Wind downscaling

[ArcticSnow]

In the current implementation, wind is simply interpolated from ERA5 pressure level wind speed, direction. Could we introduce other methods to downscale wind?

• Option 1: Winstral 2017 shows a statistical downscaling method that could be a starting base for implementing such method using WMO station network and ERA5 dataset
• Option 2: further advanced ML method between obs. and ERA5 pressure level.

What about direction? Could we derive physiographic metric reflecting the surrounding context to alter wind direction? How needed is this?

For reference the HICAR tool:

• https://github.com/d-reynolds/HICAR, https://gmd.copernicus.org/preprints/gmd-2023-16/