Blow up when MASKING is activated!

[backkom]

Hi ,
I have a weired blow-up problem when ruuning a East china sea application.
My domain is 117°-130°E, 24°-41°N, with grid resolution of 2'(about 35000m).my minimum depth is about 5m and maximum depth is 5000m.model is driven only by tidal elevation.
Minimum available timestep is 20s by CFL conditions,and my barotropic timestep is 10s. but model blows up after 900s's run .Fist I think perhaps my topograhy is too steep,so smoothing is applied,but even rx0 is reduced to 0.2,and rx1 to less 5,the result is still the same.
so some test is made for my application.

case 1.
set constant depth for my whole grid,ie 500m,but model still blows up,which means topo is not the cause.

case 2.
reduce my timestep to 5s for barotropic,model blows up!

case 3.
set constant depth ,plus constant tidal forcing with amplitude of 0.6m, model blows up!

case 4.
the blow-up point is almost nearby the land,so I carefully check my land sea mask.I confirm that no isolated island or water point exists in my grid.but I still doub the mask.so I make the fourth test:
set constant depth,constant tidal forcing,and inactivate MASKING option in my header file. Model could runs well.

I am very confused and I have no idea what trouble with my mask.I also know that masking will be applied to the whole grid after time-forward calculation is finished,and also a very large value, 1e+037, will be set to land points when masking.I wanna to know whether the large values in land will affect the nearby water points?
Could somebody give some suggestions about how to get rid of this trouble, any reply will be appreciated.
Thanks in advance !

[kate]

How did you generate the land mask? Are you sure the u-mask and v-mask are consistent with the rho-mask? I edit my land masks with editmask (matlab) and it makes sure all the masks are consistent. I've not had this sort of trouble with the land mask, even with the complicated Aleutian islands.

[backkom]

kate,thanks for your reply!
my mask is generated by uvp_mask.m,and then I manually edit using editmask.m. the code is as follows:

Code: Select all

[Mp,Lp]=size(mask_rho);
M=Mp-1;
L=Lp-1;
%
mask_v=mask_rho(1:M,:).*mask_rho(2:Mp,:);
mask_u=mask_rho(:,1:L).*mask_rho(:,2:Lp);
mask_p=mask_u(1:M,:).*mask_u(2:Mp,:);

[backkom]

Hi,kate
I carefully check the mask_rho,mask_u,mask_v,just like what you said,there are some wrong mask points in mask_u and mask_v. and model seems works well now.
Thanks again for your kind help,kate !