1. Meso-γ 3D-Var Assimilation of Surface measurements : Impact on short-range high-resolution simulations Geneviève Jaubert, Ludovic Auger, Nathalie Colombon, Véronique Ducrocq, François Bouttier CNRM/GAME Météo-France and CNRS G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France

2. Introduction Feasibility study for: the assimilation part of the next operational NWP high-resolution system of Meteo-France AROME high-resolution reanalyses of the field phase of cooperative experiments Validation on the Gard torrential event (8-9 sept.2002) and the MAP IOP08 frontal case (20-21 oct.1999) G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France

3. Hybrid assimilation system First guess: large scale analysis (ECMWF or ARPEGE) Two domains: Mesoscale domain mesh 10 km Convective scale domain mesh: 2.5 km 1 analysis or 6-hour assimilation with 1 hour step Meso-NH configuration: Two nested models mesh: 10 and 2.5 km 3 ice types Explicit convection at 2.5 km Aladin 3D-Var Analysis Meso-NH model Increments Guess Assimilation cycle Meso-NH simulation G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France

4. Experimental design 06 UTC 12 UTC 12 UTC 24 hours Simulation REF12 ANA 12 UTC mesoscale analysis and 12 UTC convective scale analysis 12 UTC Large scale analysis ASSIM REF06 06 UTC Large scale analysis Assimilation step 1 hour 070810111209 06 UTC mesoscale analysis and 06 UTC convective scale analysis G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France

5. Data used in the analyses MAP IOP8 frontal case G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France Gard torrential event Surface data (3000 observations) surface pressure 2m temperature and humidity, 10m wind hourly measurements Radiosounding data (60 messages) Temperature, humidity, wind, geopotential 06 UTC and 12 UTC Mesoscale domain Convective scale domain

6. 2800 km The background error statistics Based on the background errors covariances used for the ALADIN analyses Mesoscale domain (mesh 10 km) Mesoscale domain (mesh 10 km): ALADIN statistics except at low levels: larger variances except for the rotational wind  geostrophic balance reduced at low levels G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France 720 km Convective scale domain (mesh 2.5 km) ALADIN horizontal length scales reduced Larger variances at low levels Vertical profiles of background errors stdev

7. METEOSAT IR 20020908 18 UTC ASSIM REF12 ANA DATA Gard torrential rain event 12h precipitation statistics T+0  T+12 Simulation REF12 ANA ASSIM Correlation 0.20 0.62 0.13 Bias -3.80 -4.81 -2.31 G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France 12h accumulated rain Convective scale domain 8 sep 2002 12 UTC  9 sep 2002 00 UTC T+0  T+12

8. ASSIM REF12 ANA DATA Gard torrential rain event 12h precipitation statistics T+0  T+12 Simulation REF12 ANA ASSIM ETS 0.1 mm 20 mm 0.14 0.18 0.16 0.31 0.12 0.19 Correlation 0.20 0.62 0.13 Bias -3.80 -4.81 -2.31 G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France 12h accumulated rain Convective scale domain 8 sep 2002 12 UTC  9 sep 2002 00 UTC T+0  T+12

9. MAP IOP8 frontal case METEOSAT WV 991021: 00 UTC 12h accumulated rain Convective scale domain 20 oct 18 UTC  21 oct 06 UTC T+6  T+18 DATA ASSIM REF12 ANA G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France 12h precipitation statistics T+6  T+18 Simulation REF06 REF12 ANA ASSIM Correlation 0.61 0.66 0.68 0.70 Bias 3.31 4.10 1.98 0.76

10. MAP IOP8 frontal case METEOSAT WV 991021: 00 UTC 12h accumulated rain Convective scale domain 20 oct 18 UTC  21 oct 06 UTC T+6  T+18 DATA ASSIM REF12 ANA G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France DATA ASSIM REF12 ANA ETS 0.1 mm 20 mm 0.66 0.30 0.65 0.34 0.65 0.35 0.63 0.36 12h precipitation statistics T+6  T+18 Simulation REF06 REF12 ANA ASSIM Correlation 0.61 0.66 0.68 0.70 Bias 3.31 4.10 1.98 0.76

11. Comparaison with the surface measurements (convective scale domain) MAP IOP8 frontal case G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France Gard torrential event Bias Stdev Bias Stdev --- REF12 --- ANA --- ASSIM 2 m Temperature --- REF12 --- ANA --- ASSIM Bias Stdev 2 m Humidity Bias Stdev

12. G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France MAP IOP8 frontal case Comparaison with the 13 radiosoundings included in the convective scale domain at 18 UTC (T + 6 hours) Mixing ratio (g/kg) Bias Stdev Temperature (K) Bias Stdev

13. Conclusion The assimilation of the surface measurements reduces the bias of temperature and humidity at low levels Improvement only if the background error covariance matrice is tuned to limit the contribution of the balanced part of the flow at low levels Better results if both the surface measurements and the altitude data are used G.Jaubert Météo-France WSN05 5-9 september 2005 Toulouse France The mesoscale assimilation improved:  the forecast of precipitation  the forecast of temperature in the first 3 km