© 2023 Elsevier B.V.Surface soil moisture (SSM) is an essential component of the water cycle on earth and has a substantial role in the assessment of agriculture, drought, ecology, and climate science. Accurate estimation of SSM is critical for strategic management of precise farming, and climate change and drought monitoring. In particular, microwave remote sensing sensors can estimate SSM over large areas. The objective of this study is to propose a framework for retrieving soil moisture employing an Artificial Neural Network (ANN) for a semi-arid environment, using the data acquired from multi-frequency (X-, C-, and L-band) Synthetic Aperture Radar (SAR) sensors. In the experimental analysis, soil parameters (i.e., SSM, soil roughness, and altitude) and sensor characteristics (i.e., frequency, incidence angle, and backscatter coefficient) were used for the fallow land and, an additional feature as Normalized Difference Vegetation Index (NDVI) was included for a field cultivated with wheat. The resampling of data was tested over two windows as 3 × 3 and 5 × 5. The obtained results show that in general the window size 5 × 5 provides better results. Amongst the wavelengths, X-band Kompsat-5 achieved the highest correlation in both cultivated (R=0.85) and fallow lands (R=0.79), and L-band ALOS-2 gave the lowest RMSE as 1.68 of (vol%). Furthermore, Radarsat-2 provided slightly better results (RMSE=3.34–3.73) than Sentinel-1 (RMSE=3.20–3.77) in ANN analysis in the C-band.