Case studies

Storm Xaver

A large and violent cyclonic storm hit the North Sea region and several adjacent countries on 5 December 2013. Problems were caused both by the high wind speeds and the related storm surge. The surge reached 6 metres in Hamburg for example and was the highest along the England east-coast for 60 years. In the aftermaths of the cyclone a blizzard hit Sweden. The storm system was named Xaver by Berlin's Free University; other names assigned elsewhere include Bodil, Sven and St. Nicholas (Hewson et al., 2014). The cyclone developed around 00 UTC on 4 December northeast of Newfoundland and it was situated between converging northerly and southerly airstreams. Due to the westerly wind jet accelerated by the convergence, the cyclone moved to northeast and east, deepening explosively. It had an intense meso-vortex hanging back to west, which enhanced the strong wind (see Figure 1). The cyclone was presented in the operational forecasts 8-9 days before the event and the forecasts indicated the very strong wind gust 3-4 days in advance. (Although some strength overestimation over Germany as well as timing error in surface pressure were concluded.)

Storm Desmond

Storm Desmond caused severe flooding, travel disruption and a power outage across northern England, parts of Scotland and Ireland on 5 December 2015. Cumbria in northwestern part of England is one of the worst affected regions with more than 200 mm of rain in 24 hours recorded in that area. Storm Desmond broke the United Kingdom's 24-hour rainfall record, with 341.4 mm of rain falling in Honister Pass, Cumbria. On Saturday, 5 December, UK Met office issued a red warning of heavy rain for Cumbria. The cyclone also led to flooding in southern Norway.

Orographical enhancement of precipitation played a major role in the event and the operational model of the ECMWF picked well the highest rainfall amounts over the orographical barriers. However, the forecast underestimated the peak values of about 100 mm in 24 hours in Cumbria and overestimated the precipitation amount in lee of the hills (Figure 2).

Model experiments

Several experiments have been conducted with OpenIFS for both cases with the aim to test the effect of starting date and forecast length, initial condition as well as spatial resolution to the forecast quality. The details of experiments are summarized in Table 1.

!! The input data and the namelists needed to run the experiments can be downloaded from the ECMWF FTP server: download.ecmwf.int/test-data. !!

The files are packed in .tar.gz files and structured into directories named after the case studies (i.e., Xaver_201312, Desmond_201512) and subdirectories indicating the main experiment characteristics (e.g., T255L91_ERA5). The archive files were prepared by starting dates (e.g., gs0c_2015120300.tar.gz). Typical content of a .tar.gz file is as follows:

The namelist file highlighted in green in the box above is needed to run OpenIFS. It controls the necessary settings (e.g., time step, experiment ID) as well as the post-processing. The most important namelist elements are listed below with their explanation:

In the evaluation, we would like to investigate the following variables:

• 2-meter temperature: its GRIB code number is 167;
• precipitation: it is composed from large-scale and convective precipitation with code numbers 142 and 143, respectively;
• mean sea level pressure: its code number is 151;
• 10-meter wind gust: its code number is 49;
• temperature at 850 hPa level: its code number is 130;
• relative humidity at 700 hPa level: its code number is 157;
• geopotential at 500 hPa level: its code number is 129;

• u and v wind components at 250 and 100 hPa with code numbers 131 and 132, respectively.

The listed variables have to be included in the namelist among the post-processing variables (see the code numbers and levels highlighted with bold characters in the box above). More information about the namelist settings and GRIB field codes can be found in the how-to articles: How to control OpenIFS output.

!! To run the model, the paths of the input data and the namelist have to be set in the job. ..... Acceptance testing OpenIFS after installation.

Preparation of data for visualization

After the job run is completed, the outputs are split into 2 files (located in the working directory): ICMGG for the gridpoint fields and ICMSH for the spectral ones. The two files should include all the necessary variables set in the namelist. However, before visualisation of the results, some steps are still needed.

Post-processing of model outputs

The Metview macros prepared for visualisation of experiment results desire the meteorological variables in GRIB files separated by variables and day with the following names:

• t2_${day}.grib for 2-meter temperature,
• p_${day}.grib for precipitation,
• mslp_${day}.grib for mean sea level pressure,
• gust_${day}.grib for 10-meter wind gust,
• t850_${day}.grib for temperature at 850 hPa,
• q700_${day}.grib for relative humidity at 700 hPa,
• z500_${day}.grib for geopotential at 500 hPa,
• u250_${day}.grib for horizontal wind components at 250 hPa,
• u100_${day}.grib for horizontal wind components at 100 hPa,

where day is the given date in format of yyyymmdd (e.g., 20151203). All the files should contain data for 8 timesteps per day (i.e., in every 3 hours).

The 2-meter temperature, the precipitation, the mean sea level pressure and the wind gust are expected in gridpoint representation, so will be taken from the ICMGG files. To prepare the needed input files for the Metview macros, the next operations are needed on the raw ICMGG outputs:

where expID is the 4-digit experiment ID and step is the post-processing step (every 3 hours). For precipitation, both the convective and large-scale precipitation components have to be in the same file:

The pressure level data are required in spectral representation, so they will be taken from ICMSH files. To prepare the needed input files for the Metview macros, the next operations are needed on the raw ICMSH outputs:

For wind, both the u and v components have to be in the same file:

            % grib_copy -w shortName=u/v,level=250 ICMSH${expID}+00${step} u250_${day}${step}.grib  #to get the u and v components at 250 hPa
% grib_copy -w shortName=u/v,level=100 ICMSH${expID}+00${step} u100_${day}${step}.grib  #to get the u and v components at 100 hPa

The size of the resulted files varies by the spatial resolution and the representation of the data. For instance, the file size at T255L91 resolution is 10 MB and 8 MB per variables for gridpoint and spectral fields, respectively, whereas these values increases to 35 MB and 26 MB at T639L137, to 233 MB and 179 MB at T1279L137.

Preparation of reference data

As reference data, we will apply the ECMWF re-analyses. Both ERA-Interim and ERA5 datasets are available for the users and can be downloaded from the ECMWF MARS (Meteorological Archival and Retrieval System) system. Re-analyses are created by optimal combination of available observational information and short-range numerical weather predictions using data assimilation techniques. They provide the most comprehensive description of the past and current states of the 3-dimensional atmosphere or the Earth system.

ERA-Interim dataset (Berrisford et al., 2011) was prepared on 79 km horizontal resolution with 60 vertical levels starting from 1979. Analysis fields were constructed in every 6 hours using variety of observations (conventional measurements, remote sensing data, air craft measurements etc.), the 4D-Var data assimilation technique and the IFS model version which was operational in 2009 (cycle 31r2). The forecasts initialized from the analysis produced 3 hourly outputs up to 24 hours. 

ERA5 (Hersbach and Dee, 2016) is being constructed on higher, 32 km horizontal resolution with 137 vertical levels from 1950. Analysis fields are is being prepared hourly with inclusion of newly reprocessed observational data, using the 4D-Var data assimilation technique and the IFS cycle 42r1 model version. ERA5 forecasts initialized from the hourly analyses produce hourly outputs up to 18 hours and give an estimation of forecast uncertainty. There is an important difference between ERA-Interim and ERA5 in handling of the accumulated parameters: in ERA5 the accumulation is calculated from the previous post-processing step (i.e., along one hour), while in ERA-Interim it is from the beginning of the forecast – this feature will have relevance in evaluation of the precipitation amount and wind gust. (More information about the characteristics of ERA-Interim and ERA5 can be found in the Copernicus Knowledge Base: What are the changes from ERA-Interim to ERA5?)

The Metview macros prepared for visualisation of re-analysis data desire the meteorological variables in GRIB files separated by variables for a time period with the following names:

• ${dataset}_t2_${period}.grib for 2-meter temperature,
• ${dataset}_p_${period}.grib for total precipitation,
• ${dataset}_mslp_${period}.grib for mean sea level pressure,
• ${dataset}_gust_${period}.grib for 10-meter wind gust,
• ${dataset}_t850_${period}.grib for temperature at 850 hPa,
• ${dataset}_q700_${period}.grib for relative humidity at 700 hPa,
• ${dataset}_z500_${period}.grib for geopotential at 500 hPa,
• ${dataset}_u250_${period}.grib for horizontal wind components at 250 hPa,
• ${dataset}_u100_${period}.grib for horizontal wind components at 100 hPa,

where dataset is a 2-digit identifier of the re-analysis data, ei for ERA-Interim and eafor ERA5; period is the investigated time period in format of yyyymmdd-yyyymmdd (e.g., 20151201-20151206 for Desmond).

The 2-meter temperature, the precipitation, the mean sea level pressure and the wind gust are expected in gridpoint representation, while the pressure level data are required in spectral representation. Total precipitation and wind gust as accumulated parameters derive from forecasts, all the other variables are real analyses. Consequently, the daily quantities for precipitation and wind gust are composed of 8 and 24 timesteps from ERA-Interim and ERA5, respectively, the other variables have 4 and 8 timesteps (recall that output frequency of the forecast experiment is 3 hours). Besides the two (large-scale and convective) precipitation components, total precipitation is also available for direct retrieve both in ERA-Interim and ERA5, with GRIB code 228.

To download the necessary data from MARS, the following steps have to be accomplished:

1. To have access to the ECMWF public datasets, you will need to have an account on ECMWF web site: https://apps.ecmwf.int/registration/.
2. To retrieve data from MARS, you will need to download and install an ECMWF key. This page shows that step by step: Access ECMWF Public Datasets.
3. Having your key, you can run the scr_download_re-analysis shell script available in the OpenIFS repository (for cycle 40r1) and using getmars for retrieve.

The re-analysis source (i.e., ERA-Interim or ERA5), the surface and pressure level variables to be retrieved, the period of the data and the output directory have to/can be specified for the scr_download_re-analysis script. Calling it with -h option, it gives a detailed help with some examples at the end (or calling it without any option, it also gives short instructions to its configuration):

% ./scr_download_re-analysis -h
 ----------------------------------------------------------------------
        This script downloads surface and pressure level ERA-Interim or ERA5 re-analysis data for a given time range from MARS.
        Usage:
        -c[class],-s[surface_variables],-p[plevel_variables],-f[firstdate],-l[lastdate],-o[output_directory]
        -h/-help

        Examples:
       ./scr_download_re-analysis -cei -s"t2 p mslp gust" -p"t850 q700 z500 u250 u100" -f20151203 -l20151205 -o"../reference"
       ./scr_download_re-analysis -cea -sall -p"t850 q700 z500 u250 u100" -f20151203 -l20151205 -o"/home/rd/digs/metview/paper_OIFS/input/reference"
       ./scr_download_re-analysis -ce5 -s" " -p"t850" -f20151203 -o"/home/rd/digs/metview/paper_OIFS/input/reference"
 ----------------------------------------------------------------------

The retrieved ERA-Interim fields occupy approximately 54 MB, while ERA5 fields take 700 MB for the case Desmond (i.e., for 1–6 December 2015).

Visualization with Metview

The visualization package contains the following folder and file:

Using the Metview macros

The visualisation is done by the Metview macros ( *.mv files) interactively using a dialogue box or in batch mode.

Interactive dialogue box

With right click on the macro and then selecting the Execute  option from the menu, the settings can be seen in a dialogue box:

Batch mode

In batch mode the macro can be executed following the next syntax:

            % metview -b macro option1 option2 option3 ...

where macro is the macro to be run ( plot_raw_IC.mv, plot_forecastrun.mv or plot_ERAI_ERA5.mv). A detailed help together with some useful examples is provided with simple execution of the macro:

            % metview -b macro

The shell script scr_run_macros executes the macros from the Unix/Linux shell and it can be tailored for the own needs.

External functions and macros

To reach these functions and colour definitions, the path of the definitions directory has to be added to the METVIEW_MACRO_PATH (e.g., in .bashrc ).

Input data

Expected input data:

• Macro plot_raw_IC.mv expects the raw ICM* files as input: ICMCL${expID}INIT, ICMGG${expID}INIT, ICMGG${expID}INIUA, ICMSH${expID}INIT, where expID is the 4-digit experiment ID.
• Macros  plot_forecastrun.mv and plot_ERAI_ERA5.mv expect grib files as input with the following file names: ${variable}_${date}.grib where variable can be t2, mslp, p, gust, t850, q700, z500, u250, u100; date is day in format yyyymmdd.

Output figures

All the macros produce figures in single-page .ps files. The file name can be as follows:

• plot_raw_IC.mv: ${variable}_${level}_ERAI-ERA5${date}+${timestep}.ps,
  where variable can be stl2, lnsp, z, t, cc, u, q; level can be 0 (in case of surface variables) or from 1 to 91; date is day in format yyyymmdd; timestep is forecast lead time in hours, e.g., 0, 3, 6 etc.
• plot_forecastrun.mv: ${variable}_${level}_${expID}_${date}+${timestep}.ps,
  where variable can be t2, mslp, p, gust, t, q, z, u, q; level  can be 0 (in case of surface variables) or 850, 700, 500, 250, 100; expID is the 4-digit experiment ID; date is day in format yyyymmdd; timestep is forecast lead time in hours, e.g., 0, 3, 6 etc.
• plot_ERAI_ERA5.mv: ${variable}_${level}_${reference}_${date}.ps,
  where variable can be t2, mslp, p, gust, t, q, z, u; level can be 0 (in case of surface variables) or 850, 700, 500, 250, 100; reference can be ERAI or ERA5; date is day in format yyyymmdd.

References

Berrisford, P., Dee, D.P., Poli, P., Brugge, R., Fielding, K., Fuentes, M., Kållberg, P.W., Kobayashi, S., Uppala, S., Simmons, A., 2011: The ERA-Interim archive Version 2.0. ECMWF ERA Report Series, 27 p. [PDF]

Hersbach, H., Dee, D.P., 2016: ERA5 reanalysis is in production. ECMWF Newsletter 147, p. 7.

Hewson, T., Magnusson, L., Breivik, O., Prates, F., Tsonevsky, I., de Vries, H.J.W., 2014: Windstorms in northwest Europe in late 2013. ECMWF Newsletter 139, 22–28. [PDF]