New URL for NEMO forge!   http://forge.nemo-ocean.eu

Since March 2022 along with NEMO 4.2 release, the code development moved to a self-hosted GitLab.
This present forge is now archived and remained online for history.
ticket/1851/General (diff) – NEMO

Changes between Version 17 and Version 18 of ticket/1851/General


Ignore:
Timestamp:
2017-02-24T14:50:24+01:00 (7 years ago)
Author:
frrh
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • ticket/1851/General

    v17 v18  
    143143   
    144144 
     145 
     146== Coupled model CRUNs == 
     147 
     148We compare a 2day NRUN with a 2x1-day CRUN in u-ak282debug and u-ak282debugCRUN.  
     149 
     150Inspecting the solver.stat file, this reveals that NMEO seems to diverge on the first TS immediately after the restart i.e. TS 33! 
     151Why? 
     152We have other models (e.g. the stand-alone NEMO-CICE-MEDUSA which don't. 
     153 
     154So this could arise from the coupling fields coming from the atmos at TS0, but the atmos dumps are identical at day 1 and checking the most likely contenders (PCO2 and dust) shows they're fine at teh start of day 2.  
     155 
     156Comparing the ocean.output files at the start of day 2 leads us to fishy looking things related to TOP/MEDUSA since we have a message 
     157' New shortwave to sample for TOP at time kt = ' occurring at different timesteps for each run. In the NRUN case this occurs at TS 36 and in the CRUN case it occurs at TS 38!   
     158This leads us to trcstp.F90 which contains a flaky looking  
     159bit of code: 
     160{{{ 
     161     iseclast = nsec_year + nsec1jan000 
     162     llnew   = ( iseclast - isecfst )  > INT( rdt_sampl )   !   new shortwave to store 
     163 
     164     IF( kt /= nittrc000 .AND. llnew ) THEN 
     165          IF( lwp ) WRITE(numout,*) ' New shortwave to sample for TOP at time kt = ', kt, & 
     166             &                      ' time = ', (iseclast+rdt*nn_dttrc/2.)/3600.,'hours ' 
     167          isecfst = iseclast 
     168          DO jn = 1, nb_rec_per_days - 1 
     169             qsr_arr(:,:,jn) = qsr_arr(:,:,jn+1) 
     170          ENDDO 
     171          qsr_arr (:,:,nb_rec_per_days) = qsr(:,:) 
     172          qsr_mean(:,:                ) = SUM( qsr_arr(:,:,:), 3 ) / nb_rec_per_days 
     173      ENDIF 
     174}}} 
     175  
     176That can never give the same results for NRUNS and CRUNS! 
     177This condition arises at the 6th timestep after a restart and every 5 timesteps thereafter.  
     178So in the restart case it occurs at TS=38, 43, 48, 53 etc while in the NRUN case is arises at TS=36, 41, 46, 51 etc. 
     179 
     180This comes from the vn3.6 NEMO_stable branch via the GO6 branch, and is not specific to the MEDUSA branch. But it looks utter garbage.  
     181 
     182The description says "In coupled mode, the sampling is done at every coupling frequency", which is not what happens at all because,  
     183with a 45 minute TS and a coupling frequency of 3 hours which we have in our model: 
     184  
     185   * The first sample is taken at TS 6 after a start i.e. 4.5 hours 
     186   * subsequent samples occur at 3.75 hours not the 3 hours it should do! 
     187 
     188I think this should be sampling, consistently at TS = 0, 4, 8.... 32.... (though 32 shouldn't matter in the case of day 1 of the CRUN) or something like that.  
     189 
     190 
     191 
     192 
     193 
     194 
     195 
     196 
     197 
     198 
     199