wiki:Users/SetupNewConfiguration

Setting up a new configuration

Last edition: 2017-05-16 22:45 CEST by nicolasmartin

Starting from an existing one

There are three options for creating new configurations based on an existing one:

Option 1: Duplicate an existing configuration

Create and compile a new configuration duplicating a reference configuration (ORCA_LIM3 in the following example)

makenemo –n 'ORCA_LIM3_MYCONFIG' -r ORCA_LIM3

Option 2: Duplicate with differences

Create and compile a new configuration based on a reference configuration (ORCA_LIM3 in the following example) but with different pre-processor options. For this either add (add_key) or remove (del_key) keys as required; e.g.

makenemo –n 'ORCA2_LIM3_MYCONFIG' -r ORCA2_LIM3 del_key 'key_iomput' add_key 'key_xios'

Option 3: Use the SIREN tools to subset an existing model

Define a regional configuration which is a sub- or super-set of an existing configuration.

This last option employs the SIREN software tools that are included in the standard distribution. The software is written in FORTRAN95 and available in the NEMOGCM/TOOLS/SIREN directory of nemo_v3_6_STABLE (since revision 6468). SIREN allows you to create your own regional configuration embedded in a wider one.

SIREN is a set of programs to create all the input files you need to run a NEMO regional configuration. As a basic demonstrator, a set of GLORYS files (GLObal ReanalYSis on the ORCA025 grid), as well as examples of namelists are available ​here.

SIREN documentation

Any questions or comments regarding the use of SIREN should be posted in the corresponding forum.

Creating a completely new configuration

From NEMO version 4.0 there are two ways to build configurations from scratch. The appropriate method to use depends largely on the target configuration. Method 1 is for more complex/realistic global or regional configurations and method 2 is intended for simpler, idealised configurations whose domains and characteristics can be described in simple geometries and formulae.

Option 1: Create and use a domain configuration file

1 Even when starting from scratch it is advisable to create the directory structure to house your new configuration by duplicating the closest reference configuration to your target application. For example, If your application requires both ocean and ice components then run

makenemo –n 'MY_NEW_CONFIG' -r ORCA_LIM3

where 'MY_NEW_CONFIG' can be substituted with a suitably descriptive name for your new configuration. The purpose of this step is simply to create and populate the appropriate WORK, MY_SRC and EXP00 subdirectories for your new configuration. Other choices for the base reference configuration might be

  • GYRE - If your target application is ocean-only
  • AMM12 - If your target application is regional with open boundaries
  • ORCA2_LIM_PISCES - If your target application requires ocean, sea-ice and passive tracers

2 All the domain information for your new configuration will be contained within a netcdf file called domain_cfg.nc which you will need to create and place in the

NEMOGCM/CONFIG/'MY_NEW_CONFIG'/EXP00

sub-directory. Firstly though, ensure that your configuration is set to use such a file by confirming that

ln_read_cfg = .true.

in

NEMOGCM/CONFIG/'MY_NEW_CONFIG'/EXP00/namelist_cfg

3 Create the domain_cfg.nc file which must contain the following fields

int    ORCA, ORCA_index                  /* configuration name, configuration resolution                             */
int    jpiglo, jpjglo, jpkglo            /* global domain sizes                                                      */
int    jperio                            /* lateral global domain b.c.                                               */
int    ln_zco, ln_zps, ln_sco            /* flags for z-coordinate, z-coordinate with partial steps and s-coordinate */
int    ln_isfcav                         /* flag  for ice shelf cavities                                             */
double glamt, glamu, glamv, glamf        /* geographic position                                                      */
double gphit, gphiu, gphiv, gphif        /* geographic position                                                      */
double iff, ff_f, ff_t                   /* Coriolis parameter (if not on the sphere)                                */
double e1t, e1u, e1v, e1f                /* horizontal scale factors                                                 */
double e2t, e2u, e2v, e2f                /* horizontal scale factors                                                 */
double ie1e2u_v, e1e2u, e1e2v            /* U and V surfaces (if grid size reduction in some straits)                */
double e3t_1d, e3w_1d                    /* reference vertical scale factors at T and W points                       */
double e3t_0, e3u_0, e3v_0, e3f_0, e3w_0 /* vertical scale factors 3D coordinate at T,U,V,F and W points             */
double e3uw_0, e3vw_0                    /* vertical scale factors 3D coordinate at UW and VW points                 */
int    bottom_level, top_level           /* last wet T-points, 1st wet T-points (for ice shelf cavities)             */

There are two options for creating a domain_cfg.nc file:

  • Users can use tools of their own choice to build a domain_cfg.nc with all mandatory fields.
  • Users can adapt and apply the supplied tool available in NEMOGCM/TOOLS/DOMAINcfg. This tool is based on code extracted from NEMO version 3.6 and will allow similar choices for the horizontal and vertical grids that were available internally to that version. See NEMOGCM/TOOLS/DOMAINcfg/README for details.

Option 2: Copy and modify user-defined configuration modules

This method is intended for configuring easily defined idealised configurations which are often used as demonstrators or for process evaluation and comparison. This method can be used whenever the domain geometry has a simple mathematical description and the ocean initial state and boundary forcing is described analytically. From version 4.0, the NEMO release includes a TEST_CASES subdirectory of the reference CONFIG directory which includes several examples and will, over time, include further user-contributed examples. These will not necessarily be fully supported but should provide a source of raw material and ideas for constructing your own examples.

For now, consider the case of starting a completely new ocean-only test case based on the LOCK_EXCHANGE example. [Note: we probably need an even more basic example than this with only one namelist and minimal changes to the usrdef modules]

Firstly, construct the directory structure, starting in the CONFIG directory:

./makenemo -a TEST_CASES -n 'MY_NEW_TEST' -r LOCK_EXCHANGE

where the -a option has been used to locate the new configuration in the TEST_CASES subdirectory (it is recommended practice to keep full configurations and idealised cases clearly distinguishable). This command will have created (amongst others) the following files and directories:

TEST_CASES/'MY_NEW_TEST':
BLD     MY_SRC     cpp_MY_NEW_TEST.fcm
EXP00   WORK
#
TEST_CASES/'MY_NEW_TEST'/EXP00:
context_nemo.xml        domain_def_nemo.xml
field_def_nemo-opa.xml  file_def_nemo-opa.xml
iodef.xml
namelist_cfg
namelist_ref
#
TEST_CASES/'MY_NEW_TEST'/MY_SRC:
usrdef_hgr.F90       usrdef_nam.F90    usrdef_zgr.F90
usrdef_istate.F90    usrdef_sbc.F90    zdfini.F90

The key to setting up an idealised configuration lies in adapting a small set of short fortran90 modules which should be dropped into the MY_SRC directory. Here the LOCK_EXCHANGE example is using 5 such routines but the full set that is availablein the NEMO/OPA_SRC/USR directory is:

../NEMO/OPA_SRC/USR:
usrdef_closea.F90   usrdef_istate.F90   usrdef_zgr.F90
usrdef_fmask.F90    usrdef_nam.F90
usrdef_hgr.F90      usrdef_sbc.F90

Before discussing these in more detail it is worth noting the various namelist controls that engage the different user-defined aspects. These controls are set using two new logical switches or are implied by the settings of existing ones. For example, the mandatory requirement for an idealised configuration is to provide routines which define the horizontal and vertical domains. Templates for these are provided in the usrdef_hgr.F90 and usrdef_zgr.F90 modules. The application of these modules is activated whenever:

ln_read_cfg = .false.

in any configuration's namelist_cfg file. This setting also activates the reading of an optional nam_usrdef namelist which can be used to supply configuration specific settings. These need to be declared and read in the usrdef_nam.F90 module.

Another explicit control is available in the namsbc namelist which activates the use of analytical forcing. With

ln_usr = .true.

Other usrdef modules are activated by less explicit means. For example, code in usrdef_istate.F90 is used to define initial temperature and salinity fields if

ln_tsd_init   = .false.

in the namtsd namelist. The remaining modules, namely

usrdef_closea.F90   usrdef_fmask.F90

are specific to ORCA configurations and set local variations of some specific fields for the various resolutions of the global models. They do not need to be considered here in the context of idealised cases but it is worth noting that all configuration specific code has now been isolated in the usrdef modules. In the case of these last two modules, they are activated only if an ORCA configuration is detected. Currently this requires a specific integer variable named ORCA to be set in a domain_cfg.nc file. [Note: this would be less confusing if the cn_cfg string is read directly as a character attribue from the domain_cfg.nc]

So, in most cases, the set up of idealised model configurations can be completed by copying the template routines from NEMOGCM/NEMO/OPA_SRC/USR into your new NEMOGCM/CONFIG/'MY_NEW_TEST'/MY_SRC directory and editing the appropriate modules as needed. The default set are those used for the GYRE reference configuration. The contents of MY_SRC directories from other idealised configurations may provide more convenient templates if they share common characteristics with your target application.

Whatever the starting point it should not require too many changes or additional lines of code to produce routines in NEMOGCM/NEMO/OPA_SRC/USR that define analytically the domain, the initial state and the surface boundary conditions for your new configuration.

In summary, the base set of modules is:

  • usrdef_hgr.F90 : define horizontal grid
  • usrdef_zgr.F90 : define vertical grid
  • usrdef_sbc.F90 : provides at each time-step the surface boundary condition, i.e. the momentum, heat and freshwater fluxes
  • usrdef_istate.F90 : defines initialization of the dynamics and tracers
  • usrdef_nam.F90 : configuration-specific namelist processing to set any associated run-time parameters

with two specialised ORCA modules (not related to idealised configurations but used to isolate configuration specific code that is used in ORCA2 reference configurations and established global configurations using the ORCA tripolar grid):

  • usrdef_fmask.F90 : only used in ORCA CONFIGURATIONS for alteration of f-point land/ocean mask in some straits
  • usrdef_closea.F90 : only used in ORCA CONFIGURATIONS for specific treatments associated with closed seas
Last modified 4 months ago Last modified on 2017-05-16T22:45:34+02:00

Attachments (4)

Download all attachments as: .zip