Modelling

Some advances have been done for the gribing process of the atmospheric variables. With these GRIB headers, it should be possible to write the products directly into MARS from a post-processing task and avoid incurring a major ECFS store and restore overhead. This means that the archiving on MARS might be done "on-line" by the modelling partners. It would be preferable to do the archiving in "real-time", soon after the data are produced to avoid the ECFS overheads where possible.
An up-to-date version of the GRIB encoding software (gribex) and the file-handling routines (pbio) is avaliable in:
/home/ma/emos/public/gribex_13032.tar.gz
There is a description of updates in:
/home/ma/emos/public/gribex_13032.readme
A marsgen-type template has been developed to help partners with archiving into MARS, and rules and guidelines will be given for file organisations suitable for input to MARS. The template may be fit to every partner requirements. Some archiving tests should be carried out by Metapps.
The ocean GRIB files with the right headers may also be generated. A general purpose routine is provided to perform the task.

Modelling

The full set of wind and SST perturbations for DEMETER is already available at ECMWF. This set has been created for all the dates corresponding to the three streams of ERA40, i.e., 1957 to 2001. It is made by:

SST: four perturbations per initial date in DEMETER, that is, four ensembles of four perturbations per year. Data are in a 1x1 degree resolution.

wind: monthly files containing daily data for one perturbation of the U and V component of the wind stress in a T85 global grid. This set starts in January 1957 in order to be able to force the ocean analysis.

The data are in the DEMETER ECFS filesystem ec:/DEMETER/pert. The files have been written in ieee format and the grids have been organized from the dateline eastward. A description of the file contents is provided in this text. A documentation provided by Frederic Vitart (who also computed the perturbations) and describing the way the perturbations have been computed is available here. An example of the perturbations has been plotted: SST perturbations for 1st November 1986, u-wind and v-wind perturbations for 1st January 1986.

Modelling

Some results on the impact of data assimilation in runs using wind perturbations for creating the ocean initial state are provided here. Jerome Vialard did the experiment, provided the figures and prepared the description that you can find below.
The ECMWF strategy for generating the ensemble of ocean initial conditions for DEMETER follows. The SST of ocean analysis and the wind forcing used during the oceanic data assimilation will be perturbed. Some DEMETER partners, however, are not planning the use an oceanic data assimilation scheme. They will have at their disposal a set of wind perturbations for creating the initial state. The goal of this note is to assess the impact of the wind perturbations in terms of the ensemble spread with and without data assimilation (a strong relaxation of -400 W/(m^2K) to observed SST beeing applied in both cases). The use of data assimilation colapses the perturbation induced by the wind in the thermal field, as can be seen in Figure 1. This is more clearly appreciated over the Pacific ocean, where a large amount of data is available. This results in a larger spread of the forecast when data assimilation is not used, as shown in Figure 2. This simply reflects that the ocean state is known with a better accuracy in the case of data assimilation. We think that the wind perturbations can be used in the absence of oceanic data assimilation, and will not result in an unrealistic spread.

Modelling

An initial strategy for the use of the SST and wind perturbations can be consulted here. Please, take into account that this is just a suggestion, every modelling group having his own preferences.The content has already be sent by email to every modelling partner.

Modelling

J. Slingo (Univ. of Reading) asked for the inclusion of skin temperature in the common list of atmospheric variables, following the demands of the PROMISE users. However, this variable is available just for IFS and the Met Office models. Thus, this field will be available from ECMWF, METO, and LODYC, under request.

Modelling

Modelling partners should be registered as users at ECMWF. By doing that, they will obtain a secure-id card with which they will have access to a complete documentation about ECMWF resources, among other things. Links are available at ECMWF where the centre's computer system and tools are described (secure-id card or internal access needed). Please, to get information on all of these topics, contact your corresponding User Support contact person.

07.12.2001