Machine Learning Insights into Türkiye’s Climate Variability: Predictive Modelling and Spatial Analysis

This study focuses on predicting climatic parameters, including average temperature, relative humidity, and total precipitation, across various ecological zones in Türkiye using machine learning techniques. The study area encompasses Türkiye’s diverse climatic regions, including the Black Sea, Mediterranean, and continental climates as well as transitional zones. Utilizing meteorological and topographic data from the Turkish State Meteorological Service and digital elevation models, predictive models were developed using Random Forest (RF), Gradient Boosting Regressor (GBR), and Gated Recurrent Unit (GRU) deep learning methods. The performance of these models was evaluated based on R² and RMSE values, highlighting GRU as the most reliable method for predicting the average temperature and relative humidity, particularly when utilizing ten years of training data. Geostatistical methods were then employed to map the spatial variability of the predicted climatic parameters, illustrating temperature gradients across different regions of Türkiye and revealing trends in relative humidity and total precipitation. Overall, this study provides valuable insights into the application of machine learning techniques for climate prediction and offers a comprehensive understanding of climatic patterns across Türkiye’s ecological zones.