(Jean-François GELEYN)
1. Recent operational changes in ARPEGE
- On 23/2/2000 the compacting characteristics were modified with
a 20% reduction of the file's size.
- On 4/5/2000 the operational library was switched to a
Cy22T1/AL12-based release (for ALADIN-France as well). At the same time a
dozen or so of modifications were made to the observation handling (bug
corrections, use of NESDIS data as background for the SST analysis, retuning
of the screening for too dense observations and technical modifications to
enable the use of the ATOVS radiances as well as a 4D-Var parallel suite).
This very cumbersome process had in fact started at the end of November 99!
- On 20/6/2000 two major milestones went operational in a joint mode:
- the use of ATOVS radiances (hence the possibility
to process the data of two NOAA satellites, while we were left with a single
possibility since the end of NOAA11 in March 99). Of course the ATOVS data are
richer than the TOVS ones (thanks to the Advanced Microwave Sounding Unit
-AMSU- new instrument) and would in principle warrant a direct processing of
raw-radiances; for the time being we are however still processing, via a
1D-Var procedure, both HIRS and -SU data for the TOVS and ATOVS instruments
alike, i.e. starting from so-called «cleared radiances».
- the replacement of the 3D-Var intermittent data
assimilation by a 4D-Var continuous data assimilation (second such
implementation world wide, after the ECMWF one, two and a half year ago). The
characteristics of this implementation are as follows: similar screening and
quality control procedures as for 3D-Var, but with hourly windows for the
timing of data (against 6 hourly before); multi-incremental version of the
minimisation T42/T63/T95 (all unstretched), the «regularised MJ
physics» being activated only at T95; 6 hours time windows, centred
around analysis times, with the interface with surface assimilation (still
CANARI-based) roughly unchanged; high frequency wave filtering done without
any NMI computation, through the use of a DFI-like procedure as a weak
constraint of the minimisation.
2. Forthcoming operational changes in ARPEGE
- The parallel suite of ATOVS+4D-Var revealed a slight default
of incrementality over the Saharan region (too much difference between the
resolution of the forecasting model and that of the control variable in a
region where the usual geostrophic and moist-physical balance mechanisms are
not fully at work). It was therefore decided to test a new 4D-Var
configuration with (i) a doubling of the weak DFI constraint, (ii) the
introduction of a non-zero level for moisture handling in the latter and (iii)
the introduction of a semi-external incremental DFI, to better control noise
and humidity increments. The last point is obtained «for free» when
replacing the 3D-Var-linked «double filtering» DFI option by a
«launching» DFI option applied on two already existing 6 hours runs,
i.e. the «observation screening» one and the «ultimate
trajectory for surface analysis» one. First results of the new parallel
suite are encouraging and an operational switch in August is likely.
- The next plans concern the resolution of the forecast model.
All ARPEGE configurations and ALADIN-France will move from 31 vertical levels
to 41 (top at 1hPa, bottom at 17m above the surface). Furthermore there will
be a change of the link between spectral and grid-point resolutions: for all
forecast models (at yet unchanged situation for the control variable of
4D-Var) the quadratic grid approach will be replaced by the linear grid one,
but with the orography still computed in the quadratic mode (i.e. unchanged in
practice). Hence this change means 2.25 more degrees of freedom in spectral
space with unchanged physical and orographic forcing. Finally the grid mesh of
ALADIN-France will go down from 9.9224 to 9.5090 km at constant domain size,
in order to put it back to half the minimum value for ARPEGE. The consequences
of these resolution upgrades on the coupling strategy for deported ALADIN
applications will have to be assessed as soon as possible, but the switch
itself will be technically quasi-transparent in this respect.
3. Planned operational changes in ARPEGE
- A package of changes in the physics;
- A revision of all data bases (orography, soil, vegetation and
soil moisture climatology) to be strongly coordinated, like always, with the
ALADIN coupled runs;
- A major revision of the snow cover description (including in
the soil data assimilation procedure); here also the ALADIN coordination might
be complex;
- The use of scatterometric data;
- ????