...

...

This page describes how OpenIFS/AC v1 (derived from OpenIFS 43r3v2) can be obtained, built and used for forecast experiments. The documentation shows examples of how to use the model on the ECMWF Atos Sequana XH2000 HPCF (hpc2020). Adjustments will need to be made for local installations. 

OpenIFS/AC compared to the standard OpenIFS model

...

The standard OpenIFS model is described in detail in the OpenIFS 43r3 User Guide. The user guide introduces the model's components and supported model grids. It informs about additional required software packages and how to build the OpenIFS model. It further describes how to carry out an acceptance test after a successful model build in order to verify the model's functionality.

...

ftp ftp.ecmwf.int
(login with username and password for OpenIFS users)
cd src/openifs-ac
get openifs-ac.43r3.v1.tgzgz
bye

Installing the OpenIFS/AC sources

...

• Create a new home for the OpenIFS/AC installation. Here we call this location /home/openifs-ac. 
• Change into your new directory and unpack therein the original OpenIFS model sources:   tar xvzf oifs43r3v2.tgz
• In a second step, in the same folder unpack gunzip the OpenIFS/AC sources :  tar xvzf and patch the original OpenIFS:
  
  • gunzip openifs-ac.43r3.v1.
  tgz
  • gz
  • patch -p1 <openifs-ac.43r3.v1

...

• Change into your $OIFS_HOME directory where OpenIFS is installed:  cd $OIFS_HOME
• In this directory extract the downloaded source package:   tar xvzf
  • gunzip openifs-ac.43r3.v1
  .tgz
  • .gz
  • patch -p1 <openifs-ac.43r3.v1

...

• Unpack oifs-ac-43r3-initcond.tgz into /data/openifs-ac which results in a directory /data/openifs-ac/INITIAL-CONDITIONS.
  The files therein are initial data files for the first day of month for all 12 months of 2010.
• Unpack ICMCL-2010.tgz into /data/openifs-ac which results in a directory /data/openifs-ac/ICMCL.
  This directory contains monthly files with surface boundary conditions (e.g. SSTs, soil temps, albedos, LAI etc).
• In the tgz archives above, the data for the year 2013, provided on a Tl255L60 grid, is required for the T21 acceptance tests. The data for year 2010 is provided on a Tl255L91 grid and can be used for the example forecasts. 
• From your conventional OpenIFS installation's OIFS_DATA_DIR copy the directories ifsdata and rtables to /data/openifs-ac/43r3 as they are also needed here. If this is your first model installation, these data directories are also available from the ftp server. 
• Further, create a directory /data/openifs-ac/43r3/wam which needs to contain wave model initial data for 2010-01-01. 

Building the model

...

Building the model

This section describes how to build the This section describes how to build the binary model executable for OpenIFS/AC. Most problems related to building the model executable are due to incorrect environment settings. Therefore we recommend that particular care should be given to the environment settings to ensure that they match the local hardware platform. 

...

• To do this we run the standard t21test for experiment epc8 (T21L19) with the setup-exp.sh new script.
• The name for the hardware platform needs to be set (l.89, e.g. platform=ecmwf-hpc2020)
• The test experiment runs in the RUNDIR directory specified in setup- test experiment runs in the RUNDIR directory specified in setup-exp.sh (l.94, the default is $SCRATCH/43r3/epc8/).
• We also need to use the command line option -x and give as argument the absolute full path to the master.exe binary executable.
• In setup-exp.sh you should change the value for RUN_NUMBER otherwise any pre-existing model output from this test will be overwritten. 

...

Test result:  As before an error calculation file res_021_0072 is created in the run directory and the model output is found in the subfolder output1 (this name depends on the RUN_NUMBER variable). 

With chemistry:  Now repeat the test with enabled model chemistry. This is a short 1-day test run that uses the newly added source code of OpenIFS/AC. It has a very limited set of options and should not be used for forecast experiments. Note that you cannot increase the experiment duration beyond one day because the boundary conditions are only supplied from 2013-07-01 to 2013-07-03. 

With chemistry:  Now repeat the test with enabled model chemistry. This is a short 1-day test run that uses the newly added source code of OpenIFS/AC. It has a very limited set of options and should not be used for forecast experiments. Note that you cannot increase the experiment duration beyond one day because the boundary conditions are only supplied from 2013-07-01 to 2013-07-03. 

This This experiment is named oiac and uses a Tl255L60 grid and runs for 24 hours. The default run directory is $SCRATCH/43r3/oiac/. 

...

Test result:  This experiment calculates budgets of atmospheric mass and ozone which are stored in the , along with overall production and loss tendencies, as well as more detailed reaction budget terms relevant for the ozone production and loss tendency. These are printed in the NODE.001_01 file and also in afterwards extracted to the files $RUNDIR/output{$RUN_NUMBER}/massdia_chem_oiac.txt and ozonbud_chem_dia.txt.

These diagnostics can be compared with pre-calculated results in files massdia_chem__oiac_2013070100.txt and ozonbud_chem__oiac_2013070100.txt which are located in $OIFS_HOME/t21test_xois_t255ac/ctrl/. For example this could be done with a small shell script such as illustrated in the box below. The differences should be below 5%. 

EXPID=oiac; RUNDIR=${SCRATCH}/${OIFS_CYCLE}/${EXPID} ;  RUN_NUMBER=3  #checkoifs
fref1=${OIFS_HOME}/t21test_xios_ac/ctrl/massdia_chem__oiac_2013070100.txt
fref2=${OIFS_HOME}/t21test_xios_ac/ctrl/ozonbud_chem__oiac_2013070100.txt
fnew1=${RUNDIR}/output${RUN_NUMBER}/massdia_chem_oiac.txt
fnew2=${RUNDIR}/output${RUN_NUMBER}/ozonbud_chem_oiac.txt
diff $fref1 $fnew1
diff $fref2 $fnew2

As a rule of thumb, differences between the reference budget numbers and those from the t21test should be only small (below 5%), although this depends on the exact variable that is assessed. 

...

Single forecast experiments

This section describes the required steps to carry out a single forecast experiment (no restarts) with chemistry. The sequence of essential three steps can be summarised as
(1) generate initial conditions or use existing initial data,
(2) set up experiment parameters in config.h and select the appropriate batch job wrapper script for the HPC hardware,
(3) submit the experiment and monitor progress.

Preparing an experiment

AC-experiments directory overview

:

...

1. Edit paths and variables in $OIFS_HOME/AC-experiments/config.h. This controls model switches and the chemistry scheme. 
2. Prepare initial conditions using

1. $OIFS_HOME/AC-

...

There the following scripts can be found:

...

1. epxeriments/prep-ic-icmcl-compo.sh

...

1. .
2. Customise a job wrapper script for the relevant hardware platform (a *.job file) in order to submit the

1. oifs-run.sh

...

1. script

...

1. to

...

1. the batch scheduler

...

1. .

Overview of the

...

AC-experiments

...

• it is necessary to create a machine config file:  oifs-config.ecmwf-atos.sh

The above ecmwf-atos.job wrapper script to submit the batch job is based on templates in the platform directory; the following templates exist:

• ecmwf-hpc2020.job.tmpl    –    run oifs without XIOS
• oifs-config.ecmwf-hpc2020.sh  –  template for the machine config file

 Generating initial data overview

Initial conditions for the standard OpenIFS model can be generated using the OpenIFS Data Hub or alternatively they can be requested from the ECMWF OpenIFS support by email openifs-support@ecmwf.int. This provides initial experiment data and surface boundary conditions for the physical/meteorological NWP forecast model. In order to simulate atmospheric composition further initial and boundary conditions are needed. These can be created using the prep-ic-icmcl-compo.sh script. 

• You need to edit variables inside the file config.h to correspond with the chosen settings for your experiment before running the script to generate the initial data.

At first the following environment variables need to be set to the appropriate locations. In our example described in the Section above these locations can be found in the OIFS_DATA_DIR under /data/openifs-ac: 

OIFS_ICMCL_DIR
This points to a directory containing initial conditions for the first day of all months of the year 2010 at Tl255L91. The ICMCL files contain boundary conditions (e.g. SSTs) for the entire month. 

OIFS_INIDATA_DIR_TM5
This points to a directory containing initial conditions required for the 'tm5' chemistry (two versions for Tl255L91 are available: one with AER aerosol and one without) for 2010-01-01. It also contains boundary conditions (emission fluxes and deposition velocities) for 2010-01-01 to 2010-01-02.

OIFS_INIDATA_DIR_CBA
As above but for the 'bascoetm5' chemistry. 

• Further settings that need changing in config.h are described below in the following section.
• Once all the changes have been made in the configuration file, the prep script can be run. Running the prep script will take a few minutes and will result in new data which will be written to folder chem_ic. This folder serves as a staging area for the initial experiment data. 
• After the prep script has completed it's run, the platform-specific job wrapper script can be executed to submit the model experiment job to the batch scheduler.

Example configuration settings for single forecast experiments

Select one of the following four T255L91 forecast configurations for your experiment.  The tested and supported configurations are:

...

91

...

Using TM5 chemistry without aerosol

We will run a 2 day forecast starting from 2010-01-01 with TM5 chemistry without aerosol.

Use the following settings in config.h

PLATFORM='ecmwf-atos'
RUNDIR=${SCRATCH}/43r3/experiments\${EXPID}\/runs  # where your experiment will run
AC_IC_DIR=${SCRATCH}/43r3/experiments\${EXPID}/chem_ic  # where your chemical initial conditions will be created
run_end_date="2010-01-03"

rst_freq='3 day'  # run for 3 days without restarts (no multi-leg exp)
USE_RESTART=false   # Do not write restart files (no multi-leg exp)
LCHEM=true
CHEM_SCHEME="tm5"
CHEM_VER="ver15"
OIFS_INIDATA_DIR=$path_to_directory_T255L91-tm5-none
AC_IC_ONCE=${OIFS_INIDATA_DIR}/ICMUAoiac.ac_rst'
MET_IC_DIR='${OIFS_ICMCL_DIR}/aabg/2010010100'

LAERO=false
LAERCHEM=false
LCHEM_AEROI=false
LINJ_CHEM=false
LINJ_AER=false
LAEROSFC=false

LWAM=true" config.h
LRLXG=false" config.h
LAC4IC=true" config.h

Then run the prep script for initial conditions

./prep-ic-icmcl-compo.sh 2010-01-01 2010-01-03

Using TM5 chemistry with AER aerosol

We will run a 2 day forecast starting from 2010-01-01 with TM5 chemistry and the AER bulk aerosol scheme.

Use the following settings in config.h

PLATFORM='ecmwf-atos'
RUNDIR=${SCRATCH}/43r3/experiments\${EXPID}\/runs  # where your experiment will run
AC_IC_DIR=${SCRATCH}/43r3/experiments\${EXPID}/chem_ic  # where your chemical initial conditions will be created
run_end_date="2010-01-03"

rst_freq='3 day'  # run for 3 days without restarts (no multi-leg exp)
USE_RESTART=false   # Do not write restart files (no multi-leg exp)
LCHEM=true
CHEM_SCHEME="tm5"
CHEM_VER="ver15"
OIFS_INIDATA_DIR=$path_to_directory_T255L91-tm5-aer
AC_IC_ONCE=${OIFS_INIDATA_DIR}/ICMUAoiac.ac_rst'
MET_IC_DIR='${OIFS_ICMCL_DIR}/aabg/2010010100'

LAERO=true
LAERCHEM=true
LCHEM_AEROI=true
LINJ_CHEM=true
LINJ_AER=true
LAEROSFC=true

LWAM=true" config.h
LRLXG=false" config.h
LAC4IC=true" config.h

Then run the prep script for initial conditions

./prep-ic-icmcl-compo.sh 2010-01-01 2010-01-03

Using BASCOE-TM5 chemistry without aerosol

We will run a 2 day forecast starting from 2010-01-01 with combined BASCOE and TM5 chemistry without aerosol.

Use the following settings in config.h

PLATFORM='ecmwf-atos'
RUNDIR=${SCRATCH}/43r3/experiments\${EXPID}\/runs  # where your experiment will run
AC_IC_DIR=${SCRATCH}/43r3/experiments\${EXPID}/chem_ic  # where your chemical initial conditions will be created
run_end_date="2010-01-03"

rst_freq='3 day'  # run for 3 days without restarts (no multi-leg exp)
USE_RESTART=false   # Do not write restart files (no multi-leg exp)
LCHEM=true
CHEM_SCHEME="bascoetm5"
CHEM_VER="ver2f"
OIFS_INIDATA_DIR=$path_to_directory_T255L91-bascoetm5-none
AC_IC_ONCE=${OIFS_INIDATA_DIR}/ICMUAoiac.ac_rst'
MET_IC_DIR='${OIFS_ICMCL_DIR}/aabg/2010010100'

LAERO=false
LAERCHEM=false
LCHEM_AEROI=false
LINJ_CHEM=false
LINJ_AER=false
LAEROSFC=false

LWAM=true" config.h
LRLXG=false" config.h
LAC4IC=true" config.h

Then run the prep script for initial conditions

./prep-ic-icmcl-compo.sh 2010-01-01 2010-01-03

Using BASCOE-TM5 chemistry with AER aerosol

We will run a 2 day forecast starting from 2010-01-01 with combined BASCOE and TM5 chemistry without aerosol.

Use the following settings in config.h

PLATFORM='ecmwf-atos'
RUNDIR=${SCRATCH}/43r3/experiments\${EXPID}\/runs  # where your experiment will run
AC_IC_DIR=${SCRATCH}/43r3/experiments\${EXPID}/chem_ic  # where your chemical initial conditions will be created
run_end_date="2010-01-03"

rst_freq='3 day'  # run for 3 days without restarts (no multi-leg exp)
USE_RESTART=false   # Do not write restart files (no multi-leg exp)
LCHEM=true
CHEM_SCHEME="bascoetm5"
CHEM_VER="ver2f"
OIFS_INIDATA_DIR=$path_to_directory_T255L91-bascoetm5-none
AC_IC_ONCE=${OIFS_INIDATA_DIR}/ICMUAoiac.ac_rst'
MET_IC_DIR='${OIFS_ICMCL_DIR}/aabg/2010010100'

LAERO=true
LAERCHEM=true
LCHEM_AEROI=true
LINJ_CHEM=true
LINJ_AER=true
LAEROSFC=true

LWAM=true" config.h
LRLXG=false" config.h
LAC4IC=true" config.h

Then run the prep script for initial conditions

./prep-ic-icmcl-compo.sh 2010-01-01 2010-01-03

Multiple forecasts with restarts

Further information

----- THE REMAINING TEXT IS TO BE DELETED PRIOR TO PUBLICATION -----

EADME

...

directory

AC-experiments/
├── config.h                   # MAIN experiment config file
├── ctrl/
│   ├── context_ifs.xml        # xios resource
│   ├── ifs_xml/               # xios resource
│   ├── iodef.xml              # xios resource
│   ├── namelistfc 
│   ├── namelist.ifs-epc8.sh   # namelist for t21test without chemistry
│   ├── namelist.ifs-oiac.sh   # namelist for t21test with chemistry
│   ├── namelist.ifs.sh        # namelist template for forecast experiments
│   ├── Table/                 # dir with AC tracers
│   └── wam_namelist
├── ecmwf-hpc2020.job          # job wrapper script for oifs-run.sh
├── oifs-run.sh                # MAIN run script
├── platform/                  # dir with platform-specific templates
├── prep_emis/                 # dir with emission utilities
├── prep-ic-icmcl-compo.sh     # MAIN script to prepare initial/boundary conditions
└── scripts/                   # dir with low-level scripts
    ├── add_nrt_fire_chem
    ├── gaussgr
    ├── get_alb_lai_intLimits
    ├── get_tablecol
    ├── grib_def.h
    ├── lib_chem_setup.sh
    ├── lib_general.sh           # lib of utilities
    ├── lib_icmcl_clima.sh       # lib of functs to create ICMCL boundary cond. files
    ├── lib_icmcl_compo.sh       # lib of functs to create ICMCL-COMPO files
    ├── lib_initcond_offline.sh
    └── lib_initcond.sh          # lib of functs to create initial conditions files

The experiment controls are in $OIFS_HOME/AC-experiments/. 

There the following essential files can be found:

• prep-ic-icmcl-compo.sh  - a script to generate initial conditions for chemical fields for the selected run time. Begin and end date need to be given as an argument.
• config.h  - settings for model installation, paths, and experiment configuration.
• oifs-run.sh  - this script controls the workflow for the experiment: reads experiment config, sets up run directories (in $SCRATCH), loads platform configuration, edits namelists, links to initial conditions and executes OIFS_RUNCMD
• ecmwf-hpc2020.job  wrapper script for oifs-run.sh, required settings for the batch scheduler

In $OIFS_HOME/AC-experiments/platform:

• it is necessary to create a platform-specific machine config file:  oifs-config.ecmwf-hpc2020.sh

The above ecmwf-hpcs2020.job wrapper script to submit the batch job is based on templates in the platform directory; a number of templates exist and for the ECMWF HPCF the following templates are relevant:

• ecmwf-hpc2020.job.tmpl    –    run oifs without XIOS
• oifs-config.ecmwf-hpc2020.sh  –  template for the machine config file

Editing the experiment configuration file

For the forecast experiment to succeed and in order to generate the initial data correctly, you will need to edit variables inside the file config.h to correspond with the chosen settings for your experiment.

In config.h you should edit the following variables. The example is for a single forecast (without restarts or multiple forecast legs), starting from 1st January 2010 having 48 hours duration. 

PLATFORM='ecmwf-hpc2020' # this needs to match the name of the platform configuration file
EXP=abcd                 # you should select here an experiment ID (4-letter alphanumeric word, must not start with '0')

RUNDIR=${SCRATCH}/43r3/experiments/${EXP}/runs         # this is the experiment directory
AC_IC_DIR=${SCRATCH}/43r3/experiments/${EXP}/chem_ic   # this is the staging directory for initial and boundary conditions data files

# Simulation start and end date. If setting time (not mandatory), **put it before** the date
run_start_date="00:00 2010-01-01"
run_end_date="${run_start_date} + 2 days"
 
rst_freq='6 day'        # Run for 6 days without restarts. Our experiment will finish after 2 days, so this will be ignored. 
USE_RESTART=false       # Do not write restart files.

Path variables need to be set to the appropriate locations. In our example described in the Section above these locations can be found in the OIFS_DATA_DIR under /data/openifs-ac.

OIFS_INIDATA_DIR=${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-tm5-aer   # initial and boundary conditions for the chemistry scheme
ICMCLDIR=${OIFS_DATA_DIR}/ICMCL                                        # physical/meteorological boundary conditions
MET_IC_DIR=${OIFS_DATA_DIR}/ICMCL/aabg/2010010100                      # path for meteorological initial data in case it needs to be generated
AC_IC_ONCE=${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-tm5-aer/ICMUAoiac.ac_rst    # initial conditions for chemical tracers

Further variables and settings that need changing in config.h are described below in the next section for four possible forecast experiment configurations.

Once all the changes have been made in config.h, the experiment initial and boundary condition files can be prepared as described in the following section. 

After the data has been prepared the platform-specific job wrapper script can be executed to submit the model experiment job to the batch scheduler.

Example configuration settings for single forecast experiments

Select one of the following four T255L91 forecast configurations for your experiment.  The tested and supported configurations are:

Using TM5 chemistry without aerosol

Example:  Run a 2-day forecast starting from 2010-01-01 with TM5 chemistry without aerosol.

Use the following settings in config.h:

PLATFORM='ecmwf-hpc2020'

RUNDIR='${SCRATCH}/43r3/experiments/${EXP}/runs'        # where your experiment will run
AC_IC_DIR='${SCRATCH}/43r3/experiments/${EXP}/chem_ic'  # where your chemical initial conditions will be created
run_end_date="2010-01-03"
rst_freq='6 day'
USE_RESTART=false

LCHEM=true
CHEM_SCHEME="tm5"
CHEM_VER="ver15"
OIFS_INIDATA_DIR='${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-tm5-none'

LAERO=false
LAERCHEM=false
LCHEM_AEROI=false
LINJ_CHEM=false
LINJ_AER=false
LAEROSFC=false

LWAM=true
LRLXG=false

Using TM5 chemistry with AER aerosol

Example:  Run a 2-day forecast starting from 2010-01-01 with TM5 chemistry and the AER bulk aerosol scheme.

Use the following settings in config.h:

PLATFORM='ecmwf-hpc2020'

RUNDIR='${SCRATCH}/43r3/experiments/${EXP}/runs'        # where your experiment will run
AC_IC_DIR='${SCRATCH}/43r3/experiments/${EXP}/chem_ic'  # where your chemical initial conditions will be created
run_end_date="2010-01-03"
rst_freq='6 day'
USE_RESTART=false

LCHEM=true
CHEM_SCHEME="tm5"
CHEM_VER="ver15"
OIFS_INIDATA_DIR='${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-tm5-aer'

LAERO=true
LAERCHEM=true
LCHEM_AEROI=true
LINJ_CHEM=true
LINJ_AER=true
LAEROSFC=true

LWAM=true
LRLXG=false

Using BASCOE-TM5 chemistry without aerosol

Example:  Run a-2 day forecast starting from 2010-01-01 with combined BASCOE and TM5 chemistry without aerosol.

Use the following settings in config.h:

PLATFORM='ecmwf-hpc2020'

RUNDIR='${SCRATCH}/43r3/experiments/${EXP}/runs'        # where your experiment will run
AC_IC_DIR='${SCRATCH}/43r3/experiments/${EXP}/chem_ic'  # where your chemical initial conditions will be created
run_end_date="2010-01-03"
rst_freq='6 day'
USE_RESTART=false

LCHEM=true
CHEM_SCHEME="bascoetm5"
CHEM_VER="ver2f"
OIFS_INIDATA_DIR='${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-bascoetm5-none'

LAERO=false
LAERCHEM=false
LCHEM_AEROI=false
LINJ_CHEM=false
LINJ_AER=false
LAEROSFC=false

LWAM=true
LRLXG=false

Using BASCOE-TM5 chemistry with AER aerosol

Example:  Run a-2 day forecast starting from 2010-01-01 with combined BASCOE and TM5 chemistry without aerosol.

Use the following settings in config.h:

PLATFORM='ecmwf-hpc2020'

RUNDIR='${SCRATCH}/43r3/experiments/${EXP}/runs'        # where your experiment will run
AC_IC_DIR='${SCRATCH}/43r3/experiments/${EXP}/chem_ic'  # where your chemical initial conditions will be created

...

run_end_date="2010-01-

...

03"
rst_freq='6 day'
USE_RESTART=false

LCHEM=true
CHEM_SCHEME="bascoetm5"
CHEM_VER="ver2f"
OIFS_INIDATA_DIR='${OIFS_DATA_DIR}/INITIAL-CONDITIONS/T255L91-bascoetm5-none'

LAERO=true
LAERCHEM=true
LCHEM_AEROI=true
LINJ_CHEM=true
LINJ_AER=true
LAEROSFC=true

LWAM=true
LRLXG=false

Generating initial data

Initial conditions for the standard OpenIFS model can be generated using the OpenIFS Data Hub or alternatively they can be requested from the ECMWF OpenIFS support by email openifs-support@ecmwf.int. This provides initial experiment data and surface boundary conditions for the physical/meteorological NWP forecast model.

In order to simulate atmospheric composition with OpenIFS/AC additional initial and boundary conditions are needed. These can be created using the prep-ic-icmcl-compo.sh script. 

The prep-ic-icmcl-compo.sh script (henceforth: prep script) carries out 4 tasks:

• Creates a destination directory $AC_IC_DIR which serves as a staging area for all initial and boundary conditions required for the forecast experiment. Symbolic links will point later from inside the experiment run directory to these files.
• Copies the physical/meteorological initial conditions files that must be specific to the experiment start date from $OIFS_INIDATA_DIR to the new $AC_IC_DIR. If these files cannot be found then an attempt is made to create them from files found in the $MET_IC_DIR directory. 
• Copies the physical boundary conditions file ICMCL{exp}INIT (SSTs, albedos, etc) from $OIFS_INIDATA_DIR to $AC_IC_DIR and if necessary truncate to length of experiment duration. 
• Copies the chemistry initial and boundary conditions file (emission fluxes, deposition velocities) ICMCL-INIT-COMPO from $OIFS_INIDATA_DIR to $AC_IC_DIR. If this file cannot be found it will be created. 

The syntax to run the prep script is:

./prep-ic-icmcl-compo.sh <start_date> <end_date>

with start_date and end_date having the format YYYY-MM-DD.

In the examples above for a 2-day forecast, the syntax should be  ./prep-ic-icmcl-compo.sh 2010-01-01 2010-01-03  to generate the data for 48 hours from 1st January 2010. 

Running the prep script will take a few minutes and will result in new data that will be written to $AC_IC_DIR, by default this location is $SCRATCH/43r3/experiments/{exp}/chem_ic.

This folder serves as a staging area for the initial and boundary conditions data and if the target files in $AC_IC_DIR already exist then they they will not be overwritten. 

Creating initial conditions at the end of an experiment run

When setting  LAC4IC=true an AC-for-IC grib file is created at the end of a run. It is a snapshot of all chemical/atmospheric composition tracers. This file can be used by the pre-processing script (prep script) to create new initial conditions. 

With the AC-for-IC file, initial conditions for a new experiment can be generated which will provide a continuation of the chemical evolution of the tracer fields from the previous experiment. The meteorological fields to complement the initial conditions can originate from an archived experiment or can be generated with the OpenIFS Data Hub. This workflow is desirable when aiming to reinitialise the model integration with an updated weather forecast or analysis data (which does not provide AC tracers).  

In order to restart with atmospheric composition tracers from the AC-for-IC grib file select the following settings in config.h:

• Either leave the variable OIFS_INIDATA_DIR empty, delete it or comment it out.
• Set the variable AC_IC_ONCE to point to the generated AC-for-IC grib file from the previous experiment. This file is typically found in this location in the previous experiment:
  ${RUNDIR}/output/${RUN_NUMBER}/ICMUA${EXP}.ac_rst
  
• Set the variable MET_IC_DIR to the relevant path for the new experiment's initial conditions or modify the "variables to retrieve METEO initial conditions from an archived experiment" section as appropriate.

Running the forecast experiment

Now the actual forecast experiment can be submitted to the batch scheduler using the platform-specific job wrapper script. Examples for this are found in $OIFS_HOME/AC-experiments/platform. An example template for the ECMWF HPCF Atos Sequana XH2000 is the file ecmwf-hpc2020.job.tmpl. 

The actual forecast experiment will be carried out inside the run directory which is defined in config.h with variable RUNDIR. The default on the ECMWF HPCF is $SCRATCH/43r3/experiments/${exp}/runs/.

Due to the much increased computational requirements for simulating atmospheric composition, forecast experiments are often broken into individual parts (sometimes referred to as "legs"). For each leg a restart of the forecast experiment needs to be carried out. 

Model Output
The oifs-run.sh script for OpenIFS/AC has been designed to be able to cope with restarts or "legs", and the model output from each experiment leg is stored in individual subfolders named by leg number. In the case of a single forecast experiment without restarts the output will be written to $RUNDIR/output/001.

Log files
Log file and experiment-related configuration data (namelist, NODE file, ifs.stat), as well as text files with diagnostic model outuput (e.g. mass and ozone budgets) are copied to $RUNDIR/log/001.

Restart Data
Model data relevant for restarts (not discussed here on this page) will be stored in $RUNDIR/restart/${next_leg_number}.

XIOS output
Files relevant to XIOS server output are stored in $RUNDIR/ifs_xml (not discussed here on this page). 

...

Multiple forecasts with restarts

There are several reasons why a user may wish to restart a forecast experiment. The increase in computational costs for simulating atmospheric composition may require to break a longer forecast experiment into individual sections ("legs"). Also, a restart will allow to initialise the meteorology in the model with data taken from an analysis of the current modelled time. 

Due to the time scales associated with the different chemical processes and the lifetime of chemical tracers with respect to their loss processes, it is often convenient to retain the chemical tracer fields in the model simulation and update only the driving meteorology at the time of the restart. In some applications, the model could even be initialised every single day with an up-to-date meteorological analysis field while the chemical fields are left to evolve.  The previous section described how chemical initial conditions can be created at the end of an experiment run, thus retaining the chemical tracer information from the previous experiment leg. 

We will here provide some examples of the principle steps required for restarting experiments.

Example 1:  Longer experiment with manual stop and restart

The experiment parameters need to be defined by editing config.h. The following settings are used for a 2-month experiment with a stop-and-restart after the first month. This will result in two experiment legs of 1 month duration each.

run_start_date="2010-01-01"
run_end_date="${run_start_date} + 2 months"
rst_freq="1 month"
USE_RESTART=true

The initial conditions for each leg need to be either created or gathered from a prepared repository.

prep-ic-icmcl-compo.sh  2010-01-01 2010-02-01
prep-ic-icmcl-compo.sh  2010-02-01 2010-03-01 

These processes can be run simultaneously. The meteorological fields and initial conditions can be requested from the OpenIFS Data Hub. The wrapper script needs to be submitted for the first experiment leg. Once the experiment has completed, config.h needs to be edited to use the initial experiment data from the date and time when the first leg has stopped. Then the wrapper script for the second month should be submitted.

Example 2: Using a workflow script

The 4 jobs in the previous example, can be all triggered automatically and in a timely fashion with a single script. The  $OIFS_HOME/AC-experiments/platform/<platform-name>-workflow.sh templates provide examples that need to be adjusted for the local hardware and file paths. In this case the config.h file only needs editing once and instead of the batch job wrapper script the workflow script is submitted. "1 month"
    └────
  • Create or gather ICs:
    ┌────
    │ prep-ic-icmcl-compo.sh  2010-01-01  2010-01-11
    └────
    It is recommended to run it on a compute node, if the ICMCL-COMPO
    has to be created.
  • Then submit a job to launch an oifs executable.
    ┌────
    │ {qsub, sbatch,...} wrapper.sh
    └────
3.4.2 Example 2 - Longer experiment with stop-and-restart
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
  • Define your experiment by editing /config.h/. To run a 2-month
    experiment with a stop-and-restart after the 1st month (ie 2
    legs/chunks of 1 month), just set:
    ┌────
    │ run_start_date="2010-01-01"
    │ run_end_date="${run_start_date} + 2 months"
    │ rst_freq="1 month"
    │ USE_RESTART=true
    └────
  • Create or gather ICs for every leg:
    ┌────
    │ prep-ic-icmcl-compo.sh  2010-01-01 2010-02-01
    │ prep-ic-icmcl-compo.sh  2010-02-01 2010-03-01
    └────
    These can be run in parallel. But if ICMCLDIR is not provided, then
    ICMCL climatology needs to be created, and you have to wait that the
    first job is finished before starting the others (which can then run
    in parallel).
  • Submit your wrapper-around-oifs-run once:
    ┌────
    │ {qsub, sbatch} wrapper.sh
    └────
    Wait until it is finished, then resubmit it again to simulate the
    second month. No need to edit your /config.h/, it will pickup where
    it stopped.
3.4.3 Example 3 - Using a workflow script
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
  The 4 jobs in the previous example, can be all triggered automatically
  and in a timely fashion with a single script. The
  /platform/ecmwf-cca-workflow.sh/ provides an example that works on
  cca@ECMWF.
  • Define a 2-month experiment with a stop-and-restart after the 1st
    month:
    ┌────
    │ run_start_date="2010-01-01"
    │ run_end_date="${run_start_date} + 2 months"
    │ rst_freq="1 month"
    └────
  • Create and queue jobs with correct dependencies:
    ┌────
    │ ./workflow-cca.sh
    └────
3.4.4 About performance - Long pre-processing
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
  ICMCL-COMPO could be prepared for the entire run at once. In the
  example 2 above, it would be done with:
  ┌────
  │ prep-ic-icmcl-compo.sh 2010-01-01 2010-03-01
  └────
  Then you simulate the two legs one after the other, like above. There
  is no advantage in doing that. Indeed, the creation of the ICMCL-COMPO
  can be quite expensive (~15 min for two days!). Better to limit it to
  monthly or daily files that can be stitched together afterwards if needed.
3.5 Creating ICs at the end of run
──────────────────────────────────
  When setting LAC4IC=true, an AC-for-IC grib file is created at the end
  of a run. It is a snapshot of all AC tracers, which can be used by the
  pre-processing script to create Initial Conditions. The main
  requirement is:
  • no netCDF output (LXIOS=false)
  With the AC-for-IC file, you can create IC for a new experiment, which
  will continue from the previous experiment as far as AC tracers are
  concerned. The meteo fields to complete the IC can come from an
  archived experiment or a set of standard meteo IC. This workflow is
  desirable when you want to reinitialize your integration with an
  updated weather forecast (which does not provide AC tracers) for
  example.
3.5.1 Example - A series of daily runs
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
  • Configure your 1st day experiment:
    ┌────
    │ run_start_date="2010-01-01"
    │ run_end_date="2010-01-02"
    └────
    With a restart frequency equal or longer than the run, you can
    switch off the creation of restart:
    ┌────
    │ rst_freq="1 month"
    │ USE_RESTART=false
    └────
    • Create ICs
      ┌────
      │ prep-ic-icmcl-compo.sh  2010-01-01  2010-01-02
      └────
    • Then submit a job to launch oifs executable
      ┌────
      │ {qsub, sbatch} wrapper.sh
      └────
      This will create an extra file we will use in the next day
      experiment:
      ┌────
      │ AUXFILE=${RUNDIR}/output/001/ICMUA${EXP}.ac_rst
      └────
  • Configure your next day experiment, with a new name (or after
    removing the previous rundir - after moving away the AUXFILE!):
    ┌────
    │ EXP=bcde
    │ run_start_date="2010-01-02"
    │ run_end_date="2010-01-03"
    └────
    In the /config.h/, be sure to:
     - set OIFS_INIDATA_DIR to nothing, or comment it.
     - set the variable AC_IC_ONCE to the AUXFILE from the previous exp.
     - set the MET_IC_DIR or modify the "variables to retrieve METEO initial
       conditions from an archived experiment" section as you see fit.
    • Create ICs
      ┌────
      │ prep-ic-icmcl-compo.sh  2010-01-02  2010-01-03
      └────
    • Then submit a job to launch oifs executable
      ┌────
      │ {qsub, sbatch} wrapper.sh
      └────
      This will create an extra file for another run:
      ┌────
      │ AUXFILE=${RUNDIR}/output/001/ICMUA${EXP}.ac_rst
      └────
  • Etc.
4 Output
════════
  Output data are on the reduced grid - which is an unstructured grid in the
  netcdf files. To go further, you may want to remap the output.  This can be
  achieved with /cdo/. The procedure is slightly different for grib and netCDF
  files. Check the scripts in the /postproc/ directory for examples dealing
  with grib files. For netCDF files, you must set a grid (grid descriptions
  are available in the /ctrl/ directory) before remapping:
  ┌────
  │ cdo -L setgridtype,regular -setgrid,${OIFS_HOME}/t21test_xios_ac/ctrl/TL255-griddes.txt \
  │      AMIP_6h_reduced_ml.nc oiac_6h_reduced_ml.nc
  └────
  Note that available UV-processor output is an experimental product, which is
  currently not filled (just zeroes). It concerns two variables: uvp2drad
  (gribcode 210055) and toaer340 (gribcode 210056)