machine learning and hybrid models for assessing climate change impacts on runoff in the kasilian catchment, northern iran

This study evaluates the performance of artificial neural networks (anns), gene expression programming (gep), and the hec-hms models in assessing the impacts of climate change on runoff in the kasilian catchment, northern iran. daily data from 2007 to 2021 were divided into calibration (2007–2018) and validation (2018–2021) periods. the results indicate that gep and ann models surpassed the hec-hms model across all performance metrics, including rmse and nse, when applied individually. furthermore, hybrid models, integrating hec-hms with gep and hec-hms with ann, exhibited superior performance compared to individual machine learning (ml) or hec-hms models. input variables (temperature and rainfall) were generated using lars-wg software, incorporating five climate models and the ssp585 scenario for future climate change studies. additionally, these hybrid models were used to forecast runoff changes for the observed period (2007-2018) and future periods (2031-2050 and 2051-2070). the results show a rise in average annual precipitation, extreme precipitation events, and precipitation intensity, implying a higher likelihood of flooding and erosion in the future for the kasilian catchment and similar small catchments in north of iran.