1. PP CONSENS Merging COSMO-LEPS and COSMO- SREPS for the short-range Chiara Marsigli, Tiziana Paccagnella, Andrea Montani ARPA-SIMC, Bologna, Italy

2. Aim of the work Compare COSMO-SREPS with COSMO-LEPS for the short-range prediction of precipitation Decide if and how merge the two ensemble systems From the scientific point of view: assess the validity of the multi-model approach to provide initial and boundary conditions with respect to using a single model ensemble (EPS)

3. COSMO-LEPS Downscaling of some selected members of the ECMWF EPS with the LAM COSMO Selection is made using a Cluster Analysis and Representative Member selection technique Perturbations of the model physics parameters are applied to the COSMO runs Operational since November 2002 12 UTC 7 km, 40 levels 16 members 132 h

4. COSMO-SREPS multi-analysis multi-boundary approach IC and BC are provided by 3 operational deterministic global models (IFS, GME, GFS) Perturbations of the model physics parameters are applied to the COSMO runs Running regularly for testing, last set-up since November 2010 00 and 12 UTC 7 km, 40 levels 16 members 48 h

5. Ensemble mixing m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12 m13 m14 m15 m16 IFS GME GFS IFS (ctrl) m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12 m13 m14 m15 m16 m13 m14 m15 m16 m1 m6 m11 m16 m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12 m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12 m1 m6 m11 m16 COSMO-SREPSCOSMO-LEPSMIX16 MIX20

6. Verification method Winter 2010/2011 (20 Nov 2010 - 28 Feb 2011) Northern Italy high-density network average precipitation over boxes

7. Comparison among ensemble generation techniques Average precipitation on 0.5 x 0.5 deg boxes > 1mm/6h BSSROC 1.0 0.9 0.8 cleps csreps mix 20 mix 16 612182436424830 forecast range (h) 612182436424830 forecast range (h) fc. range (h)612182430364248 number of occurrences563502529618577465511584 0.5 0.4 0.3 0.2

8. Comparison among ensemble generation techniques Average precipitation on 0.5 x 0.5 deg boxes > 5mm/6h 612182436424830 forecast range (h) 612182436424830 forecast range (h) BSSROC 0.5 0.4 0.3 0.2 0.1 0.0 1.0 0.9 0.8 0.7 cleps csreps mix 20 mix 16 fc. range (h)612182430364248 number of occurrences107171151170136156134144 COSMO-LEPS better than COSMO-SREPS for the short-range Mixing has a positive impact

9. Comparison among ensemble generation techniques Average precipitation on 0.5 x 0.5 deg boxes > 1mm/6h 612182436424830 forecast range (h) 612182436424830 forecast range (h) BSSROC 0.5 0.4 0.3 0.2 0.1 0.0 1.0 0.9 0.8 0.7 cleps csreps mix 20 mix 16 fc. range (h)612182430364248 number of occurrences282213227232305225211219 Spring 2011

10. multi boundary effect Average precipitation on 0.5 x 0.5 deg boxes > 1mm/6h csreps 16 ifs 5 gme 5 gfs 5 mix 5 Use of multi-model boundaries has a positive impact even if models have different qualities 0.5 0.4 0.3 0.2 -0.1 0.0 -0.2 -0.3 -0.4 0.1 BSSROC 1.0 0.9 0.8 0.7 0.6 0.5 612182436424830 forecast range (h) 612182436424830 forecast range (h)

11. multi boundary effect Average precipitation on 0.5 x 0.5 deg boxes BS reliability 1mm/6h5mm/6h csreps 16 ifs 5 gme 5 gfs 5 mix 5 612182436424830 forecast range (h) 612182436424830 forecast range (h) 0.05 0.04 0.03 0.02 0.01 0.00 0.05 0.04 0.03 0.02 0.01 0.00

12. boundary diversity wrt population Average precipitation on 0.5 x 0.5 deg boxes > 1mm/6h cleps 5 csreps 5 With multi-model boundaries several models are needed to get a performance similar (or better) to a downscaling from a well constructed ensemble (like EPS) BSSROC 612182436424830 forecast range (h) 612182436424830 forecast range (h) 1.0 0.9 0.8 0.7 0.5 0.4 0.3 0.2 0.1 0.0

13. boundary diversity wrt population Average precipitation on 0.5 x 0.5 deg boxes > 5mm/6h cleps 5 csreps 5 BSSROC 612182436424830 forecast range (h) 612182436424830 forecast range (h) 1.0 0.9 0.8 0.7 0.5 0.4 0.3 0.2 0.1 0.0

14. boundary diversity wrt population Average precipitation on 0.5 x 0.5 deg boxes > 1mm/6h cleps 3 csreps 3 BSSROC 612182436424830 forecast range (h) 612182436424830 forecast range (h) 1.0 0.9 0.8 0.7 0.5 0.4 0.3 0.2 0.1 0.0

15. boundary diversity wrt population Average precipitation on 0.5 x 0.5 deg boxes > 5mm/6h cleps 3 csreps 3 BSSROC 612182436424830 forecast range (h) 612182436424830 forecast range (h) 1.0 0.9 0.8 0.7 0.5 0.4 0.3 0.2 0.1 0.0

16. Conclusions Generally COSMO-LEPS outperforms COSMO-SREPS The multi-model approach for i.c. and b.c. proves valuable even if model with different qualities are used For the multi-model approach to be effective, several models are needed to get a performance similar (or better) to a downscaling from a well constructed ensemble (like EPS) Combining the 16 COSMO-LEPS members with 4 COSMO- SREPS members this 20-member ensemble outperforms both systems Combining 12 COSMO-LEPS members with 4 COSMO-SREPS members, this 16-member ensemble outperforms COSMO- LEPS for the first day (from 18 to 30 h)

17. Proposal for COSMO Stop running the extra COSMO-SREPS members (nested on the same sets of IC and BC but with different physics) The COSMO-SREPS members which receive i.c. and b.c. by different global models are anyway available, being part of the BC-EPS suite run by DWD at ECMWF (for COSMO-DE-EPS) Create a new COSMO-SREPS system? –by merging the 16 COSMO-LEPS runs with the 4 BC-EPS runs, creating an additional product for the short-range with no extra BUs required WG7 recommendation: –Not easy to use one system for medium-range and one for short- range only –Consider the possibility to enhance COSMO-LEPS for the short-range by adding few multi-model driven COSMO members for the whole forecast range (run as part of the COSMO-LEPS suite)

18. COSMO-SREPS suite set-up convection scheme: 0 Tiedtke 1 Kain-Fritsch maximal turbulent length scale length scale of thermal surface patterns scaling factor of the laminar layer depth ratio of laminar scaling factors for heat over sea minimal stomata resistance