This repository contains the notebooks used for creating test data for the packages developed within the MLLAM community.
For all datasets the intention is that they overlap in both space and time. The
spatial domain is set to be roughly 100x100
grid points centered on
Copenhagen, Denmark. The temporal cover is set to be the first two weeks of
April 2022.
All test datasets are stored in the mllam-testdata
bucket on the European Weather Cloud (EWC): https://object-store.os-api.cci1.ecmwf.int/mllam-testdata/
Currently, two datasets are available:
danra_cropped
: reads DANRA zarr dataset directly from European Weather Cloud (EWC) and stores into the testdata bucket on EWC. These are stored in zarr format and are used for testing both themllam-data-prep
andneural-lam
packages.meps_example_reduced
: relies on MEPS example files for April 2022 which are included with neural-lam. There are stored as.npy-files
and stored into.tar.gz
file stored in the testdata bucket on EWC. These are used for testing theneural-lam
package.