ppt - COSMO model

New PP:
Consolidation of COSMO ensemble
(CONSENS)
Chiara Marsigli
ARPA-SIM, Bologna, Italy
Aim of the Project
Consolidating the ensemble forecasting systems for the
mesoscale built within COSMO in the past years
 COSMO-LEPS and COSMO-SREPS have been designed for
different forecast ranges (day 3-5 and 1-3, respectively) and with
different perturbation strategies
 The aim of the project is to build a unique “multi-perturbationstrategy” COSMO ensemble system, benefiting of perturbations
able to produce appropriate spread for the entire forecast range
(i.e., day 1-5)
 The ensemble should have a wide set of model perturbations to
guarantee a good description of the COSMO model error
 A calibration strategy should also be developed, to be applied
to the ensemble output
Actions (I)
1. COSMO-SREPS suite availability: COSMO-SREPS will be
maintained as a system running regularly, also providing initial
and boundary conditions to the COSMO-DE-EPS system
2. Model perturbations: Testing of perturbations of parameters of
the COSMO model physics will be carried on, analysing also the
impact of combining perturbations (positive impact?). It will be
studied a methodology to perturb also the model lower
boundary. This could be done by switching among fields
obtained with different methodologies or by applying
perturbations directly to the fields (define how!)
3. Ensemble merging: Define the “multi-perturbation-strategy” to
be used in the COSMO ensemble system, which perturbations
appropriate for both the short and medium range (from day 1
to 5) -> comparison of the multi-model approach with the
single-model ensemble or even multi-model ensemble ones for
the different forecast ranges
Actions (II)
4. Calibration: Development and testing of a calibration strategy
for the high-resolution precipitation output. The reforecasts run
by MeteoSwiss for the COSMO-LEPS system will be used for this
action. Different methodologies should be compared:
 CDF calibration using model climatology (following
Fundel et al., 2008), which is the technique already
applied by MeteoSwiss to COSMO-LEPS precipitation
forecasts
 standard bias correction as described in Hamill and
Whitaker (2006)
 logistic regression
 analog technique (Hamill and Whitaker, 2006)
collection of precipitation data on the whole domain is essential
to perform a complete ensemble calibration
Links
 Link with TIGGE-LAM and with the SRNWP Programme
Interoperability as regards the possibility of nesting the
COSMO model on other global models
 Link with model development for getting new tunable
parameters or meaningful combinations of the existing ones
in the parametrisation schemes, or even for obtaining new
alternative schemes
 PP COLOBOC for obtaining reasonably perturbed lower
boundary fields or techniques for perturbing them
 COSMO-DE-EPS
 PP KENDA
Main risks
 Lack of Billing Units for running at ECMWF. Since a robust
statistical analysis of the alternative options is needed in
order to decide about the future ensemble set-up, it is
necessary to run COSMO-SREPS for the whole period and
parallel runs on some periods (already accounted for in the
BU request)
 Difficulty of calibrating high-resolution precipitation
forecasts. Most of the works found in literature deal with
lower resolution systems. Furthermore, a dense and
homogeneous precipitation network providing data over a
very long period is needed for calibration and it seems
difficult to obtain this over the whole COSMO ensemble
area
People and FTEs
 2-year Project
 Estimated resources needed in total: 5.0 FTEs
 Estimated resources needed for 2009:
2.45 FTEs
 Estimated resources needed for 2010:
2.55 FTEs
 Institution/People involved:
 ARPA-SIM (Cesari, Diomede, Marsigli, Montani: 3.4 FTE)
 HNMS (Gofa, Louka: 1.4 FTE)
 MeteoSwiss (Fundel: 0.2 FTE)