Impact Analysis of Spatial Distribution Scenarios of Ultra-Short-Term Rainfall Observations on Streamflow Simulation Using the SWAT Model

Recent changes in rainfall patterns under climate change have increased the frequency and intensity of short-duration heavy rainfall, making short-lead-time flood forecasting increasingly important. This study aims (i) to incorporate spatial rainfall heterogeneity into the Soil and Water Assessment Tool (SWAT) by constructing 50 non-overlapping subbasin rainfall allocation scenarios using the Korea Meteorological Administration (KMA) 5-km ultra-short-term precipitation analysis grid data, and (ii) to compare parameter-calibration methods for hourly simulations using Grid Search and a Genetic Algorithm (GA). The GA, an optimization technique inspired by the principles of biological evolution, was employed to identify optimal parameter combinations. The study area is the Doma-cheon watershed in Wangsan-myeon, Gangneung-si. The SWAT setup comprised 11 sub-basins, with precipitation inputs taken from four KMA grid points (IDs 9, 14, 15, 16). Hourly runoff simulations and calibrations were performed for the top three rainfall events observed by in-situ gauges in August, September, and December 2023. Across all events, GA calibration yielded equal or superior performance relative to Grid Search. For Rainfall Event 1, the Nash–Sutcliffe Efficiency (NSE) reached 0.808 with GA versus 0.605 with Grid Search. While certain allocation scenarios consistently showed high performance across multiple events, the optimal scenario differed by event, indicating that event-specific spatial rainfall patterns substantially influence simulation performance. This study demonstrates the effectiveness of scenario-based spatial inputs and hourly calibration in event-scale SWAT simulations and highlights the need for continued research on integrating spatial rainfall heterogeneity into hydrologic modeling under increasingly frequent localized heavy rainfall events driven by climate change.