Introduction

This page describes the basic steps of running a forecast experiment using OpenIFS.

files typically found in an OpenIFS 43r3 experiment. Experiment files are provided by the OpenIFS team at ECMWF as gzipped tarfile. These files provide initial and boundary conditions and currently are provided on request to the user for download via ftp.ecmwf.int.

Note that the experiment ID forms part of the name for all binary ICM* GRIB files. The experiment ID is also stored inside each GRIB file. Grib tools can be used to modify the experiment ID if required. Users should not just copy the files to a new name.

Other Files in the Experiment Directory

The experiment directory contains the initial data files described above. Moreover, other files and symbolic links are created when the model is run for the first time.

OpenIFS run script:
It verifies the required environment variables and file paths and creates symbolic links to other climatological and grid-related data directories.
It calls the binary model executable to start the experiment.
Once the experiment has completed the script carries out housekeeping procedures and initiates first post-processing steps of the Grib output files.

This page can only provide the basic steps to run OpenIFS as the requirements for at user sites vary considerably depending on the hardware, batch scheduling system and so on.

Set the environment

Use the oifs-config.sh file created when building OpenIFS to correctly set the run environment for the model.

% source ./oifs-config.sh

How to run the forecast: oifs_run

The recommended way to run the OpenIFS model to perform a forecast is to use the run script 'bin/oifs_run'. This has a number of arguments to control the model. It

How do I start the experiment?

If all the environment variables are set and the symbolic file path links exist and are valid, the experiment could in principle be started by executing the model binary master.exe with appropriate command line parameters, as it is done in the oifs_run script:

master.exe -r -e -n

In practice this method is not recommended.  It is far better to make use of the provided service script oifs_run which verifies all the settings and if necessary creates the missing links in the experiment directory to the ifsdata required. It calls the executable by invoking the MPI environment (which may well be hardware dependent) and will initiate postprocessing of the experiment results by moving the Grib GRIB model output to a target directory and concatenates and concatenate fields according to time step. oifs_run can be called with command line parameters which are passed on to the model executable. If no command line parameters are provided oifs_run will substitute with default values for experiment ID, grid resolution, number of tasks and threads, etc. which are likely to be of no use. Therefore the command line parameters need to be set.

By calling oifs_run the model will run interactively in the login shell.:

% oifs_run -r -e -n

Passing on the appropriate command line parameters can further be simplified by writing a brief batch job script which itself calls oifs_run with the required parameters . This job script is either run.ccb or run.ws.  run.ccb is a batch script which can be submitted to the Cray PBS schedule using the qsub ./run.ccb command.  run.ws is a shell script which can be called interactively from the command line and simply calls oifs_run.

The experiment is started by editing the job script of choice (either run.ccb for batch scheduler jobs or run.ws for interactive shell) with the appropriate parameters.

Either call or submit the appropriate job script:    qsub ./run.ccb    or   ./run.ws

(or to run interactively).