SURFEX

Practical changes

Minvielle — 2026-01-29T08:52:49Z

On a practical point of view, the following differences are noted between version 8_0 and version 8_1:

New developments in SURFEX V8.1.0

Minvielle — 2026-01-29T08:50:05Z

This version contains the following novelties:

1) Optimization and MPI parallelisation of the PGD and PREP steps.
2) The general OPEN-MP in the OFFLINE driver is removed.
3) By default, the output written fields are all 1D-fields (there is no longer the dimension PATCH in output files).
4) The types introduced in version 8_0 are used to simplify the writing of several subroutines of the code.
5) Some obsolete options are suppressed.
6) The modifications of P. Marguinaud for the parallelization of the PREP with FULLPOS in AROME.
7) The modifications of J. Escobar and M. Moge for the MPI parallelization of the PREP_REAL, PREP_NEST_PGD and SPAWNING.
8) The parallel I/O server XIOS implemented in Surfex by S. Senesi.
9) The fortran interface for NETCDF4 replaces the previous NETCDF3 interface.
10) A first draft of the new coding related to the further building of ECOCLIMAP-SG.
11) The GRIB-API library is updated to version 1.17.0.
12) The output SNOW fields can be written 2D (number of points / number of snow layers), by M. Lafaysse (CEN).
13) The NETCDF forcing file can be partially read, by M. Lafaysse (CEN).
14) The definition of the OFFLIN output fields to be written is suppressed from files init_outfn_... .
15) The OFFLIN output fields are adapted to plot their fields with EPyGrAM.
16) The possibility to give CONDSAT, MPOTSAT, BCOEF, WWILT, WFC and WSAT, by soil layer, in the namelist NAM_DATA_ISBA, is added.
17) A correction for the forcing radiation is introduced, under key (C. Szczypta, P. Le Moigne).
18) Technical news for the assimilation part SODA (C. Albergel, S. Munier).
19) The XRD and DR_HOOK library are merged and updated to version 44.

19) The XRD and DR_HOOK library are merged and updated to version 44.

2017-03-27T14:49:43Z

The XRD39 library from the GMAP is replaced by the XRD44. The DRHOOK library is included in the XRD44 library.

18) Tehcnical news for the assimilation part SODA (C. Albergel, S. Munier).

2017-03-27T14:49:16Z

These changes include:

the names of the observation fields can be given in the namelist (new entry COBS_M in NAM_OBS).
The EKF scheme can run with ISBA-DF, and can have WG1, WG2... WG8 as variable fields.
No more limitation on Jacobians values (was [-0.1,1]) but on analysis increments instead ([-1,1] for LAI, [-0.1,0.1] for SM)
new diagnostics: Jacobian values, innovations and residuals (additionally to the analysis increments)

17) A correction for the forcing radiation is introduced, under key (C. Szczypta, P. Le Moigne).

2017-03-27T14:48:34Z

If LINTERP_SW (new key in NAM_IO_OFFLINE) is .TRUE., a new time interpolation for the short-wave radiation forcing data is done, taking into account the zenith angle, and allowing not to lost energy anymore.

- Surfex V8.1

16) The possibility to give CONDSAT, MPOTSAT, BCOEF, WWILT, WFC and WSAT, by soil layer, in the namelist NAM_DATA_ISBA, is added.

2017-02-15T07:38:46Z

It's now possible to set these paramaters at the PGD step, via the namelist NAM_DATA_ISBA.
They are defined by soil layer and don't depend on the vegetation type.

- Surfex V8.1

5). An OFFLINE test of performance for XIOS

2016-11-30T15:50:59Z

Version of SURFEX: V81

Version of XIOS: r986

Test case:

global 1/2 degree. 720*300 points.
ISBA-DIF, 19 patches, NCB, ISBA-ES.
Number of written output fields: 3466
Duration of the simulation: 1 month of 31 days.

Performances:

Without XIOS and with none output field: 2 nodes 80 MPI tasks
- Time Elapsed: 7 min 19 s
- READ_SURFX1_NC is in 3^rd position with 8% of the total time against 9.93% for HYDRO_SOILDIF in 1^st position.
  => the reading begins to restrain the scalability.
  ===> will take around 1h18min for 1 year.

Without XIOS and with all output fields: 1 node 40 MPI tasks
- Time Elapsed: 36 min 20 s
- 3 routines linked to the output writing are at the 3 first positions with a total of 57% of the total time.
  => the output writing highly constrains the scalability.
  ===> around 7h15 for 1 year.

With XIOS and with all output fields: 2 nodes, 1 XIOS server and 60 MPI tasks for the OFFLINE
- Time Elapsed: 10 min 29s
- READ_SURFX1_NC in 3^rd position, 5.93% of the total time against 10% for HYDRO_SOILDIF in first position.
  => the reading begins to constraint the scalability when using 2 nodes.
  ===> around 2h06 minutes for 1 an.

Remarks:

With r986 of XIOS, it's recommended to use only 1 XIOS server (not more) because of a problem being treated by the XIOS team.
For the moment, XIOS can't handle output files with too many fields defined in them. So it's better to define in the XIOS .xml input file several output files with not a too big amount of fields in each of them.

Conclusion:

Even with one only XIOS server, the performances for SURFEX OFFLINE are very satisfying.

Input .xml and OPTIONS.nam files, batch script to run it:

15) The OFFLIN output fields are adapted to plot their fields with EPyGrAM.

2016-11-08T14:46:55Z

The documentation to plot SURFEX OFFLIN output fields with EPyGrAM is available here.

To make it possible, several changes are applied to the ....OUT.nc output files:

Each field to be plotted has a new attribute, grid_mapping = "Projection_Type".
The variable Projection_Type is added to each OUT.nc file and filled according to:
- the projection of the run,
- the needs of EPyGrAM to plot it correctly.
This implies the creation of the new file ol_write_proj.F90.

14) The definition of the OFFLIN output fields to be written is suppressed from files init_outfn_....

2016-10-04T10:11:22Z

When CTIMESERIES_FILETYPE = "OFFLIN", the init_outfn_... files allow to define which fields will be written in the output OUT.nc files.

They force to define two times the variables to write:

in the write_... routines, when the write_surf routine is called for each field to write.
in the init_outfn_.. routines.

They are hard to maintain and sources of errors.

So, according to what was done for the "NC" I/O format, they are kept only to create the output files and the definition of output fields is replaced by the following mechanics:

at the first OFFLINE writing time step, the routines write_surf_atm and write_diag_surf_atm are called two times, one with LDEF = T, one with LDEF = F.
the routine mode_write_surf_ol is adapted so that, if LDEF = T, the call to write_surf leads to the call of def_var_netcdf, and not to the writing of the field yet.
The init_output_ol and init_oufn_... routines are kept only to create the output OUT.nc files.
The routine init_io_surfn takes in optional argument the name of the file where the following fields are to be written.

Remark:

In the init_outfn_... files, the attributes long_name and units were more precisely defined that in the write_... routines.
If they want to keep this level of precision, the users are invited to take the old init_outfn_.. files and to put the attributes long_name and units in the HCOMMENT argument of the write_surf routine, with the following syntax:
YCOMMENT = "long_name (units)"
Thus mode_write_surf_ol will separate long_name from units and put them in the appropriate NETCDF variable attributes.

13) The NETCDF forcing file can be partially read, by M. Lafaysse (CEN).

2016-10-04T10:00:58Z

Two keys are added in NAM_IO_OFFLINE:

LDELAYEDSTART_NC:
If .TRUE., the variable time in the NETCDF FORCING file is considered. This variable TIME is defined for each forcing time step, so the forcing time step corresponding to the date of beginning of the simulation (coming from the PREP file) is kept as the first forcing time step for the run.

NDATESTOP:
Integer array of 4 elements: year, month, day and hour (in s) for the end of the simulation.
If NDATESTOP is not 0, it's converted in number of forcing time steps and used to stop the simulation at this ending date.