Soil Moisture Retrieval Under Vegetation Cover Using Multi-source Remote Sensing Data

Soil moisture is an important parameter that affects the balance of water in nature.It plays a decisive role in the material and energy exchange between the surface-atmosphere interface,and it’s also the basic condition of vegetation growth and development.This study aims to develop soil moisture retrieval models over vegetated areas based on Sentinel-1 SAR,Landsat 8 OLI and FY-3C/MWRI data.In order to remove vegetation effect,the vegetation indices obtained from Landsat 8 data and the Microwave Polarization Difference Index obtained from FY-3C/MWRI data were applied to establish the vegetation water content model respectively.The models were combined with the original water-cloud model,and two soil moisture retrieval models were developed by combining Multi-source remote sensing data.Finally,in order to evaluate the retrieval accuracy of the models,the experiment of the soil moisture retrieval was conducted in China and Spain,and the retrieval accuracy was validated by in-situ data.The main conclusions are:(1)Compared with the Sentinel-1 VH polarization,the backscattering coefficient at VV polarization was more suitable for soil moisture retrieval and obtained a higher accuracy.(2)In contrast with NDVI and EVI,the semi-empirical model developed by NDWI1 and VV data was the best for soil moisture estimation in the study of combining optical and microwave data to retrieve soil moisture over vegetated surface.In Spain study areas,the correlation coefficient between the estimated soil moisture by the model and measured value was 0.865,and the root mean square error and mean bias were 0.045 cm3/cm3,0.036 cm3/cm3 respectively.In China study areas,the correlation coefficient between the estimated soil moisture by the model and measured value was 0.627,and the root mean square error and mean bias were 0.044 cm3/cm3,0.036 cm3/cm3 respectively.(3)In the study of combining active and passive microwave data to retrieve soil moisture over vegetated surface,the semi-empirical model based on Sentinel-1 VV and MPDI from FY-3C/MWRI could also estimate the soil moisture well.The correlation coefficient between estimated soil moisture and measured value was 0.563,and the root mean square error and mean bias were 0.046 cm3/cm3,0.037 cm3/cm3 respectively.(4)According to the sensitivity analysis of each parameter in the model,the soil moisture is most sensitive to the radar backscatter coefficient,and the sensitivity of the radar wave incidence angle is the lowest.(5)A regional soil moisture mapping can be realized at 20m resolution by the model with Sentinel-1 and Landsat 8 data while the resolution was 1km in model with Sentinel-1 and FY-3C data.