The scientific highlights of the fourth ALATNET year are very much in line with those of the second and third years.
In the part concerning the non-hydrostatic formulation, most of the remaining problems have been solved and the iterated two-time level scheme (also called predictor-corrector) together with all novelties announced last year shows exceptional performances in terms of stability and accuracy. The problem of the bottom boundary condition has also been solved, even if the new framework has still to be extended to other operators than the semi-Lagrangian advection, for a completely consistent approach. A strategy for interfacing physical parametrisations having very specific small-scale features (short time response, memory of falling condensates, …) has also been prepared and concrete work on the compressibility problem for diabatic sources finally started.
For the ALADIN variational tools, a lot of experimentation took place, on the one hand in order to consolidate all the previously reported work and on the other hand to prepare the software for a more flexible approach to the inclusion of new data sources at high resolution (radar measurements for instance), a domain of activity where new promising results start to come by.
For reasons of overall priorities, work on the physics concentrated on the continuation of the work targeted on convection at scales where it is neither fully parametrised nor fully resolved as well as on "larger-scale" type parametrisations like mountain drag and lift effects and radiative forcing. The remainder of the work was put a bit on stand-by in order to study a possible convergence with micro-physics and turbulence parametrisation developed in the upstream research groups of Météo-France. The tail end of the already reported third year progress on "classical" parametrisation schemes got also translated in operations, early in the fourth year period.
Finally one should mention the development of an original way of diagnosing how the coupling information has best to be passed from the coupling to the coupled model, a breakthrough with potentially wide ranging applications.
Looking now at the four years in one go, through an evaluation of the execution of the initial ALATNET scientific plan, the situation is also rather satisfying. Among the 44 sub-items listed in the contract, only eight correspond to real misses, three were not tackled because of a change of the scientific context, fifteen can be considered as fully completed while work is still on-going on the remaining eighteen ones. On the other hand, there were thirteen closely related sub-topics that were added for consistency to the ALATNET efforts and further tackled as things progressed. This of course helped reorienting the plan in a more effective direction (some new items were indicated in a modification of the plan at mid-term, a few came even later). Among the thirteen added topics, five can even already be considered as successfully solved. Thus on a total of 54 still relevant sub-items we have twenty full completions, twenty-six partial ones and eight misses, well above the "three time one third" usual in scientific endeavours on such a time scale. If one analyses the reason for the eight "misses", we see three times too big ambitions in the original plan for a four year target, twice a lack of manpower for reasons of change of internal priorities, twice a lack of progress for too little effort put on the topic and once a clearly too late start of the effective work.
Going into more details one can try to view the outcome on the basis of the 12 items of the plan:
As already hinted at in the report about the fourth year, this item, mainly driven by Partners 3 and 1 and involving two young researchers, is concluded with positive results (Bénard, 2003 ; Brozkova et al., 2003 ; Bénard et al., 2004 a & b), the only yet unfinished work being on the upper boundary condition. Interesting extensions to the initial plan are the new approach to the problem of constructing the reference linear model of a semi-implicit scheme (Bénard et al., 2003 ; Bénard, 2004) and the harmonisation of the new NH dynamics around the concept of a "predictor-corrector" time-marching scheme (Vivoda and Bénard, 2002 ; Vivoda et al., 2003). All this can be considered as one of the main satisfactions of the ALATNET effort and would not have been possible without the EU support.
This aspect, driven by Partners 5, 1 and 3 and involving one ALATNET student is more contrasted, despite its proximity with the previous one. On the one hand the panoply of testing tools was spectacularly increased and consolidated (this including the novelty of the development of a semi-academic test-bed that mixes nature-like orographic forcing with a regular atmospheric flow, the so-called ALPIA environment) but on the other hand the full-scale validation of new developments in dynamics remained partial and not much was done concerning developments of the physical part (Banciu et al., 2002). On the whole it was nevertheless a rather successful effort.
This topic, mainly tackled by Partners 3 and 1, had to be refocused owing to the rapid progress on the Item 1 (see above). In fact the main emphasis was shifted from the properties of the adiabatic calculations studied at the beginning (Gospodinov et al., 2002) to the issue of how to get a flow-dependent lateral mixing in a spectral model. Some progress was made following the idea to rely on the more or less damping properties of semi-Lagrangian algorithms, but work is still ongoing at the end of the ALATNET period.
This item started according to the plan and was solved by Partner 1 for the case of a hydrostatic atmosphere. The impact was found small at scales where the hydrostatic assumption prevails (without surprise). But then a paper from British colleagues (Staniforth and Wood, 2003) indicated that a similar enhancement to the basic equations in the non-hydrostatic case may be obtained with a far more direct method, especially in the case of our own version of the NH dynamics. It was thus decided to freeze the issue until completion of the work on the three previous Items. Formally speaking the ALATNET target has been reached, half by work, half by default.
This item involved all Partners as well as one young researcher and has a rather mixed-type outcome. Partial progress was achieved on the problem of "transparent" lateral boundary conditions (Mc Donald, 2002) and on the issue of domain size and resolution jumps. The development of a spectral coupling code was also achieved as an addition to the original plan. But the part about bi-directional coupling was abandoned and the tackled issues are still on going at the end of the ALATNET period. On the other hand, one got the addition of a new strategy for an event-dependent transfer of information between coupling and coupled models (Termonia, 2003 & 2004), something fundamental for the distant networking aspect of the ALADIN project.
Although closely related to the previous one, this item (that can be considered as the most transversal one since it was tackled by all teams and one ALATNET student in a quite synergistic way) was rather successfully completed. Even if with a somewhat surprising negative conclusion, the work on tendency coupling for the orography-related surface pressure was done. The blending technique was tested, brought to operational use in its most advanced version (Brozkova et al., 2001) and also became the basis for more fundamental research actions (Beck et al., 2004 ; Guidard et al., 2004). The other coupling issues in data assimilation are also now better understood and discussed through the community. All in all a really collegial effort meeting some deserved success.
Work on this topic was conducted by Partners 2, 1 and 3 and involved one young researcher. Some progress was achieved but mainly not where it was initially intended to put the focus on. The work on the exact introduction of diabatic forcing had to be displaced and started late while the only (but rather original and with on-going promises for further work) new results came from the addition of a topic on diagnostic of nonlinear instability and stiffness in the practical implementation of a given set of parametrisation algorithms.
This originally wide ranging topic, tackled by Partners 2, 1, 5 and 3 and involving one young researcher has been translated in a very advanced proposal for addressing the problem of moist convection at scales where it is neither fully parametrised nor fully resolved. But it remained inconclusive or even untouched on most other sub-items. The circumstances were in fact not favourable, with a rapid evolution of the background to which most of this work should have been fitting, but one may still speak here of a too much targeted initial approach that could not be steered in other directions quickly enough.
All five Partners worked on this topic and one ALATNET student was also involved. Like for Items 2 and 5, we have a contrasted picture. New ideas (about mountain drag-lift forces and about intermittent radiation computations) were introduced with apparently good progress. A strategy for profiting from the work done elsewhere on turbulence and micro-physics has been put forward, but most of the original targets of the plan are at an "in-between" state.
A bit alike the previous topic, this one, mainly tackled by Partners 1, 4 and 3 and involving one ALATNET PhD has met success on some issues but not systematically where it was envisaged first. In fact this is more normal in the present case since reactivity is of paramount importance in this very peculiar scientific trade and the additional work on data banks and quality control makes the picture even more favourable in terms of concrete achievements. There is however still the general impression of a shift of calendar with respect to the original ambitions, those being probably too high and the nicest results being obtained on more traditional data sources (microwave radiances, aircrafts measurements).
This item, where Partners 4, 1, 3 and 2 were active together with two young researchers, is part of the three main satisfactions of the ALATNET programme and this success would surely have been of less scope without the TMR programme. Apart from still missing definitive statements about the scientific validation of the 3D-Var application, all foreseen objectives have been reached (Siroka et al., 2001, Berre et al., 2002 ; Brozkova et al. 2002 ; Siroka et al. 2003, Sadiki and Fischer, 2004) and there was the bonus of unforeseen enhancements to the work on the background error term (Deckmyn and Berre, 2004).
Given its long-term character the topic was always considered rather as a feasibility study. But even this ambition had to be scaled down when priority was put on the FGAT version of the 3D-Var for an earlier operational target (Fischer at al., 2000). Some very interesting work was nevertheless produced mainly through the two already defended ALATNET PhDs, the work being mostly supervised by Partners 1 and 4 (Soci et al., 2002 & 2003). In particular the 2D-Var analysis of surface moisture represents a world-class development that shall have important applications in the future (Balsamo et al., 2004). A number of late extensions to the plan also started to be tackled (predictability issues, a general approach to the use of humidity data, …).
Concerning the 9 registered PhD students among our young researchers, two have been defended at the time of writing this final report (G-P Balsamo, 4-4-2003 and C. Soci, 19-4-2004), four are "on track", two will need a good additional writing effort to be concluded and one is still in an uncertain status.
Finally, even if it goes beyond the pure scope of the ALATNET endeavour, a very important event happened right in the middle of the fourth year, but with such consequences on a longer time scale that it is thought worth putting right at the end of this Section. The HIRLAM group (the five Nordic Countries -with some association to Baltic states-, the Netherlands, Ireland and Spain) decided to try and join forces with the ALADIN community for the development of an even wider pan-European collaboration in Numerical Weather Prediction at high resolution. Among the main motivations for their step, there were the trust they put on the performances of the ALADIN non-hydrostatic dynamics, the similitude of their approach to variational data assimilation (where they are in fact more advanced on the 4D-Var side) with ours and the common aims to achieve ambitious goals on the coupling problem as well as to tackle the physical problems of the so-called "grey-zone", in order to be able to significantly progress in the simulation of real weather without waiting to reach the (expensive) scales or resolved convection and advected turbulence. All this is indeed closely linked to the success of the ALATNET scientific effort and to the entrainment effect it had on the ALADIN and LACE projects, be it in terms of new results or of more structured research and teaching actions. The participants to the ALANET effort are therefore proud to have contributed to such a promising widening of existing European co-operations in their field of activity, and wish to thank the TMR programme of the EU for giving them the chance to show their potential and to trigger the above-mentioned ambitious and far-reaching convergence with HIRLAM.
References (only those not provided in Section A.2 below) :
Fischer, C., M. Siroka, W. Sadiki and C. Soci, 2000 : Boundary problems and scale selection in a limited area NWP model. Proceedings of the workshop "On Adjoint Applications in Dynamic Meteorology", 26-30 June 2000, Moliets, France.
Brozkova, R., D. Klaric, S. Ivatek-Sahdan, J.-F. Geleyn, V. Cassé, M. Siroka, G. Radnoti, M. Janousek, K. Stadlbacher and H. Seidl, 2001 : DFI blending: an alternative tool for preparation of the initial conditions for LAM. Research activities in atmospheric and oceanic modelling, Report N°31 of CAS/JSC Working Group on Numerical Experimentation , 1, 7-8.
Siroka, M., G. Bölöni, R. Brozkova, A. Dziedzic, C. Fischer, J.-F. Geleyn, A. Horanyi, W. Sadiki and C. Soci, 2001 : Innovative developments for a 3D-Var analysis in a Limited Area Model: scale selection and blending cycle. Research activities in atmospheric and oceanic modelling, Report N°31 of CAS/JSC Working Group on Numerical Experimentation , 1, 53-54.
Banciu, D., S. Alexandru, E. Bazile, L. Gerard and K. Stadlbacher, 2002 : Precipitation, resolution and orography, LAM Newsletter 30.
Berre, L., G. Bölöni, R. Brozkova, V. Cassé, C. Fischer, J.-F. Geleyn, A. Horanyi, M. Raouindi, W. Sadiki and M. Siroka, 2002 : Background error statistics in a high resolution limited area model. Proceedings of the HIRLAM Workshop on "Variational Data Assimilation and Remote Sensing", 21-23 January 2002, Helsinki, Finland.
Brozkova, R., G. Bölöni, C. Fischer, J.-F. Geleyn and A. Horanyi, 2002 : Recent experiments with data assimilation in ALADIN/LACE model, LAM Newsletter 30.
Mc Donald, A., 2002 : A step toward transparent boundary conditions for meteorological models, Mon. Wea. Rev., 130, 140–151.
Soci, C., C. Fischer and A. Horanyi, 2002 : Simplified physical parameterization in the computation of the mesoscale sensitivities using ALADIN model, LAM Newsletter 30.
Vivoda, J. and P. Bénard, 2002 : Iterative implicit schemes for non-hydrostatic version of model ALADIN, LAM Newsletter 30.
Siroka, M., C. Fischer, V. Cassé, R. Brozkova and J.-F., Geleyn, 2003 : The definition of mesoscale selective forecast error covariances for a limited area variational analysis. Meteorol. Atmos. Phys., 82, 227-244.
Staniforth, A. and N. Wood, 2003 : The Deep-Atmosphere Euler Equations in a Generalized Vertical Coordinate. Mon. Wea. Rev., 131, 1931–1938.