tuning ice model ...

Discussion about modeling ice with ROMS

Moderators: arango, robertson

Post Reply
Message
Author
turuncu
Posts: 128
Joined: Tue Feb 01, 2005 8:21 pm
Location: Istanbul Technical University (ITU)
Contact:

tuning ice model ...

#1 Unread post by turuncu »

Hi,

I am running the ROMS-ice model in Caspian Sea. The problem is that the model generally underestimates the ice thickness and coverage when i compare with the ARCLAKE dataset. Based on my previous experience, the ice formation is closely linked with the depth of the sea and in my case, i set the minimum depth as 5 m for the northern part of the sea which is relatively deeper than the real bathymetry (around 2-3 m). I tried to decrease the minimum depth of the model to 3 m but the model crashed after couple iteration with the high current speeds. Of course, this could be triggered by the vertical grid and pressure gradient error. So, decreasing the minimum depth is not an option for now. I just wonder that is it possible to tune the ice model to increase the ice thickness? I could play with the ice emissivity to increase the ice thickness but it might be a better solution to do that with playing other parameters. So, i am looking forward to your suggestions about that.

Regards,

--ufuk

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: tuning ice model ...

#2 Unread post by kate »

I would try to see why the minimum depth of 3 m didn't work. Do you need a shorter timestep?

turuncu
Posts: 128
Joined: Tue Feb 01, 2005 8:21 pm
Location: Istanbul Technical University (ITU)
Contact:

Re: tuning ice model ...

#3 Unread post by turuncu »

Actually, i did not try to reduce the time step because i did not change the horizontal resolution. Currently, i am using following parameters to define the vertical grid,

Vtransform == 2 ! transformation equation
Vstretching == 4 ! stretching function

! Vertical S-coordinates parameters (see below for details), [1:Ngrids].

THETA_S == 7.0d0 ! surface stretching parameter
THETA_B == 0.2d0 ! bottom stretching parameter
TCLINE == 5.0d0 ! critical depth (m)

in this case DT is 300 s and NDTFAST is 30. Do you think that reducing time step can help to prevent the speed problem? If this is the case, i could perform a test simulation with 3 m.

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: tuning ice model ...

#4 Unread post by kate »

Some processes are limited by the horizontal grid while others are limited by the vertical grid. Since it is such a simple test, I would try it.

User avatar
shchepet
Posts: 188
Joined: Fri Nov 14, 2003 4:57 pm

Re: tuning ice model ...

#5 Unread post by shchepet »

Your choice Vtransform=2 ; Vstretching=4 in combination with such small value for critical depth

TCLINE=5.0d0

would yield an very tight packing of sigma levels toward the surface: suppose you have 40 layers
total. Then your uppermost grid box will be only 12.5cm thick, and so do other grid boxes just below.
Is this what you actually want?

I suggest you plot a relevant cross-section of your vertical grid using matlab and decide whether
you like it or not. Then choose parameters accordingly.

turuncu
Posts: 128
Joined: Tue Feb 01, 2005 8:21 pm
Location: Istanbul Technical University (ITU)
Contact:

Re: tuning ice model ...

#6 Unread post by turuncu »

Hi,

i tried to decrease the time step up to 180 s but i have still problem with the model and it is also running much more slower. It gives error related with ice frazil in the 27th day of the simulation.

13106 27 07:18:00 5.470609E-03 2.530466E+03 2.530471E+03 7.354573E+13
(37,088,32) 1.099318E-03 7.474667E-04 1.391403E-01 7.686928E-01
trouble in ice_frazil 56 34 -2.303653101454643E-003
3.451582526012355E-004 -1.719989149099419E-009 2.670898819416213E-005
-1.643950240595160E-004

and in some tests i got NaN errors in the output. So, the solution is not stable with this configuration.

I also updated the ROMS+Ice code to the latest version but still no success. I could change the vertical grid structure but in any case the grid will be dense in the surface with 3 m minimum depth.

--ufuk

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: tuning ice model ...

#7 Unread post by kate »

For code that's never supposed to be executed, I seem to be checking that ice_frazil print statement pretty often. The statement in question is:

Code: Select all

          IF (wfr(i,j) .lt. 0.0_r8) THEN
            print *, 'trouble in ice_frazil', i, j,                     &
     &         t(i,j,N(ng),nnew,itemp), t(i,j,N(ng),nnew,isalt),        &
     &         wfr(i,j), gamma_k, Hz(i,j,N(ng))
          END IF
The problem is a negative value for Hz, the thickness of a cell. Do you have WET_DRY turned on?

turuncu
Posts: 128
Joined: Tue Feb 01, 2005 8:21 pm
Location: Istanbul Technical University (ITU)
Contact:

Re: tuning ice model ...

#8 Unread post by turuncu »

No, i did not activated WET_DRY is this case but i'll plan to activate it in the production run. This could be also effect of the storm surge but i am not sure. i could change the code to write the zeta value in this cell. it might be a negative value greater than the cell depth. According to my previous runs, the zeta oscillates too much in a small time scale over the northern Caspian Sea but i have no idea about that. it might not be realistic.

Update 1:

i modified that part of the code to write the zeta values (OCEAN(ng)%zeta(i,j,:), all three of them),

trouble in ice_frazil 56 34 -2.303653101454643E-003
3.451582526012355E-004 -1.719989149099419E-009 2.670898819416213E-005
-1.643950240595160E-004 -3.01092344094394 -3.01088860733177
-3.01092344092898

the zeta values goes more than -3 meter and this triggers the ice_frazil check.

Update 2:

if i enable WET_DRY then model works without any problem and it passes that point but it gave " DIAG speed trouble 38.7647803685003" error in 173th day of simulation.
Attachments
zeta_in_ocean_rst_zeta.png

User avatar
kate
Posts: 4091
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: tuning ice model ...

#9 Unread post by kate »

I'm glad the WET_DRY worked for you. Moving on, we have insufficient information to help with the extreme velocities.

I have told you that Paul Budgell has run a Caspian Sea with ice, haven't I? You might ask him for advice.

turuncu
Posts: 128
Joined: Tue Feb 01, 2005 8:21 pm
Location: Istanbul Technical University (ITU)
Contact:

Re: tuning ice model ...

#10 Unread post by turuncu »

Yes, i know him. I'll look at the output to find the source of the problem. In any case, i could write to him about the problem. Thanks for your help.

Post Reply