Spatial-Resolution Harmonization and Bias Decomposition of Open Global Climate Datasets over European Built-Up Areas: the Case of ERA5-Land and TerraClimate

Open raster climate datasets are increasingly used as background climate layers for urban monitoring, exposure assessment, and decision-support applications. However, different products may yield different values within the same built-up domain due to differences in spatial resolution, grid geometry, land-surface representation, temporal aggregation, and modeling assumptions. This study proposes a geospatial workflow to compare ERA5-Land and TerraClimate over homogeneous European built-up domains. Berlin, Milan, and Madrid were selected as representative metropolitan areas along a latitudinal and climatic gradient. Adaptive domains were generated in Google Earth Engine using constraints on built-up surface, low-relief topography, and Köppen-Geiger climatic coherence. TerraClimate was harmonized to the ERA5-Land reference grid geometry prior to extracting monthly descriptors of Tmin, Tmax, Tmean, and precipitation over 2000-2020. For each variable, area-weighted mean and area-weighted spatial standard deviation descriptors were computed using the GHSL built-up surface. Results show high consistency for Tmean, with RMSE below 1 °C and Pearson correlation above 0.997 in all domains. Tmin and Tmax show opposite systematic differences, suggesting a compression of the thermal range in TerraClimate relative to ERA5-Land. Precipitation is more domain-dependent, with the largest mismatch observed in Milan. The proposed workflow supports uncertainty-aware use of open climate raster datasets in geospatial applications for built-up areas.