multivariate gard implementation

[orianac]

as committed here ran a test case for western US using three features to predict precip. note that this will break any caching schemes previously set up as the run_id now includes the feature names.

flow is functional at this point with a few remaining TODOs:

• read in u/v winds from era5
• read in u/v winds from cmip
• read in pressure from era5
• read in pressure from cmip
• accommodate different transformations/detrending of different variables
• attr adjustment is hardcoded inside observations.py - add the additional variables (and consider reading those attr spec from somewhere else or maybe moving that attr adjustment to somewhere else)
• confirm that variables are passed correctly (order, shape) to sklearn
• why are there negative values in bias-corrected precip (see example in tasks.py)

@jhamman

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Updated
cmip6-downscaling	⬜️ Ignored (Inspect)		Jun 30, 2022 at 4:34PM (UTC)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[norlandrhagen]

pre-commit.ci autofix