Satellite information excels at rapid, large-scale data acquisition, which is crucial for disaster response, waterresource management, etc. However, satellites alone cannot continuously capture temporal changes in reservoirs.To address this limitation, this research integrates modeling approaches to estimate water level (storage), inflow,and outflow dynamics.
The study focused on six small- to medium-sized reservoirs in South Korea. Time-series data on water surfacearea variations were extracted from Landsat and Sentinel-2 optical satellite imagery on Google Earth Engine,employing techniques such as the Normalized Difference Water Index (NDWI), Canny edge detection, and Otsu'sthresholding. To bridge gaps in satellite-derived observations, this study supplemented discontinuous data withexisting relationships among water level, storage volume, and surface area, combined with watershed andreservoir operation models to enable continuous monitoring.
For watershed modeling, the HEC-HMS model was employed to simulate rainfall-runoff. A Python-basedReservoir Operation Module (ROM) was developed to simulate water level and outflow. ROM uses a storagerouting function and incorporates site-specific empirical relationships between water level and storage. TheNewton-Raphson numerical method was applied to support continuous daily simulations. The module supportsboth gauged and ungauged conditions, allowing estimation of key variables even in the absence of observationaldata.
The module was validated using observational data from the study reservoirs. Simulations were conducted for thefollowing periods: 2005?2023 for Gwangdong, Daegok, and Sayeon Dams; 1997?2022 for Idong and Gamdonreservoirs; and 2007?2022 for Cheontae reservoir. The results showed strong agreement between simulated andobserved values, with simulation accuracy largely depending on the reliability of the water level?storage?surfacearea relationships...