1. AROME WMED, a real-time mesoscale model designed for the HyMeX Special Observation Periods N. Fourrié,, E. Bresson, M. Nuret, C. Jany, P. Brousseau, A. Doerenbecher, M. Kreitz, H. Bénichou, M. Amodei and F. Pouponneau M. Nadia.Fourrie@meteo.fr CNRM-GAME, Météo-France and CNRS Outline: –Characteristics of AROME WMED –Evaluation against AROME FRANCE –Reanalyses and data targeting system (DTS) impact –Conclusions and perspectives

2. 16-17 Septembre 2010AG GMAP 16-17 Septembre 2010 SOP2: Intense air-sea exchanges (severe winds, dense water formation) 1 Feb- 15 March 2013 SOP1: Heavy precipitation and flash-flooding 5 Sept-6 Nov 2012 2012-2013 : 2 field campaigns (SOP) in Northwestern Mediterranean HyMeX SOP1 and SOP2 More than 200 instruments deployedMore than 200 instruments deployed About 300 scientists on the fieldAbout 300 scientists on the field

3. Introduction AROME-WMED (west mediterranean) developped for the real-time analyses and forecasts during the SOPs of HyMeX and delivered to the Hymex Operational Centre each day a 48-hour forecast. Tool for some studies of atmospheric data assimilation task Based on the operational AROME-FRANCE model, horiz. resolution 2.5 km, 60 vertical levels from 10m to 1 hPa, time step 60s Domain: 34°N -9°W/ 48°N-20°E, AROME_WMED = +11% points vs AROME_FRANCE sea=47%, land=53%sea=41%, land=59% (nx:960,ny:640) (nx:750, ny:720) AROME WMED AROME FRANCE

4. AROME-WMED: Description Atmospheric initial conditions : 3D-Var with an assimilation window of 3h Background error covariance matrix computed with DA ensemble (Brousseau et al, 2011) over the rainy period from 2010-09-10 to 2010-10-14 (analysis times 00 et 12UTC). Lateral boundary conditions provided by the french global model ARPEGE Pronostic variables: T,q, vort,div, Ps in the analyses, vert div, pressure departure from hydrostatism, turbulent kinetic energy, Ql,Qi,Qr,Qs,Qg 48h forecast range

5. AROME-WMED: Data usage Assimilated observations Conventional data : surface data, wind profilers,radiosondes Ground-based GPS stations Satellite radiances from geostationnary and polar-orbiting satellites Wind derived from satellite imagery, Ascat winds Radar doppler winds and reflectivities Assimilation of additional Spanish surface observations Surface initial conditions: surface analysis (T and water content of the 2 reservoirs of the surface scheme) with optimal interpolation every 3h from T2m and Hu2m. SST: OSTIA analysis from UKMO

6. Evaluation of AROME-WMED during SOP1 24-h accumulated precipitation from raingauges data available for the 2 month period of SOP. Equitable threat scores for AROME- WMED and AROME-France similar. Same pattern for the 24-48 hour accumulated precipitation but the ETS score is lower. ETS Obs. number used for the score Number of data per station available during SOP1

7. Evaluation: temperature and Humidity at 2 m temperature For T2m :Bias and standard deviation larger for AROME-WMED, diurnal cycle. For Hu2m: Standard deviation similar, bias larger Statistics slightly larger for the 24-48h forecast range. humidity

8. IOP 8: 28 September 2012 Heavy precipitation in Southern Spain with strong damages Andalucia, Murcia and Valencia areas. Simulation of the 3 maxima (even if not well located) at the 48-hour forecast range. More than 150mm/24h simulated. 48h-forecast24h-forecast 24h-precipitation observations

9. First AROME-WMED reanalysis Period : SOP1 (05/09/2012  05/11/2013) = 2 months / +54hr range (+48) unique software version : observations new-extraction (cut-off) implies more data available coupling: ARPEGE « long » cut-off (instead of ARPEGE « short » version  more observations in the coupling flux) assimilation of some experimental mobile soundings not assimilated in real-time (not assimilated before 10/10 in « oper ») activation of several wind profilers Boundary Layer Pressurized balloons (Alex Dorenbecher et al.), humidity data (Q) assimilated, ( wind and T already used in « oper ») more data assimilated, but overall ~ neutral impact in the forecast quality Available in the HyMeX database

10. Impact of Data Targeting System (DTS) observations Objective: to improve short-range forecast of high impact weather events in the mediterranean area (heavy precipitation,…) From 11 September 2012 to 30 October 2012: 53 cases, 403 additional observations supported by EUCOS launched at 06UTC and 18UTC for selected days (373 additional radiosoundings in the AROME-WMED domain from 2012- 09-11 18UTC to 2012-10-30 06UTC) = DTS soundings Blue: operational network / Orange: HyMeX research mobile radiosounding (CNRM) Red: HyMeX research radiosounding (KIT) / xx: additional DTS soundings

11. DTS radiosonde impact: Results BS_NO (54km) BS_NO=1-BSS_NO, threshold 0.2mm/24hr (rain / norain separation) September 2012 October 2012 IOP8 Good Bad DTS data represents 10 to15% of the radiosonde data flux in the experiment, about 0.7 % of the assimilated data a neutral to very weak positive impact of the extra DTS sounding on short term forecast of precipitation Impact on 24hr cumulated rainfall HYMEX vs HYMEX_NODTS (48 simulations from 20120912 to 20121030) 00UTC runs (fcst range 30-06)

12. Towards a new AROME-WMED reanalysis of the HyMeX SOP1 (5 September 2012-5 November 2012) New background error matrix, computed over a 2 week period of October 2012. Improved orography (same as AROME_France oper. which is new in cy38) enhanced observations flux vs the first reanalysis AROME-WMED:  high resolution radiosounding (France, Espagne, RS/4M, KIT, L’Aquilla, Biscarosse)  dropsondes F20  reprocessed GPS and biais correction scheme adapted  reprocessed wind profilers  ground-based lidar (Candillargues, Minorque) + airborne  research aircraft measurements (F20, ATR, Dornier)  spanish radar  surface stations (Candillargues, …)

13. Conclusions and future works  Similar performance of AROME WMED and AROME France on the common area in terms of routine scores for SOP1 and SOP2. Feeding of the Hymex database in near-real time. both models were similar over the common area of their domain The (almost) neutral impact of additional radiosondes could be partly explained:  by the relative small size of the extra DTS sounding dataset (when compared to the whole size of all data assimilated)  by the non-optimal use of radiosonde information in the present assimilation system (ballon drift, time lag not taken into account)  A reanalysis of HyMeX SOP1 available mid-2015. More informations can be found in the paper Fourrié et al 2014, AROME- WMED, a real-time mesoscale model designed for the HyMeX Special Observation Periods, submitted QJRMS.