THIRD MEDIUM-TERM (2002-2004) RESEARCH PLAN FOR ALADIN : summary
« marching towards
very high resolution and continuous data assimilation
while preserving if not improving
the current level of response to operational problems »
WHAT'S NEW ?
an old (10 years) and mature (?) project
+
less external financial support
ê
a new organization of research is required !
increasing responsibilities for each partner
a higher part of local research
a more intensive networking
a research plan elaborated through
a close cooperation between all partners
TRAINING
Local basic training : the present effort must be preserved
Advanced training : alternative solutions are to be designed
PhD theses : all partners must be aware of their importance for
local research & training
ALADIN / ALATNET PhD theses
defended :
1995 : Radmila Bubnova (CZ), Elena
Cordoneanu (RO)
1996 : Mihaela Caian (RO),
Andras Horanyi (HU)
1997 :
1998 : Marta Janiskova (SK)
1999 :
2000 : Mark Zagar (S I),
Ilian Gospodinov (BU)
2001 : Doina Banciu (RO),
LoÏk Berre (FR)
Francois Bouyssel (FR), Luc
Gérard (BE)
à
11
in progress :
AU : Klaus Stadlbacher
CZ : Martin Janousek, Filip Vana
FR : Jean-Marcel Piriou
IT : Gianpaolo Balsamo
MO : Radi Ajjaji, Wafaa Sadiki
PL : Malgorzata Szczech
PT : Margarida Belo-Pereira
RO : Steluta Alexandru, Raluca Radu,
Cornel Soci
SK : André Simon, Jozef Vivoda
TU : Karim Bergaoui
à
15
MAINTENANCE
ESSENTIAL FOR THE LIFE
AND PROGRESS OF THE ALADIN PROJECT
INCLUDING :
· phasing,
· optimization,
· documentation,
· a few other
technical tasks of common interest
EMERGING PROBLEMS !
· increasing
part of data assimilation Þ heavier phasing
· maintenance of scripts
· very (too)
large discrepancies between local versions
Þ
efforts required from local teams and also from the Toulouse one
· increasing
portability problems
EWGLAM/SRNWP initiative
for common model interfaces
_________________________________________________________________________________________________
OPERATIONS
Main objectives,
to be further discussed after the technical overview of the project
· reduced
discrepancies between local versions
· harmonized
operational schedules (forecast, assimilation)
· extended
production of coupling files (frequency, range, assimilation, ...)
· implementation
of local databases for observations
· exchange
of local observations between partners
(no longer any transmission of full observation
files from Météo-France !)
APPLICATIONS
POST-PROCESSING :
§
forecast : new fields on demand
¨ research
configuration (assimilation) : implementation
AVAILABLE TOOLS FOR OPERATIONS AND RESEARCH :
ª
re-launching the exchange of applications
¨ ensuring portability
¨ design of a
management procedure
INTERFACE TO DOWNSTREAM APPLICATIONS :
¨ documentation
of existing tools
STATISTICAL ADAPTATION
¨ more
networking, within ALADIN and SRNWP
DYNAMICAL ADAPTATION
ª
further research required
MODEL TO SATELLITE APPROACH
§
a recent and promising tool
DIAG-PACK
¨ enhanced
exchanges with nowcasting teams
¨ further
research required
_________________________________________________________________________________________________
VERIFICATION
« This 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.»
(Second medium-term research plan)
BUT THIS TOPIC IS NEARLY AT THE SAME POINT AS IN 1999 !
Building a coordinated
procedure
for objective
verification
at synoptic scales
|
Definition of rules, implementation of
the database
Routine update of the database
Improvements and diffusion of results
|
Defining a
verification procedure
for high resolution forecasts
|
Definition of a working group, involving
modellers and forecasters
Use of satellite and radar data (precipitations)
Safe exchange of local observations
between ALADIN partners
|
COUPLING
An emerging priority :
· severe
operational failures (1999 Christmas storms)
· results
from sensitivity studies based on variational tools (to prepare 4d-var)
· results
from case studies at very high resolution
Interaction with orography
|
"Surface-pressure tendency" coupling
"Orography" coupling
|
Spectral coupling
|
Use of large-scale spectral information
Interpolation of amplitude and phase-angle
Combination with Davies'scheme
Case studies and tuning
|
Time-interpolation
|
Further comparison of the present schemes
|
New prognostic variables
|
Design of a strategy for new variables
from physics
Introduction whenever required
|
Towards higher resolution
|
Pseudo-radiative scheme
Two-way nesting
Non-hydrostatic variables
Case studies
|
Choices for data assimilation
|
Comparison of present choices
Spectral coupling and 4d-var
|
_________________________________________________________________________________________________
DYNAMICS
Hydrostatic dynamics :
improved semi-Lagrangian schemes
|
Uniformly accelerated scheme
Predictor-corrector (P/C) scheme
|
NH dynamics :
three-time-level semi-Lagrangian (3TL) schemes
|
Optimal choice of model variables
P/C scheme
|
NH dynamics :
two-time-level semi-Lagrangian (2TL) scheme
|
Properties of P/C scheme
Refinement in the choice of model variables
Use of decentering
|
Bottom boundary
condition (NH) & related discretisation problems
|
Optimal discretisation
General improvement of the current scheme
|
Diabatic forcing)
|
Strategy for the diabatic forcing
Adaptation to the final choice of
prognostic variables (NH)
|
Orographic forcing
|
Optimal filtering of the orography
Resonance problem in NH
|
Relaxation of the thin
layer hypothesis
|
Implementation and test in ALADIN
Extension to NH
|
Radiative upper
boundary condition
|
Feasibility study : analysis, academic
2d tests
Adaptation to ALADIN NH
Control of the hydrostatic version
|
Horizontal diffusion
|
Horizontal diffusion using
semi-Lagrangian interpolators
Gridpoint treatment of humidity
|
--- high priority,
important and/or urgent action
--- intermediate priority
or "medium-term" action
--- low priority or
"long-term" action
PHYSICS
Solving identified problems at the present operational
resolutions & Going towards higher resolution : ~ 5 km
USE
OF NEW
|
Convection
|
Introduction of a prognostic convection scheme
Management of the 4 new variables
Validation over an extended set of situations
Investigating problems in the triggering
of convection
Analysis of the closure and hysteresis problem
|
PROGNOSTIC
VARIABLES
|
Microphysics
|
Management of 2 or 3 new variables :
condensed water
Further analysis of the "Functional
Boxes" approach
Introduction / choice of a semi-complex microphysics
Interface with convection
Prognostic treatment of falling
condensates or not?
|
AND
CHANGES
IN
CONCERNED
|
Vertical diffusion,
low cloudiness,
PBL, ...
|
Introduction of a
prognostic TKE scheme (1 new variable)
Interaction with other developments concerning PBL :
* link between top of PBL fluxes and
cyclogenetic activity
* noise in shallow convection
* PBL-height dependent mixing lengths
* developments in the anti-fibrillation scheme
* improvement of low-level cloudiness
(diagnostic scheme)
|
PARAMETERIZATIONS
|
General problems
|
Update of thermodynamics
Consistency with the other parameterizations
Interface with coupling, dynamics and
data assimilation
Consistency with regular physics
Validation at various horizontal and
vertical resolutions
|
IMPROVEMENT
|
Radiation
|
Refinements of optical depths
Move (choice, development) to an
intermediate scheme
|
OF
BASIC
|
Orography
|
Improved smoothing of very small scales
Management of the extension (and
coupling ?) zone
Tuning of the envelope
Better description of roughness length
Investigation of feed-backs with other parameterizations
Study of local circulations
Development of new diagnostics
|
PARAMETERIZATIONS
|
Surface
|
Parameterization of lakes
Improved description of evaporation over sea
Revisit of the z0h/z0m ratio over land
Improved databases for soil and vegetation
|
Simplified regular
physical parameterizations
|
Tuning of diffusion
Improved description of humidity
Validation at high resolution
Consistency with the "full" physics
|
Physics / dynamics interface
|
Introduction of the new variables
Interaction with the predictor /
corrector approach
Interface with "externalized"
parts of the physics
|
|
Case studies
|
Identification and study of
"strange behaviour" cases
Selection and documentation of extreme situations
Validation on a wider range of situations
|
VALIDATION
|
New observations
|
Comparison to satellite data
Comparison to radar or lidar data
Interfaces to new field experiments
|
|
New methods
|
Design of new scores or criteria
Use of expert systems to identify fine
scale structures
|
DATA ASSIMILATION : METHODS
Main objectives for upperair assimilation
Prototype 3d-var á
Operational 3d-var á
Operational 3d-FGAT á Prototype
4d-var
Cautions !
· increased
local and centralized maintenance
· management
of scripts, maintenance of a reference version
· need for
new observations to feed data assimilation systems
· mostly
local choices and tunings
Algorithmic aspects
3d-var
|
Use of observations at the
borders of the domain
New minimization
algorithms
Design of an explicit
spectral blending and combination with 3d-var
Improvement of observation
operators (vertical interpolations)
Choice of the time-window
for the selection of observations
|
3d-FGAT
|
Implementation of a 4d screening
Choice of lateral boundary conditions
Choice of the time-window
|
4d-var
|
Maintenance of the TL/AD
code (for various research purposes)
Coding TL/AD of
semi-Lagrangian schemes
Definition of coupling
strategies for the various elements
Adaptation of Jc-dfi (to
high resolution, to new variables)
Improvement of simplified physics
|
Simplified physics
|
Evaluation through
sensitivity studies
Evaluation and tuning at
high resolution
Solving incrementality problems
Adapting observation
operators to new variables
|
A-posteriori validation
|
Further tuning of
statistics (observations + background) for 3d-var
Extension of diagnostic
tools to 4d-var
|
TL/AD tools
|
Maintenance of the TL/AD
code (reminder)
Use in the design of the TL/AD code
(e.g. LBC, NH, new variables)
Use to study nonlinearity problems (e.g.
in simplified physics)
Predictability studies
|
Var-Pack
|
Watch
|
Modelisation of background / forecast errors
Sampling methodology
|
Evaluation of the
different contributions to error covariances
Ensemble analyses and
forecasts with perturbed observations
Singular vector approach
|
Diagnostics
|
Heterogeneity and anisotropy
Time-dependence
Nonlinear effects
|
Jb formulation
|
Approaches based e.g. on
diagonal blocks and wavelets
|
New variables
|
Taking into account new
prognostic variables (NH, cloud water, ...)
|
Cycling
Blending
|
Maintenance of a reference
version of dfi-blending
Adaptation to the main
changes in the model
Development of
double-nested blending
Comparison of
"dfi" and "explicit" blending for spectral fields
|
Assimilation cycle
|
Investigating the various combinations
between 3d-var, dfi and blending
Combination with surface analysis or
surface blending
Moving to 3d-FGAT
Maintenance of a reference version
|
Frequency of 3d-Var
|
Evaluation through
sensitivity studies
|
DATA ASSIMILATION : OBSERVATIONS
Neglected along the last year but more and more essential
The first step, not explicitly mentioned here, is of
course the implementation of local databases.
ODB
|
Maintenance and documentation
Development of new tools
|
Satellite data
|
IASI / AIRS: Improved
description of surface emissivity
Raw ATOVS data : use of
local data
Cloudy ATOVS data :
observation operator, Jb
GPS
SSM/I
Profiler data
|
Surface observations
for upperair analysis
|
From the less difficult or most
important ones to new ones: surface pressure, 2m-relative humidity, 10m-wind, ...
|
Aircraft data
|
Use of local data
|
Radar
|
Winds
Reflectivities
|
Pre-analysed data
(or pseudo-observations)
|
Pseudo-TEMP for relative
humidity : case studies
Pseudo-TEMP for relative
humidity : regular use
Surface data bogus
|
Screening
|
Evaluation for
high-density data
New data types
Time window
Time dimension
PBL fields
|
Space consistency of
the quality control
|
Combination with the use
of the CANARI quality control
Variational quality control
|
_________________________________________________________________________________________________
DATA ASSIMILATION : SURFACE
Surface analysis is used in two ways :
â
to initialize soil and surface variables
(temperature, moisture, snow characteristics, ...) : assimilation purpose
â
to provide an analysed state as close as
possible to all available observations : nowcasting purpose
The Optimal-Interpolation version will be
maintained for the next three years. The combination with upperair
assimilation (within the problems of scripts and coupling) is to be considered
at each main step.
Analysis of PBL
fields, for :
* Diag-Pack
* the correction of
soil fields
|
Retuning of statistics
(forecast and observation errors)
Geographical dependent
error statistics (orography, coasts, ...)
Analysis of new fields
(precipitations, visibility, cloudiness, ...)
|
SST analysis
|
Retuning
Use of pre-processed
satellite data
|
Snow analysis
|
Retuning of statistics,
for large and small scales
Estimation of the vertical
correlations for errors on snow depth
Calculation and use of a
snow mask derived from satellite data
Use of pseudo-observations
from local networks
Analysis / correction of
new fields (albedo)
Improved climatological fields
|
Assimilation of
soil moisture and temperature
|
Reduction of the
horizontal heterogeneity of soil moisture
Retuning and
implementation in ALADIN
Combination with dfi-blending
Combination with 3d-var
Moving to a variational assimilation
Use of satellite data
Improved climatological fields
|
Diag-Pack
|
Improvements in
observations operators (vertical interpolations)
Use of aircraft, profiler,
radar-wind data
Diagnostics fields
(smoothing, new ones)
|
Quality control
|
Retuning screening for
surface observations
|
PREDICTABILITY
Ensemble Prediction Systems (EPS) for
Limited-Area Models is an emerging topic.
A large range of solutions may be considered,
e.g. :
· Initial state
uncertainties sampling, using :
¨ singular vectors,
¨ breeding
methods, based on perturbations on the assimilation cycle, ...
· Model
uncertainties sampling, using :
¨ a stochastic physics,
¨ several models,
¨ changes in
namelist parameters, within a reasonable range of variation, ...
· Perturbed
lateral boundaries, using :
¨ several
coupling models,
¨ forecasts from
a global EPS, ...
· Several LAM
with their own lateral boundary conditions, ...
An EPS project focusing on small scales but
using ARPEGE has just started.
There is a general agreement on the necessity to
investigate EPS with ALADIN, but with a low priority, focusing on :
· defining the
aim(s) or requirements,
· keeping a watch,
· performing
preliminary case studies (for instance in the framework of a PhD ?)
Note : the cost of EPS is not negligible at all !
|