Soil moisture retrieval from Sentinel-1 Synthetic Aperture Radar using a water cloud model-constrained physics-informed neural network

This repository contains the analysis-ready data, source code, and retrieved soil moisture products supporting the manuscript "Soil moisture retrieval from Sentinel-1 Synthetic Aperture Radar using a water cloud model-constrained physics-informed neural network," The study develops a Physics-Informed Neural Network that embeds the Water Cloud Model (WCM) into the network loss function (PINN-WCM) to retrieve high-resolution (10, 30, and 50 m) surface soil moisture from Sentinel-1 SAR over nine in situ monitoring sites on the Korean Peninsula, covering cropland, forest, and grassland. The repository is organized as follows: data/ — A quality-controlled, analysis-ready table (Data_s1_rescale_QC_results_VF.xlsx) containing the model inputs (normalized VV/VH backscatter, local incidence angle, DpRVI), the matched in situ soil moisture at 10 cm depth, and the predicted soil moisture from all three networks (FFNN, PINN-LR, PINN-WCM), with one record per site, spatial resolution, and Sentinel-1 acquisition date. Site metadata (site_info.xlsx) for the nine stations is also provided. code/ — MATLAB scripts that reproduce the figures of the manuscript from the data and the retrieved products. SM_PINN_RESULT/ — Retrieved 10 m surface soil moisture maps (GeoTIFF) produced by the integrated PINN-WCM, organized into one subfolder per site. Third-party inputs are not redistributed here but are openly available from their original providers: Sentinel-1 SAR data from the European Space Agency / Copernicus (e.g., via the Alaska Satellite Facility, https://search.asf.alaska.edu/), and in situ soil moisture observations from the Rural Development Administration (RDA) of the Republic of Korea (https://weather.rda.go.kr/). Unless otherwise noted, the contents of this repository are released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.