Description of the upgrade

This cycle produced a remarkable improvement over the previous version (Cycle 36r4 implemented on 9 November 2010). Cycle 37r2 combined a number of significant scientific contributions with the instalment of GRIB-2 that permits the encoding of data on a larger number of model levels as required by the increased vertical resolution planned for 2012 The scientific components of the Cy37r2 cycle enhanced the accuracy of both the analysis system and forecast model.

Implemented: 18 May 2011  




Atmospheric (unchanged)

  • HRES: TL1279
  • ENS: Leg A T639, Leg B T319, Leg C T255

Wave (unchanged)
  • HRES: 0.25 degrees
  • ENS: 0.5 degrees

Atmospheric (unchanged)

  • HRES: L91
  • ENS: L62

Meteorological content

  • Background error variances from the ensemble of data assimilations (EDA) used by the deterministic 4D-Var assimilation
  • Improvements to cloud scheme formulation
  • Changes to all-sky assimilation of microwave data
  • Reduction of AMSU-A observation errors and adjustments to MODIS AMVs
  • Accounting for tangent point drift in GPS radio occultation

Meteorological impact

The impact of the new cycle on the performance of the forecasting system has been tested in research mode during the period June to December 2010, and in pre-operational runs during the period since 1 January 2011.

The new cycle shows clear benefit in terms of objective upper-air scores in the medium range in both hemispheres: temperature and winds are improved throughout the troposphere. Tropical wind scores are also improved. The improvements to the cloud scheme increase the humidity in the upper troposphere. This results in some increase in bias, but overall provides a better fit to observations and improves humidity scores in the extra-tropics in the early forecast range.

There are no specific changes to the ENS configuration apart from the model changes described above. The impact on the ENS is a small reduction in spread throughout the forecast range. The change is neutral in terms of probabilistic scores.


See Newsletter No. 128 - Summer 2011

New and changed parameters

New parameters


Changes to existing parameters


Technical content

Migration to GRIB edition 2

From cycle 37r2, ECMWF uses WMO FM-92 GRIB edition 2 (GRIB-2) encoding for its model level fields for the deterministic forecast and the EPS, including monthly forecasts. GRIB-1 model-level data will no longer be produced and disseminated. This applies to model level products only; all other operational data will remain in GRIB-1.

See Migration to GRIB edition 2


In order to be able to interpolate GRIB-2 fields, users must upgrade to version 000381 (released on 22 March 2011) or higher of EMOSLIB.

Second-order packing

Encoding of second order packed data in GRIB-1 and GRIB-2 has been implemented in version 1.9.8 of grib_api.

  • GRIB-1 second order data encoded with grib_api can be decoded by GRIBEX.
  • Encoding of GRIB-2 second order data is not fully compliant with the WMO FM-92 specifications and therefore a local template has been designed for it. For further details, see GRIB API keys for GRIB edition 2.
  • Decoding of second order packed data produced by GRIBEX has been fully implemented in previous grib_api versions.

Latitude and longitude values for reduced Gaussian grids in GRIB-2

Latitude and longitude values of the first and last point are encoded in millidegrees in GRIB-1 and in micro-degrees in GRIB-2. As those values are encoded with the highest possible accuracy in each GRIB edition users should expect to get different values in GRIB-1 and GRIB-2 fields.

The latitudes and longitudes for each grid point can be obtained from grib_api with high accuracy using the keys latitudes, longitudes, distinctLatitudes, distinctLongitudes or using the grib_get_data Fortran functions or the grib_iterator C functions.

As grib_api is computing latitudes and longitudes from the order N of the Gaussian grid the values are identical in both editions and they are very accurate. Definitions of the various Gaussian grids in use at ECMWF can be found here.

Coding of negative longitude values in GRIB-2

According to the WMO FM-92 GRIB edition 2, only positive values of longitudeOfFirstGridPoint, longitudeOfLastGridPoint, longitudeOfSouthernPole are allowed. Negative values of longitudeOfFirstGridPointInDegrees, longitudeOfLastGridPointInDegrees, longitudeOfSouthernPoleInDegrees are converted by grib_api to positive values by adding 360 to the value before encoding.


Please see GRIB edition 1 to 2 conversion. GRIB-2 to GRIB-1

PrepIFS experiments

PrepIFS experiments for IFS cycle 37r2 and above produce data in GRIB-2 for model level parameters and GRIB-1 for the rest.


  • The trajectory module in Metview 3 and the corresponding trajectory database have not been migrated to GRIB-2. The trajectory database has been discontinued.
  • The 'multi-analysis suite' producing forecasts from other global NWP centres' analyses has been discontinued. Such data was stored in MARS as stream = AMAP / MAED / MAWV.

Model ID and experiment number

Model ID = 37r2

E-suite experiment number = 0053


ECMWF Newsletter: See Newsletter 128