This document describes the required tasks to carry out the control experiment to simulate the extra-tropical evolution of Tropical Cyclone Karl.
In contrast to yesterday's Acceptance Test experiment this will not require an interactive session on the Cray's postprocessing nodes. We will submit the control experiment as a batch job.
The Control Experiment
- We will use a pre-compiled and pre-installed version of OpenIFS 43r3 on the ECMWF's HPCF.
- We will carry out a 6-day (144 hours) control experiment starting on 25 Sep 2016 at 00h00 UTC.
- We will use OpenIFS with a cubic octahedral grid at horizontal resolution Tco399 and 137 vertical levels.
- The control experiment will be initialised from ERA-5 reanalysis data.
Preparing the Model
In your working group you should login to the RACC cluster and login to the ECMWF HPCF with your training account ID.
- Start the Mobaxterm application and open a local terminal
- Login to the RACC using
ssh -X cluster.act.rdg.ac.uk
On the cluster type the command:
ssh troifsX@ecaccess.ecmwf.int
Note: Instead oftroifsX
you should use your ECMWF training user-ID- You will be prompted for the hostname with a choice between
ecgate
,cca
andccb
. You should selectccb
. - When you have completed your work you can disconnect from
ccb
by typingexit
at the command prompt.
In a previous exercise you have called the my43
script which in turn has copied further scripts to your local ~/bin
directory. If you have not participated in yesterday's practical please get in touch so we can ensure your account is correctly set up.
You should now use the script getkarl
to copy the NAWDEX control experiment to your training account. This script can be called from any directory.
- Type the following command:
getkarl
The script copies all the files required for the control experiment to the following location: ~/scratch/tc-karl_h7cc
Note that this directory will contain the precompiled master.exe
file which already includes the software branch which we later will use to perturb the physics tendencies. In the control run we will leave everything at the standard setting.
Directories for climatological input files are mostly included through symbolic links to a central storage location.
Model Experiment Parameters
The model experiment directory should now contain the following files:
% ls 399_4@ ICMSHh7ccINIT master.exe sfcwindin wam_namelist ICMCLh7ccINIT cdwavein oifs_run specwavein wam_subgrid_0 ICMGGh7ccINIT fort.4 rtables/ uwavein wam_subgrid_1 ICMGGh7ccINIUA ifsdata@ run.ccb wam_grid_tables wam_subgrid_2
This directory contains initial files for the atmospheric model (ICM*INI*
) and also input files for the wave model.
399_4 links to a directory which contains data on sea surface temperature and radiatively active greenhouse gases (e.g. ozone).
rtables contains input files for the model's radiation scheme.
run.ccb is the model run script which will be submitted to the PBS job scheduler. It calls oifs_run
and loads all the module files and defines shell variable required for the runtime environment of the model. Therefore the oifs-config.ccb.sh
file is not needed in this case.
oifs_run is a generic run script for the model. It is possible to pass arguments to this script which is done by run.ccb
. At the end of the model experiment this script moves the model output to a specific output directory (which is numbered according to the OIFS_RUN
variable). Also, command line grib tools are used to extract selected model fields into separate files.
All settings and parameters which define the model experiments are set in namelists which need to be copied to fort.4
. We will explore here some of the values set in the namelists:
UTSTEP=900.0 ! model time step is 900 seconds CUSTOP='h144' ! experiment for 144 hours (6 days) CNMEXP="h7cc" ! experiment ID NFPLEV=137 ! model grid has 137 vertical levels NFPMAX=399 ! horizontal resolution is Tco399 NPOSTS=0 ! regular output every 24 steps, NHISTS=0 ! with UTSTEP=900 this means every 6 hours NFRPOS=24 NFRHIS=24
Submitting the Control Experiment
We will not run this experiment in an interactive node because this would not provide sufficient resources for a timely completion of the experiment. We will instead submit a job to the batch scheduler, requesting a larger number of computing nodes.
- Type the following command:
qsub ./run.ppn
- Verify the status of your submitted job by typing:
qstat -u $USER
or use the aliasqstatu
The qstat command indicates the current status of your job in the batch scheduling system. The status is indicated by a capital letter. At the start the letter 'Q' indicates that your job is waiting in the queue to be processed. 'R' indicates that the job is now running. 'E' indicates that the job has stopped and is exiting.