An AWOC meeting took place during the ALADIN workshop in Bucarest. This is the report from Claude Fischer.
The extended activities of the AWOC group have been recalled during the discussions at the Aladin workshop in Bucarest (devoted to fixing installation or operation problems, code, compiler or related bugs).
In this frame, it was noticed that the auxiliary library had missing routines (xrd19). Also, non-standard non-ANSI code was detected (to be in general reported to the gourous). Duplicated declarations of variables cause some compilers to fail. Also, the order of the declarations should be carefully checked. Some shortcomings have been reported recently (march) by Luc Gerard and the gourous have sworn some of them will be fixed over xrd21.
Some news on CY20 were given (see the detailed report in this newsletter). Also, the cleaning for CY21 was mentioned: IMPLICIT NONE, free source format, fortran 90 generic types used for the definition of real expression accuracies (see Eric Sevault's description in the previous newsletter). It was recalled that the surface reservoir for frozen water in the soil parametrization will not become operational before april 99. The preparations for the switch have started and the date of the change should be May, 17th.
The status of the documentation was mentioned. For the reports, the Aladintheque is regularly updated with the latest ''stagiere's`` reports. Nothing was started to translate the documentation of the physics into english. The existence of a rather complete IFS documentation was mentioned. This documentation can be installed on the Prague webserver for the LACE countries but does not concern the Arpege physics. For non-LACE countries, only a delivery via mail, on their specific demand, is possible.
Nothing new this quarter.
The main events of the first quarter of 1999 are :
- Quasi-operational implementation of CANARI
Since the end of January the CANARI optimal interpolation scheme is executed quasi-operationally providing inputs for the nowcasting applications at the Hungarian Meteorological Service. At the moment CANARI is running every 6 hours using local SYNOP and TEMP observations (the observations are read from the local data base then it is converted into ASCII format for Mandalay and then Mandalay creates the direct CMAFOC input for CANARI).
The quality of the CANARI analysis is planned to be monitored and improved continously. In the near future it is planned to have analysis every 3 hours, later on we might continue with hourly analysis.
It is emphasised that at this stage CANARI is only running for nowcasting purposes, however later on we plan to perform some data assimilation experiments as well.
- Migration to SGI ORIGIN2000
In the beginning of 1999 a new SGI ORIGIN 2000 machine with 6 processors was delivered to our Service (now there are two more additional processors available). The speed of the processors is 0.4 Gflop each.
In the beginning of February we started the adaptation of the ALADIN model to the SGI platform using the distributed memory version of the code. After the successful adaptation some tests were carried out for showing the scalability of the machine for the ALADIN/HU operational domain. The speed-up of the machine with 8 processors is 6.7 (see the speed-up curves enclosed).
In the near future we will port the operational ALADIN/HU application to the SGI platform and after the stabilization of the operational suite a domain and resolution increase is planned.
- Verification
Finally the necessary softwares were completed for the relative comparison of NWP models applied in an operational manner at the Hungarian Meteorological Service. The software is able to calculate simple statistical scores of the models with respect to SYNOP and TEMP observations. At the moment the ECMWF model, the ALADIN/LACE and the ALADIN/HU models are compared using the observations available on the ALADIN/HU domain. The visualisation of the results are carried out using the same software as it is in Toulouse (using gnuplot).
News from the Council:
The 9th RC LACE Council held in Prague on 25th February 1999 approved the new shape and candidates of the Management Group as follows:
- Project Leader: Radmila Bubnová, CZ
- Deputy Project Leader: Dijana Klaric, HR
- Prague Team Leader: Martin Janousek, CZ
- Deputy Prague Team Leader: Mária Siroká, SK
- Project Scientific Officer: Thomas Haiden: AU
- ALADIN/LACE System Co-ordinator: Gábor Radnóti, HU
This Management Group will start its work on 1st May 1999 and has been nominated for 1 year.
Evolution of the ALADIN/LACE application.
In the beginning of the year the Prague Team concentrated on the validation of the cycle AL09/CY19T1/XR19, which repatriation started still before Christmas. The validation in a parallel suite revealed a bug affecting the use of the 8th lateral boundary conditions file. The bug was immediately reported to all ALADIN sites.
The cycle AL09/CY19T1 has been put into the operations on 10/02/1999 for 12 UTC network time (cycle AL09/CY19T1/XR19) with neutral results with respect to the previous cycle AL08T3/CY18T3/XR04, containing the switch LSETTLS.
Research & Development, Parallel Suites
The Prague Team has developed the necessary tool of automated parallel suite environment and agreed with COMPAS and TTI at Météo-France on the possibility to launch ALADIN/LACE parallel suites coupled with ARPEGE parallel suites and thus to see an impact of the change in ARPEGE on ALADIN/LACE. The mechanics was first tested on the ''radiances & screening” ARPEGE parallel suite in the beginning of March. Thus it was ready to test the foreseen ''LFGELS” suite.
The Prague Team completed the ''verif.pack” procedure in the following terms: generality (use for the operational & parallel suite & research experiments, enabling more general combination of experiment dates and choice of the samplings; graphical presentation of scores, based on NCAR. The first version of this package is being documented and will be available for any ALADIN site (the portability requires: configuration 701, NCAR).
The Prague Team launched the parallel tests to assess the impact of all the modifications before they entered the operational application:
- Test of the cycle AL09/CY19T1.
- Research suite testing the orographic lift and gravity wave drag schemes.
- Test of the pre-cycle AL10/CY20 for R & D purposes only, revealing neutral scores.
- Research ''USURIC” test, which goal is to tune the vertical exchange coefficients of turbulence. This suite has been one of an helpful element in the study of last December storm.
Other R & D developments:
3D VAR: work on Jb and Jo; CANARI: validation for the cycle AL09/CY19T1 & start of the work on diag.pack; physics: tuning of the new convection scheme.
Other technical developments:
Port of the linking script SLD, GRIB tool for getting extreme values of fields over a defined period, test of MPI version, different tests of MANDALAY functionalities (proved to be quite a painful exercise), preparation of the local LACE-domain SYNOP database.
Since 25th of February '99 high-speed telecommunication link beetwen IMWM and Cracow Supercomputer Center works in operational service. Link is based on Ericsson Mini-link radioline capable of 2x2Mb/s transfer rate. At present only one channel is being used. In near future our NWP operational system will be able to switch automatically from radioline to ISDN in case of any problems with radio-link.
(more details mario.almeida++at++meteo.pt )
The ALADIN/Slovakia model is running once per day ( LBC from 00:00 UTC ALADIN/LACE run) in operational suite with horizontal resolution 7,18 km.
Graphic outputs (NCAR) are included into same graphical menu at meteorological workstation as products from ALADIN/LACE. Outputs products from ALADIN/Slovakia are following : Clouds (total, low-level, middle, high), 2m temperature, 10 m wind, precipitation (stratiform, convective, liquid, solid).
Also verification surface fields (ALADIN/LACE =96 ALADIN/Slovakia =96 observations) are running and results are available on Intranet environment.Some successful results was done in field of visualisation pseudoTEMPs.
Information about pre-operational ALADIN/SI model:
Two workstations are used: ee927 and visualisation on "calvus" (DEC AlpaStation 600 5/333, 192 MB RAM, 333 Mhz 21 164 CPU, 4MB secondary cache, DigitalUnix 3.2).
The model integration on "castelanus" (BIT Alpha 500, 256 MB RAM, 500 Mhz 21164 CPU, 1 MB secondary cache, DigitalUnix 4.0d).
- space resolution :11.16 km (LELAM), 31 levels.
- integration domain: number of grid points x*y (72x72 for C+I, 8 points in E zone), latitude of reference (46.0), longitude of reference (13.5), longitude and latitude of the 4 corners (SW corner : 8.6864, 42.3278; NW corner: 8.0257, 49.4358; SE corner : 18.3136, 42.3278; NE corner : 18.9743, 49.4358)
- coupling files from ALADIN/LACE, 6 hours coupling frequency
- range of forecast: two runs daily: 00 UTC and 12 UTC +48 hours
- duration of integration until the maximum forecast range : optimally 110 minutes (14s for one time step of integration 2TL-SL (400s))
The address of the Slovenian WWW server containing:
- information about Slovenian version of ALADIN : /www.rzs-hm.si/napoved/aladin_opis.html (in the Slovene language)
- operational ALADIN forecast : /www.rzs-hm.si/napoved/aladin.html
- high resolution wind forecast : /www.rzs-hm.si/napoved/DADA.html
Applications on ALADIN/SI forecast:
- meteograms for 19 stations (model points)
- amount of precipitation for 19 stations (model points) and 6 selected areas (catchment) used mainly for hydrology forecast
- creation of pseudotemps for 3 stations (model points)
- creation of file with model data for Vis5d interactive visualisation
- sequence of the pseudosatelite pictures
- dynamical adaptation of surface wind for 3 selected areas and 5 forecasted ranges
- vertical cross-sections for 3 pre-selected slices (in testing)
ALADIN/SI model availability in 1998
The overlook of the availability of ALADIN/SI products has been prepared for last year. The starting and the ending time of the integration was analysed from 11th November 1997 (when also afternoon run was introduced) until the end of the year 1998 (see figure).
Run |
morning run |
afternoon run | ||
|---|---|---|---|---|
Completed |
373 |
90% |
331 |
80% |
without intervention |
348 |
84% |
324 |
78% |
rerun |
25 |
6% |
7 |
2% |
operational not finished |
68 |
16% |
92 |
22% |
no results |
43 |
10% |
85 |
20% |
Table1: Availability of ALADIN/SI model results from 11/11/1997 to 31/12/1998 (416 days)
It can be seen from the table that one out of ten morning model runs is missing and app. one is very late; in the afternoon, two out of ten are missing.
The managing of the operational suite is still in our group responsibility. If something goes wrong in the afternoon run, there's usually nobody available to fix the problem and to rerun the model. The situation is slightly better for the morning run, when model has been rerun quite often. Longer "blank" time periods appeared when there was nobody there to fix the suite (due to holidays, absence, etc.).
The most often objective reason for integration failure is incomplete transfer of coupling files. In the first part of the year the files have been transferred via InterNet from Toulouse. After the introduction of the transfer of coupling files via IBM-GN from Vienna this problem disappeared.
Often integration stopped because a coupling file was not ready in time and the model was not waiting for it. This has repeatedly appeared for time periods after the new libraries had been introduced (end of March-begin of April, end of July- begin of August).
With availability like the one presented here, we can not claim that our model is really an operational one. But on the other hand, from September 1998 conditions improved and number of failures was significantly reduced.
