Test grid in upwelling example

[RubenDiez-Lazaro]

Hi!

I am trying to do a working application for a real scheme, but I always get a BLOWUP termination.

Now I don't sure about my input data, thus I have decided to try my data steep by steep: first of all, i must probe the grid...

At the beginning I compiled and run the upwelling example (with 6x7 size partition) with good results.

Next, I modified the upweling.h file, changued the "#define ANA_GRID" directive with the "#undef ANA_GRID" one, so I have specified my test grid on the ocean_upwelling.in file (GRDNAME)...

The only others changes that I have done in the ocean_upwelling.in file were:
- The Lm and Mm values, according to my test grid.
- The NtileI and NtileJ (6x7), according with my cluster characteristics,

Then I prepared some test grids for doing probes...

The result was that if the test grid has some land (masking or marked by a depth of 0), the application BLOWUPs. But if the test grid has not land, the application runs ok...

I have tried to add some grid and mask related definitions in "upwelling.h", like "MASKING", "ANA_MASK" or "CURVGRID", and probe change the "THETA_B", "THETA_S" and "TCLINE" parameters in the "ocean_upwelling.in" file, with identical results....

The test grid I have used is a slope in the west-east direction...
You can see a representation of the bathymetry of the test grid without mask at http://web.usc.es/~rdlazaro/slope_without_land.png and the grid with mask at http://web.usc.es/~rdlazaro/slope_with_land.png

When the application does a blowup termination (land test), the energy (kinetic and potential) appears as NaN...

Others strange info from the roms output where:

Code: Select all

[...]
 Minimum Z-grid spacing, DZmin =             NaN m
 Maximum Z-grid spacing, DZmax =             NaN m
[...]
Maximum grid stiffness ratios:  rx0 =            NaN (Beckmann and Haidvogel)
                                                rx1 =            NaN (Haney)

Initial basin volumes: TotVolume =                   NaN m3
                                MinVolume =                   NaN m3
                                MaxVolume =                  NaN m3
                                Max/Min =  0.00000000000000E+00
[...]

Do someone know where my mistake is???
How can I do for use a grid with masking land in the upwelling example??

Best regards.

[jcwarner]

what are the grid dimension?
Does h == 0 anywhere ?

[RubenDiez-Lazaro]

The masking grid was made with constant slope from -1000 m to 0m... then masking as land until the end of domain (East side)

The grid without land was made with a constant slope from -1500 (west side) to -100 (east side).

Do you suggest that may be depths with an value of 0 and without masking???

I will chack it...

[RubenDiez-Lazaro]

Yes...

When i change sentences of type

Code: Select all

if(depth>0.0) then depth=LAND

by other of type

Code: Select all

if(depth>0.5) then depth=LAND

all my test grids works fine

Thanks you very much...

Now i will continue with my step by step probe...

[arango]

The depth (h) in ROMS can never be less or equal than zero. It is always positive Neither can be masked to a value of zero

Recall that the vertical coordinates in ROMS are terrain-following and the number of vertical levels are distributed within the depth of each grid point. The thickness of the model level at each grid point need to have a finite value greater than zero. This implies that you need to choose a depth value for the shallowest grid and has to be greater than zero and set land regions to that value. The land/sea masking is done on the model state variables and not on the grid variables. This shallowest value depends on you application and resolution. The smaller the depth that you choose, the smaller the time-step that you need to not violate vertical CFL critiria.

[jcwarner]

Unless, of course, you activate
#define WET_DRY
Then the model can have h = to any value, including 0, any + value, or any - value.

[RubenDiez-Lazaro]

Thank you very much for the answers....
I'm still working on it...