Opened 6 years ago

Closed 5 years ago

Last modified 5 years ago

#143 closed defect (fixed)

1+1

Reported by: jgipsl Owned by: jgipsl
Priority: major Milestone: IPSLCM6.v1
Component: Model architecture Version:
Keywords: Cc:

Description

When running ORCHIDEE 1 month then restarting the 2nd month, the results must be identical of a run of 2months starting with the same conditions. This is not the case currently. (ORCH trunk rev 2149)

Change History (5)

comment:1 Changed 6 years ago by jgipsl

  • Status changed from new to accepted

Some enhancements in revision [2435] and [2436]

comment:2 Changed 6 years ago by jgipsl

Enhancement : correction in routing rev [2580], see ticket #169
See also previously correction in condveg rev [2279], see ticket #145

comment:3 Changed 6 years ago by jgipsl

Correction in slowproc see rev [2654]. There is still more to do.

comment:4 Changed 5 years ago by jgipsl

Lots of corrections in rev [2868]. Comments:

  • CALL hydrol_tmc_update removed from initialization phase
  • kfact_root now calculated at each time step, see ticket #192
Last edited 5 years ago by jgipsl (previous) (diff)

comment:5 Changed 5 years ago by jgipsl

  • Resolution set to fixed
  • Status changed from accepted to closed

Correction in LMDZ trunk rev 2356 done to have 1+1=2 in coupled mode LMDZOR. Note that this revision is not working due to other problems in LMDZ.

Following has been tested with success:

  • ORCHIDEE trunk revision 2892
  • LMDZ5/branches/testing revision 2314 + local modification as in the commit above (2356 correction on restart variable sollwdown)

These to components are correct restarted (1+1=2) for the following cases (no test case failed) :

  • ORCHIDEE with CWRR or Choisnel
  • Coupled LMDZOR configuration is tested using grid 96x95x39 and old physics (AP) in LMDZ, tested at ada/IDRIS only
  • ORCHIDEE offline global grid using NCC forcing, tested at ada, curie and obelix
  • The routing scheme in ORCHIDEE has been activated in some of the tests
  • The tests are done running 2 simulations, one over 2 days in one execution, the other running 1day and restarting the 2 days in 2 executions. The two simulations are compared and give the same restart files after the 2nd day. The output during the 2nd day are compared and give the same results for the both simulations. Some longer simulations are also done running one year complete comparing to monthly executions.
  • Multiple year simulation in one execution has not been tested as we do not provide forcing files over several year.
Note: See TracTickets for help on using tickets.