Part B - Comparison with the Joint Programme of Work

The programme of work is split in 12 main topics. Each partner is responsible for 1 to 4 topics, even if the basic work is shared between all teams. The thematic reports correspond to the revised work plan. Since (in this combined yearly and final report) emphasis will be given after B.3 on the final status of the various research actions, this B.1 part will be kept short, only insisting on the most important aspects of last year's work.

The task of developing a new radiative upper boundary condition (RUBC) for hydrostatic and non-hydrostatic dynamics has started but is not yet completed. An analysis was done in order to assess the impact of the semi-implicit treatment on the efficiency of RUBC, with an optimistic answer. There is no theoretical obstacle to apply a RUBC together with a semi-implicit approach, but practical experimentation must await the stabilization of the code.

Concerning the bottom boundary condition, problem extensively mentioned in the previous reports, a new solution was proposed that keeps the advantages of advecting vertical velocity (w) rather than divergence, while avoiding the complexity linked to the interpolation of semi-Lagrangian trajectories and to the change of vertical scaling along them. In short, the idea is to have a semi-Lagrangian treatment of the lower boundary condition, keeping the spirit of the w-advection near the surface while returning to the usual method immediately above. However this doesn't completely solve "chimney" problems. Furthermore, it was shown that horizontal diffusion may also lead to such problems.

The most important breakthrough of this fourth ALATNET year was however the validation of an alternative formulation of the linear constant-coefficient solver used to apply semi-implicit time-marching schemes. Until now this solver was always derived by applying a linear approximation around a very simplified unique atmospheric reference state. In contrast with this "classical" approach, the reference temperature is now chosen to be different between the vertical momentum equation and the other ones. With two distinct reference temperatures, the set of linearized equations cannot be seen any more as some "physically simplified" system but it provides quite an efficient numerical control of all kinds of waves, a crucial step for the implementation of a semi-implicit treatment for the fully elastic equations.

Taking into account the above progress, it was clearly shown, both through a stability analysis and by experiments, that the iterative approach with the two-time-level semi-Lagrangian advection scheme substantially enhances the stability with respect to the three-time-level one when the iterative procedure converges. Further, when convergence exists, the scheme converges to a unique solution. A satisfying convergence rate is now practically always reached with one additional iteration, which makes this method the best contender for future developments (improved upper and lateral boundary conditions, physics-dynamics interfacing, …).

Comparison of the behaviours of "linear" and "quadratic" spectral truncations at high resolution showed that the "linear" option (i.e. higher spectral resolution) seems a proper choice for running the model with mesh-sizes below 10 km. However it requires a modification of the spectral representation of orography, which has to be smoothed or at least truncated to the "quadratic" resolution. Several tests were performed at high resolution with several levels of smoothing for the model orography. Smoother, more realistic, precipitation patterns are obtained, while keeping significant extrema (i.e. managing to get rid of the noise). This is detailed in the report of K. Stadlbacher.

The impact of the dynamics of the coupling model, at 5 km for experiments at 2.5 km, was also re-examined with the very new NH dynamics. While previous results were neutral, a significant improvement when using similar (NH) dynamics in both models was shown in this study.

The so-called ALPIA framework (a semi-academic test-bed that mixes nature-like orographic forcing with a regular atmospheric flow) was used for the first time in the study of a parametrisation scheme (see Item 9).

Work on this issue during the fourth ALATNET year concentrated on the further validation of the so-called SLHD scheme (semi-Lagrangian horizontal diffusion), based on the dissipative properties of semi-Lagrangian operators and acting as a flow-dependent 3d grid-point diffusion. Several implementation details were improved, and the code was extended to the non-hydrostatic Euler equations as well as to the geometry of a global variable mesh. However, testing at high resolution with a "linear" spectral truncation (see Item 2) showed that the previously advocated method for making SLHD independent of horizontal and temporal resolutions was not satisfying in such new conditions. Optimisation of the scheme is now performed at one given resolution through diagnosis of the flow deformation, which is the main driving component of SLHD and preliminary results are promising. The scale-dependency issue will be addressed again afterwards, in light of all recent findings.

As explained in the previous annual report, the topic was temporarily frozen, waiting for the availability of a "clean" source code containing the latest NH developments.

In the search of a better temporal scheme it was finally proven that it does not make much sense to try to improve the time-interpolation of lateral boundary conditions. Instead, one turned to a strategy of monitoring the transfer of the coupling data. Two methods were developed, one based on a correction of the linear interpolation and one using a recursive digital filter. The latter turned out to be the most attractive, the most secure, and the most practical one, and will be used for the future dissemination of the coupling data to the ALADIN partners.

Concerning the definition of "transparent and well-posed" lateral boundary conditions, tests in the spectral framework of ALADIN showed that we have to pay extra attention to the extension zone. Simplified experiments in a one-dimensional model showed that a bi-periodicization of fields is required at all time-steps and not only at the occasion of the coupling updates.

The problem of the jump in resolution between coupling and coupled models was addressed in the framework of regional re-analyses that are computed from low-resolution global datasets. An optimal coupling chain for regional and local re-analyses was determined objectively on a MAP-SOP period and connections with the corresponding NWP problematic are currently investigated.

Following successful completion of the basic work in the previous years, no more developments happened on the dfi-blending technique (based on digital filter initialization). But this (locally) operational tool was used for further research, e.g. in climate-type downscaling or in high-frequency provision of surface fluxes for other application.

However it now faces two alternative approaches in data assimilation : either using an additional cost-function in 3d-var to take into account the large-scale information from the coupling model, or the simpler "explicit" blending. It relies on a direct combination / blending of the respective spectra. However comparisons to dfi-blending showed that some additional filtering has to be applied and that no real simplification of the whole procedure can therefore be achieved.

On a complementary topic, the choice of an optimal strategy for estimating forecast errors remains open, with the recent evaluation of an Ensemble method for ALADIN. Here also, the question of large-scale coupling has to be addressed : whether only analysis and observations should be perturbed to obtain the ensemble or whether perturbation of the large-scale lateral forcing should be added as well. Finally the scale-selection problematic that was successfully addressed within the NMC method by the so-called lagged statistics will have to be tackled in the new framework and some thinking about this problem is now on-going.

The study of two topics restarted. The first topic concerns the evaluation of different strategies for the numerical treatment of the diabatic terms : where should the several components of the physical forcing be applied along the semi-Lagrangian trajectory ?

The second topic is the treatment of the diabatic terms in non-hydrostatic mode : which should be the partition of the heating / cooling impact of diabatic changes between temperature and pressure evolutions ? Preliminary results have been obtained, but the problem is still open.

The methodology developed during the third ALATNET year for diagnosing "sleeping" problems of nonlinear instability and / or stiffness has been applied to the new state of the operational physics package (see Item 9). At this occasion the culprit for a long-standing problem (spurious oscillations at temperatures just around the triple point) was found and corrected : a too high ratio between the assumed fall-speeds of liquid and ice precipitations.

The design of a prognostic and scale-independent parametrisation of deep moist convection was pursued by including a prognostic micro-physical scheme. Efforts concentrated on the integration of the differing sources of precipitation and cloudiness (subgrid versus resolved) to address a list of pending issues, all more or less related to the difficulties at the "grey-zone" horizontal resolutions, where convective systems are partially resolved by the model grid but still need to be partially parametrised too.

An extended investigation was pursued of the way the traditional exchange coefficients can be parametrised in terms of turbulent kinetic energy and of stability. A numerical integration scheme for solving the prognostic second-order nonlinear differential equation for turbulent kinetic energy is under ongoing development.

The study about the cyclogenetic impact of parametrisation schemes in the planetary boundary layer was reoriented into two directions : that of the so-called symmetric instability and that of a more physically based definition of mixing lengths and stability functions for the computation of the (purely diagnostic in that case) exchange coefficients. The results are compared with the current ALADIN computations of momentum and heat fluxes, but the complexity of the issue and its interaction with cloud and radiative aspects make the operational outcome very hard to steer.

Last year it was reported that a succession of two packages deeply modified the basic version of the ARPEGE / ALADIN parametrisation set, thanks to the joint efforts of many participants, not all from ALATNET centres but all working within the ALATNET spirit and roughly following its work plan on the issue. In fact the second of these packets was implemented in the fourth ALATNET year (but before the publication of the third annual report). The positive impact of those changes was confirmed by numerous a-posteriori studies during the fourth ALATNET year.

Since then, work on parametrisations has concentrated on two issues : the reformulation of the description of drag and lift effects from unresolved orographic features, in order to be able to suppress the currently used artefact of an enhanced topography (the so-called envelope orography), and the search for a new solution to the problem of the balance between complexity of the radiation computations (for the sake of accuracy) and frequency of their calling (for a better interaction with cloud processes).

In the first case the scheme for mountain effects was revised following most recent findings reported in the literature and the lift effect was made orthogonal to an estimate of the geostrophic wind and not any more to the wind itself. Furthermore a new set of diagnostics based on the ALPIA framework was prepared and extensively used to steer the tuning of the five or so relevant parameters. The work, that involved people from four of the Partners' teams, has now reached the state where the envelope orography can be removed with equal quality scores and a better description of the flow in the mountains' vicinity, with positive consequences on the rainfall forecast.

In the second case, the design and coding work is nearly completed and some preliminary tests with a simplified option for the application (statistical computation of the quantities that are later supposed to be stored in-between more complex computations) have shown a very nice potential. Work on the target application has just started, people from three of the Partners' teams being involved in what promises to become again a complex validation process.

Besides the significant progress in the PhD work of M. Szczech, there was an acceleration of the work on observations along this last year.

The ODB (Observation Data Base) management was regularly improved, in order to take into account new observation types or new parameters (such as observation height or the statistical model for observational errors). ODB is now available by most partners and problems related to model boundaries may be considered as solved.

The quadratic variational quality control was updated and successfully tested. The remaining work is described hereafter, since highly dependent on observation-types.

The study of the impact of (SYNOP) observations of temperature and relative humidity at 2 meters on 3d-var increments was pursued in Budapest using the parallel assimilation suite, and started in Toulouse, with more focus on diagnostic analyses for nowcasting (using mainly conventional data and assumed to be closer to observations). One of the challenge is to answer whether applications based on optimal-interpolation analysis might be ported to 3d-var.

The observations of wind at "10 meters" received significant attention. Over sea, corrections were brought in order to use the real altitude of data (on average 7 m for buoys and 24 m for ships !). On land, the design of a careful black-listing procedure (since attempts to use directly the model information on orography for observation selection failed) allows to use them now. Besides the observation operator was improved in order to better take into account the vertical structure of the boundary layer.

The use of aircraft observations (AMDAR data) was further evaluated in Budapest, with a thorough analysis of data availability and a retuning of the selection procedure, especially the thinning distance (defining the geographical density of the observations effectively provided to the model). A positive impact was noticed at times where SYNOP and sounding observations are fewer (06 h, 12 h, 18 h UTC). The interaction with the background-errors representation, a refinement of the statistical model and the optimal length of the assimilation cycle (3 h rather than 6 h ?) were discussed.

Significant effort was devoted to ATOVS (Advanced TIROS Operational Vertical Sounding, on board NOAA15/16/17 satellites), in a global framework in Toulouse and with more focus on LAM specificities in Budapest :
- improvement of channel selection, bias corrections and control of rainy observations, evaluation of the impact of the thinning distance (120 against 80 km, for an initial resolution of ~45 km) and the statistical model for AMSU-A (Advanced Microwave Sounder Unit A, providing information on temperature mainly);
- improvement of data selection (control of cloudy observations) and impact studies for HIRS (High resolution Infra-Red Sounder, providing informations at a higher resolution, ~20 km);
- design of quality control, definition of a statistical model and a bias correction for AMSU-B data (providing more informations on water, at a resolution of 16 km).
Unluckily, the impact on 3d-var assimilation in ALADIN, though positive, keeps small in the first experiments, with interactions with the choice of the background-error model, as in the case of aircraft observations.

The most innovative work concerned the newly available MSG/SEVIRI data, provided by Meteosat Second Generation at very high spatial and temporal resolutions (5 km and 15 min), and including informations on the variations of temperature and humidity. The whole procedure to use such observations was designed, and impact studies performed using ALADIN 3d-var assimilation, at various spatial resolutions (10 km, 2.5 km) and frequencies (from every 6 h to hourly analyses). No significant impact on the forecast of temperature was found, but the humidity and cloudiness fields are improved up to 24 h.

The work on radar reflectivities started only in March 2004, due to the late availability of some members of the corresponding working group. Only the research plan was designed along the last year of the ALATNET project.

The fourth ALATNET year was mainly used to consolidate and extensively test the 3d-var developments successfully performed in the previous years. A parallel suite is routinely running for more than one year now in Budapest, and the design of another one has started in Toulouse, with different options.

Research initiatives were numerous and promising anyway this year. The work on background-error statistics was pursued, with the first use of the Ensemble method for a limited area model (and thanks to the previous work of M. Belo-Pereira at the global scale), further research on a wavelet representation, re-examination of the univariate versus multi-variate approaches for humidity, ... Coupling issues were also examined, with a new proposal to introduce information from the coupling model in the analysis process, based on the new Jk cost-function (described in the last ALATNET Newsletter), and the test of explicit blending (see Item 6).

The use of the adjoint model for sensitivity studies, initiated by C. Soci, attracted followers(including another young researcher, A. Simon) and supported analyses of problems in physics. The work on a-posteriori retuning of model and observation error specification was finalized, and refined. And the initial effort on singular vectors led to the emergence of a team specialized on predictability issues in Budapest.

Apart from the continuation in Toulouse of the work of G.P. Balsamo on 2d-var soil moisture assimilation and the finalization of the work of C. Soci in a PhD defence, there was no further step forward. The effort on assimilation issues was mainly devoted to 3dvar and observations.