Comparing the estimation of soil surface moisture in agricultural lands using SAR data: ALOS PALSAR-2 and Sentinel-1A data

 Aq Qala plain is one of the fertile plains of Golestan province. Runoff accumulation and flood formation cause a lot of human and financial losses. Understanding the temporal and spatial distributions of soil surface moisture is a key variable in simulating runoff and reducing flood hazards in the research area. In this study, the soil moisture content was investigated using data from ALOS PALSAR-2 radar satellites in the L band and Sentinel 1 in the C band.

 In this research, three series of ALOS PALSAR-2 radar data, Sentinel 1, and optical data of Sentinel 2A were used. To calculate the soil surface moisture, first soil samples were taken and weighed at a depth of 5 cm in each cluster by weight method. After drying the samples in the oven, the dry and wet weight was calculated. Then two important parameters of soil moisture, including dielectric constant and soil surface roughness were measured. The dielectric constant was calculated using TDR and the roughness was calculated using two digital cameras. Using Agisoft photo scan and Arc GIS software, a digital elevation model and surface roughness profiles were extracted. Radar data were processed with SNAP software from the European Space Agency (ESA). The processing steps included radiometric calibration, sigma note extraction in decibels, filtration of Lee filter and 5x5 kernel size, geometric correction and finally geocoding of all three data. With Sentinel-2 optical data, while preparing vegetation and moisture indices, a land-use map was prepared with the Forrest random algorithm. Then, with the values of roughness, dielectric constant and plant and moisture indices in MATLAB 2019 software environment, two models Oh, and WCM were coded and the soil moisture values were calculated and taken out. Then the moisture obtained from the surface redistribution model with gravity, moisture at the site, and previous studies was evaluated.

 In terms of radar signal sensitivity to soil surface moisture in Oh model in HH and HV polarization ALOS PALSAR- 2, in  with R2 = 0.86,  with R2 = 0.905 and for Sentinel-1 images, in  with R2 = 0.91 and  with R2 = 0.93 are sensitive to soil surface moisture. The surface backscattering coefficients sensitivity of the Oh model in ALOS PALSAR-2 in HH polarization and in HV polarization and for Sentinel-1 data in VV polarization and in VH polarization is sensitive to surface roughness. Also, the surface backscattering coefficients sensitivity of the Oh model in ALOS PALSAR-2 in HH polarization with R2 = 0.96 and in HV polarization with R2 = 0.94 and for Sentinel-1 data in VV polarization with R2 = 0.72 and in VH polarization with R2 = 0.74 is sensitive to surface roughness. Relationship between total surface backscattering coefficients obtained from the Water cloud model and soil surface moisture measured in HH and HV polarization PALSAR-2 images, in  and  with R2 = 0.73 and R2 = 0.399, respectively, and for Sentinel-1 data in polarization VV and VH, in  and  with R2 = 0.073 and R2 = 0.13, respectively, are sensitive to soil surface moisture. In the soil moisture estimation section based on Oh model for PALSAR-2 and Sentinel-1 data on bare lands in HH and HV polarization for PALSAR-2,  and  with R2 = 0.88, RMSE = 1.924  and R2 = 0.93, RMSE = 1.45  for Sentinel-1 data, respectively,  and  are with R2 = 0.81, RMSE = 2.57  and R2 = 0.97, RMSE = 0.90 , respectively. The results of Oh model showed that in HV polarization, Pulsar-2 images and VH sentinel-1 polarization images have stronger soil moisture results. Soil moisture results in WCM model in vegetated lands for PALSAR-2 images in HH and HV polarization in  and , respectively, with R2 = 0.81, RMSE = 2.44  and R2 = 0.78, RMSE = 2.64 , R2 = 0.88, RMSE = 1.924  for Sentinel-1 in  and  with R2 = 0.75, RMSE = 2.74  and R2 = 0.76, RMSE = 2.69 , respectively. In the discussion of estimating soil moisture with WCM model, the results of PALSAR-2 in polarization HH with R2 = 0.81  in lands with vegetation density of more than 0.4% have better results.

In the Oh model, the results showed that the sensitivity of HV polarization in PALSAR-2 images and VH polarization in Sentinel-1 in an area with vegetation density less than 0.4 is much higher than the polarization sensitivity of HH PALSAR-2 and polarized VV Sentinel-1. Also, the HH polarization sensitivity of PALSAR-2 and VH Sentinel-1 polarized images to surface roughness is higher than other polarizations. In the Water cloud model, the results obtained from Sentinel-1 data by eliminating the effects of vegetation and simulating the surface backscattering coefficients are very poor. Has reduced the penetration of the wave. The results showed that the polarization sensitivity of HH in areas with vegetation with a density of more than 0.4 was much higher than the polarization sensitivity of HV, also the sensitivity of the radar signal to noise was low in polarized HH but high in polarized HV. This showed that the polarization of the HH matrix is sensitive to vegetation, so the model will be more stable in the presence of vegetation. A comparison of two SAR data in the range showed that PALSAR-2 images are more efficient for estimating soil surface moisture in the study area and similar areas due to their long wavelength and depth of penetration into soil and vegetation. In line with the research topic, other researchers, including Alishan et al. A study has attempted to estimate soil surface moisture in barren lands using PALSAR-2 and Sentinel-1 data, which concluded that the WCM model eliminates plant water content. And surface moisture estimation using the inverse model was high in both PALSAR-2 and Sentinel-1 data. A study by Zaribi et al. Showed that the OH model in both polarized HV and HH with lower RMSE is able to estimate soil surface moisture. This research is suitable for environmental management, agriculture, natural resources, and water resources management and contributes significantly to hydrological modeling.