|
SECOND MEDIUM-TERM (1999-2001) RESEARCH PLAN FOR ALADINOBJECTIVES AND MEANSIntroductionHere is a revised version of the Second Medium-Term Research Plan for ALADIN (old version), coming at a turning point when the model is operational or nearly by most Partners. The scientific content is the same as for the initial version which was presented and approved (for its purely scientific aspects) at the third ALADIN Assembly of Partners (Prague, 6/11/98). Only some priorities have changed, according to the problems recorded last winter. However the presentation is completely different so as to underline the different steps and the required work and means. Three main axes of research were identified for ALADIN. The first one concerns inevitably the maintenance and the improvement of the current version, addressing problems encountered in operational exploitation. The second one is related to high resolution modelling, the natural evolution of any LAM, especially when global models steadily go to finer and finer grids. The third one focusses on data assimilation, the most debated, since expensive, issue. Preliminary studies show that 4d variational assimilation in ALADIN is quite promising, but the other assimilation tools must be maintained as well. The last part of this document addresses the delicate but essential question of means, involving the problems of dedicated manpower but also training and source code maintenance. A. Maintenance and improvement of the operational versionsA.1 Model verificationA.2 DynamicsA.3 PhysicsA.4 CouplingA.5 ApplicationsB. High resolution modellingB.1 Non-hydrostatic dynamicsB.2 CouplingB.3 PhysicsB.4 ValidationC. Data assimilationC.1 Observations managementC.2 Optimal Interpolation analysis (CANARI)C.3 Variational analysisC.4 CouplingD. MeansD.1 Local ALADIN teamsD.2 TrainingD.3 MaintenanceA. Maintenance and improvement of the operational versionsA.1 Model verificationThis point is not exactly a topic of research but is to be mentioned here since it provides the basis for the definition of priorities. Routine subjective and objective verification, including the comparison to observations, to other ALADIN models, to other forecasting systems, and between operational and test suites, is crucial to track deficiencies and steer further developments.
A.2 DynamicsConsiderable progress has been achieved in this domain along the last years, leading to an enhanced stability and efficiency of dynamics. However current studies show that significant improvements can be introduced at a quite reasonable cost, and help in the further evolution towards very fine grids. In the meantime it is worth underlining the potential use of ALADIN for some small scale applications.
A.3 PhysicsDevelopments in this domain are now essential to improve the forecast of sensible weather. Changes are required anyway to go to higher resolution. The set of topics is quite large and part of the work can be easily deported. However the importance of validation is to be emphasized : at the local scale (using ALADIN in its 1d and 3d versions), at the global scale (to check balances and meet a larger set of situations, at least), with the combination of several developments, and in assimilation mode. Furthermore the corresponding modifications in other domains (e.g. handling and initialization of new fields, post-processing, assimilation, ...) must not be neglected.
A.4 CouplingNo major problem arises with the present coupling method, only some occasional problems in digital filter initialization and the need for some further sensitivity studies to define the "best" strategies, especially in the framework of higher resolution models, have been mentioned so far. The choice of another method is not excluded but must be worth the significant effort required for such a change. These are anyway rather long-term oriented projects. The diffusion of results is crucial here to avoid a useless scattering of experiments. A.5 ApplicationsA close cooperation between Partners is strongly advised for the exploitation of ALADIN outputs, from the computation of derived indices to the coupling to other models (as statistical, air pollution, snow or hydrological ones) or to ensemble forecasting (as a long-term project). B. High resolution modellingB.1 Non-hydrostatic dynamicsAs the model resolution continuously increases the use of non-hydrostatic dynamics will be required at one stage anyway. Though the general background has been implemented some years ago, some basic work is still required to allow it run at an acceptable cost, solving the current instability problems. This is one of the main issue to be addressed in the march to high resolution NWP.
B.2 CouplingAs previously there is no major problem to underline here. Sensitivity studies will be required to define the best choices for the relative horizontal and vertical discretizations of the coupling and coupled models, as well as coupling frequencies. The coupling of the surface pressure tendency instead of surface pressure itself seems also worth to be studied. Some refinements may also be brought to the current coupling method between a hydrostatic and a non-hydrostatic model. B.3 PhysicsMoving to higher resolutions will require at least some more validation and tuning of physical parameterizations, but more likely deeper changes. A close cooperation is required to avoid an irrealistic increase in the number of parameterization schemes and make cross-validations easier.
B.4 ValidationTo complete the set of individual tests performed by developers and specific case studies, it is strongly advised that a "neutral" team assumes the responsibility of a very careful validation of the changes required in the march to high resolution applications. This is a huge but essential task, that can be considered as a project itself. This includes the following tasks :
This implies significant computational resources and a close coordination with other research teams. C. Data assimilationC.1 Observations managementThis is the key point for a further use of ALADIN in data assimilation mode.
C.2 Optimal Interpolation analysis (CANARI)Though efforts in high resolution data assimilation should focus on 4d-variational assimilation, it is really important to maintain and improve the optimal analysis code, CANARI. First CANARI is already used by some partners in data assimilation mode, which implies its maintenance till better solutions are available. Second it will be still be required afterwards for surface data assimilation, since studies on surface variational assimilation are just beginning. Third it has proved useful in other applications : objective verification of forecasts (through the Verif-Pack package), monitoring of observations, diagnostic-oriented analyses (i.e. using available observations to provide an improved representation of the atmosphere to forecasters).
C.3 Variational analysis3d-variational analysis is to be considered here as a crucial step towards 4d-variational analysis rather than a future operational data assimilation scheme. Coordination with developments in dynamics and physics must be ensured.
C.4 CouplingThe problem of coupling in data assimilation mode is still an open one, some of the issues addressed here are still debated while new strategies are to be designed.
D. MeansD.1 Local ALADIN teamsThe existence of operational (or pre-operational) ALADIN suites among almost all Partners gives the opportunity for a new burden of research, with the emergence of deported actions. In the meantime the maintenance of operational applications is an heavy task, which may easily suffocate research and thus prevent further improvements of the model if means (mainly the size of ALADIN teams) are not increased accordingly or a closer cooperation between teams not established. This is true if even the prospect is limited to the minimum (axis A), but the problem is all the more acute since more ambitious projects are considered, especially for data assimilation. D.2 TrainingThe necessary widening and renewal of NWP teams implies a recurrent need for a basic ALADIN training. However needs may strongly differ between Partners, e.g. in time or level, depending on local recruitment policies. In the meantime experience is now widespread among ALADIN teams and the code can be run nearly everywhere. This militates in favour of a local training of newcomers on (pre-)operational ALADIN suites. This would be quite beneficial for both research and operations. Research actions would draw already trained and maybe more motivated people, well informed of the needs and constraints of operational suites. Experimented persons would receive some help in the maintenance of operational applications and get some free-time for prospective. ALADIN training schools will remain necessary, at least to teach the basic knowledge for new research axes. As an example such an exercise would be useful for high resolution modelling, where the previous team is to be renewed. But these operations are expensive and can succeed only if each Partner agrees to send experts as teachers and certifies that trainees will effectively work on the corresponding topics afterwards. D.3 MaintenanceMaintenance is essential to ensure a sensible evolution of the code, allowing everyone to benefit from any new improvement. It covers the following tasks : a. Phasing exercises, when code developments from the different Partners and ARPEGE/IFS are merged together, usually twice a year. Due to the emergence of deported developments, of new research axes, and the necessity to ensure portability and consistency with ARPEGE, these operations will remain heavy. To face the recurrent problems encountered along the last years, the following rules are suggested :
b. Code optimization, to improve portability, efficiency or solve identified problems. The development of diagnostic tools or simplified research versions, the elaboration of "benchmarks" to check the portability of future technical changes, have also to be mentioned here. c. Documentation, with its several facets :
|