Inconsistent values of variables in TL initial condition

Discussion about tangent linear and adjoint models, variational data assimilation, and other related issues.

Moderators: arango, robertson

Post Reply
Message
Author
User avatar
Biswamoy_Paul
Posts: 6
Joined: Sat Oct 18, 2014 4:11 pm
Location: INCOIS, Hyderabad, INDIA

Inconsistent values of variables in TL initial condition

#1 Post by Biswamoy_Paul » Thu Jul 19, 2018 1:32 pm

I have setup IS4DVAR for the Indian Ocean with 8~9 km resolution and 40 vertical levels. I am testing a one day run with 30 inner loops and 1 outer loop using MetOffice EN4 T and S observation data.
It is running fine till the adjoint run in the last inner loop. After the corresponding Descent Algorithm (in inner loop# 30) the values of free-surface, u and v-momentum component, potential temperature and salinity in the input tangent linear model initial conditions file (namely, *_itl.nc) increase abruptly as seen in the following log:

Code: Select all

 TLM: GET_STATE - Read state initial conditions,             t =  9710 00:00:00
                   (Grid 01, File: io_1x12_itl.nc, Rec=0001, Index=1)
                - free-surface
                   (Min = -1.05473263E+03 Max =  8.77788746E+02)
                - free-surface, open boundaries
                   (Min = -6.16861539E-01 Max =  2.13182573E-01)
                - vertically integrated u-momentum component
                   (Min =  0.00000000E+00 Max =  0.00000000E+00)
                - vertically integrated u-momentum component, open boundaries
                   (Min =  0.00000000E+00 Max =  0.00000000E+00)
                - vertically integrated v-momentum component
                   (Min =  0.00000000E+00 Max =  0.00000000E+00)
                - vertically integrated v-momentum component, open boundaries
                   (Min =  0.00000000E+00 Max =  0.00000000E+00)
                - surface u-momentum stress, adjusted tl_ustr
                   (Min = -4.47085764E-01 Max =  8.82287023E-01)
                - surface v-momentum stress, adjusted tl_vstr
                   (Min = -2.79389044E-01 Max =  3.06903690E-01)
                - u-momentum component
                   (Min = -2.21421791E+06 Max =  1.21455456E+06)
                - u-momentum component, open boundaries
                   (Min = -1.00039186E-01 Max =  3.04749362E-02)
                - v-momentum component
                   (Min = -1.45533555E+06 Max =  1.45103002E+06)
                - v-momentum component, open boundaries
                   (Min = -1.13609944E-01 Max =  4.29629471E-02)
                - potential temperature
                   (Min = -4.40319194E+03 Max =  7.84010994E+04)
                - salinity
                   (Min = -3.24198687E+04 Max =  4.41747228E+04)
                - potential temperature, open boundaries
                   (Min = -6.09766327E-02 Max =  1.63820702E-01)
                - salinity, open boundaries
                   (Min = -2.45726672E-02 Max =  4.26380653E-03)
                - surface net heat flux, adjusted tl_tflux
                   (Min = -7.60638358E-07 Max =  1.92845377E-02)
                - surface net salt flux, (E-P)*SALT, adjusted tl_tflux
                   (Min = -8.47218473E-07 Max =  9.61237175E-03)
      WRT_INI   - wrote initial  fields (Iter=0032, Index=1,1, Rec=0001)
Hence the following forward nonlinear model run blows up at the first time step.
I have tried to locate the problem with increasing the friction, changing the decorrelation scales, increasing the # of the inner loop, enabling/disabling ADJUST_BOUNDARY, ADJUST_WSTRESS, ADJUST_STFLUX, and also with relaxing T and S. But the problem still persists.
I have attached the figures of tangent linear u and v-momentum component at the surface. I don't have any clue what causes this sudden increment of these values. Any kind of help or advice would be highly appreciated. Thanks in advance.
Attachments
tangent linear v-momentum component at the surface
tangent linear v-momentum component at the surface
tangent linear u-momentum component at the surface
tangent linear u-momentum component at the surface

User avatar
kate
Posts: 3702
Joined: Wed Jul 02, 2003 5:29 pm
Location: IMS/UAF, USA

Re: Inconsistent values of variables in TL initial condition

#2 Post by kate » Thu Jul 19, 2018 4:55 pm

I suppose you are focused on the big blobs of extreme values, but my eye is drawn to Y=380 and the weird line there.

User avatar
Biswamoy_Paul
Posts: 6
Joined: Sat Oct 18, 2014 4:11 pm
Location: INCOIS, Hyderabad, INDIA

Re: Inconsistent values of variables in TL initial condition

#3 Post by Biswamoy_Paul » Fri Jul 20, 2018 6:38 am

Thanks, Kate for your reply.
Yes, I was more focused on those extreme values in velocity and tracer fields. Those high anomalies are found even in the subsurface.
I assume that the line of sharp gradient in velocity field, which is exactly on the Equator (my model domain ranges from 30S to 30N), is the source of this trouble. This weird line is found throughout the depth as seen in the attached figure, a vertical slice of v-momentum at X=458. Although no such line is found in tracers and free-surface field.
I wonder what causes this sharp changes in velocity field at the Equator.
Attachments
TL v-momentum component at X=458
TL v-momentum component at X=458

wavewatch
Posts: 7
Joined: Thu Nov 14, 2013 1:51 am
Location: institute of oceanology

Re: Inconsistent values of variables in TL initial condition

#4 Post by wavewatch » Wed Apr 24, 2019 2:55 am

Hi,
May I ask if your problem has been solved ?
I had the same problem :( :( :(

Post Reply