ERA Interim meteo

From tm5
Jump to: navigation, search



General information for ERA-Interim reanalysis data can be found at, which mentions that "an open-access [journal article] describing the ERA-Interim reanalysis is now available from the Quarterly Journal of the Royal Meteorological Society."

ECMWF continues the Era-Interim stream on a monthly basis, with a delay of about 2 months. Pre-processing for TM5 regular grids is done with TM5 onto the global 1x1 resolution and the native 60 layers (ml60), by Ph. Le Sager (KNMI). Data are processed as soon as they are made available by ECMWF. In addition, data coarsened to 25 model levels (ml25) for selected regions and resolutions are also made available (see table below for details), by A. Segers (TNO).

Archived data sets are available in two formats (HDF and netCDF) and can be found at ECMWF (on their ECFS archive system) below these top directories:

  ec:/nlh/TM/meteo/ec/ei/      ; for HDF format
  ec:/nlh/TM/meteo-nc/ec/ei    ; for netCDF format

See below for more details. Note that the HDF archive stopped in 2015-12-31. It is recommended to use the netcdf for all time periods.

The volume of the ml60/glb100x100 dataset is about:

Availability of the netCDF archive

The netCDF files for the 1x1 resolution are in these directories:


where YYYY specifies the year. Note that the lsm and oro datasets are in YYYY=0000. All fields at 1x1-60L are available from 1979/01/03 to 2016/08/30 included. All data can be downloaded through ftp (account required). See Getting input data for details.

Availability of the HDF archive

For the HDF files, look into:


The table below summarises the years for which processed ERA-interim data are available.

  L60 T255/N128
  levels sfcml60 ml60/tropo25
grid  glb100x100 glb100x100 glb600x400 eur300x200 eux100x100 nam300x200 nam100x100 glb300x200
size 12 Gb/yr 288 Gb/yr 8 Gb/yr 4 Gb/yr 9 Gb/yr 5 Gb/yr 8 Gb/yr 26 Gb/yr
1979/01/03-1988 x x - - - - - -
1989-2003 x x x x x x x -
2004-2012 x x x x x x x x
2013-2015/12/30 x x - - - - - -
  Legend:      x  available      - not planned or not started yet

Using ERA-Interim convective fluxes and diffusion coefficients

For ERA-Interim, two versions of convective etrainament/detrainment fluxes are available.

SUB files

The 'sub' files contain fluxes computed from ECMWF 3D temperature/humidity/wind fields based on Tiedtke et al..

CONVEC files

The convec' files with fluxes created from the convective fields archived by ECMWF for ERA-Interim (Berrisford et al, 2011, section 2.7.1) .

These files were first produced november 2013 after experiments comparing north-south gradients in CH4 and SF6 between TM5, C-IFS, EcEarth/TM5, and TM5 versions configured as close to EcEarth/TM5 as possible. The results showed that the convective fields were one of the key parameters that could explain the differences: if TM5 uses the convective fields dumped by EcEarth/TM5, the computed north-south gradients are much closer to the observations than originally. After this result it was deceided to make the convective fields archived for Era-Interim should be made available for TM5 too. Code to do this was already present from experiments with ERA40 meteo described in (Olivié et al, 2004a), for which the convective fluxes were archived too.

Note that for ERA-Interim accumulated fluxes and rates are archived, thus integrated in time. Units are therefore 'kg/m2' for the fluxes and 'kg/m3' for the entrainment/detrainment rates. For use in TM5 the fields are converted to temporal averages, thus units 'kg/m2/s' and 'kg/m3/s' and valid for 3-hourly intervals [00,03], [03,06], etc.

A zipfile with figures of zonal averages of the entrainment/detrainments for various model experiments is availble; after unpacking, load the index page in your browser.

To use the new convec files in TM5-MP, make sure that the following key is set to F in your main rc file, otherwise the aforementioned 'sub' files are used:

convective.fluxes.tiedtke : F

To use the new convec files in TM5-zoom, ensure that:

   ! convec fields are 3 hourly averages:
   meteo.tinterp.convec             :  aver3
   ! define that the convective fluxes should be read from 'convec' files, instead of from the default 'sub' files;
   ! the extra 'filekey' setting is the essential key to enforce this, keep the rest close to your original settings;
   ! TIP: for testing, first use a dummy name (for example 'XconvecX), the model should now complain about missing files ...
   tmm.sourcekey.glb300x200.convec  :  tm5-hdf:mdir=ec/ei/fc012up2tr3/tropo25/glb300x200;tres=_00p03;namesep=/;filekey=convec
   ! if necessary, ensure that the 'convec' files are installed instead of the 'sub' files:
   tmm.setup.003.apply    :  T
   tmm.setup.003.files    :  sp mfuv mfw t q cld convec
   tmm.setup.003.sources  :  tm5-hdf:mdir=ec-ei-fc012up2tr3-tropo25-glb300x200;tres=_00p03   :  ec:ecfs[nlh]:TM/meteo

DIFFUS files

Apart from convective fluxes, also 'turbulent diffusion coefficient for heat' is archived. Defined on half levels, archived from model top (half level zero) to top of surface layer (half level L), implicitly zero at the surface (half level L+1). See (Olive et al, 2004b) for evaluation of similar fields available from ERA40.

For use in TM5 these fields are gridded to regular resolutions. Archived fields are accumulated in time and stored at 3 hourly resolution. Native units are 'm2' (per 3 hour), in the meteo processing these are averaged in time to 'm2/s'. The value at the surface is explicitly set to zero.

To use these diffusion files files, ensure that:

   ! diffusion fields are 3 hourly averages:
   meteo.tinterp.diffus             :  aver3
   ! if necessary, ensure that the 'diffus' files are installed:
   tmm.setup.003.apply    :  T
   tmm.setup.003.files    :  sp mfuv mfw t q cld convec diffus
   tmm.setup.003.sources  :  tm5-hdf:mdir=ec-ei-fc012up2tr3-tropo25-glb300x200;tres=_00p03   :  ec:ecfs[nlh]:TM/meteo



netCDF production was temporarily on hold

With the change of hardware at ECMWF (from an IBM power7 to a CRAY XC30) in September 2014, it was not possible to pre-process the met fields in netCDF format out-of-the-box. The gribex and udunits v1 (i.e. fortran bindings) libraries are not available any more. TM5 code has been modified to use the grib_api module and an interface with the udunits v2 (C binding) has been implemented. These new libraries work fine when converting GRIB files to HDF files, and when coarsening netCDF files at 1x1 to netcdf files at 3x2. However, conversion from HDF to netCDF, and from GRIB to netCDF failed with the netCDF4-parallel-IO library. Commenting all the code that has to do with udunits does not change this conclusion. It turned out that the netCDF library was conflicting with both the grib_api and the hdf4 library. Reverting to the netCDF lib without parallel IO solved the problem.

problem with g10m after March 2014

It has been reported "that there was a change in the g10m files between April and March 2014 in the headers for time1 and time2. In March they are both the same, whereas in April there’s a 3 hour difference". The issue has yet to be investigated.

Corruped convection fields

Fields pre-processed on c1a at ECMWF, with openMP, and with a version of TM5 older than revision #4378 (2012-04-20), have been corrupted. This is an openMP issue, and it concerns only the 'SUB' met fields for the 03/01/1989-31/01/2012 period. The conclusion:

As far as convection fields are concerned, the differences are not large enough to consider reprocessing the meteo.

For details, see the Bug in Processed Convection Fields page.

Personal tools