62 z ale remapping #188
62 z ale remapping #188
Conversation
|
The changes to MOM_regridding lead to NaNs in several of the ALE tests. I think this is why I had set the default minimum to that value (for stability). Can you revert that default value and add a method setMinThickness that you can use from within the diagnostics to set it to the tiny value you need? |
|
Doh! There is a method already - setRegriddingMinimumThickness() which is how MOM_ALE.F90 sets it for the model at run-time. |
there are many thin layers. mom-ocean#62
nichannah
force-pushed
the
62-z-ALE-remapping
branch
from
e9cf620
to
d38f434
Compare
nichannah
force-pushed
the
62-z-ALE-remapping
branch
from
a3c1d00
to
9ecfe58
Compare
|
OK, thanks I have fixed that. I'm no longer getting any results (timestats) changes from master. I still don't completely understand why remapping conservation checks are failing when using minthickness of 1e-3). |
| @@ -1920,6 +1918,10 @@ subroutine initialize_MOM(Time, param_file, dirs, CS, Time_in) | |||
| endif | |||
| call callTree_waypoint("state variables allocated (initialize_MOM)") | |||
|
|
|||
| ! Initialise the diag mediator, this needs to occur _after_ CS%h has been | |||
| ! allocated. | |||
| call diag_mediator_init(G, CS%h, param_file, diag) | |||
There was a problem hiding this comment.
How about providing a method to associate h, instead of moving this call so far?
…, avoids conflict with existing Z level diagnostics. mom-ocean#62
diagnostic remapping. The format is now:
DIAG_REMAP_Z_GRID_DEF = "FILE:OM3_zgrid.nc,zw"
Also use library to read netCDF files. mom-ocean#62
nichannah
force-pushed
the
62-z-ALE-remapping
branch
from
14aa761
to
d556620
Compare
|
Thanks for your comments, I've fixed those things. There's one thing I'm not sure about. I've changed the grid definition config to be like: DIAG_REMAP_Z_GRID_DEF = "FILE:," e.g. This means that when we move to other vertical coords there will need to be another config e.g.: DIAG_REMAP_SIGMA_GRID_DEF = "FILE:OM3_sigmagrid.nc,si" Also if we want to use a generated grid, can do: I'm not sure if this is what you had in mind. |
|
@nicjhan That's exactly what I had in mind. |
De-allocate Kd_bkgnd and Kv_bkgnd
* Change dumbbell initialization * Change in Dumbbell Layer Mode * Fix sponge diagnostics * Fix ALE Sponge Diagnostics * A minor style change removing spaces around = in optional. function arguments. * Fix ALE sponge diagnostics * character declaration fix
This pull request will change results. The line that causes the change is:
src/ALE/MOM_regridding.F90 122
Without this conservation checks fail when there are lots of 'squashed' layers. In model coords these have thickness 1e-10, whereas the new grid uses 1e-3