WindWing data archive: prepared ERA5 wind profiles for migration inference

This archive provides prepared weather data for WindWing, a Python package for numerical inference of wind-assisted migration routes and multi-year corridors. The WindWing package is maintained separately on GitHub. Coverage. The archive contains all twelve months for all 62 research weather years, 1961–2022 inclusive. Training and test-year labels are retained as metadata and do not restrict data availability. Archiving these input fields does not fit, tune or select a migration model. Wind fields. Global zonal and meridional wind components are provided at six-hour UTC intervals on the original WeatherBench 2 64 × 32 longitude–latitude grid (5.625° spacing), at eight heights above ground: 100, 500, 1000, 1500, 2000, 3000, 4000 and 5000 m. The prepared float32 profiles are exported without further temporal or spatial aggregation or precision reduction. Processing and files. The existing profiles were derived from ERA5 pressure-level wind and geopotential fields, using surface pressure to exclude underground levels and a 10 m wind anchor for linear interpolation to geometric heights above ground. Unbracketed heights are not extrapolated. Annual archives contain monthly NumPy wind arrays, coordinates and timestamps; a separate static archive contains terrain and dataset metadata. The release manifest supplies archive checksums and processing provenance for reproducible downloads and use with WindWing. Source and attribution. ERA5 is produced by ECMWF for the Copernicus Climate Change Service and accessed through the WeatherBench 2 archive. Contains modified Copernicus Climate Change Service information (2026). Neither the European Commission nor ECMWF is responsible for any use of the information. The upstream ERA5 licence remains applicable. See the WeatherBench 2 data guide and Hersbach et al. (2020), The ERA5 global reanalysis.