Comparative Performance of Global Datasets and Ground‐Based Precipitation and Temperature Products in the Eastern Mediterranean Basin: The Case of Türkiye

ABSTRACT Accurate climate data underpin water resource management, agricultural planning, and disaster risk reduction globally, yet obtaining reliable temperature and precipitation estimates in mountainous regions remains challenging due to sparse monitoring networks and complex topography. This data gap is particularly acute where stations are concentrated in accessible lowlands, leaving high‐elevation terrain poorly monitored. Global climate datasets integrate satellite observations, numerical models, and station records to address these deficiencies, but their reliability in topographically complex regions requires rigorous validation, particularly where orographic effects dominate precipitation patterns. This study evaluated four global datasets—Climatologies at High Resolution for the Earth's Land Surface Areas (CHELSA v2.1), WorldClim Global Climate Data (WorldClim v2.1), Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS v2.0), and ECMWF Reanalysis v5–Land (ERA5‐Land)—against data from 1005 Turkish State Meteorological Service stations across Türkiye, spanning seven geographic regions including 315 long‐term records (≥ 25 years) and 690 shorter records (12–25 years). Lin's Concordance Correlation Coefficient served as the primary agreement metric. Temperature estimates demonstrated high concordance (CCC > 0.82; r > 0.88; MAE: 0.5°C–1.55°C), with WorldClim (CCC = 0.940) and CHELSA (CCC = 0.944) performing optimally. Precipitation estimates revealed a substantial regional variability. CHELSA ranked highest nationally (CCC = 0.824), followed by ERA5‐Land (0.760), CHIRPS (0.742), and WorldClim (0.712). In orographically complex settings, WorldClim exhibited severe deficiencies, including sign‐reversed elevation–precipitation relationships in the Mediterranean region (CCC = 0.081) and underestimation exceeding 60% at high‐precipitation sites. Critical sampling bias emerged: long‐term stations concentrated at 681 m mean elevation versus 1131 m national average, constraining validation in mountainous terrain. These findings demonstrate that localised bias correction, high‐resolution topographic integration, and hybrid downscaling approaches are essential for improving precipitation estimates in heterogeneous landscapes, with direct implications for hydrological forecasting and climate impact assessments in mountainous regions worldwide.