Watershed Runoff Correlation Analysis through the Development of a Synchronization Model for Satellite and Ground-Based Soil Moisture Data in Korea

Recently, Korea and the rest of the world have been experiencing significant changes in rainfall patterns, leading to an increase in sudden runoff caused by abrupt rainfall events. This phenomenon, referred to as “Abnormal flood,” has resulted in accumulating damage. Runoff responses to rainfall events vary depending on surface cover conditions. In the mid- and upstream regions of rivers with relatively low impermeable layers, antecedent soil moisture saturation plays a significant role. In Korea, soil moisture observations are characterized by much lower observation density and shorter data records compared to variables such as rainfall, river water levels, and discharge, making statistical estimation challenging. Additionally, uncertain soil moisture data from unmeasured watersheds have been used in traditional watershed runoff estimation models to adjust assumed values of spatial hydrological characteristics within watersheds, thereby correcting and back-calculating runoff. While runoff calculated based on such assumed soil moisture saturation values demonstrated high reliability under traditional hydrological conditions, the reliability of runoff prediction results has relatively decreased in the face of vast data produced by advanced real-time monitoring technologies and hydrological changes due to climate change. To address these issues, since 2022, this study has developed a model to produce relatively reliable watershed antecedent soil moisture saturation observation data by synchronizing satellite and ground-based soil moisture data. In Korea, ground-based soil moisture sensors measure soil moisture at depths of 0–10 cm, while satellite observation data measure soil moisture at depths of 0–5 cm. Therefore, synchronization of these two data sources is essential. This study developed a synchronization model for satellite and ground-based soil moisture data using machine learning and analyzed the correlation between watershed soil moisture variations estimated by the model and the occurrence of sudden runoff within watersheds. Although soil moisture saturation of watershed soils significantly impacts runoff hydrologically, soil moisture has shown relatively low importance in runoff estimation models due to the lack of accurate data. The high-resolution watershed soil moisture data produced by this study are expected to enhance the accuracy of runoff analysis results, thereby improving the reliability of anomalous flood prediction and analysis models.