SED_MORPH causes blow up after restart

Sediment modeling collaborators: issues, applications, information exchange

Moderators: arango, robertson, rsignell

Post Reply
Message
Author
c.drinkorn
Posts: 110
Joined: Thu Mar 08, 2018 2:47 am
Location: German Research Centre for Geosciences

SED_MORPH causes blow up after restart

#1 Unread post by c.drinkorn »

HI all,

I have been looking into this issue for quite some time now and wrote about pre-steps of it in other threads. At this point I am not getting any further and desperately hope for help by the community.
Here is a short outline:
I am running a high-resolution (delta x ~ 20km) Arctic application including sedimentation, also allowing the bathymetry to evolve with time. The simulation is supposed to run for ten years with a timestep of 10 minutes. Barotropic timesteps are 40 each baroclinic, Every year there should be a restart from the previous one.
The problem is:
When switching on SED_MORPH, I get a blow up at the fifth timestep. The diagnostics are NaNs for the energies, maxspeed and volume but zeros for the Courant numbers. The values before look just fine (I set ninfo to 1) and there is no obvious increase or so. I can't really look into previous fields because no history nor average fields have been written out until then. In the restart file the momentum fields are NaNs, the bathymetry looks fine but the bed thickness is zeros everywhere.
The Error comes from main3d after the massfluxes and diagnostics are called (line 298). I kind of have the feeling Hz is the cause of the issue but I couldn't get behind it yet. In set_depth Hz is calculated from h and for SED_MORPH there is an extra line replacing the pre-step bed thickess by the actual. Maybe something goes wrong here with the restart bathymetry?
THe Courant numbers in diags.F can become zero if u2v2 is zero. This will also cause maxspeed to become a NaN since it would be the sqrt of zero. But why are the energies NaNs? They can only become that if one of the input vars are NaNs...
But these are all just guesses... :?:
I just found out that after restart settling velocity and erosion stress both jump to 1 for the blow up timestep. Also mud_frac becomes zero (should be one). The parameters are read in from the sediment.in correctly at the beginning, though. The value of one could be due the calculation. With a mud_frac of 0 leading to a mud-mass of zero, too, erosion stress and settling velocity probably fall to the alternative values of MIN/MAX functions. But I'm not sure...
I am really thankful for any hint! :idea:

Post Reply