Projecting and Downscaling Future Temperature and Precipitation Based on CMIP6 Models Using Machine Learning in Hatay Province, Türkiye

Abstract Projections for future changes in precipitation and temperature are essential for decision-makers to understand climate change impacts on any region in the world. General circulation models (GCMs) are widely used tools to assess the future impacts of climate change. However, since they are produced at global scales, they cannot provide reliable information at local scales. For this reason, downscaling applications have been applied in recent years. In this study, support vector regression (SVR), random forest (RF), and multiple linear regression (MLR) methods were evaluated to improve the forecast accuracy of EC-EARTH3 CMIP6 GCM outputs for the Hatay province of Türkiye. The results obtained from the models were compared with meteorological observation data on a monthly time scale. As a result of the study, RF (RMSE = 19.19–45.41) for precipitation projections and SVR for maximum temperature (RMSE = 1.49–2.23) and minimum temperature (RMSE = 1.44–1.69) projections were found successful compared to other methods. These methods were applied to GCM’s future outputs. According to the results, it was determined that there could be a significant increase in the annual average temperature in Hatay province under the SSP2-4.5 and SSP5-8.5 scenarios. It is also estimated that there may be an increase in temperature between 2.1 and 2.9 °C for the SSP2-4.5 scenario and 2.4 °C and 5.2 °C for the SSP5-8.5 scenario in the near (2020–2060) and far (2060–2100) future periods, respectively. It is also estimated that by the end of the 21st century, annual precipitation in Hatay province may decrease by approximately 10% for SSP2-4.5 and by approximately 20% for SSP5-8.5 scenarios.