하천 유량 예측 정확성 향상을 위한 강우-유출 상관성 비교 분석

Forecasting streamflow is essential for preparedness against flash floods triggered by intense rainfall and for establishing reliable early-warning systems. Streamflow is primarily influenced by rainfall within watershed, which has led to the development of various streamflow prediction techniques. Conventionally, streamflow prediction has been preformed using total rainfall and total runoff data. However, this approach has some limitations in that it includes rainfall events on dry soil or that do not substantially contribute to runoff. In this study, we attempted to investigate more suitable data for streamflow prediction by performing correlation analysis for total rainfall-total runoff sets and effective rainfall-direct runoff events. As a result, the effective rainfall-direct runoff combination relatively achieved higher correlation coefficients and lower error metrics (RMSE, MAE) compared to the total rainfall-total runoff combination, indicating it could improve the accuracy of streamflow prediction. Furthermore, the initial loss ratio and antecedent soil moisture conditions exerted significant control on both the correlation structure and error magnitude. In this study, we investigated the potential of using effective rainfall-direct runoff sets to improve the reliability of streamflow prediction. The findings are expected to contribute to the future development of data-driven river flow prediction models.