I’m trying to reproduce AIFS forecasts. I’m wondering which timestamp to download for the static fields.
In particular I try to run AIFSv1.1 which requires static variables lsm, slor, sdor, and z_sfc, essentially all about orography. But the root of the question is independent of model version.
I see multiple ways to get slightly different static fields, especially over poles and mountain ranges:
- CDSAPI for dates 1940 (ERA5), 1979 (training start date), 202x (training end or recent)
- opendata
Which static variables timestamp were used for training and hence should be used for inference? Found no exact mention in the papers or docs.