Assessment of seven different global climate models for historical temperature and precipitation in Hatay, Türkiye

Global climate models are important tools for estimating the possible future impacts of climate change and developing necessary adaptation strategies. This study assessed the suitability of global climate models for local climate projections in Hatay, Türkiye. Temperature and precipitation data from different Coupled Model Intercomparison Project Phase 6 climate models were compared with ground-based observations. For stations lacking historical data, multilayer perceptron artificial neural networks were used to generate data. These networks were trained with data from neighboring stations from 1980 to 2014. The most suitable global climate model was determined using a multi-criteria decision-making approach. As a result of the study, it was determined that the multilayer perceptron models effectively generated long-term temperature data with a normalized root mean square error of less than 0.50. Precipitation estimates, while less accurate, achieved reasonable accuracy with a normalized root mean square error of less than 0.70. The evaluation of global climate models revealed a tendency to underestimate minimum temperatures and overestimate maximum temperatures and precipitation. Specifically, the EC-EARTH3, CMCC-ESM2, and MPI-ESM1-2-HR models excelled in maximum temperature estimations; the CMCC-ESM2, GFDL-CM4, and TAIESM1 models were superior for minimum temperatures; and the EC-EARTH3, GFDL-CM4, and MPI-ESM1-2-HR models performed best for precipitation. The findings of this study will provide a framework for the assessment and selection of appropriate climate models for local regions and will help to develop targeted adaptation strategies.