High-resolution (100 m) daily temperature and relative humidity dataset for Belgium for 2020 from UrbNet

This dataset provides high-resolution (100 m) meteorological information covering Belgium for the year 2020, generated using the UrbNet machine-learning framework (Souverijns et al., 2026). It includes daily minimum, mean and maximum near-surface (2 m) air temperature and relative humidity.

UrbNet is a convolutional neural network trained to reproduce high-resolution simulations from the UrbClim urban climate model. By combining large-scale meteorological forcing from ERA5 reanalysis with detailed land-surface characteristics, including land use, vegetation cover, building density and topography, UrbNet generates spatially consistent meteorological fields at 100 m resolution across regional and national domains.

The dataset captures fine-scale spatial variations in meteorological conditions, including urban heat island effects, temperature differences between built-up and vegetated areas, coastal influences and topographic variations. It thereby provides substantially more spatial detail than conventional meteorological reanalysis products.

Dataset characteristics

• Geographical coverage: Belgium
• Temporal coverage: 1 January – 31 December 2020
• Spatial resolution: 100 m
• Temporal resolution: Daily
• Variables: Minimum, mean and maximum 2 m air temperature; Mean relative humidity
• Data source: UrbNet, driven by ERA5 meteorological forcing and high-resolution land-surface information

This dataset is released as a sample of the UrbNet high-resolution meteorological product, demonstrating its potential for urban climate analysis, heat exposure and heat stress assessments, climate adaptation planning, and environmental research.

Data developed under C3S National Collaboration Programme, funded by EC and implemented by ECMWF.

Methodological reference

Souverijns, N., Wouters, H., Veldeman, N., Francis, D., and Broeckx, J. (2026). UrbNet: A machine-learning framework to replicate and upscale 100 m UrbClim urban climate simulations to regional domains. Manuscript under reivew.