COSMOSREPS Priority Project C Marsigli ARPASIM Hydro Meteorological

COSMO-SREPS Priority Project C. Marsigli ARPA-SIM - Hydro. Meteorological Service of Emilia-Romagna, Bologna, Italy

COSMO-SREPS methodology larger scale errors driving model error analysis error at the global model scale LAM ensemble limited-area model error: schemes, parameters, surface forcings smaller scale errors v i. c. and b. c. perturbations -> INM multi-model multiboundary (SREPS) v model perturbations -> parameter perturbation (fixed value throughout the integration to test the impact of the different parameter perturbations)

System applications v COSMO Priority Project -> Short-Range Ensemble v 1 D-Var DA of satellite data to provide retrievals to be nudged within 7 km COSMO-I 7: test of the use of COSMOSREPS to estimate a flow-dependent B matrix v COSMO-SREPS provides boundary conditions for EELMK, the ensemble system at 2. 8 km under development at DWD

Involved COSMO people v Marsigli, Montani, Paccagnella (ARPA-SIM): 2. 0 FTE v Flora Gofa, Petroula Louka (HNMS): verification, 0. 5 FTE Collaboration: v Susanne Theis, Christoph Gebhardt (DWD): EELMK v Uli Schaettler (DWD): COSMO model code

Task 0. Preliminary actions for the system set-up: completed in 2006. Task 1. Definition of the methodologies to perturb model trajectories: completed (set-up on the basis of the autumn 2006 results + test of al lot of parameters on the regular testing period (JJASON 2007) Task 2. Definition of the methodologies to perturb initial conditions: suspended Task 3. System set-up: completed (set-up for the JJASON 2007 period during the MAP D-PHASE DOP) Task 4. Test on case studies: on-going (21 runs in autumn 2006 done, analysis on-going, verification by HNMS on-going) Task 5. Test on a regular basis: on-going (running on the JJASON 2007 period during the MAP D-PHASE DOP) Task 6. Verification: on-going

IFS – ECMWF global COSMO at 25 km on IFS System set-up P 1: control (ope) by INM Spain GME – DWD global COSMO at 25 km on GME UM – UKMO global COSMO at 25 km on UM AVN – NCEP global COSMO at 25 km on NCEP P 2: conv. scheme (KF) P 3: tur_len=1000 P 4: pat_len=10000 16 COSMO runs 10 km hor. res. 40 vertical levels

intra-group distance 14 days (00 UTC runs) 2 m. T whole domain

intra-group distance 7 days (12 UTC runs) 2 m. T whole domain

Daily precipitation ROC area Northern Italy + Switzerland observations Average over 0. 5 x 0. 5 deg boxes - +42 h fc range

Daily precipitation ROC area Northern Italy + Switzerland observations Average over 0. 5 x 0. 5 deg boxes - +66 h fc range

Relationship between error and spread observations COSMO analyses 00 UTC runs 12 UTC runs

Relationship between error and spread Northern Italy observations Nearest grid point 12 UTC runs

Relationship between error and spread Northern Italy observations Nearest grid point 00 UTC runs

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 12 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 12 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 12 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 12 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 00 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h 00 UTC

Deterministic scores – ave 0. 5 x 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h

Conclusions v the initial and boundary condition perturbations contribute to the spread for an amount approximately double with respect to the parameter perturbations, statistically over the whole domain and the whole period in terms of 2 m. T. Locally (in space and time) the situation can be quite different v the selected parameters permit to have a detectable amount of spread among members with the same father v the rms error is grater that the rms spread for all the events (less than double if error is computed with respect to analysis -> without model bias)

On-going activities v run of COSMO-SREPS with model perturbations for the DOP MAP D-PHASE (6 months, June-November 2007) : v statistical analysis of the system v provide b. c. for EELMK (DWD) tests v test a flow-dependent B-matrix for 1 D-Var of satellite data v comparison with the other available mesoscale ensemble systems v test of different parameters (16 runs using only the IFS initial and boundary conditions) v verification - system analysis (HNMS)

Test of parameter perturbations (same father) IFS – ECMWF global LM at 25 km on IFS P 1: control (ope) P 2: conv. scheme (KF) by INM Spain P 3: parameter 1 P 4: parameter 2 P 5: … 16 LM runs at 10 km

Fine

Verification TP - N. Italy – 18 -42

Verification TP - N. Italy – 18 -42

Verification TP - N. Italy – 18 -42

Deterministic scores – max 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h

Deterministic scores – max 0. 5 IT+SV 1 mm/24 h 5 mm/24 h 10 mm/24 h