Soil Moisture Inversion Using Microwave Remote Sensing Data In Sparse Vegetation Covered Areas

Soil moisture is an important part of water cycle, carbon cycle and atmospheric circulation of regional terrestrial system, which affects the energy balance and material exchange of land surface. The study area, Uxin Banner of Erdos city, is located in the arid district of China, with little precipitation, strong evapotranspiration, especially precious water resources. Soil moisture monitoring on regional scale is helpful to provide decision support for local crop production, drought prevention, regional development planning and ecological monitoring. Microwave remote sensing that has the characteristics of all day and all-weather and is not affected by weather and light conditions has become the main method of soil moisture monitoring. Compared with the traditional measurement method, it can efficiently acquire large-scale, real-time soil moisture information. Inversion of soil moisture by microwave remote sensing is affected by many factors, such as radar parameters, surface roughness as well as vegetation, etc. The effect of vegetation makes the relationship between microwave signal and soil moisture more complicated, which makes it difficult to obtain soil moisture information. How to effectively eliminate the influence of vegetation on soil moisture retrieval has been a hot research topic at home and abroad.In this paper, Uxin banner of Erdos city in Inner Mongolia was selected as study area, RADARSAT-2 radar data and optical data were comprehensively utilized to develop soil moisture retrieval models that are suitable for sparse vegetation covered surface in arid area, the vegetation indexes of NDVI and NDWI were applied to correct the vegetation effect on backscattering coefficient through water-cloud model. In addition, the empirical coefficients of water-cloud model were discussed and were then determined by the inversion precision of soil moisture. By analyzing the reason of lower inversion precision of water-cloud in this study area, combined with AIEM model, an improved method was developed, then, the precision of inversion methods that ignored the influence effect and only considered the correction of effective correlation length as well as the corrected backscattering coefficient were compared, and the optimal inversion algorithm of soil moisture in the study area was eventually determined. The distribution information of soil moisture in the study area was mapped through the optimal inversion model. Moreover, the corresponding relationship between the soil moisture content and NDVI was analyzed.The main results of this paper are as follows:?1? The vegetation effect can be removed from the backscattering coefficient through the vegetation water content that respectively calculated by the NDVI and NDWI using water-cloud model, and the corrected result by NDVI is better than NDWI.?2? The empirical coefficient A of water-cloud model almost has no effect on the inversion precision, while the value of coefficient B was different in different site. The inversion mode of Mv^lh_??vv??1? acquired the optimal precision in site 1 and site 4 when B is 0.02, with the IA reaches to 0.928, and the value of IA reaches the optimal value of 0.744 in site 3 and site 6 when B is 0.03.?3? The precision of effective correlation length inversed by backscattering coefficient is related to the polarimetric mode and root mean square height, when the root mean square height was 2.0cm, the correlation coefficients between backscattering coefficient of HH and inversed effective correlation length as well as and backscattering coefficient of VV and inversed effective correlation length reached the highest values, which both were higher than 0.8.?4? The precision of inversed soil moisture was different in different polarimetric mode. As for the blown-sand area, the precision of inversed soil moisture by VV mode is better than that by HH mode.?5? The proposed inversion method Mv^lh_??vv??1? that only corrected the backscattering coefficient in this paper was more suitable in blown-sand region, among of them, the inversion mode that combined the backscattering coefficient of VV that has corrected the vegetation effect using NDVI with effective correlation length of HH that including the vegetation effect was more suitable to the soil moisture retrieval in arid region, with R=0.852, IA=0.902 and RMSE=6.178%.?6? The Soil moisture over the study area is generally low, the soil moisture content are mainly concentrated in the region of 9%-19%. The soil moisture content and NDVI is linear fitting relationship, R2=0.5809; when the soil moisture content is less than 15%, the NDVI value is mostly lower than 0.3.