General Circulation Models (GCMs) represent a contemporary and advanced tool designed to simulate the response of climate systems to alterations in greenhouse gas levels. Increasing spatial resolutions of the outputs of GCMs on a regional scale requires a downscaling process. This study applied six Machine Learning (ML) models, named decision tree regression (DTR), support vector regression (SVR), artificial neural networks (ANN), K-nearest neighbors (KNN), Light Gradient-Boosting Machine (LightGBM), and Stochastic Gradient Descent Regressor (SGDRegressor), to downscale daily temperature data from CMIP6 models in Kohgiluyeh and Boyer-Ahmad, Iran. Observations from Nazmakan station were used for training (1995 -2009) and testing (2009 -2015). In addition, future temperature projections during 2015 -2045 were made under SSP2-4.5 and SSP5-8.5 scenarios. Results showed that LightGBM and KNN developed the most reliable results. Mann-Kendall\\\\\\\'s analysis confirmed a significant upward trend, predicting cooler summers and warmer winters. The predicted data was also validated against observations from the period 2015 -2022. This study highlights the strengths and limitations of nonlinear ML techniques and emphasizes the need for further research to enhance predictive accuracy and spatial resolution in statistical downscaling.

Keywords