search

JCSDA-internal / soca

Sea-ice Ocean Coupled Assimilation
Apache License 2.0
12 stars 4 forks source link

Rename Field mask metadatum #1078

Open svahl991 opened 5 days ago

svahl991 commented 5 days ago

Description

Changes Field metadatum key due to https://github.com/JCSDA-internal/oops/pull/2759

Dependencies

build-group=https://github.com/JCSDA-internal/oops/pull/2759 build-group=https://github.com/JCSDA-internal/fv3-jedi/pull/1273

travissluka commented 4 days ago

tagging @hga007 , not sure if you're using interp_source_point_mask anywhere in your code as well

hga007 commented 3 days ago

@travissluka: No, I am not using the interp_source_point_mask. Thank you! I am still determining what type of mask it is. ROMS has land/sea masks at every staggering grid cell point (center, corner, east-west sides, and north-south sides). Our experience with high-resolution coastal applications is that linear or nearest-point interpolation/extrapolation of the land/sea mask is not a good idea because it can alter the river outflow in estuaries and bays, which may change the coastal buoyancy-driven currents.

Now, @fmahebert mentioned that this mask is used in oops for unstructured interpolators, which I think is or will be exclusively done with ATLAS, which only allows the A-grid (grid cell center). We specify the appropriate mask from the application geometry in such FunctionSpace. Even if we are horizontally interpolating for fine-to-coarse and vice versa, why do we need to interpolate the mask between grids? We could specify the appropriate masks for fine and coarse grids to determine the unmapped points and perform a second-step extrapolation via different methodologies (nearest point, nearest inverse distance, creep fill, or other). This topic is critical in coupled atmosphere-ocean applications with incongruent grids and land/sea masks. It gets more complicated when the atmosphere model has partial masks.