You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
I have implemented multi-LAM support for anemoi-graphs, extending its functionality to handle multiple Limited Area Models (LAMs) with hierarchical masking and varying resolutions. This feature allows the graph to handle different overlapping (or not) LAM domains, such as MEPS and AROME-Arctic.
Since all development now takes place in the consolidated anemoi-core repository, I would like to discuss the next steps for merging this feature, if appropriate.
Describe the solution you'd like
If this is of interest to others than MET Norway, I propose integrating the multi-LAM support functionality into anemoi-core. The implementation is compatible with the latest anemoi-datasets updates, which include multi-LAM hierarchical masking.
Describe alternatives you've considered
None at the moment, as this implementation addresses the need for managing multi-LAMs in the graph.
Additional context
Here are some key points for discussion:
Need for Discussion or Direct PR:
Should we open a PR directly for the implementation, or is a preliminary discussion needed to align on design and integration details?
When is good timing?
Compatibility with anemoi-core:
The implementation is compatible with the latest anemoi-datasets multi-LAM updates.
Are there additional requirements specific to anemoi-core that should be addressed?
Testing:
Do we need additional testing steps within the anemoi-core context?
Below is an example of how the mesh looks with different refinements and hierarchical cutout for ERA5, AROME-Arctic, and MEPS.
Hi Paulina, thank you for bringing this up and already working on an implementation! I think the best approach would be to open a draft PR so we can review and discuss any necessary changes directly there. I'm also happy to meet if you'd like to discuss the work further. Just a note that in anemoi-core, we'll need to wait for the CI pipelines to be set up and successfully passed before merging the changes (this is a work in progress)
Thank you, Mario, for your feedback! Opening a draft PR sounds like a great plan, but since there’s no rush on my end, I’m happy to wait until the CI pipelines are fully set up and operational before submitting it.
Is your feature request related to a problem? Please describe.
I have implemented multi-LAM support for
anemoi-graphs
, extending its functionality to handle multiple Limited Area Models (LAMs) with hierarchical masking and varying resolutions. This feature allows the graph to handle different overlapping (or not) LAM domains, such as MEPS and AROME-Arctic.Since all development now takes place in the consolidated
anemoi-core
repository, I would like to discuss the next steps for merging this feature, if appropriate.Describe the solution you'd like
If this is of interest to others than MET Norway, I propose integrating the multi-LAM support functionality into
anemoi-core
. The implementation is compatible with the latestanemoi-datasets
updates, which include multi-LAM hierarchical masking.Describe alternatives you've considered
None at the moment, as this implementation addresses the need for managing multi-LAMs in the graph.
Additional context
Here are some key points for discussion:
Need for Discussion or Direct PR:
Compatibility with
anemoi-core
:anemoi-datasets
multi-LAM updates.anemoi-core
that should be addressed?Testing:
anemoi-core
context?Below is an example of how the mesh looks with different refinements and hierarchical cutout for ERA5, AROME-Arctic, and MEPS.
@JPXKQX
Organisation
Norwegian Meteorological Institute
The text was updated successfully, but these errors were encountered: