model blow-up due to abnormal temperatures

[leonjld]

I was running a tidal simulation when it blew up after 1 day. This blow-up is really strange: I have checked ubar, vbar, u, v, w and zeta of the last time step at blow-up (in restart file), and found nothing abnormal. However, the temperature has extreme values at two locations, one at the southwest corner of the domain and the other in an inlet. Since temperature is not of concern for our study, it was originally set to be uniformly 10 degree throughout the whole domain, and there are no temperature flux anywhere. The only forcing file is the tidal netcdf forcing which forces tidal harmonics of elevation and velocity at all three boundaries. I have tried reducing the time steps and bottom frictions, but had no luck.

The same settings have been applied to the model to simulate other areas with no problem. I am really puzzled what is the reason for the abnormal temperature generated this time?

[kate]

Do you need the 3D flow? Have you tried it without SOLVE3D?

On the river input, you have to specify the tracers on influx. You can set it to ten, but you have to set it or you will be doing the unstable downstream advection.

[jacopo]

try using MPDATA

[leonjld]

Do you need the 3D flow? Have you tried it without SOLVE3D?

On the river input, you have to specify the tracers on influx. You can set it to ten, but you have to set it or you will be doing the unstable downstream advection.

I use the 3D flow. Since other cases also use SOLV3D and no temperature variation was found in those cases. and I don't have river influx.

[drews]

This problem sounds like one I had last year:

Water temperature from -492 to 602 degrees Celsius?
viewtopic.php?f=17&t=995

As Kate noted, the model involves 3D flow in shallow water. My fix was to #define MY25_MIXING in myModel.h, which turns on Mellor-Yamada vertical mixing. Try that and see what happens.

Carl

[leonjld]

thanks for all your suggestions.

it seems that the abnormal temperatures are caused by the discontinuity of the tidal forcing at the boundary. The figure shows where the tidal forcing data are (red area) after interpolating the adcirc data to ROMS grid. The blue area are water cells without adcirc tidal data. Since the area with no tidal data is given zeros amplitudes and phases for all harmonics, discontinuity occurs at the southern boundary (southwest corner, to be specific). Inner grids are ok since only the data at boundaries are utilized by ROMS.

The discontinuity of the tidal forcing seems to be the villain for causing a false temperature fluctuation at the southwest corner (due to some instability?), which was then spreaded to the inner grid and cause false temperatuer fluctuation in an inlet farther north.

After I masked out the blue area at the southwest corner, the blow-up problem has now gone.