;YAN Song-hua;YANG Yong-li;TIAN Mao(College of Information Science and Engineering,Wuhan University of Science and Technology,Wuhan 430080,China;Beidou College,Wuhan Qingchuan University,Wuhan 430204,China;Institute of Aerospace Science and Technology,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉科技大学信息科学与工程学院,武汉430080 [2]武汉晴川学院北斗学院,武汉430204 [3]武汉大学宇航科学与技术研究院,武汉430072

出　　处：《科学技术与工程》2021年第11期4540-4549,共10页Science Technology and Engineering

基　　金：国家自然科学基金(42071406);国家重点研发计划(2018YFB2100500)。

摘　　要：近年来使用国产光学遥感数据反演土壤湿度已成为业内研究热点,其中利用高分一号(GF-1)影像反演土壤湿度是一种潜在的新兴手段,但应用中模型参数需要根据地域调节。为提高GF-1反演土壤湿度的准确度,将豹澥试验区2019年12月至2020年6月的GF-1多光谱宽幅覆盖(WFV)影像和地面观测站点实测数据作为数据源,以垂直干旱指数(PDI)、基于归一化植被指数(NDVI)的修正型垂直干旱指数(MPDI N)、基于两波段增强型植被指数(EVI2)的修正型垂直干旱指数(MPDI E)以及植被调整垂直干旱指数(VAPDI)这四种干旱指数为基准建立土壤湿度反演模型并进行精度评估。实验结果表明:在无植被区域,四种模型的反演结果大致相同,决定系数均在0.7350左右、平均相对误差均在4.50%左右、均方根误差均在1.10%左右,具有较高的精度;在有植被区域,VAPDI的反演效果最优,MPDI N与MPDI E次之,PDI效果不佳。与PDI相比,VAPDI由于考虑了混合像元的影响,不仅适用于稀疏植被区域,也适用于密集植被区域,应用范围更广;与基于两种不同植被指数的MPDI相比,VAPDI由于克服了植被覆盖度和植被像元反射率等因素的影响,基于该指数的土壤湿度估计值与实测值的决定系数达到0.7277以上,具有较高的反演能力。因此,针对豹澥试验区的实际情况,VAPDI指数具有精确反演土壤湿度的潜力。In recent years,using domestic optical remote sensing data to retrieve soil moisture has become a hot research topic in the industry.Among them,using GF-1 images to retrieve soil moisture is a potential emerging method,but the model parameters need to be adjusted according to the regional situation during application.In order to improve the accuracy of soil moisture inversion using GF-1,the GF-1 WFV images and ground observation site measured data from December 2019 to June 2020 in Baoxie experimental area were used as the data sources.The soil moisture inversion model was established and its accuracy was evaluated based on these four types of drought index.They were perpendicular drought index(PDI),normalized difference vegetation index(NDVI)-based modified perpendicular drought index(MPDI N),two-band enhanced vegetation index(EVI2)-based modified perpendicular drought index(MPDI E),and vegetation adjusted perpendicular drought index(VAPDI).The following experimental results are got from the research.The inversion results of the four models are roughly the same in the area without vegetation,with the determination coefficient around 0.7350,mean relative error around 4.50%,and root mean square error around 1.10%,indicating high inversion accuracy;in areas with vegetation,the inversion effect of VAPDI is the best,followed by MPDI N and MPDI E,PDI doesn t work very well.Compared with PDI,VAPDI is not only suitable for areas with sparse vegetation,but also for areas with dense vegetation due to the consideration of the effects of mixed pixels,with a wider range of applications.Compared with MPDIs which are based on two different vegetation indexes respectively,VAPDI has a high ability to invert soil moisture due to overcoming the influence of vegetation coverage and vegetation pixel reflectivity,and the coefficient of determination is above 0.7277.Therefore,VAPDI has the potential to accurately invert the soil moisture according to the actual situation in the Baoxie test area.

关 键 词：高分一号 土壤湿度 干旱指数 反演模型 空间分布 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]