1. In Toulouse (France)
The work of the two ALATNET PhD students in Toulouse, Gianpaolo Balsamo and Cornel Soci, is described in separate reports. The summary hereafter corresponds to the joint efforts of the other visitors and the permanent staff.
1. Theoretical aspects of non-hydrostatism (NH)
Significant progress was obtained concerning the stability of semi-implicit semi-Lagrangian advection schemes in NH dynamics. An overview of the latest results from the Toulouse (T) and Prague (P) teams is available in the proceedings of the 10th ALADIN workshop : "Progresses in NH numerics", by Pierre Bénard (T) , Radmila Brozkova (T,P), Jozef Vivoda (P) , Christopher Smith (P) , Petra Smolikova (T) , and Jan Masek (T) .
4. Removal of the thin layer hypothesis
The current state of research is described in the proceedings of the 10th ALADIN workshop : "Relaxation of thin layer hypothesis in the model ARPEGE / IFS : application to Eulerian scheme", by Karim Yessad. Abdelwaheb Nmiri began to examine the specific ALADIN features during its stay.
5. Coupling and high resolution modes & 6. Specific coupling problems (6b: Tendency coupling for surface pressure and other technical variations around Davies' technique of field coupling in a buffer zone)
Jean-Marc Audoin has just joined the small team working on coupling problems : Gábor Radnóti, Tamás Szabó and Piet Termonia. In consequence the main effort during this six months consisted in discussions with Piet Termonia and Gábor Radnóti during their stays in Toulouse, and collecting/reading documentation.
6. Specific coupling problems (6a: Blending of fields in data assimilation for preserving high resolution forecast details)
The "blending" test-suite for ALADIN-France was refined (Olivier Pédoussaut, Adam Dziedzic, Claude Fischer) and some sensitivity experiments performed, as described in he proceedings of the 10th ALADIN workshop : "DFI-Blending dans le modèle ALADIN/France, étude d'impact des filtres digitaux", by Adam Dziedzic. A detailed documentation was written by Dominique Giard, starting from the sparse previous reports, and is available via the ALATNET and ALADIN web sites.
8. Adaptation of physics to higher resolution (8b: Test, retuning and improvement of the various physical parameterisation in the framework of a very high resolution)
Till the first "CYCORA" package (autumn 1999), the problem of the horizontal resolution dependency of the convection scheme, i.e. the partition between resolved and unresolved precipitations, was partially cured, following the results of Jean-Marcel Piriou (1991), by introducing a modulation of the dynamical part of the humidity convergence by a factor depending on the model resolution. Afterwards the stratiform or resolved precipitations were computed first and then the humidity convergence available for convection was reduced correspondingly, in order to avoid a double count. However recent studies, mainly:
showed severe inconsistencies. "Stratiform/resolved" precipitations decreased and "convective/ unresolved" ones increased as the mesh-size was reduced.
A mixed solution with a lowering of the transition scale from 17 to 10km was tested on a wide set of situations and will be part of the "CYCORA-ter" package, to be tested in a parallel suite next autumn. Let us recall that CYCORA is an alias for CYclogenis + COnvection + RAdiation.
8. Adaptation of physics to higher resolution (8c: Improved representation of boundary layer)
André Simon, with some help from Jean-François Geleyn and Jean Pailleux, analysed the reasons of the exaggerated cyclonic activity in ARPEGE (and ALADIN, indirectly) mainly over sea and at fine scales. The study rapidly evolved towards considerations of increased mixing length for the turbulent vertical diffusion parameterisation scheme. A lot of different solutions were tested, including retuning of the most crucial parameter of the CYCORA-bis package (Rid, related to the modified formulation of the Richardson number), in order not to lose benefits of past work on the 1998 and 1999 Christmas' storms. The proposal, put in a temporary parallel suite, was excellent for the surface part of the problem, with a better structure of pressure lows. But its too simple character (proportional increase of mixing length everywhere) led to strong degradations at the tropopause level and around, where fields were over-smoothed. He addressed afterwards the (sometimes really old) problems of consistency in the definition of the mixing lengths for heat and momentum (computation and respective usage). This study was pursued by Jean-Marcel Piriou, Eric Bazile and Jean-François Geleyn. After intensive validations, a new set of modifications, with different tunings for heat and momentum, is ready to enter "CYCORA-ter".
8. Adaptation of physics to higher resolution (8d: Improved representation of orographic effects)
Steluta Alexandru designed new formulations for the two cost functions used in the computation of spectral orography for ALADIN. The fraction of sea is now taken into account, not only the height, so as to better preserve contrasts along coasts. A summary of her work is available in the proceedings of the 10th ALADIN workshop : "Tuning the orography representation in the E923 configuration", by Steluta Alexandru. This work also gave the opportunity to correct a very old bug, as old as E923 : the height values in the extension zone where taken into account in the cost function, though meaningless. The two formulations were also harmonized, to make further tuning easier.
9. Design of new physical parameterisations (9b: Use of liquid water and ice as prognostic variables, implementation of a new microphysics parameterisation)
A modular approach for the parameterization of moist processes, the so-called "Functional Boxes", is under study for some years. It must :
A first version is now available and was tested in the 1d version of ARPEGE/ALADIN, using GATE observations. Results are promising. This work was presented at the 10th ALADIN workshop and the corresponding paper is available on the ALATNET and ALADIN web sites : " "Functional Boxes" approach for moist processes ", by Eric Bazile.
9. Design of new physical parameterisations (9c: New parameterisation of exchanges at sea and lake surface)
Taking advantage of the short stay of Mihaela Caian in Toulouse, a closer collaboration between the French and Romanian teams was organized and a new experimental framework designed to evaluate the impact of a parameterization of lakes.
9. Design of new physical parameterisations (9d: Improved representation of land surface, including the impact of vegetation and snow)
Discussions with Mihaela Caian also led to new proposals for a simple description of the impact of urbanisation within the current land surface scheme. However this cannot be tested before the new, high resolution, land-cover map is fully available. And significant modifications were brought to this database till June, delaying its adaptation to the ARPEGE/ALADIN framework.
The work on snow cover progressed in two directions. First Mohamed El Haiti and Eric Bazile refined the previous modifications, i.e. introducing the albedo of snow as a new prognostic variable and taking into account the vegetation (coverage, leaf area index, albedo) in the formulation of the snow coverage and the global surface albedo. The age of snow, estimated from the albedo, is now considered in the definition of the snow fraction, and a saturation of the dependency on the leaf area index is introduced. However some more retuning, e.g. of the formulation of thermal inertia, and more extensive sensitivity experiments are required. Second a simplified version of this new scheme contributed to the international SNOWMIP experiment, for the intercomparison of snow-cover parameterizations, and proved as good as several more sophisticated schemes.
9. Design of new physical parameterisations (9e: Refinements in the parameterisations of radiation and cloudiness)
Janko Merse tried to solve the problem of the underestimation of cloudiness below very stable layers using the data produced by a cloud-resolving model (CRM) as the basis for validation. This study is described in : "Relations between water particles content, vertical stability and cloudiness in cloud resolving models", published in the proceedings of the 10th ALADIN workshop. But it failed because of a conceptual problem in the CRM used here.
10. Use of new observations (10a: Yet unused SYNOP observations) : see the ALADIN report
10. Use of new observations (10b, 10d : satellite observations)
As a preliminary to the use of new satellite data at high resolution over continental domains, Malgorzata Szczech began to implement a more detailed description of surface emissivity in the model, with dependency on vegetation, season, and wavelength, and consistent with the model land cover. A more detailed paper is available in the proceedings of the 10th ALADIN workshop : "Improved description of emissivity over land".
Roger Randriamampianina got an intensive training on the treatment and the use of satellite data, analysing the impact of local ATOVS observations, available sooner and at higher resolution over France, on short-range forecasts over Europe.
Thanks to Philippe Caille, new types of observations are now routinely pre-processed and available to modellers. The impact of some of them was studied along the last months, but in a global framework. A summary is available on the ALATNET and ALADIN web sites : "Use of ATOVS and SSMI observations at Météo-France", by Florence Rabier, Elisabeth Gérard, Zahra Sahlaoui, Mohamed Dahoui and Roger Randriamampianina.
11. 3D-Var analysis and variational applications (11a: Definition and calculation of new background error statistics, impact of domain resolution and extension, identification of horizontal relevant scales)
Mohamed Raouindi examined, with Loïk Berre, the basic homogeneity hypothesis used in the computation of background error statistics for ALADIN up to now, so as to keep only diagonal terms. He exhibited a latitudinal dependency of spatial covariances for ALADIN-Maroc : thinner vertical structures and wider horizontal ones as the latitude decreases. Sensitivity experiments indicated the possibility to select some important couples of meridional wavenumbers to be related to each other (using a block-diagonal covariance matrix), in order to represent the identified latitudinal variability in an economical way.
11. 3D-Var analysis and variational applications (11b: Scientific investigation of the problem of the extension and coupling zone, analysis of the impact of initialization)
A detailed study of the respective impacts of 3d-var analysis and digital filter initialization on spectral fields in ALADIN was performed by Simona Stefanescu. She compared large-scale versus mesoscale background error statistics for 3d-var, with varying weights for the differents parts of the cost-function, and incremental versus full versus no initialization. Her main conclusions are : initialization should be used incrementally, its impact is noticeable at the smallest scales and sensitive to the weight of vorticity in the background cost function, and coupling is a significant source of noise.
11. 3D-Var analysis and variational applications (11c: Management of observations in 3d-var)
Wafaa Sadiki and Sandor Kertesz studied how to adapt the operational selection of observations to the case of a high-resolution limited-area model.
11. 3D-Var analysis and variational applications (11f: Development of variational type applications (adjoint methods for sensitivity studies, singular vectors), as a project itself and to provide more insight into the coupling problem for 4d-Var)
The configuration 601, adapted to ALADIN and carefully updated for several years by Claude Fischer, produced its first results last spring under the hands of two French students, Vincent Guidard and Pierre-Luc Payet. It uses a Lanczos algorithm to produce the first singular vectors of the model, i.e. the most unstable perturbations of the initial conditions, for a given time-span and a given situation. It was run for large ALADIN domains at a low resolution (30 km) and at the full ALADIN resolution (10km) for small targeted areas. The identified structures look quite sensible at first. Some examples may be found in the devoted paper : "Singular vectors in ALADIN", by Claude Fischer, Vincent Guidard and Pierre-Luc Payet, in the proceedings of the 10th ALADIN workshop. More details may be asked to Claude Fischer.
12. 4D-Var assimilation (12d: Improvement of the treatment of humidity in data assimilation)
The problem noticed in the management of humidity by 4d-var assimilation in ARPEGE was identified and temporarily solved. More details may be found in the report from Radi Ajjaji.
2. In Bruxelles (Belgium)
5. Coupling and high resolution modes
A description of the corresponding work is available in the proceedings of the 10th ALADIN workshop : "Tests of some temporal interpolation schemes for the coupling mechanism", by Piet Termonia.
6. Specific coupling problems (6a: Blending of fields in data assimilation for preserving high resolution forecast details)
Alex Deckmyn designed a test blending suite for ALADIN-Belgium
8. Adaptation of physics to higher resolution (8a: Parameterisation of the small-scale features of convection)
A summary of the corresponding work is available in the proceedings of the 10th ALADIN workshop : "Some results using a prognostic convection scheme in ARPEGE-ALADIN", by Luc Gérard.
8. Adaptation of physics to higher resolution (8d: Improved representation of orographic effects)
Philippe Nomerange started analysing the representation of the small-scale orography features in ALADIN. After a bibliographic research phase aimed at learning and understanding the various methods that can be used to represent the small-scale effects of the orography - envelope orography, improved roughness length, ... - a comparison phase between those methods has been started using a test case of ALADIN on a domain covering the Pyrénées.
3. In Prague (Czech Republic)
1. Theoretical aspects of non-hydrostatism
1a) Top and bottom boundary conditions (C. Smith)
When making the experiments with the so-called non-hydrostatic nonlinear regime (in 2d vertical plane), we noticed that in case of the semi-Lagrangian advection treatment there is a small spurious standing wave above the mountain. This spurious pattern does not appear when using the Eulerian scheme. In fact, it is sufficient to have relatively high mountain and some mean flow to reproduce the spurious pattern. A defect of this type occur even in case of the potential flow regime. In order to find the cause, the semi-Lagrangian discretisation was examined especially for the 3d continuity equation (treatment of the 3d divergence) and for the bottom boundary condition. The spurious patterns depends on the length of the time-step, though it is not very clear at this stage whether it has to do something with the length of trajectory or not. At the moment the influence of the individual terms in the equations on the spurious pattern is investigated. There are also a few ideas how to improve the current discretisation of the bottom boundary condition.
1b) Predictor-corrector scheme (J. Vivoda)
The predictor-corrector scheme for the 2TL non-hydrostatic semi-Lagrangian scheme is basically working. However, in presence of a steep orography 3 iterations of the scheme are necessary in order to obtain a stable integration. This is in contradiction with the linear stability analysis results where only one iteration of the scheme would have been fully sufficient. To understand this result, one has to keep in mind that the basic linear stability analysis does not evaluate the effects of purely nonlinear terms. For example, one part of the 3d divergence linked to the coordinate transformation from z to h and denoted as X is not evaluated (the term X is equal to "¶V / ¶j dot grad(j)", hence closely depending on the orography gradient ). There are all reasons to think that the term X is responsible for the instability linked to the orography: an enhanced linear stability analysis taking into account this term by introducing a constant orographic slope over the domain showed the instability (as demonstrated by Pierre Bénard). However results of the enhanced analysis has to be taken with some precautions since the constant slope does not fulfil the periodic condition, for example.
During the successive corrector steps the semi-Lagrangian trajectories can be recomputed to achieve higher accuracy. It was expected that improved accuracy will accelerate the rate of convergence and thus minimise the number of needed iterations. The following methods of trajectories re-computations were tested: i) first and second order accuracy approximation during the predictor step followed using the second order approximation during the corrector steps; ii) the second order approximation was used in the predictor step and the third order approximation was used in the corrector step. It was experimentally found that these trajectory re-computations have not improved the speed of convergence.
2. High resolution runs
2.a) Experimental framework (D. Cemas)
Thanks to the availability of the ee923 configuration (Jure Jerman) the 1km mesh-size domain with sufficiently fine orography was created. Hence, the very first tests of ALADIN on 1km domain are ready to start this summer. These tests are meant to discover the major difficulties to run the model on such a mesh. The domain includes a very sharp orography (the Julian Alps) and will serve for benchmarking ALADIN NH versions as they will evolve.
2.b) Lipschwitz versus CFL criteria (Alena Trojakova, Radmila Brozkova)
The study of the applicability of Semi-Lagrangian schemes in NWP models with respect to the dependency on the increase of spatial resolution of the computational grid” shall be based on the pseudo-academic SCANIA-type” of tests. The model will be launched at finer and finer resolution, going from 10km to 156m horizontally and from 867m (30 levels) to 108m (240 levels) vertically, by keeping the same depth of the atmosphere. The specially prepared fine mesh orography from the Alpine region shall be used (with help of CNRM/GMME). However, the nesting of domains requires a procedure of the EE927-type in order to prepare the initial and lateral boundary conditions for each inner domain. Here the standard Full-Pos approach does not satisfy the specific academic conditions, therefore a simple procedure is currently being developed and validated, using only the core interpolating routines (HORIBLE, APACHE). This academic procedure may be also useful in future for any kind of similar needs.
3. Data assimilation related coupling issues
3.a) The blending technique operational in ALADIN/LACE (F. Vana, M. Siroka, D. Klaric)
Starting from 6th June 2001, the ALADIN/LACE initial conditions are computed using the blending technique instead of a pure dynamical adaptation method. The first operational version of the blending algorithm comprises the treatment of both spectral and surface fields. The tuning was retained as established during the period of the extensive testing: The cut-off spectral truncation represents the mesh of about 30 km, the internal digital filters work with the stop-band edge of 5 hours and there is no external initialization within the long cut-off cycling procedure. Only a weak external digital filter with the stop-band edge of 1 hour and a half is applied in the 48 hour production forecasts. the blending of the surface fields, these are blended only over the land surface (the sea points are left without any change). The cycling procedure is fully in phase with the data assimilation cycle of ARPEGE, at least for what concerns the long cut-off cycle. The production blending and following forecasts are done only twice a day for the time being. Within the spring 2001, the blending algorithm was phased with the operational library AL12_op6 and the source code was cleaned. In May, a bug in the handling of the surface monthly constants was discovered and corrected (see also the Prague Team short report in the ALADIN Newsletter). The last validation tests before the operational application of the blending technique were done at the end of May, beginning of June. The next research and development work using the blending algorithm is scheduled for July 2001, where the incremental digital filter initialization should replace the remaining external initialization in the production forecasts. The combination of the blending technique with the linear grid (higher spectral resolution) will also be a R & D topic in the second half of 2001.
3.b) 3d-var strategy in ALADIN (M. Siroka, G. Boloni)
The results of the family of 3d-var algorithms run last November on LACE domain were examined. To remind, three basic algorithms were tested:
- "STANDARD" using the classical Jb (background term) penalty function and ALADIN +6h forecast as a guess. The same observations are used like in ARPEGE but ATOVS/TOVS, which are not used. There is no Jc (noise control) penalty function in ALADIN.
- "BLENDVAR" using the "mesoscale" Jb penalty function and ALADIN blended state as a guess. Otherwise the rest (observation file and missing Jc penalty function) is like in the STANDARD case. This algorithm is the closest one to the 3d-var multi-incremental approach. The blending step might be considered like the first outer-loop fixing the large-scale part of the spectra consistently with the ARPEGE analysis. The analysis step behind is supposed to analyse something else, hopefully the finer part of spectra. Of course, the long waves might be modified a bit as well but this is true also in the classical multi-incremental approach.
- "VARBLEND" using the "mesoscale" Jb penalty function and ALADIN +6h forecast as a guess. The blending procedure is applied on the result of 3d-var like if the analysis is taken as a guess of the blending. Here the blending step should bring the large scales back to the ARPEGE analysis.
The algorithms were run for couple of days in the assimilation cycles of 6h. Then the results were examined particularly within the forecast window 0-6h in order to see how big are the differences in the first guess of the next analysis step.
The role of the large-scale forcing at the initial time of the forecast
It was noticed that the role of the "0h coupling file" is quite important, especially in the STANDARD case. There were two solutions: i) either to make the lateral boundary and initial conditions identical (the result of the analysis becomes its own lateral boundary forcing at the initial time); ii) or to use the ARPEGE analysis as the lateral boundary condition. Since both BLENDVAR and VARBLEND algorithms fix the long waves according to the ARPEGE analysis, there is not really any noticeable difference between the two solutions. In the STANDARD case the usage of the ARPEGE analysis as the "0h coupling file" was bringing the solution quickly back to the one of BLENDVAR and VARBLEND. The usage of the ALADIN analysis as the "0h coupling file" originated a bit different trajectory (at least up to +6h). Therefore the STANDARD case assimilation cycle was rerun with this option, giving to the STANDARD case the most "independent" algorithm from the ARPEGE assimilation result. On the other hand the fact that the lateral boundary forcing at the initial time dominates the ALADIN trajectory within the first 6 hours of the forecast might be a simple consequence of the relatively small size of the ALADIN/LACE domain.
The balance of the ALADIN analysis and the role of DFI initialization
The STANDARD case analysis was always quite noisy. This result confirmed the experiments of Adam Dziedzic done with the ALADIN/FRANCE configuration in Toulouse. The use of the external initialization is then necessary. On the other hand the BLENDVAR and VARBLEND analyses are already well balanced. The application of the external DFI does not seem to change much the results. Unfortunately, these experiments were not yet run using the incremental DFI technique (where only increments are initialized).
The increments
The shape (broadness) and amplitude of the analysis increments were looked at, also by plotting the charts of differences between the results of the tested algorithms (or between the internal steps within one algorithm). One can say that the increments were rather dependent on Jb: in case of BLENDVAR or VARBLEND with the meso-scale Jb the increments were more "local", sharper, and with smaller amplitude. In the STANDARD case there were large and strong increments, however most of their signal was wiped out like a noise in the following integration.
The spectra
The analysis spectra were examined with the help of "ECTOPLASM" tool. There were not really any big differences between the examined 3d-var algorithms themselves. The "largest" impact on the spectra (taken in the relative sense) was due to the blending step (namely in the wave number interval from 10 to 30). The 3d-var itself was not bringing further change. This result is less encouraging since it could be interpreted as that the ALADIN 3d-var does not analyse more structures than ARPEGE already does (in none of the algorithms). This could be caused by the fact that the ALADIN analysis does not work with denser observation network than the one used by ARPEGE. Perhaps also the Jb term is not really well adapted and tuned.
The scores
Finally the scores of the 6h forecasts (in fact of the guess within the data assimilation cycle) were computed over a short period (against SYNOP and TEMP stations). There were not really any significant differences. Forcing oneself to find the "winner", one could choose the BLENDVAR case.
This first experiment using various 3d-var algorithms brought some interesting results and also it showed us the limits of the current ALADIN 3d-var set-up. At the moment it seems that the STANDARD case does not provide a promising solution if nothing else is done. The most logical and promising is the BLENDVAR case. However, as the whole, the ALADIN 3d-var analysis is not showing much of the performance, in none of the cases.
Unfortunately, the experiment was run at the time when the Prague centre did not have a stable archiving device which limited the storage capacity to be used for these experiments. By consequence, the comparison to the pure blending algorithm (without the use of the observations) could not have been done for that concrete period. Therefore the whole experiment was re-run for another period in May 2001. The results of this period were not yet examined in details. The more recent experiment shall be also completed by the use of the incremental DFI technique. We expect to analyse the results in the second half of 2001 and to provide everybody with a complete report. Before any further important stage of experiments it will be needed to work out better the Jb term or to use significantly more observations than ARPEGE or both.
4. Coupling of surface pressure tendency (T. Szabo)
The main development phase of the surface pressure tendency coupling was successfully finished this spring. A major difficulty was to conciliate this method with the semi-implicit treatment of the large-scale forcing. Finally, the solution was to code the tendency coupling as the correction of the "classical" coupling step though this solution also has some weaknesses. The first experiments were done in the 2d vertical plane using the idealized mountain put to the lateral boundary in order to create the difference in orography between the "large-scale" and "high-resolution" model. The new method showed encouraging results. Then one test was done in the full 3d case with ALADIN/LACE. This time one cannot really say which method provided a better result. Perhaps the LACE domain is not the best one for this kind of tests, since the lateral boundary is mostly over sea or flat regions where we cannot expect so significant differences between ARPEGE (large-scale) and ALADIN (fine-scale) orography. A detailed report by Tamas is available in this Newsletter.
5. Spectral coupling (G. Radnoti)
This work was motivated by the case of 26 December 1999 storm (denoted as French T1 storm) when the 6h coupling refreshment interval was too long to pass correctly the signal of the fast moving cyclone from ARPEGE to ALADIN/LACE domain. From the experiments done in the beginning of 2000 it is known, that the increased time frequency of the lateral boundary refreshment (to 3h for example) is sufficient to capture the T1 storm by the limited area model. Since this simple solution would be costly in terms of the telecommunications, it is interesting to look for other methods, which could supply the missing time-frequency information. One of this alternative is the so-called spectral coupling, which may provide sufficiently good spatial information (the T1 storm is present in the large scale solution though it passed behind the "coupling belt"; hence it cannot be captured any more by the classical grid-point coupling). The hope is to find a way to combine successfully the grid-point method with the spectral one (the spectral coupling alone cannot be applied). The amplitude-angle-phase method of the spectral coupling was developed and successfully tested in the shallow water model. The difficulties came when generalizing the method to the 3d model. Please, refer to the detailed contribution by Gabor Radnoti in this Newsletter.
6. Horizontal diffusion of humidity (Thomas Haiden)
The basic development of this method is finally over, though there remains still one weak conceptual point to be corrected. Despite this point the method seems to work, at least from the point of view of the coherent behaviour, as expected.
To remind the main idea: due to the terrain-following coordinate the diffusion in the model follows the coordinate surfaces, which ceases to be quasi-horizontal in presence of steep orography. Hence the spectral horizontal diffusion operator may create spurious vertical diffusion. In the model a correction exists for the diffusion of temperature, while nothing is done for moisture. The idea tested here consists in making a correction to the moisture convergence fields, assuming that the worse manifestation of the problem lies in the link between dynamics and convective parameterisation.
The method was tested on some pre-selected cases of summer convection in the region of the Alps. The observed effects were rather mild. In order to judge whether the method brings the results closer to the reality, a finer verification will have to be done. It will be necessary to work with radar data (for verification). It will be also useful to verify the assumptions used in the method (the existence of moist bias over the mountains caused by the spurious vertical diffusion of humidity). These two issues are proposed for future work around this subject. A detailed report (written by Thomas Haiden) is available on request.
4. In Budapest (Hungary)
INTRODUCTION
In the first half of 2001 efforts were concentrated on the one hand to the filling up of the ALATNET position already advertised in 2000 and on the other hand progress could be identified in the key ALATNET topics for the Hungarian Meteorological Service. From administrative point of view the originally planned post-doc ALATNET stay (after two unsuccessful call for candidates) was converted into a 20 months pre-doc stay in Budapest and finally was filled thanks to the last call of candidates in April (the stay will start in November). Hereafter this short report will concentrate on the main progress on the different scientific topic.
SCIENTIFIC ACTIVITIES
The main scientific achievements will be mentioned with special emphasis on coupling, surface analysis and variational data assimilation (3DVAR).
COUPLING
The started work had been continued in this year as far as surface pressure tendency coupling is concerned. The code developments were completed and tested together with the experimentation of the new scheme (see more details in the same Newsletter).
SURFACE ANALYSIS
The work had been also continued using a new soil texture data base over Hungary. The sensitivity tests carried out this year were concentrated on the CANARI surface analysis scheme based on optimal interpolation. The CANARI scheme was used in data assimilation mode hoping that the accumulated effect of the better surface description will improve the forecasts of the near surface characteristics of the ALADIN model. For the tests some interesting cases were used, when it was expected that significant differences will be found between the experiments using the old and new data sets. The preliminary results show that there are no clear improvements identified using the new data, so further investigations are need in the interpretation of the results.
3DVAR
The main emphasis was put on the background error statistics computed for the ALADIN/HU model version of the ALADIN model. Background error covariance matrices were computed using the NMC method for different integration lengths and integration differences. The aim of the investigations is finding the optimal background errors to be used in the mesoscale data assimilation process for ALADIN. It was shown that longer integration lengths favour the shift of the variance spectra towards larger wavenumbers (smaller scale), however at the same time the energy of the spectra is decreased. It can be seen that an optimal compromise between these two effects should be found for the optimal background error statistics. The investigations should be continued and completed until the end of this year.
SUMMARY
A candidate was successfully found for the ALATNET position in Budapest (stays to be started in November) and reasonable progress can be shown on the ALATNET scientific topics as well. Finally it is remarked that Hungarian colleagues were also working at other different ALADIN centres (mainly in Prague and Toulouse) in the above mentioned topics.
5. In Ljubljana (Slovenia)
The ALATNET actions in Ljubljana or involving the Slovenian team during the first 6 months of 2001 are reported here :
Case studies aspects of NH
Klaus Stadlbacher was concentrating his efforts mainly on finding proper use of orography on high resolution. He was mainly focusing on the methods of removal unrealistic wavy structures in precipitation field. It turned out that the most realistic structures could be obtained with the use of orography on coarser resolution than the other spectral fields. Orography was produced on quadratic grid while the other spectral fields were produced on linear grid, which in fact means an increase of the resolution by a factor of 1.5. For more information see the article of Klaus Stadlbacher in the Proceedings of the ALADIN workshop in Toulouse.
Work of Danijel Cemas in Prague
He was trying to push the limit of high resolution even below 2.5 km what is currently believed the target high resolution. He was able to run the ALADIN model even at 1 km horizontal resolution. The analyses of results has still to be done but the fact that it is able to run the model in such resolution is very encouraging.
Coupling and high resolution modes
Together with Hungarian colleagues the detailed working plan of the new ALATNET student, Raluca Radu from Romania, was produced together with the list of references. Raluca Radu has started her study in Budapest in September.
 |
Home |