Check the initial conditions
The inital conditions are available under the master troifs0 training user. You do not need to copy them, but you can check their content.
troifs1@cca-login4:/perm/ectrain/troifs1> ls -ltr /perm/ectrain/troifs0/user-cray/oifs40r1/data/input/
File name |
|
Description |
Size |
2015120300/ICMCLgs0cINIT |
: |
input file containing surface and soil information (albedo, soil temperature etc.) |
9.3 MB |
2015120300/ICMGGgs0cINIT |
: |
input file containing gridpoint surface initial data |
7 MB |
2015120300/ICMSHgs0cINIT |
: |
input file containing initial data for the prognostic variables in spectral representation |
35 MB |
2015120300/ICMGGgs0cINIUA |
: |
input file containing initial data for the prognostic variables in gridpoint representation |
101 MB |
2015120300/wam_grid_tables |
: |
model grid and tables for the wave model |
52 MB |
2015120300/wam_subgrid_0 |
: |
information for model advection, including sub-grid parametrisation for the wave model |
12 MB |
2015120300/wam_subgrid_1 |
: |
information for model advection, including sub-grid parametrisation for the wave model |
25 MB |
2015120300/wam_subgrid_2 |
: |
information for model advection, including sub-grid parametrisation for the wave model |
25 MB |
2015120300/cdwavein |
: |
initial value of drag coefficient for the wave model |
63 KB |
2015120300/specwavein |
: |
initial wave spectra for the wave model |
7.6 MB |
2015120300/uwavein |
: |
initial value of wind speed for the wave model |
63 KB |
2015120300/sfcwindin |
: |
initial value of 10-meter horizontal wind components and sea ice fraction for the wave model |
2.2 MB |
Prepare the experiment
Go to the working directory where your experiment will run and clean it if it has been already used:
% cd ${OIFS_HOME}/rundir % rm -f ICM* ifsdata rtables *l_2 wam* *in fort.* master.exe MPP*
Copy or link the initial conditions from the input directory to your working directory:
% INPUTDIR=${OIFS_HOME}/data/initial_conditions/gs0c % ln -sf ${INPUTDIR}/ICM* . % ln -sf ${INPUTDIR}/wam* . % ln -sf ${INPUTDIR}/*in .
- Set the necessary environment variables:
- OIFS_GRIB_API_DIR: folder of GRIB-API
- OIFS_DATA_DIR: folder of climate data called ifsdata
- LD_LIBRARY_PATH: folder of GRIB_API libraries
- GRIB_SAMPLES_PATH: path of GRIB_API samples
Copy the namelists to your rundir:
% NAMDIR=${OIFS_HOME}/data/initial_conditions/gs0c % cp ${NAMDIR}/wam_namelist . % cp ${NAMDIR}/wam_namelist_coupled_000 . % cp ${NAMDIR}/namelistfc fort.4
Modify the fort.4 file with setting the following namelist variables:
Number of processors you use to run OpenIFS: NPROC in the NAMPAR0 group,
- Set the NPRNT_STATS (in the NAMPAR0 group) to the same value as NPROC,
- Timestep: TSTEP=2700 in the NAMDYN group,
- Number of vertical levels and spectral truncation: NFPLEV=91 and NFPMAX=255 in the NAMFPG group,
- Experiment ID: CNMEXP="gs0c" in the NAMCT0 group,
Number of latitudes and longitudes on the transform grid: NLAT=256, NLON=512 in the NAMFPD group,
- Fullpos variables in the NAMFPC group (for more details please see section 6.2),
Output file names: setting LINC="true" in the NAMOPH group leads to ICM??${expID}+00${step} output file names where step is in hour instead of timestep (here on 4 digits).
Run the model using the oifs_run.sh script which is in the OpenIFS 40r1v2 package. To run the script you have to specify the experiment ID, the executable file (named master.exe), the forecast length and you can set the spectral truncation as well as the timestep here:
% # oifs_run.sh -e${expID} -m${executable} -r${RESOUTION} -f${SCLENGTH} -s${TSTEP} % ${JOBDIR}/oifs_run.sh -egs0c -m${BINDIR}/master.exe -r255 -fh96 -s2700. > out 2>err
With these settings you run a 96-hour forecast (-fh96), where JOBDIR and BINDIR are the location of the oifs_run.sh script and the master.exe OpenIFS executable, respectively.
Before the model run begins, the oifs_run.sh script checks the inputs and ensure that the necessary climate data on the corresponding resolution are also available in your working directory.When OpenIFS starts to run, you will see the following content in your working directory:
For more information about the content of the working directory, please see section 6.2.master.exe fort.4 # executable and namelist wam_namelist wam_namelist_coupled_000 # namelists for the wave model ICMGGgs0cINIUA ICMCLgs0cINIT ICMSHgs0cINIT ICMGGgs0cINIT # initial conditions cdwavein sfcwindin specwavein uwavein # wave model initial conditions wam_grid_tables wam_subgrid_0 wam_subgrid_1 wam_subgrid_2 # wave model files ifsdata rtables 255l_2 # climate data ncf927 ifs.stat NODE.001_01 # text output (log) files out err # output and error files from the model run ICMGGgs0c+000000 ICMSHgs0c+000000 ICMGGgs0c+000003 ICMSHgs0c+000003 # output files for the first two steps
For more instructions and details about running the model, please visit the related page in the OpenIFS User Guides.
Post-processing the model outputs
The Metview macros prepared for visualization of experiment results require the meteorological variables in GRIB format separated by variables and days with the appropriate names.
The 2-meter temperature, the precipitation, the mean sea level pressure and the wind gust are expected in gridpoint representation. They are prepared from the ICMGG* outputs with the next operations:
% expID=gs0c; date=20151203 % steps="00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72" # every 3 hours from 0 UTC on 3 December to 0 UTC on 6 December % for step in ${steps} do grib_copy -w shortName=2t ICMGG${expID}+0000${step} t2_${date}_${step}.grib #to get the 2-meter temperature grib_copy -w shortName=msl ICMGG${expID}+0000${step} mslp_${date}_${step}.grib #to get the mean sea level pressure grib_copy -w shortName=10fg ICMGG${expID}+0000${step} gust_${date}_${step}.grib #to get the 10-meter wind gust done
For precipitation, both the convective and large-scale precipitation components have to be gathered in the same file:
% expID=gs0c; date=20151203 % steps="00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72" # every 3 hours from 0 UTC on 3 December to 0 UTC on 6 December % for step in ${steps} do grib_copy -w shortName=lsp/cp ICMGG${expID}+0000${step} p_${date}_${step}.grib done
The pressure level data are required in spectral representation. They are prepared from the ICMSH* outputs with the next operations:
% expID=gs0c; date=20151203 % steps="00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72" # every 3 hours from 0 UTC on 3 December to 0 UTC on 6 December % for step in ${steps} do grib_copy -w shortName=t,level=850 ICMSH${expID}+0000${step} t850_${date}_${step}.grib #to get the temperature at 850 hPa grib_copy -w shortName=r,level=700 ICMSH${expID}+0000${step} q700_${date}_${step}.grib #to get the relative humidity at 700 hPa grib_copy -w shortName=z,level=500 ICMSH${expID}+0000${step} z500_${date}_${step}.grib #to get the geopotential at 500 hPa done
For wind, both the u and v components have to be collected in the same file:
% expID=gs0c; date=20151203 % steps="00 03 06 09 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72" # every 3 hours from 0 UTC on 3 December to 0 UTC on 6 December % for step in ${steps} do grib_copy -w shortName=u/v,level=250 ICMSH${expID}+0000${step} u250_${date}_${step}.grib #to get the u and v components at 250 hPa grib_copy -w shortName=u/v,level=100 ICMSH${expID}+0000${step} u100_${date}_${step}.grib #to get the u and v components at 100 hPa done
After these operations, the timesteps belonging to the same days have to be merged into a common file and the concatenated file has to be moved to the input directory of the Metview visualization:
% expID=gs0c; date=20151203 % variables="t2 mslp gust p t850 q700 z500 u250 u100" % for variable in ${variables} do cat ${variable}_${date}_*.grib > ${variable}_${date}.grib mv ${variable}_${date}.grib ${OIFS_HOME}/data/${expID} done
More details can be found about the post-processing in section 6.3.
Download the re-analysis data
As reference for the evaluation, we use ERA5 re-analyses. The data are provided on the ECMWF download server in the proper format needed for visualization with Metview or they can be downloaded in their original format from the ECMWF MARS system. This retrieve has to be accomplished only once, if we download the data for the whole period of the case study.
To download the re-analysis GRIB files from the download server into the input directory of the Metview visualization:
% cd ${OIFS_HOME}/data/reference % wget -c http://download.ecmwf.int/test-data/openifs/reference_casestudies/data/Desmond_201512/reference/ea_20151201-20151206.tar.gz
More information is provided about the background of the different re-analyses in section 6.3.
Plotting
The visualization package is available in the casestudies directory of the OpenIFS cycle 40r1v2 release as well as in the ECMWF download server: http://download.ecmwf.int/test-data/openifs/reference_casestudies/programs/metview. Transfer the content of the definitions and macros folders into the corresponding local directories. Going to the macros folder, the visualization can be done by the Metview macros (*.mv files) in two ways: (1) interactively using a dialogue box or (2) in batch mode.
Before using the macros, the path of the definitions directory has to be added to the METVIEW_MACRO_PATH to reach the external functions and colour definitions:
% cd ${OIFS_HOME}/macros % export METVIEW_MACRO_PATH="${OIFS_HOME}/definitions:${METVIEW_MACRO_PATH}"
There is an include statement in the plot_forecastrun.mv and plot_ERAI_ERA5.mv macros and further 2 ones in plot_IC.mv taking the two colour definitions from this directory. The path of the definitions folder has to be set in the macros according to the local working tree. To do that, open these files in a text editor, search the definitions and modify its path if it is necessary (4 times in total).
(1) Interactive plotting
Forecast maps
Re-analysis maps
(2) Plotting in batch mode
To prepare maps from the T255L91 forecast initialized with ERA5 on 3 December 2015 for the validation date, i.e., 5 December 2015:
% # metview -b plot_forecastrun.mv [experiment_ID] [resolution] [initial_condition] \ % # [parameters] [pressure_levels] [start_date] [verification_date] \ % # [area] [input_directory] [output_directory] % metview -b plot_forecastrun.mv gs0c T255L91 ea "t2,mslp,p,gust" " " 20151203 20151205 "25,-35,75,50" ../data/forecast/ ../figures/ # surface parameters % metview -b plot_forecastrun.mv gs0c T255L91 ea "t" "850" 20151203 20151205 "25,-35,75,50" ../data/forecast/ ../figures/ # t850 % metview -b plot_forecastrun.mv gs0c T255L91 ea "q" "700" 20151203 20151205 "25,-35,75,50" ../data/forecast/ ../figures/ # q700 % metview -b plot_forecastrun.mv gs0c T255L91 ea "z" "500" 20151203 20151205 "25,-35,75,50" ../data/forecast/ ../figures/ # z500 % metview -b plot_forecastrun.mv gs0c T255L91 ea "u" "250,100" 20151203 20151205 "25,-35,75,50" ../data/forecast/ ../figures/ # u250, u100
To prepare maps for the validation date, i.e., 5 December 2015 from the ERA5 data:
% for date in 20151205 20151206 do metview -b plot_ERAI_ERA5.mv ERA5 "t2,mslp" " " ${date} " " ../data/ ../figures/ # surface parameters metview -b plot_ERAI_ERA5.mv ERA5 t 850 ${date} " " ../data/ ../figures/ # t850 metview -b plot_ERAI_ERA5.mv ERA5 q 700 ${date} " " ../data/ ../figures/ # q700 metview -b plot_ERAI_ERA5.mv ERA5 z 500 ${date} " " ../data/ ../figures/ # z500 metview -b plot_ERAI_ERA5.mv ERA5 u "250,100" ${date} " " ../data/ ../figures/ # u250, u100 done % metview -b plot_ERAI_ERA5.mv ERA5 "p" " " 20151205 " " ../data/ ../figures/ # precipitation % metview -b plot_ERAI_ERA5.mv ERA5 "gust" " " 20151204 " " ../data/ ../figures/ # windgust
More details about plotting with Metview are available in section 6.4.
Troubleshooting
Here some cases are collected, when the model run and the forecast evaluation can terminate with error.
Model run
- The initial conditions are prepared for fix dates and spatial resolutions. Make sure that your settings coincide with that.
- The initial conditions are prepared with fix experiment ID that you can identify from the file names (e.g., in ICMGGgs0cINIT experiment ID is gs0c). Please make sure that you use the proper expID in the job and the namelist, too.
- Further possible issues regarding running the OpenIFS and their solution are listed in the OpenIFS FAQ.
Visualization with Metview
- If you do not find the evaluation package (basically the Metview macros) in your local computing environment, you are likely to use earlier OpenIFS version than OpenIFS40r1v2. In that case, please download and install this version from the FTP server or download the Metview macros and the corresponding macro definitions from the ECMWF download server.
- Metview macros can stop with error message if not the recommended software versions are used. The supported (minimum) versions are GRIB-API 1.18.0, ecCodes 2.5.0 and Metview 4.7.1.
Using Metview 4.7.1 or 4.7.2, the next warnings are produced:Metview 4.8.0–4.8.5 give the following warnings:macro - WARN - 20180611.125356 - Ambiguous verb: 'READ' could be: macro - WARN - 20180611.125356 - READ (METVIEW) macro - WARN - 20180611.125356 - READ (TIGGEMETVIEW) macro - INFO - 20180611.125356 - Choosing READ (METVIEW) macro - WARN - 20180611.125356 - Ambiguous verb: 'READ' could be: macro - WARN - 20180611.125356 - READ (METVIEW) macro - WARN - 20180611.125356 - READ (TIGGEMETVIEW) macro - INFO - 20180611.125356 - Choosing READ (METVIEW) macro - INFO - 20180611.125358 - Warn: Magics-warning: [NONE] is not a valid value for SUBPAGE_MAP_PROJECTION: reset to default -> [cylindrical] macro - INFO - 20180611.125358 - Warn: Magics-warning: tolerance condition error for 3.14159 -1.5708
Metview 4.8.6–5.0.3 return with the following message:macro - INFO - 20180611.125358 - Warn: Magics-warning: [NONE] is not a valid value for SUBPAGE_MAP_PROJECTION: reset to default -> [cylindrical] macro - INFO - 20180611.125358 - Warn: Magics-warning: tolerance condition error for 3.14159 -1.5708
You can neglect all these warnings, they do not influence the output figures.macro - INFO - 20180611.125358 - Warn: Magics-warning: [NONE] is not a valid value for SUBPAGE_MAP_PROJECTION: reset to default -> [cylindrical]
Metview macros might terminate with the next message if METVIEW_MACRO_PATH is not set:
macro - ERROR - 20180611.143638 - Line 2729 in 'plot_IC.mv': Function not found: build_layout_2plus1(list) metview: EXIT on ERROR (line 1), exit status 1, starting 'cleanup'
Metview macros stops with error if the include statements in the .mv files are not set properly:
base_visdef: No such file or directory macro - ERROR - 20180611.144342 - Line 151: Cannot include file metview: EXIT on ERROR (line 1), exit status 1, starting 'cleanup'
- Metview macros partly require mandatory directory structure. The folder of the input data can be arbitrary, however, under this directory macros expect the re-analysis fields to be placed in the subdirectory named reference, while the forecast data in the subdirectory identified with the 4-digit experiment ID (i.e., data/reference and data/gs0c etc.). Please make sure that the input data are stored in these folders with the proper file names (more information about the requested file names can be found in section 6.4).
- Metview macros can fail if they refer to non-existing directories. Please make sure that folders of the input data, output figures and the macro definitions are created.
- Saving the outputs can cause surprise if you do not append "/" at the end of the path when specifying the output directory. Please use "/" at the end of the input data path, too.
Metview macros are constructed to process the data for given dates and parameters. If you use them for different dates and meteorological variables (it is easy to do that in batch mode), they will return with error message. To avoid that you can run the macros first in help mode, typing:
% metview -b macroname
Please be aware that Metview macros require the input parameters in pre-defined order and number. If you do not want to specify a given parameter, use " " instead, e.g.:
% metview -b plot_IC.mv Desmond "t2,mslp,p,gust" " " 20151203 "25,-35,75,50" ../data/forecast/ ../figures/
where macro plot_ic.mv is executed for surface parameters, therefore, vertical levels are not specified.