적설 및 융설 영향을 반영한 GR4J 모형의 유출 모의 평가

This study aims to evaluate and improve the reproducibility of winter–spring (December–April) streamflow using a GR4J-SN model that explicitly accounts for snow accumulation and melt. Precipitation, air temperature, and dam‐inflow data were used to estimate seasonally varying parameter sets for three regimes—December–February, March–April, and May–November—via a multi-objective function combining the Kling–Gupta efficiency (KGE) with RMSE computed separately for low and high flows. Model skill was assessed for the full period and by month using KGE, Nash–Sutcliffe efficiency (NSE), correlation coefficient (CC), RMSE, and PBIAS. In addition, standardized anomaly plots and conditional probability analysis were employed to examine temperature–runoff response characteristics. Results show that in March–April, when snowmelt strongly influences runoff, KGE improved from 0.400 and 0.471(GR4J) to 0.754 and 0.767(GR4J-SN) and NSE from 0.145 and 0.576 to 0.562 and 0.578, indicating clear gains in reproducing both peak flows and low-flow conditions. The systematic overestimation of winter flows was also alleviated, while annual KGE and NSE remained ≥0.90, demonstrating that the December–April improvements were achieved without degrading performance in other seasons. These findings suggest that a snow-aware GR4J-SN model provides sufficient reliability for dam-inflow simulation and offers practical support for spring flow forecasting and water-resources management in snow-affected mountainous basins.