Combining Climate Datasets for Temperature Mapping

Résumé fourni par la source

Accurate temperature information is essential for climate assessment, environmental modelling, and hydrological analysis; however, mountainous regions with sparse meteorological networks often lack spatially continuous observations. This study aims to identify a suitable interpolation method for observed temperature data and to develop a hybrid modelling approach to improve temperature estimation in data-scarce regions. Temperature records from eight stations in Türkiye’s Eastern Black Sea Region were analysed using Inverse Distance Weighting (IDW), Kriging, and Spline interpolation. Based on statistical performance metrics, IDW provided the most reliable spatial representation and was therefore selected as a reference surface, and a hybrid machine-learning model was subsequently developed using a Random Forest algorithm by integrating the WorldClim v2.1 and CHELSA v2.1 datasets. The hybrid model showed strong agreement with the IDW-derived reference surface (MAE = 0.32, RMSE = 0.48, R² = 0.94), highlighting the potential of integrating open-source climate datasets in data-scarce mountainous regions. Further validation using a denser and independent station network would strengthen the applicability of the approach at finer spatial scales.