|
The scientific contents of
Modelling aspects of meteorology will rely more and more on tools
developed in an operational environment, which will require more and more
« industrial » methods for their production and
maintenance. In that respect, the ALADIN
environment anticipates the working conditions that atmospheric science
will witness in the 10 coming years. Nevertheless, that environment is still
there to foster scientific and technical progress.
Common points between ARPEGE and ALADIN :
- Both models use the spectral technique for the horizontal representation
of fields. This means that special provisions have to be taken for the use of
a bi-Fourier LAM representation. The solution chosen here is that of
Machenhauer and Haugen that requires the so-called bi-periodicization of
fields through a fictitious « extension zone ».
- As an example, on the figure (colour or
bw) that features the
« extended » orography of the ALADIN-LACE application, one
notices the smoothness and isotropic character of the transition form one side
to the other through the additional zone. It is sometimes claimed that
spectral methods are unsuitable for LAM applications and/or that they cannot
represent sharp features for lack of « locality ». The
other figure (colour or bw
) representing 18 hour forecasts of 10 m winds (the colour of the arrows
depends on 2m temperature), is a good counter example : no apparent
boundary problems exist but many realistic features over land as well as over
sea at the two grid-length scale do (plotting) ;
- The vertical discretization is hydrid (going progressively from
« p » to « sigma » ) ;
- The time stepping now uses the semi-implicit semi-Lagrangian scheme with
two-time-levels solution ;
- The physics is for the time being identical with ARPEGE (which also
benefited from ALADIN work) ; (see
Geleyn et al., 1994).
- The DFI technique is used, the latest, most efficient version (Lynch,
Giard and Ivanovici, 1997) having been suggested in the framework of ALADIN ;
- The CANARI and Full-Pos applications are mirrors of the ARPEGE ones, with
specific steps linked to the LAM geometry.
The points specific to ALADIN :
- The bi-periodicization is accomplished only on the files interpolated from
the coupling model (i.e. as seldom as possible) and can thus be performed with
a quite sophisticated iterative combination of splines and filters ;
- The form of the coupling function (i.e. the relative weight taken at each
time step by the larger-scale solution near the boundaries of the LAM) that
has been optimized as much as possible in the spectral context ;
- The implementation of the semi-Lagrangian scheme in the case of
trajectories originating in the « extension zone » which
required an original treatment of the link between semi-implicit time stepping
and coupling (Radnoti, 1995
) as well as a specific adaptation of Rochas's idea about Coriolis terms
in the two-time-level algorithm ;
- The existence of a non-hydrostatic option, based on Laprise's
« hydrostatic pressure » type of vertical coordinate,
which also requires a redefinition of the
« Simmons-Burridge » vertical discretization operators (
Bubnova et al., 1995).
Finally around the project itself :
- A special study (Caian and
Geleyn, 1997) was performed to evaluate the respective merits of the
stretched solution in ARPEGE and of the coupled solution represented by
ALADIN ; the conclusion was that the combination of moderate stretching
for the global part and of local adaptation via the LAM solution is the best
combination, given the current computing constraints ;
- ALADIN was used as a tool for adjoint sensitivity studies on frontogenetic
problems (Horanyi and Joly, 1996
) ;
- A simplified physics package and its adjoint version for future mesoscale
4D variational data assimilation are currently being developed ( Janiskova,
Thépaut and Geleyn, 1996) ;
- Other important (past and present) topics of scientific activity
are : the behaviour of the model's physics at the limit of validity of
the hydrostatic assumption (and beyond) ; the conditions for a successful
dynamical adaptation process ; and the intrinsic properties of the
semi-Lagrangian time-stepping scheme.
|