Summary of changes

• an evaluation package is now included, comprising scripts and Metview macros. This uses two sets of case studies for verifying the user's OpenIFS model against OpenIFS forecasts performed at ECMWF and plot standard maps for comparison (see below).
• reset of accumulated output. New namelist variables have been added to allow accumulated variables (e.g. total precip.) to be reset at a specified frequency. See NRSACCFREQ & NRSACCOFF in yomppc.F90 and namelist namppc.nam.h.
• aquaplanet configuration. New variables introduced to set the min & max SST (with thanks to Jakob Beran, Stockholm)
• bugfix in wave model code which would cause a crash when writing a restart file.
• long-runs. A number of routines have been modified to ensure some integer variables and counters do not overflow on very long integrations (with thanks to Uwe Fladrich, SMHI and EC-Earth).
• modified SETTLS scheme for improved stratospheric warming forecasts and various improvements for the semi-Lagrangian interpolation and mass fixers (see further notes below).
• fix to bug in qnegat.F90 where negative values of moisture were not reset correctly, leading to a very slow moistening of the atmosphere. Negligible difference for medium range forecasts.
• ecCodes is now supported by a change to grib_api_interface.F90. In line with this change, environment variables OIFS_GRIB_API* have been renamed to OIFS_GRIB* and OIFS_SAMPLES_PATH has been renamed OIFS_IFS_SAMPLES_PATH to avoid confusion with the default samples. As grib-api will not be supported after the end of 2018, we recommend users migrate to ecCodes.
• sucpl0.F90 setup routine for coupled runs now added (was missing in previous version).
• spectral_arp_mod.F90 modified to add security fix to computation of noise (has potential to alter results for ensembles).
• bugfix to cptime.F for RS7K architecture (with thanks to Siham SBII, Morocco).
• new command create_ex available to create the experiment directory structure for single or ensemble forecasts.
• intel compiler options improved to prevent some crashes due to speculative execution of if branches on some hardware.
• some I/O statistics (from iostream_mix) are now reported correctly in the model log file (NODE.001).
• python/ecmwfapi updated to version 1.5.
• minor efficiency improvements.
• some surplus code removed.

Note some features from Fortran 2003 are now used, that may require more recent versions of compilers (allowing support of the ASSOCIATE construct).

We also do not recommend version 15 of the Intel compiler which has known issues with the OpenIFS code.

Technical

This version is tested with a range of compilers and known to work with the following using the configurations supplied:

Version 15 of the Intel compiler is not recommended as it has known issues with the OpenIFS code.

Version 16.0.0 of the Intel compiler may generate compile errors in sufa.F90. We recommend using a more recent version if possible.

IBM compiler. Although configuration files are provided for the IBM compiler, we can not test it at ECMWF and welcome any corrections from users.

The Known Issues page has more information about potential problems with the OpenIFS code.

New evaluation package and test-data

A new evaluation package has been added to OpenIFS that allows the user to verify the meteorological performance of their model version against the same case studies run at ECMWF. This includes a new test-data download site (see below), with a step-by-step guide on how to download the initial files, ECMWF reanalyses for comparison and a gallery of the forecasts performed at ECMWF. The OpenIFS team will use these case studies with each new release to catalogue the changes in model releases. This evaluation suite will also be useful in analysing contributions from users. A more extensive evaluation suite is used at ECMWF to assess changes for the operational IFS but this is too complex to be run outside ECMWF.

For more information

See OpenIFS User Guide : Meteorological Evaluation

After introduction of the two cases, this section guides the user step by step through running the selected experiments, post-processing the model outputs, downloading the reference (ERA-Interim and ERA5) data and visualising the model outputs with Metview. A catalogue from the prepared figures is also included. The input data and the namelists needed for the model runs, the scripts for the MARS retrieves of the re-analyses, the Metview macros for visualization and the output figures are available for download on the ECMWF download server.

See ECMWF Newsletter article : New forecast evaluation tool for OpenIFS

All figures, initial data for case studies, scripts, Metview macros and re-analysis data are available from the ECMWF download server: http://download.ecmwf.int/test-data/openifs/reference_casestudies

Modified SETTLS scheme for improved stratospheric sudden warming

In 40r1v1, during stratospheric sudden warmings (SSW), noisy flow fields can occur resulting in incorrect temperature forecasts at heights typically above 10hPa.  This improved scheme works by identifying gridpoints prone to noise and applies a non-extrapolatory scheme improving the representation of the warming. Below 60hPa, this has a neutral impact.

For reproducibility with the old behaviour in 40r1v1, add the following:

NAMDYN
  LSETTLSTF = false,

Some further passive changes (ie. not enabled by default) are:

• a modified quasi-monotone limiter which applies the limiter in all three directions at the end of the interpolation rather than at each direction separately. When applied to the specific humidity it reduces the cold bias error of long-range (climate) forecasts.
• alternate gridpoint mass-fixer for pressure, instead of the standard spectral mass fixer, which can enforce global mass conservation to near machine precision.
• global mass diagnostics for total air mass and tracers. This is extra diagnostic information that can be useful to see global mass conservation errors per timestep.

For further information on the improvements to the forecasts of SSW, please see the article by Michail Diamantakis in ECMWF Newsletter no. 141, Autumn 2014 (p.30).

Grib-api / ecCodes

This version of OpenIFS supports the use of ecCodes, the replacement for grib_api.

Grib-api will no longer updated after the end of 2018, we therefore recommend users move to ecCodes.

Acknowledgements

The OpenIFS team would like to acknowledge the research and support of ECMWF scientists involved in this release and the contribution of the OpenIFS users mentioned above.