Abstract

The development of the rainfall–runoff relationship constitutes a fundamental aspect of hydrological modeling. Given the inherent complexity of this relationship, accurate runoff prediction plays a pivotal role in water resources planning and management. This study investigates the rainfall–runoff relationship within the Khorramabad River Basin, employing simultaneous data from the Khorramabad Synoptic Station and the Cham-Anjir Hydrometric Station. Rainfall–runoff modeling was conducted using two conceptual hydrological models WEAP and IHACRES as well as three artificial intelligence (AI) approaches, namely Artificial ANN, ANFIS and SVM, to estimate runoff. The modeling period for all models extended from October 1956 to September 2024, except for the WEAP model, for which the period October 2010 to September 2023 was selected due to the large number of input parameters required. For the AI-based models, 80% of the data were used for training and 20% for testing. The performance of the models was evaluated using standard statistical indicators, including the R², NSE and RMSE. The results indicated that, among the hydrological models, WEAP outperformed IHACRES, and among the AI models, ANFIS exhibited superior performance compared to ANN and SVM. Overall, the ANFIS model demonstrated the best performance among all models employed, with R² = 0.96, NSE = 0.98, and RMSE = 2.08 during the training phase, and R² = 0.94, NSE = 0.87, and RMSE = 1.93 during the testing phase. Consequently, the findings suggest that artificial intelligence models generally outperform conceptual hydrological models in simulating the rainfall–runoff process.