Soil Moisture Change Monitoring from C and L-band SAR Interferometric Phase Observations

The soil moisture changes (Delta M-v) have a significant influence on forestry, hydrology, meteorology, agriculture, and climate change. Interferometric synthetic aperture radar (InSAR), as a potential remote sensing tool for change detection, was relatively less investigated for monitoring this parameter. DInSAR phase (phi) is sensitive to the changes in soil moisture (M-v), and thus, can be potentially used for monitoring Delta M-v. In this article, the relations between phi and Delta M-v over wheat, canola, corn, soybean, weed, peas, and bare fields were investigated using an empirical regression technique. To this end, dual-polarimetric C-band Sentinel-1A and quad-polarimetric L-band uninhabited aerial vehicle synthetic aperture radar (UAVSAR) airborne datasets were employed. The regression model showed the coefficient of determination (R-2) of 40% to 56% and RMSE of 4.3 vol.% to 6.1 vol.% between the measured and estimated Delta M-v for different crop types when the temporal baseline (Delta T) was very short. As expected, higher accuracies were obtained using UAVSAR given its very short Delta T and its longer wavelength with R-2 of 47% to 59% and RMSE of 4.1 vol.% to 6.7 vol.% for different crop types. However, using the Sentinel-1 data with the long Delta T and shorter wavelength (5.6 cm), the accuracies of Delta M-v estimations decreased significantly. The results of this study demonstrated that using the phi information from Sentinel-1 data is a promising approach for monitoring Delta M-v at an early growing season or before the crop starts growing, but using L-band SAR data and lower temporal baselines are recommended once the biomass increases.

Automated summary

This study investigated the impact of soil moisture changes on different crops and analyzed the relationship between ΔM-v and φ through regression techniques, suggesting that UAVSAR is more accurate for monitoring ΔM-v compared to Sentinel-1. The results demonstrate promising potential for using φ information from Sentinel-1 data in the early stages of crop growth, but recommend the use of L-band SAR data and shorter temporal baselines as crop biomass increases.