On Monday 24th of January 2000, the working group (Météo-France, INMH and HMIS) for the drafting of the new MoU met in Ljubljana for one day of discussion around the current MoU and its necessary modifications to be presented in the next Assembly of ALADIN Partners.
The code of AL12 was eventually completed and validated in January 2000. All the phasers present in November-December 1999 did a great job, and it is worth mentioning the appreciated contribution of Gabor Radnoti, whose experience as former LACE/ASC was very helpful. The ALADIN project for sure needs a few more people of such scientific and technical level.
Although there were as usually a few surprises coming from ECMWF, the phasing was fairly smooth, and the finalized cycle could be delivered for operations in Toulouse in February. AL12 is presently the operational ALADIN cycle at Météo-France, along with CY22T1 of ARPEGE/IFS.
- some in-depth namelist cleaning
- recoding of dummy arrays in the spectral transforms
- code cleaning specific to ALADIN, such as the merge of e(ree)(spe)sm with dm
- extra code for wavelet cost-function
A bugfix is needed to run the non-hydrostatic 3tlsl version
- the LACE team has also implemented the AL12 library, and
found incompatibilities between openMP directives and multitasking mode (see
the mail by Gabor Radnoti sent to the alabobo list).
- there is one important lesson to extract from this: since in future the
natural technical evolution goes towards the use of the standard microtasking
directives from openMP, and due to its progressive insertion into the ALADIN
code, the level of non-compliance with multitasked/macrotasked platforms will
increase. Time is now more than ripe to think about alternative operational
speed-ups to replace multitasking for those still relying on it.
- NEC implementations might switch on message passing, which is now more and
more mature in many ALADIN configurations, and other sites have already plans
to leave the historical sphere of influence of Cray computers. Morocco also
intends to move to distributed memory on its CRAY J90, till a new computer is delivered.
- those who will continue on multitasked platforms like J912 must check the
possibility to implement and test openMP libraries on them.
In April-May-June, the ALADIN code was phased to the ARPEGE cycle CY22T2. The former cycle AL12 was with respect to CY22T1. As often, the phasing was rather chaotic and far from optimal:
- ECMWF sent us its cycle CY22R2 with 3 weeks of delay, so
that the ALADIN phasing did not start in time.
- the overlap with holidays and seminars involving GMAP people caused some
extra delay.
- finally, CY22T2/AL13 and CY23 (soon after, this is the true common cycle
ECMWF/Météo-France, without ALADIN and still requiring
significant cleaning in dynamics) clashed a bit in time in Toulouse.
In this context, the phasers showed a great sense of tolerance and adapted to the situation. We were thus in a position to declare CY22T2/AL13 beginning of June, although not all the configurations had been tested, and bugs were of course still present (NHS, CANARI - 3DVAR not fully tested).
Mid-July, a thorough debugging period allowed to create a bugfix version for AL13, for which almost every configuration is working optimally:
- hydrostatic models (e001,e501,e401,e601,e801) mono-and
multi processor
- non-hydrostatic models (e001,e501,e401) mono-and multi processor
- fullpos and e(e)927 mono-and multi processor
- canari, but the multi processor versions for sure still have
"marginal" bugs
- 3d-var multi processor
- provisional skeleton for predictor/corrector advection scheme
- additional code for semi-Lagrangian TL and AD code
- a new key LMPOFF to switch off MPI in mono-proc mode (only for e001&fullpos)
- bugfix for test of TL and AD non-hydrostatic full model in multi-proc mode
- first developments for ozone as a spectral thermodynamical variable
- new output fields in fullpos: CAPE ...
- the option to use background statistics in gridpoint for ALADIN 3d-var
- removal of calls to the NAG library in 3d-var (not yet tested in Toulouse !)
This cycle is planned for phasing in November/December 2000. It will most likely be phased with CY23Tx. See the list the technical changes in CY23 and those expected in CY24 (article of Claude FISCHER in this Newsletter: "Better to know ...") for estimation of the modifications that should be introduced in AL14.
Note that CY24 is scheduled for end 2000, so ALADIN might still be lagging one cycle behind ARPEGE at the end of 2000, as it is now in July.
During the last quarter, two small workshops enabled significant progress in 3D-Var work : in Budapest with A. Horanyi, C. Fischer and M. Siroka (see Article in this Newsletter) and in Toulouse with C. Fischer, W. Sadiki and A. Dziedzic (see report on Toulouse work in this Newsletter).
Organized by Andrew Lorenc (The Met.Office), this workshop was held at The Met.Office College in Reading, England, 8-10 May 2000. ALADIN was represented by C. Fischer, E. Gérard and A. Horanyi.
One day and a half was devoted to presentations and two half days to working groups.
As an introduction, Jean Quiby, SRNWP coordinator, presented the current organization of NWP development within Europe in four groups (HIRLAM, COSMO, ALADIN and UKMO) and the role of each lead centre. The operational data assimilation methods within those groups have been specified: OI to be soon replaced by 3DVar in Hirlam, 3DVar in UKMO, OI and nudging in global and local German models respectively, and 3DVar to be soon replaced by 4DVar in ARPEGE (which was effectively done in June 2000).
Beside the presentations about the current developments within the four SRNWP groups, representatives of American models (MMM-NCAR, NCEP, RUC/MAPS) and Canadian model (CMC) gave their expertise in mesoscale variational assimilation. Five working groups were defined with reflection subjects entitled: Large scale and topographic forcing, boundary conditions (1-2); High resolution observations: their assimilation and utility (3); Assimilating clouds and precipitation observations (4); 4DVar, how can we make it affordable ? (5); Analysis techniques, 3DVar and error covariances (6).
For more informations about the issues and conclusions of this workshop,
the abstracts and working groups reports are available on The Met.Office web
site:
http://www.met-office.gov.uk/sec5/NWP/DA_MesVAR/
.
The 8th ALADIN workshop, entitled "Toward high resolution modelling" was organized at IMWM / Cracow (Poland) on 19-20 June 2000. Participants were : Yong Wang (Austria), Luc Gérard, Philippe Nomerange and Piet Termonia (Belgium), Dagmar Merkova (Czech Rep.), Elisabeth Gérard and Dominique Giard (France), Gabor Radnoti (Hungary), Evgheni Alexeev (Moldova), Marek Jerczynski, Witold Owcarz, Malgorzata Szczech and Jadwiga Woyciechowska (Poland), Vanda Maria Sousa da Costa (Portugal), Oldrich Spaniel (Slovakia), Mark Zagar (Slovenia) and Jean Quiby (Swiss). Jean Quiby was present as the SRNWP coordinator and as a representative of the COSMO group. Though invited, the HIRLAM and UKMO groups did not send anyone. The ALADIN participation was unexpectedly small, distorted, and hopefully not representative of the implication of partners in research. This was quite disappointing for the organizers, the SRNWP coordinator and those partners who made a significant effort. All the more since this workshop was expected to illustrate the effective emergence of deported research and discuss the mid-term completion of the ALADIN medium-term research plan. Moldova expressed its willingness to involve itself further into the ALADIN project.
Some of these presentations are detailed in articles hereafter.
Mark Zagar presented the first results (see Article in this Newsletter) of the operation "Towards coordinated exchange of information " launched by the Slovenian team at the previous ALADIN workshop. It was quite successful since only 3 partners have not yet answered (Bulgaria, Moldova and Morocco) and about 50 applications were described. Some redundancies already appeared. It was decided to :
The decisions of the last Assembly of Partners concerning the coordination of verification within ALADIN were recalled. Apart from the creation of the e-mail list verifala++at++meteo.fr nothing was done. Who will initiate the exchange of EWGLAM-type scores? More details on the definition of scores may be asked to the French contact points: Samuel.Westrelin++at++meteo.fr, Francis.Pouponneau++at++meteo.fr.
The importance for each team to control and eventually re-tune the orography for its domain, once again underlined by the results of the recent study of Steluta Alexandru, was recalled. Since e923 can now run on workstations, thanks to the work of Neva Pristov and Mehdi El Abed, and as the corresponding documentation was published, there is no longer any major obstacle.
The discussion on the medium-range scientific plan underlined the importance of a mid-term report, with a detailed description of the studies related to each item. Such a check-up is very useful but cannot be achieved solely by the small coordination team. Volunteers for synthesis on some topics and reports on deported research are welcome ! If this succeeds, the report will afterwards be put on the ALADIN web server and updated regularly.
Thanks to the Cracow team for the very nice welcome (in a really beautiful town) !
|
Home |