The operational ALADIN models

1. Status report of the operational ALADIN versions

ALADIN runs operationally on a wide range of computers, from single workstations to vectorial computers in shared or distributed memory, and on a cluster of workstations under Linux. 12 operational ALADIN suites are presently running in various NMS. 7 coupling domains are produced operationally at Météo-France, 1 in Prague.

See updated informations on this web server at : http://www.cnrm.meteo.fr/aladin/oper/oper.html .

2. Operational version at Austrian Meteorological Service

Implementation of MANDALAY

In order to be able to use DIAGPACK for analyzing surface parameters like surface pressure, wind, visibility, precipitations, ..., the tool MANDALAY was installed at ZAMG as the first step for implementing DIAGPACK. Some technical investigations on the adaptation of data from the ZAMG database to the data format needed by MANDALAY were also done.

Implementation of 500 hPa geopotential verification of ALADIN

A 500 hPa geopotential verification routine for ALADIN-VIENNA was developed. This routine calculates bias, mean absolute error, rms error, standard deviation, and anomaly correlation for the 500 hPa geopotential each day. The verification procedure will be used for monitoring purposes, e.g. when the model cycle is changed.

3. Operational implementation of ALADIN-Belgium

Main operational work :

• Installation of AL12/CY22T1/XR22 under MPI on a HP 9000/800 system: this work required the modification of about 290 auxiliary routines, and some corrections in ALADIN routines
• Installation of AL12/CY22T1/XR22/Cycorabis under MPI and Linux on PCs.
• Installation of all (locally developed) operational ALADIN tools on a PC under Linux.

Development

• CANARI assimilation
• Implementation of a parallel run including a data assimilation suite based on CANARI and cycle AL11T2. Unfortunately, the scores are not much better.
• Implementation and test of a new database for vegetation and soil data.

4. Operational version at Bulgarian Meteorological Service

Last news in Newsletter16.

5. Pre-Operational version at Croatian Meteorological Service

The 3rd quarter of the year has been devoted to the organization and preparation of the pre-operational version of ALADIN-HRVATSKA. Preparation steps:

• porting the AL11T2 version from Ljubljana
• parallel processing tests on 10 processors (three 2xPIII 800 MHz 256 Mb RAM and two 2xPIII 667 MHz 256 Mb RAM)
• used software: OS Linux RedHat 6.2, MPICH1.2, F90 Fujitsu compiler
• use of Slovenian configuration
• switch to the AL12 version
• writing or adapting scripts
• post processing and visualization
• quasi operational run
• defining Croatian domain: number of points 108x90 (SE corner 41.86, 22.16; NE corner 49.55, 15.00), 31 vertical levels, resolution 11 km, time step 450 sec
• operational cycle AL12T2 (since September)
• visualization using Vis5d

We would like to thank our Slovenian colleagues Neva Pristov and Jure Jerman for help and support.

6. Operational ALADIN-FRANCE in Météo-France

The last upgrade performed on SYNERGIE server has been done last year. It is now possible to look at ALADIN-France test suite as well as the ARPEGE one. This tool is helpful to validate the behavior of any new scheme, in particular for all the features not controlled by classical scores. For instance, the present test suite is about modifications on CYCORA physics for both ARPEGE and ALADIN, and a slight increase of horizontal resolution for ALADIN-France (9.9 to 9.5km in grid point). The impact on surface temperature of the new resolution is quite visible on this pictures (figure 1), directly printed from SYNERGIE. The Belle-Ile temperature is much better now. The problem here was similar to that encountered on Lake Balaton (paper of S. Kertesz in the last Newsletter). The island was ignored by the model but present on the finer post-processing grid. The increase in the model resolution solved the problem here.
 

ALADIN-France e-suite on SYNERGIE display

Figure 1: ALADIN-France 2m temperature forecasts, operational (left) and test (right) suite as seen with SYNERGIE software.
 
 

7. Workstation version at French Meteorological Service

Last news in Newsletter18.

8. Operational version at Hungarian Meteorological Service

During the period several tools were adapted and installed in Budapest, which on the one hand helps on the development work and on the other hand extends the available products based on the ALADIN model. The following tools were adapted (the work had been started during the stay of Neva Pristov in Budapest):

• Adaptation of e923 in Budapest in order to be able to create climatological files locally for any domain over Europe. This facility gives more flexibility while creating any kind of new ALADIN domain.
• Adaptation of ascs software in order to visualize vertical cross sections over the ALADIN domain. The software was also extended using the products of the ECMWF model. Ascs was put into operations during the summer.
• Further dynamical adaptation (dada) of the wind field over Hungary. A domain with high resolution (around 2,5 km mesh) was defined covering Hungary and the forecasted wind fields were adapted to the high resolution climatological data base. Dada became operational at the end of summer.

During the last quarter of 2000 the following main activities were performed around the workstation version of ALADIN:

• Introduction of new post-processed variables in the operational suite. The most interesting new variables in use are potential vorticity, total precipitable water and short wave radiation on the surface.
• During the stay of Ryad El Khatib in Budapest some tests were carried out on the formulation of the post-processed wind gusts. According to some case studies it was found that the turbulent wind gust seems a better option than the PBL gust formulation.
• At the end of the year the ARPEGE/ALADIN source code browser was installed.
• A one week visit of Metod Kozelj was devoted to the demo installation of SMS (Supervision Monitor Scheduler) in Budapest. Some operational scripts were considered and modified in order to demonstrate the applicability of SMS.
• The adaptation of the Kalman filter software package (originally written by Jozef Vivoda) was continued but not yet finished.

9. Operational ALADIN-LACE in CHMI

1. Evolution of the ALADIN/LACE application.

The new operational suite of ALADIN/LACE based on SMS (Scheduler-Monitor-Supervisor tool ©ECMWF) started on:

• 22/08/2000 for 12 UTC network time: afternoon suite starts to run under SMS
• 25/08/2000 for 00 UTC network time: complete switch to SMS

After quite a complex testing, the technical organization of the operational suite was completely changed. The suite is now handled by SMS running on a SUN server. The advantages of this technical change are the following:

• better supervision of individual suite tasks by the system and also by the operator, having a comprehensive suite diagram on the screen at any instant;
• easy operator's intervention (like reload of individual tasks) in case of problem;
• optimisation of the long distance telecommunication I/O operations, when both download and upload of data via EQUANT network is moved from NEC/SX4 to a more appropriate SUN server;
• easy incorporation of any new element of the suite (blending-suite, data assimilation type of suite, more frequent runs, etc.).

The operational switch went smoothly. The blending modules are ready by now and may be activated whenever needed.

2. Parallel Suites & Code Maintenance

The Prague Team launched the following parallel tests to assess the impact of different modifications:

• Blending suite, running in its "assimilation cycle" mode (without regular production runs). A modification was ported to the surface part of blending, allowing to blend just land points. This modification was necessary to avoid even stronger drift of surface lake temperature than it is witnessed in the e927/ee927 procedures. Files from the blending assimilation cycle are used both for specific validation tests and for R & D purposes. The standard scores of the blending suite production brought improvements of bias in the initial state conditions and also weak improvements of RMSE scores. On the other hand there is yet an unexplained worse bias of mean-sea level pressure for the network time at 12h. See, please, ALATNET Newsletter for further details on blending. As a whole, blending suite was judged to have either neutral or slightly positive impact. The scores are not presented on the web pages (the production parallel test was launched manually) but they are available on request.
• Linear grid suite "aba". It consists in using a linear grid type of spectral truncation for all variables but orography. The orography remains the same like it is when using the quadratic grid, in order to avoid 2Dx waves to get excited because of the orographic forcing. This suite reflects recent results brought by the quasi-academic experiment SCANIA. The results of "aba" suite are slightly positive. However the introduction of this modification was reported to a later stage when a more efficient data packing would take place in order not to charge the telecommunication lines and archive by voluminous history files (foreseen library AL12_op6).
• AL12_op3 library suite validation. It was intended to prepare this library for an operational switch in order to profit from new products: CAPE, moisture convergence, height of PBL and wind gusts. Unfortunately, the problems of archiving server in CHMI disabled to finish correctly the evaluation of this parallel test as well as a new trial to test ALADIN/LACE 37 levels coupled with ARPEGE 41 levels parallel test. Due to a complete collapse of the archiving server the Prague Team was forced to restrict parallel tests and to devote the remaining free space on various computers to the R&D tasks of Prague Team visitors.

  
  The situation with archive server has been treated together by CHMI and by the provider. It was an interesting lesson: when a bug appears in ARPEGE/ALADIN it is quite quickly fixed, unlike in case of some software companies which are not able within two years to provide a stable version of the software they have charged money for. The new archiving server is now under testing.
  Two new parallel tests were started on December 1st. Meanwhile the parallel test plan was redefined, mainly due to the evolution of ARPEGE parallel test.

• AL12_op6 library and CYCORA-bis tests. Since ARPEGE parallel test library evolved to incorporate CYCORA-bis modifications and more sophisticated diagnostics of wind gusts, Prague Team decided to abandon the level of op3 library and to switch to the op6 library directly. Future operational library of ALADIN/LACE is then AL12_op6. On the same library the new package of physics shall be tested (CYCORA bis). However, for the moment we stick to 31 levels taking a lesson from a few bad surprises of ARPEGE 41 levels parallel test. Hence the modified physical package is to be tested first and to enter the operational suite as soon as possible (pending the scores, of course) since it is supposed that the improved physics will behave better in winter situations, namely it should not make such an erosion of inversion layers.

The results of parallel tests may be consulted on www.chmi.cz/meteo/ov/lace/aladin_lace/partests/ pages.

3. ARPEGE back-up special event.

Since last year autumn Prague center has got a task to ensure the initial and lateral boundary conditions for all ALADIN applications in case of some major accident in Toulouse. This task resumes to receive a pre-processed observation data file from ECMWF, to run ARPEGE 3DVAR assimilation cycle (4 long cut-off times), to make ARPEGE production forecast up to 60 hours (due to 12h lagged coupling in some ALADIN applications) based on the short cut-off analysis (twice a day) and finally to compute all the LBC files for all ALADIN domains not forgetting to run the LACE suite behind all this. It can be easily understood that all the above mentioned pieces needed a lot of preparation, starting from small details such as verifying all national e-927 namelists and climatological files and ending with more tricky problems on how to properly vectorise and multitask the Canari configuration using the Japanese way of tasks numbering. Though the Prague Team has been more or less continuously working on the issue, November strike-menacing events in Lautrec came something like one week too early (or too late) to get combined with a long weekend (of course!) due to the celebration of the victorious fight for democracy in the country of Vaclav, following thus Murphy's law ""à la lettre"". Despite the enormous efforts the Prague Team managed to run safely under SMS just the long cut-off ARPEGE assimilation cycle before the above mentioned long weekend and happily enough the real intervention of Prague was not needed before the weekend got over. The remaining suite pieces were fixed the following Monday and the vitality of ARPEGE back-up was proved (in terms of still useful timeliness of ALADIN forecasts as also confirmed by SELAM colleagues). The event, ended by corks flying in RC LACE chiefs' office, was quite an useful exercise for the Prague Team to lively experience the pressure of having to handle an operational suite of data assimilation (yet just technically and without a real worry about the observation pre-processing). To say the truth, the ARPEGE back-up suite is never going to be a story of a simple "pressing of the button", for example the supply by the observations from the whole of globe is still quite a vulnerable point.

10. Operational ALADIN-MAROC in MAROC-Météo

Since July 2000, Aladin/Morocco is running Canari analysis part using AL11T2 (CY21T2) version. The forecast part has been performed with AL11T2 since February 2000.

This new Upgrade was possible after the tuning of CANARI structure functions for both altitude and surface parameters in the assimilation mode.

The validation has been done on a 10 days parallel suite; and the results obtained with this new version affected especially the surface fields analysis. Altitudes fields still have the same behaviour. A report of this work has been written and is present at GMAP.

So, for the time being, Aladin/Morocco is operational with AL11T2 in both assimilation and forecast modes.

The parallel version of Aladin has been tested on CRAY J90 plate-forme using either PVM and MPICH ; this tests were successful with use of PVM, but the mode LMESSP=.TRUE and NPROCA=4 is very expensive in terme of memory and CPU time compared to the operational mode LMLTSK=.TRUE. and NTASKS=4.

Finally, MAROC-METEO decided last November to buy a new computation machine, an IBM Super Scalar POWER III machine, a shared memory one with 48 GFLOPS of power (32 procs).

Now we plan to run Aladin/Morocco with less than 10km resolution for 60 hours forecast range twice a day. All our futur efforts will then focus on implementing a 3DVAR assimilation suite.

11. Operational version at Polish Meteorological Service

In the second half of 2000 most efforts made by Cracow Aladin team focused on shifting the operational work from IBM RS/6000 to SGI 2800 machines. New, much more enhanced model environment was prepared to reach the aim. Here are some details of current operational suite:

• machine: SGI~2800 ( 128 CPU's, 40 GB RAM, 190 GB HDD )
• resources devoted to operational runs: 8 CPU's, 2 GB RAM, 4 GB HDD
• model version: ALADIN AL09.3
• operational modes: ee927, e001, Full-Pos
• mesh size: 180x180 ( 169x169 without E zone )
• number of levels: 31
• spatial resolution: 13.5 km
• domain: 2270 km x 2270 km
• time step: 450 sec.
• forecast range: 48 hours
• coupling: Arpege, every 6 hours
• initialisation: DFI
• post-processing: Full-Pos on-line
• frequency: twice per day

Some other efforts focused on the creation of preliminary versions of two new systems. First one is designed for NWP data storage and access: operational database was built on a basis of RDBMS PostgreSQL and datastored in it are made available via HTML/PHP3 pages. The second system is devoted for qualified forecast preparation support and it is also based on WWW technology: forecaster can get numerical prediction data for given locations placed in interactive form, data can be modified and then preparation of some forecast products can be initialized.

12. Operational version at the Portuguese Meteorological Service

The e923 configuration was implemented on workstation with operational ALADIN cycle (AL11T2) following N. Pristov documentation. We are now testing the two cost functions - Jercynzsky (LNEWORO2 = T) and Bouteloup (LNEWORO = T ) - in order to verify which one represents better the orography of ALADIN/Portugal domain.

In order to run OI/CANARI with local observations some steps have already be done:

• extraction of TEMP/TEMPSHIP and SYNOP/SHIP observations from our national database;
• installation of the code to produce the ASCII files used by MANDALAY to create the CMAFOC file, which was provided by our hungarian colleagues;
• implementation of MANDALAY itself.

13. Operational version at the Romanian Meteorological Service)

Last news in Newsletter18.

14. Operationa version at Slovak Meteorological Service

• computer:

DEC Alpha Xp1000 (Compaq), EV6 processor DIGITAL UNIX V4.OD
Digital F90 compiler 640 MB memory
12 GB HDD

• model characteristics:

domain size 826 x 646 km (120 x 90 points)
corners 14.69, 46.05 SW 25.26, 51.07 NE
horizontal resolution 7.18 km
vertical resolution 31 levels
time step 337.5 s
forecast length 48 hours / twice per day (00 UTC & 12 UTC)
coupling model ALADIN/LACE (12.2 km horizontal resolution)
mode dynamical adaptation
dynamics and physics the same as ALADIN/LACE

• configurations:

ee927 creation of initial and lateral boundary conditions
e001 model forecast
postprocessing on model levels & standard isobaric levels
e701 CANARI (optimal interpolation) recently ported

• products:

surface parameters in map form (2m temperature, 2m Tmax, 2m Tmin,
2m relative humidity, l0 m wind, precipitation, cloudiness)
meteograms for 22 cities over Slovakia
input for vertical cross-sections
inputs for the Nuclear Safety Authority (RODOS program)
generation of the automatic forecasts

• visualisation:

for central forecasting office, Internet and Intranet
using NCAR graphics, GMT software, HTML

• verification:

surface parameters of ALADIN/LACE and ALADIN/SLOK forecasts verified against SYNOP observations over the territory of SLOVAKIA
pseudoTEMP messages verified against TEMP observations
24 total precipitation verified against the local rainguage network

Main news

• update of the ALADIN/SLOK domain

  
  - size of the domain increased (90x64 = 120x90)
  - horizontal and vertical resolution kept

• automatic forecast generation
• on-line verification

  
  - 2m temperature, l0m wind speed and wind direction
  - last 2 runs of ALADIN/LACE and ALADIN/SLOK compared to actual SYNOP observations

• precipitation verification (diploma thesis)

  
  - comparison of ALADIN/LACE and ALADIN/SLOK against our rainguage network observations (600 stations) - data interpolated into common grid
  - 24h total amounts used
  - contingency tables and continuous scores calculated - evaluation monthly with 3 months delay
  - software ready to be used routinely at SHMI

• update for AL11T2 + CYCORA setting

  
  - comparison of pseudoTEMP forecast
  - with CYCORA: warmer and drier planetary boundary layer, increased wind speed

• RODOS

  
  - inputs from ALADIN/LACE and ALADIN/SLOK for the RODOS program (Nuclear Safety Authority)

15. Operational version at Slovenian Meteorological Service

Last news in Newsletter16 .

Home