Fix vertical coordinate weights #317
Conversation
|
@pwolfram Please test this on your wetting/drying case. |
|
@mark-petersen, could you make sure to test this with |
pwolfram
left a comment
pwolfram
left a comment
There was a problem hiding this comment.
Reviewed by inspection-- will pull commits onto my branch for future testing.
|
@mark-petersen, this works with the drying slope case for sigma layers-- thank you!
|
pwolfram
left a comment
pwolfram
left a comment
There was a problem hiding this comment.
Tested via the drying slope test case.
|
@mark-petersen, this is great and I'm happy for a merge. Thanks! |
|
@vanroekel is out of town, removing him from reviewer list. |
src/core_ocean/Registry.xml
Outdated
There was a problem hiding this comment.
@pwolfram I realized that we don't want this alteration in general, so I am protecting it with this flag. In general in the ocean, we need to have the thickness vary in proportion to the weights*resting thickness (i.e. original distribution), not the weights only. You can run wetting/drying test cases with the
config_ALE_thickness_proportionality = weights_only
setting. But this breaks ice shelf cavities, as @xylar feared, because you need to 'squeeze the accordion' in proportion to the original thicknesses, especially for extreme compression, like we have under ice shelves.
This commit is now bit-for-bit with previous for MPAS-O and E3SM. We will need to think about this more carefully to put wetting/drying along coastlines and severe SSH alterations below ice shelves in the same domain.
|
BFB against ocean/develop for nightly test suite with gnu on grizzly. Also tested debug using gnu and intel. |
In eqn 6 of Petersen et al 2015, remove h_k^{rest}. This allows the
resting thickness to be zero, which can occur with wetting/drying and
sigma coordinates.
For the initial condition, require that
sum (restingThickness ) = bottomDepth for each column.
This is always done for z-star global ICs, but should be done for sigma
as well.
mark-petersen
force-pushed
the
ocean/fix_ALE_weights
branch
from
b2daa8c to
6975e0e
Compare
|
rebased. |
This PR does two things: Add optional new weighting in ALE SSH thickness distribution Add tapered weights options for z-star The first is used for wetting/drying, to allow bottomDepth to be zero, which can easily occur on a beach. It changes the weighting for the z-star distribution from vertCoordMovementWeights(k) * restingThickness(k, iCell) to vertCoordMovementWeights(k). The first formulation can be seen in eqn 6 of Petersen et al 2015. The new formulation stretches in proportion to the defined weights. The standard formulation stretched in proportion to the weights*restingThickness. That stretches deep layers more than shallow layers with surface gravity waves. These changes are protected by flags, so are bfb in MPAS-Ocean and E3SM. Fixes #316
This PR does two things: Add optional new weighting in ALE SSH thickness distribution Add tapered weights options for z-star The first is used for wetting/drying, to allow bottomDepth to be zero, which can easily occur on a beach. It changes the weighting for the z-star distribution from vertCoordMovementWeights(k) * restingThickness(k, iCell) to vertCoordMovementWeights(k). The first formulation can be seen in eqn 6 of Petersen et al 2015. The new formulation stretches in proportion to the defined weights. The standard formulation stretched in proportion to the weights*restingThickness. That stretches deep layers more than shallow layers with surface gravity waves. These changes are protected by flags, so are bfb in MPAS-Ocean and E3SM. Fixes MPAS-Dev#316
4 participants
This PR does two things:
The first is used for wetting/drying, to allow bottomDepth to be zero, which can easily occur on a beach. It changes the weighting for the z-star distribution from
vertCoordMovementWeights(k) * restingThickness(k, iCell)to
vertCoordMovementWeights(k).The first formulation can be seen in eqn 6 of Petersen et al 2015.
The new formulation stretches in proportion to the defined weights. The standard formulation stretched in proportion to the weights*restingThickness. That stretches deep layers more than shallow layers with surface gravity waves.
These changes are protected by flags, so are bfb in MPAS-Ocean and E3SM.
Fixes #316