Real-Time Retrieval of Daily Soil Moisture Using IMERG and GK2A Satellite Images with NWP and Topographic Data: A Machine Learning Approach for South Korea

Soil moisture (SM) is an indicator of the status land surface, which useful for monitoring extreme weather events. Representative global SM datasets include National Aeronautics and Space Administration (NASA) Moisture Active Passive (SMAP), Global Land Data Assimilation System (GLDAS), European Centre Medium-Range Weather Forecasts (ECMWF) Reanalysis 5 (ERA5), but due to their low spatial resolutions, none these well describe changes in local areas, they tend have a accuracy. Machine learning (ML)-based predictions demonstrated high accuracy, obtaining semi-real-time information remains challenging, dependence validation accuracy on data sampling method used, such as random or yearly sampling, has led uncertainties. In this study, we aimed develop ML-based model real-time estimation that can capture local-scale variabilities reliable regardless method. This study was conducted South Korea, satellite image data, numerical prediction (NWP) topographic provided within one day were used input data. For modeling, 13 variables affecting surface selected: 10- 20-day cumulative standardized precipitation indexes (SPI10 SPI20), normalized difference vegetation index (NDVI), downward shortwave radiation (DSR), air temperature (Tair), (LST), soil (Tsoil), relative humidity (RH), latent heat flux (LE), slope, elevation, ruggedness (TRI), aspect. Then, models based forest (RF) automated machine (AutoML) constructed, trained, validated using leave-one-year-out (LOYO) cross-validation. The RF- AutoML-based had significantly rates comparisons with situ (mean absolute error (MAE) = 2.212–4.132%; mean bias (MBE) −0.110–0.136%; root square (RMSE) 3.186–5.384%; correlation coefficient (CC) 0.732–0.913), while tended higher than RF-based model, used. addition, when compared highest outperforming both GLDAS ERA5 representing dryness/wetness according meteorological events (heatwave, drought, rainfall). proposed can, thus, offer aiding short-term disasters, like flash droughts floods.