NVIDIA
TempestExtremes tracking
Earth2Studio Pull Request
Description
This PR adds a connector to TempestExtremes in the form of a diagnostic model. Since TE natively only supports command-line interface, the model is implemented by spawning a sub-process call. The command for that call has to be provided by the user, providing flexibility in usage and installation of TE.
Due to the nature of track data (multiple orders of magnitude smaller than field data and unstructured), the diagnostic here does not return the tracks as tensor, but finishes with TE writing its file to disk. That leaves the question if this and potentially other trackers fit in this category. Maybe this PR can be basis for discussion.
Checklist
- [x] I am familiar with the Contributing Guidelines.
- [x] New or existing tests cover these changes.
- [ ] The documentation is up to date with these changes.
- [ ] The CHANGELOG.md is up to date with these changes.
- [ ] An issue is linked to this pull request.
Dependencies
hi @NickGeneva @dallasfoster , from my side the PR is ready to review. testing and ensemble size>1 is now supported as well.
Are there any instructions for installation or checks if the software is not available?
Are there any instructions for installation or checks if the software is not available?
yes, just added install instructions to the docs and checks if software is available is done at class initialisation. Since TE is not available as python package, users have to install it (locally) on thier system. AT class ini, they have the option to provide the path to the executable and until that path is known it is not possible to test install.
The default dependency failure does not work here for the same reason, and also cannot be hacked to write a message which is reasonable for the present case. So I provided a custom one. Hope that's fine
would a small example of using TE help? if so, where should I put it? could do at the end of the tempest_extremes.py file, just not sure if it justifies a full example.