Improving soil moisture estimation in wet soils using L-band Synthetic Aperture Radar (SAR) through polarization and filtering optimization

Accurately mapping soil water distribution is crucial for effective irrigation management and landslide risk assessment. Microwave remote sensing is an effective method for assessing soil water content over extensive areas; however, its efficacy tends to diminish in regions with high soil water content. This study aimed to evaluate the sensitivity of soil moisture estimation in areas with volumetric water content exceeding 0.3 m 3 /m 3 , focusing on methodologies of optimizing polarization, despeckling filter application, and ground truth data determination. Among the polarizations of electromagnetic wave analyzed—vertical and horizontal polarization (VH), vertical and vertical polarization (VV), horizontal and horizontal polarization (HH), and horizontal and vertical polarization (HV)—HH and VH polarization exhibited the favorable performance. We evaluated the effectiveness of various despeckling filters—Boxcar, Lee, Frost, Gamma-Map, Refined Lee, and Lee Sigma—on the accuracy of soil moisture estimation. The combination of HH polarization with the Frost filter yielded high correlation coefficients, low root mean square error, and satisfactory sensitivity. While the smallest window size (3 × 3) provided sufficient noise reduction, larger window sizes degraded estimation accuracy. A notable correlation was observed when the ground-truth data were derived from averaging soil water content across an area matching the despeckling filter window size. This suggests that SAR image pixel values effectively represent the mean soil water content within an equivalent spatial extent of the despeckling filter.These findings highlight that accurate soil moisture estimation using SAR backscatter requires careful consideration of polarization, filter type, window size, and the spatial scale of in-situ measurements.