Comparison of genetic programming and HEC-HMS as a conceptual model in simulating rainfall-runoff time series

Forecasting runoff is an essential task in hydrological modeling. The factors influencing the rainfall‑runoff process vary in space and time, introducing uncertainty and complexity that often lead to modeling errors and make accurate prediction challenging. This study employs a genetic programming (GP) model to simulate the rainfall-runoff process in the Khorramabad River basin in Iran. To validate the performance of the proposed GP approach, its results are compared with those from the conceptual HEC-HMS model. First, the HEC-HMS model was calibrated and used to simulate the rainfall-runoff process for four flood events in the study area. Subsequently, the GP model was applied as a data-driven rainfall-runoff model for the same events. A sensitivity analysis was conducted by sequentially excluding individual sub-watersheds, identifying those critical to cumulative flood discharge. Additionally, the Curve Number (CN) parameter was thoroughly investigated. The performance of the two simulation approaches was compared using several criteria: the root mean square error (RMSE) of peak discharge, mean absolute error (MAE), and the coefficient of determination (R²). For the GP model, R² values ranged from 0.88 to 0.94 across the four events, whereas HEC-HMS gave R² between 0.65 and 0.79. The results indicate that, under the data-rich conditions of this study where long-term hydrological records were available, the GP model provided more accurate simulations of the rainfall-runoff process compared to the event-calibrated HEC-HMS model in the Khorramabad River basin.