The discussions were organised along 3 main items : assimilation systems, observations and surface analysis. The 3rd point on surface analysis was eventually skipped, due to the lack of time.

1. Assimilation systems :

The blending technique was discussed. There exist now 3 different versions of the blending scripts, adapted to the local environment and usage in Prague (under SMS), Toulouse (refreshed research version in preparation) and Bruxelles (A. Deckmyn). This divergence seems presently unavoidable, but a reference version will anyhow be made available in the export version of AL25T1. It was stressed that the diverging local scripts can produce differences in the way each centre does its blending. Furthermore, a potential "bug" was detected : in pure assimilation mode, when the surface blending is switched off, the surface fields would always be those of the 6 hour ALADIN forecast, thus the surface would be kept unchanged throughout the assimilation cycle, which is undesirable for long cycles. A clear "safety" switch to restore dynamical adaptation when no blending is performed must be provided in every local installation (Prague does have such a switch in operations / SMS, but for the full blending algorithm : suppressing only surface blending is not allowed).

So far, blending is operational in Prague, and run in research mode in Toulouse and Bruxelles. The partners in Croatia plan to test and implement it, there are also some considerations in Romania and Portugal.

It was recalled that some forecast errors or shortcomings can be emphasized in the blending suite. A Deckmyn's example of the very bad scores on T2m was recalled : the blending suite simply amplifies the misforecast of T2m, when a strong inversion caps the PBL, because the errors in the forecasted fields are reintroduced during blending into the next "analysis". However, the actual problem is not a problem of blending , but one of the model forecast.

The point was raised whether surface blending was needed in addition to a genuine surface analysis. This point should be open to tests, since it is not clear whether the 2 can be complementary or overlapping. Surface blending is interesting if local adaptation to a better representation of physiographic data (orography, vegetation ...) improves the 6 hour forecast surface fields, compared to the analysis. Surface analysis is however the cleaner way to introduce new data and the relaxation to climatology.

The scientific program was reviewed, to update those items planned for 2002 : the developments around 3D-FGAT were confirmed (to start 2nd semester), continuation of the work on Jb and a posteriori validation. Issues on double-nesting were noticed as well (with blending - Belgium -, without blending - Hungary -). the continuation of some work on simplified physics is postponed to 2003 (via a short stay of C. Soci in Toulouse, to liaise with C. Loo and F. Bouyssel ?).

2. Observations :

The installation of ODB must be continued ; ODB is quite different from the CMAFOC principles, so a significant initial learning is needed in order to install the libraries the first time. An informal group for exchanging informations about SGI implementations is encouraged. Available documentation about ODB should be put on the Web.

The progressive use of observations or assimilation systems should be taken into account when new ALADIN domains are defined. So far, mostly telecom and computational arguments were considered , but "meteorological" connexity and homogeneity of the observation network should now also be addressed.

The validation of raw ATOVS data will be continued, and finished in fall 2002 (CY25T2). Specific LAM aspects for a sounder scientific use will be listed by Elisabeth Gérard. Impact studies are planned by Roger Randriamampianina.

The work on radar data is postponed to 2003. This work should be started by a thorough investigation of present use of radar data in other projects, and the possible use of the data in 3D-VAR or CANARI. The group of persons who can work on this topic should also be redefined.

Other works already done or in progress were listed : drift of RS, pseudo-profiles of relative humidity, RH2m, IASI data, aircraft data.

3. Surface analysis :

skipped from the discussions.

Home