NCAR / DART

Data Assimilation Research Testbed
https://dart.ucar.edu/
Apache License 2.0
196 stars 145 forks source link

feat: bug: fug: mom6 perturbation model specific perturbation #730

Open hkershaw-brown opened 2 months ago

hkershaw-brown commented 2 months ago

Perturbing MOM6 to get some ensembles to play with. Currently (lazily) using the default perturbation routine, which perturbs everything in the state including land.

This gives some fuzz on the land:

c.TL319_t232.001 Screenshot 2024-09-05 at 4 25 36 PM

nwa12_test.3_0_a03a Screenshot 2024-09-05 at 4 28 47 PM

Compare to non-fuzz before perturbation Screenshot 2024-09-05 at 4 33 00 PM

who cares? You'll inflate the land because mean /= 0 (adds more fuzz), get_close makes 'dry' state far away so the assimilation update doesn't change the land, but still I think dangerous to fuzz. See https://github.com/NCAR/DART/issues/359 for missing_in_state comment that I have not rewritten yet. https://github.com/NCAR/DART/issues/359#issuecomment-2332614632

hkershaw-brown commented 2 months ago

note: land is land+ocean floor

todo @hkershaw get_state_meta_data / get_close_state should be checking ocean floor?

This ocean floor check wrong/not needed? https://github.com/NCAR/DART/blob/a85e162288a90da1fc1ed28fa4d80acb1ca80920/models/MOM6/model_mod.f90#L547-L549

hkershaw-brown commented 2 months ago

I'm not sure what the MOM6 grid looks like: Screenshot 2024-09-06 at 9 52 35 AM

2D wet/dry sufficient (squishy bottom grid follows ocean basin) vs. 3D wet/dry (grid extends into ocean basin)

hkershaw-brown commented 2 months ago

https://github.com/NCAR/DART/tree/pert_plus_missing-grid has pert_model_copies & missing grid from #727 so it can run