1. In Austria
MAP orographic precipitation simulations
Within the framework of the bilateral Amadeus project, F. Bouyssel, E. Bazile, and Y. Wang performed additional sensitivity experiments with AL12 on the MAP cases previously studied (see Newsletter 20). Setting horizontal diffusion of water vapour to zero showed little effect, as did switching off digital filter initialization. An academic experiment where a certain supersaturation was allowed to occur (to mimic the cloud water phase and its advection) produced a smoother precipitation field, with smaller amounts, and a downstream shift. It generally brought the simulated field in closer agreement with observations. Unexpected was the large sensitivity of simulated precipitation to the length of the time-step. Changing from 600 to 450 and 300 sec (at Dx=10 km) gave a significant increase, coming mainly from the resolved (large-scale) part of precipitation. [Further details: Yong Wang ]
Aircraft icing prediction
Experimental daily forecast charts of aircraft icing potential are produced based on ALADIN prognostic fields. The charts can be viewed on-line at :
http://mailbox.univie.ac.at/reinhard.stepanek/icing/icing.html.
The computation of icing potential is based on an empirical formulation developed by the HIRLAM group, with cloud water content, temperature, and vertical velocity as input. It was found that due to the higher resolution of ALADIN-VIENNA compared to the Swedish HIRLAM model, the weight given to vertical velocity in the formulation has to be reduced. Pilot reports from the Austrian civil and armed forces aviation are being collected and studied for verification purposes. [Further details: Reinhard Stepanek ]
Prognostic convection scheme
The prognostic convection scheme developed by Gérard (2001) for AL12/CYCORA-bis has been implemented at ZAMG. Experiments on the 10-12 Aug 2000 mountain convection case are being carried out. The goal is to see whether the delayed onset of convection due to prognostic updraft strength and updraft area fraction brings the forecast in closer agreement with radar observations. [Further details: Stefan Greilberger]
Operational model verification
Starting 2002, a bi-annual operational verification report is being issued by ZAMG. The first report contains a comparison of ALADIN-VIENNA and ECMWF point forecast errors of surface variables for the station Vienna (Hohe Warte) over the last 2.5 years. The figure shows, as an example, results for 10m wind direction and speed. The improvement compared to the ECMWF model is most apparent in wind direction. Note that for wind speed the ALADIN forecast, in spite of a somewhat larger negative bias, manages to have smaller mean absolute error than the global model. [Further details: Klaus Stadlbacher ]
References :
Gérard, L., 2001: Physical parameterizations for a high resolution operational numerical weather prediction model. University of Brussels, Ph.D. Thesis.
2. In Belgium
Porting to the new computer (Olivier Latinne)
Three months of work were necessary to finalize the migration of the operational ALADIN suite on SGI Origin 3400. Because scripts of the old operational suite were written in standard Kornshell, only some minor modifications were brought, those carried primarily on some parameters of different environment, as well as the suppression of necessity to convert some data files, from "CRAY" to "IEEE 64 bit" formats. The parameters of compilation and execution of ALADIN were optimized, allowing a 48 hour ALADIN run (with 41 levels) to be executed on 16 processors in less than 20 minutes. This number of processors allows a good compromise between the speed of execution of ALADIN and the degradation of performances connected to the increase in the number of processors used. Two comparative tests with the operational chain running on CRAY were successfully realized, for September 19 and 26, 2001. The date of September 19 corresponds of a cyclonic situation, therefore more sensitive to the numerical differences (singularity point corresponding to the cyclonic centre). For the whole of the computed fields only small differences (for instance about 0.1 m/s for the speed of the wind, 1 degree for direction of the wind, 0.1 K for two meters temperature, 1% for humidity) were observed. The purely numerical differences are however probably much smaller, considering that the ALADIN executables on SGI and CRAY used for these tests correspond to two different cycles (and physics). We have however faced many problems during the validation tests, linked to the commercial software of queuing management PBS-PRO. These problems have been corrected by SGI and seem now entirely solved.
Pollution forecasts (Piet Termonia)
This corresponds to the completion of a study demanded by the Brussels Institute for Environmental Management (BIM/IBGE), with the aim to find out whether extreme peaks of pollution in the Brussels region can be predicted using ALADIN. BIM/IBGE provided data of concentrations of the pollutants NO, NO2, PM10, O3. These concentrations where compared to the ALADIN forecasts of a number of meteorological parameters. Two parameters were retained for an operational use: the presence of inversions and a kind of horizontal transport index (wind velocity divided by the Brunt-Vaisala frequency). The results of this research will most probably be applied by the three collaborating institutes RMI, BIM/IBGE, CELINE/IRCEL and the Brussels Ministery of Environment.
3. In Bulgaria
During the stay in Toulouse some experiments were performed over a subdomain of ALADIN-BG with a step of 3 km. The correspondant climatological files were created, and two successful 12 hours runs of the model were realized.
4. In Croatia
See the joint paper.
5. In Czech Republic
See the RC-LACE report.
6. In Hungary
Most of the ALADIN related activities at the Hungarian Meteorological Service were concentrated on ALATNET topics, therefore more details can be found in the ALATNET report of HMS.
At the end of 2001 efforts were done in order to finish the public procurement of the high performance computing facility for the Hungarian Meteorological Service. The IBM company was selected and a 32 processors Regatta server was chosen as future computer of the Service for numerical weather prediction. More details about the machine itself and the migration of the ALADIN code will be given in the next Newsletter.
Roger Randriamampianina started the feasibility study for an implementation of ATOVS data in the ALADIN model. TOVS data were introduced in the ALADIN model (AL11) in summer 1999 (Sadiki and Soci, 1999), but at this stage only the screening was possible. As it was proved (Randriamampianina and Rabier, 2001), the locally received and pre-processed ATOVS radiances have positive impact on subsequent ARPEGE model forecasts and analyses. Moreover, the impact increases by increasing the resolution of the radiances. Based on these results the implementation of the ATOVS radiances in the limited area model ALADIN has special interest. We started this work in autumn 2001. The operational cycle at the Hungarian Meteorological Service is AL12. AL13 is also installed but for research and development only. We apply the one dimensional variational 1d-var assimilation scheme to the pre-processed radiances to obtain quality control flags, vertical profiles of temperature, humidity and surface temperature. We found out, however, that no radiances were injected into the 1d-var. Thus, screening of TOVS/ATOVS radiances is not possible with AL13. The next operational cycle is going to be AL15. Since AL15 with ODB can handle ATOVS radiances (screening) (Kertesz and Fischer, 2001), further implementation of ATOVS radiances should be done with AL15.
References :
Kertesz S. and Fischer C., 2001: Observation management for ALADIN. Internal report, available from Météo-France or HMS.
Randriamampianina R. and Rabier F., 2001: Use of locally received ATOVS radiances in regional NWP, SAF NWP report, available at Météo-France, HMS or U.K. Met. Office.
Wafaa S. and Soci C., 1999: The observation screening in ALADIN 3D-Var data assimilation system. Internal report, available from Météo-France.
Regarding other topics of interest the following activities can be mentioned:
-- A special interface was written in order to provide forecasted radiation components of the model. The forecasted data are compared to the real observations and the throughout analysis also concluded some properties of the radiation parameterisation scheme of the model (article to be found also in this Newsletter).
-- The work on the application of ALADIN precipitation forecasts for hydrological purposes has been continued. The obtained results provide verification of the precipitation fields of the ALADIN model (a separate paper on that topic is planned for the next Newsletter).
7. In Moldova
Nothing new.
8. In Morocco
Most of the effort of the ALADIN team was devoted to the implementation of the operational suite and of a 3d-var assimilation suite on the new computer. See the report on operations for more details.
9. In Poland
Yet another dark side of NWP operational work
In the end of 2001 the Cracow ALADIN Group network was attacked by hackers. Three machines were compromised including our group server. In two cases it was possible to reveal most of elements of the intrusion anatomy. The likely way to hack the machines was exploitation of bugs in versions of ssh or wu-ftpd installed on our computers - in the end of the year there were exploits widely available in the Internet. After gaining root privileges hackers installed rootkits to hide their presence in the systems and to prepare easy further access to the machines. In one case "bobkit" rootkit was used and in another one a clone of "t0rn v8". Both rootkits provide hacker with replacements of some commands (among them ls, find, du, ps, pstree, top, ifconfig, netstat), Trojan versions of other (ssh, login, ...) also with backdoors, log cleaner, sniffer and some additional libraries.
The presence of the intruders was discovered due to some suspicious behaviour of the systems (e.g. some logs were incomplete). Quick search through catalogues with mc showed presence of strange names as "..." or ".. " (of course not seen with replaced ls ) and proved that somebody installed his malicious stuff. The scanning of tripwire logs and usage of installed rpms database revealed more. In latter stage of search we also used chkrootkit tool. Analysis of hacker's software made possible to prepare a little trap. Knowing number of port which was being used by hacker backdoor we put it into ports entry configuration file and we waited what would happen. After several days the hacker tried to go back to our server. Firstly he tried to use Trojan ssh and he failed, next he tried to use installed backdoor and the trial was automatically blocked. In logs we found out information which made possible to establish that a computer of an institution in Seoul was used to perform the attack.
As a consequence of intrusions two systems had to be installed from scratch and another one was refreshed. We were forced to build a firewall to make protection of our network more reliable and take some additional security measures which increase overall cost of our network administration. Actually we test LIDS package, which is likely to be installed on group server in near future.
10. In Portugal
At 30th of July 2001 (12 UTC run), after a parallel suite during one month and a half and an objective verification for the same period, we introduced AL12_bf02 version (with CYCORA-bis included) in our operational suite.
An objective verification for six months (24th of January - 24th of July), using AL11T2_03 and AL12_bf02, showed that the switching on of CYCORA-bis physical package in ALADIN-Portugal had a nearly neutral impact on the forecasts.
On the analysis/assimilation field the CANARI (AL12) / DIAGPACK is being installed.
The developed diagnostic tools, divergence of Q vector, low-level moisture convergence (1000-850hPa), K-index, Modified K-index, Total-Totals index, Modified Total-Totals index and SWEAT index, were tested on two severe weather events of deep convection, over Portugal. One associated with a cut-off low and the other associated with a strongly precipitating cold frontal system.
A major effort was also devoted to the organization of the 11th ALADIN workshop.
11. In Romania
Simulation of convective systems at different resolutions (Doina Banciu)
The experiments on the simulation of the Cleopatra have been continued by testing the mixed solution (included in CYCORA-ter) for the resolution dependency of the convective scheme: a combination of the explicit dependency method used before CYCORA package and the limitation of the available humidity convergence for the convective scheme by substracting the large scale precipitation, method used in CYCORA packages. The results have shown that the problems encountered in the evolution of the convective and stratiform precipitation when increasing resolution have been cured by using the mixed solution (more details can be found in "Precipitation: resolution and orography" by D. Banciu, S. Alexandru, E. Bazile, L. Gérard and K. Stadlbacher, to appear in the EWGLAM Newsletter, no. 31/2001)
Comparison between CYCORA and CYCORA-bis packages (Simona Stefanescu)
The test carried out before the implementation of CYCORA-bis package in operational have not shown dramatic changes in comparison with the simulation using CYCORA (operationally for the test period in September). The areas covered by precipitation, the speed and the partition between convective and stratiform precipitation are quite similar. However some unrealistic areas of precipitation disappeared by using CYCORA-bis package. Also few differences in the maximum of precipitation amount were noticed.
Implementation of the Aladin model on a SUN platform (Doina Banciu)
The ALADIN Model (version 12) have been implemented on a SUN E4500 server having 8 UltraSparc processor of 400 MHz, with the generous help of Gabor Radnoti. To be able to run the model on this platform it was necessary to re-configure the system for 64 bits environment and to change the compiler Forte 6 HPC update 1 to Forte 6 HPC update 2. Even the SUN platform performance is far from our needs, it is an important improving regarding our old platform DEC and it will allow us to enlarge our operational domain and to integrate the next Aladin versions.
Aladin Meteograms (Steluta Alexandru and Simona Stefanescu)
Meteograms for the main 10 cities over Romania are available on the internal web, for forecaster use, twice per day, till 48 h range. The surface and upper air fields represented are: mean sea level pressure, 2 m temperature, 10 m wind, total precipitation, total and low cloudiness, temperature and wind for 850, 700 and 500 hPa levels.
An example of such meteogram is presented for Oradea.
Experiments on the background error statistics for the ALADIN 3d-var data assimilation (Simona Stefanescu under the supervision of Loïk Berre)
The experiments are based on the work of Mohamed Raouindi and Loïk Berre, which have carried out a study on the spectral representation of the latitudinal variability of the bi-dimensional forecast error covariances for temperature at level 18, derived using the NMC method from the ARPEGE coupling files for the ALADIN/Morocco integration domain over a 87 days winter period.
Continuing their experiments, the longitudinal variability of the temperature forecast error covariances was studied. There were realized pseudo-single-observation experiments including calculation and plotting of the latitudinal and longitudinal variation of the correlation function for different locations of the observation point. The latitudinal and longitudinal variation of standard deviation and length scale of the correlation function was investigated. Different evolutions for the latitudinal and longitudinal variability have been observed. The correlation length in "x" direction decreases when increasing the latitude and it can be seen an anisotropy of the correlation function in "y" direction at the northern part of the domain explained by the limits between two regions with very different features (tropics and mid-latitudes). The longitudinal variability study revealed a decrease of the correlation length in "y" direction when decreasing the longitude. In "x" direction, the anisotropy of correlation function is not so obvious like for "y" direction. This will be deepened by the calculation of the spatial variations of the length scales of the covariance functions.
12. In Slovakia
See the papers by Jan Masek and Martin Bellus.
13. In Slovenia
The main developments are described in the section about operations and in a paper by Mateja Irsic and Neva Pristov.
14. In Tunisia
Most of the effort was devoted to the setup of the environment for ALADIN-Tunisia (visualization for forecasters, verification, benchmark and administrative tasks to buy a dedicated computer, ...). See the corresponding paper.
 |
Home |