Comparison of Remote Sensing Soil Moisture Algorithms Across Optical, Thermal and Radar Classes

As coastal river deltas experience specific climate change and management pressures, monitoring programs are necessary for proper agroecosystem management strategies. This study evaluates the performance and temporal dynamics of multiple remote sensing-derived moisture products and indices against in-situ soil volumetric water content (VWC) measurements collected from January 2021 to December 2022 at Vidrice polder (Neretva River Valley, Croatia). We compared Thermal-Optical Trapezoid Model (TOTRAM), Optical Trapezoid Model (OPTRAM) estimates derived from Landsat-8/9 (L8/L9) and Sentinel-2 (S2), Synthetic Aperture Radar (SAR) based moisture from Sentinel-1 and spectral indices including Normalized Difference Moisture Index (NDMI), Normalized Difference Water Index (NDWI) and Moisture Stress Index (MSI) from both L8/L9 and S2. All datasets were harmonized to a common temporal scale and merged for direct comparison. Results demonstrate distinct seasonal patterns across most products with S2 OPTRAM showing highest correlation (r=0.631) while L8/9 OPTRAM (r=0.413) and L8/9 TOTRAM (r=0.512) showed moderate correlation with in-situ measurements. SAR-based estimates showed weak correlation likely due to C-band sensitivity to surface roughness and vegetation biomass. Differences in spatial resolution between L8/L9 and S2 were also reflected in simple moisture indices, with ones based on S2 data showing higher alignment (r=-0.466 – 0.505) with soil moisture measurements. These results suggest that physically-based trapezoid models, particularly when applied to high-resolution optical data like Sentinel-2, provide better representation of soil moisture dynamics than simpler vegetation indices, a desirable factor in soil moisture monitoring.