Opened 11 years ago

Closed 10 years ago

Last modified 10 years ago

#13 closed defect (fixed)

Time steping of the driver

Reported by: jpolcher Owned by: somebody
Priority: major Milestone: ORCHIDEE 1.9.7
Component: Driver files Version:
Keywords: Cc:

Description

The driver does not start on the first forcing date of the file when there is a restart file. The information of the restart file is not well carried over.

A first attempt to correct this has been implemented in revision 944 of the hydrological branch.

As it is unlikely that this error has affected all, there must be some other work-around being used and we need to ensure that all this remains compatible.

Attachments (1)

test_basic.pdf (69.9 KB) - added by jpolcher 10 years ago.

Download all attachments as: .zip

Change History (5)

comment:1 Changed 10 years ago by dsolyga

  • Milestone changed from orchidee_1_9_6 to ORCHIDEE 2.0

Changed 10 years ago by jpolcher

comment:2 follow-up: Changed 10 years ago by jpolcher

The driver cannot be run with the TIME_LENGTH variable to set to default. This causes a bug. In run.def for a simulation of one year one has to set TIME_LENGTH=1y.

To verify this the time evolution over 1 year were compared after 4 years of simulation with WFD as forcing. The following versions are compared in the attached file :

  • version LMD : the driver had been patched in order to be able to use TIME_LENGTH=DEF
  • Version HydroMerge? : with the same patch transferred to this version of the driver and thus TIME_LENGTH=DEF
  • Version Trunk : without any change to the driver or model and TIME_LENGTH=1y

The chronologies of Version HydroMerge? and Version Trunk match perfectly. There is a small difference with version LMD but that does not matter any more.

The ticket will be closed once more years were performed and a few leap years passed.

comment:3 in reply to: ↑ 2 Changed 10 years ago by nvuilsce

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

By default (when the TIME_LENGTH parameter is not set), the time length of a simulation corresponds to the length of the forcing file (tm). This is true since revision r254. Before this revision, the default time length was the length of the forcing minus 1 forcing time step (tm-1) to be in agreement with some forcing files such as the NCC reanalysis that contain an additional time step unused by the driver. For instance a non-leap yearly NCC forcing file at 6-hour contains 1461 time steps (and not 1460).
We decide to keep the default time length equals to the length of the forcing file. To easily use forcing files such as NCC that contain an additional time step, set the TIME_LENGTH to "1Y". This setting works correctly since revision r999.

Replying to jpolcher:

The driver cannot be run with the TIME_LENGTH variable to set to default. This causes a bug. In run.def for a simulation of one year one has to set TIME_LENGTH=1y.

To verify this the time evolution over 1 year were compared after 4 years of simulation with WFD as forcing. The following versions are compared in the attached file :

  • version LMD : the driver had been patched in order to be able to use TIME_LENGTH=DEF
  • Version HydroMerge? : with the same patch transferred to this version of the driver and thus TIME_LENGTH=DEF
  • Version Trunk : without any change to the driver or model and TIME_LENGTH=1y

The chronologies of Version HydroMerge? and Version Trunk match perfectly. There is a small difference with version LMD but that does not matter any more.

The ticket will be closed once more years were performed and a few leap years passed.

comment:4 Changed 10 years ago by dsolyga

  • Milestone changed from ORCHIDEE 2.0 to ORCHIDEE 1.9.7
Note: See TracTickets for help on using tickets.