Climate change in Iran is significant, as reduced rainfall adversely affects both biological and social systems. This study aims to long-term predict rainfall changes based on social and economic scenarios from the sixth climate change report (HistSSP126SSP245SSP585) in the Kermanshah synoptic station. Different machine learning models, have been employed to analyze data from three CMIP6 public circulation models. These models are well-established for classification and prediction tasks. The ML-based downscaling models will estimate monthly rainfall for three time periods: 2026–2050, 2051–2075, and 2076–2100. These predictions will be made under three different scenarios: SSP1, SSP2, and SSP5. Historical monthly rainfall data from a Kermanshah station (1990–2014) have been divided for model training and testing. The models were checked and adjusted using MAE, MSE, RMSE, R², and NSE to see how well they performed. Results show no significant changes in the prediction results for SVR and RF models, with the best climate models varying by region. In all scenarios, the CANESM5 model closely matches the Random Forest predictions. Projected declines in annual rainfall range from 31% to 33% across scenarios and periods, with a multi-scenario average of 32% by 2100.
Almasi et al. (Fri,) studied this question.