Experiences with SMHI local ALARO DA suite LACE Data Assimilation Working Days, Budapest, 14-16 June, 2011 Magnus Lindskog, Ulf Andrae, Lisa Bengtsson, Lars Meuller, KarlIvar Ivarsson, Martin Ridal Outline • • • • • • Introduction SMHI local ALARO data assimilation set-up Results from pre-operational system SMHI data assimilation impact studies Some recent general HARMONIE data assimilation developments Conclusions and future plans for SMHI ALARO data assimilation PRE-OPERATIONAL HARMONIE DOMAINS 2011 AEMET DMI FMI KNMI Met Eirann met.no SMHI Veðurstofa HARMONIE DOMAINS AEMET DMI FMI KNMI Met Eirann met.no SMHI Veðurstofa 3dvar/can/oim NWP models at SMHI Operational HIRLAM C22/C11 (4D-VAR) HIRLAM E11 (3D-VAR) HIRLAM G05 (3D-VAR) Pre-Operational E05 (3D-VAR) ALARO (3D-VAR) Daily AROME (Downscaling) SMHI Pre-operational ALARO system General System Design SMHI HARMONIE 2010: • 35h1.3 • 5.5 km horisontal resolution • 60 vertical levels (HIRLAM definitions) • 3 hourly LBC from ECMWF fc • Forecast length: 0-36 h • ALARO with 2-L ISBA (not SURFEX) • Hydrostatic forecast model • Surface analysis and 3DVAR • IDFI SMHI HARMONIE 2011: • 36h1.3 • ALARO-0 physics with surfex scheme SMHI ALARO is run under mini-SMS system Background error statistics • Background error statistics from ensemble of downscaled ECMWF 6h forecasts (20060920-2061031, 00UTC) • REDNMC=0.6 • REDZONE=250 km • Background error statistics also derived also utilising ensemble DA (not used) • (Shiyu at DMI has derived background error statistics based on downscaling for different seasons and time of day (201001-201012)) Observation usage • SYNOP/SHIP (Z) • DRIBU (Z) • AIREP/AMDAR (u,v,T) • TEMP (u,v,T,q) • PILOT (u,v) • ATOVS AMSU-A (NOAA 18 and METOP) (Tb ch 6-10 and VarBC) Surface data assimilation (SYNOP T2m H2m observations over land) ECMWF SST, temperature over sea ice from surface temperature in boundary field, LST from FA file surface temperature climatology Spatialisation of screen level data Surface data assimilation (CANARI OI) OImain a b T T 1 b x x BH ( HBH R ) ( y Hx ) T1 T2m T2m T2 2 T H w T H 1 1 2 m 1 2 m T H w T H 2 2 2 m 2 2 m T1 , w1 T2 , w2 Scores for verification against observations April 2011 RMS/BIAS as function of forecast range ALARO Surface Pressure (hPa) E11 E05 T2m (K) E05(7.3) 10 m Wind speed (m/s) Scores for verification against observations April 2011 T2m BIAS (K) averaged over forecast lengths ALARO E05 E05(7.3) 00 UTC 12 UTC Scores for verification against observations April 2011 T (K) RMS/BIAS averaged over forecasts length as function of vertical level Wind speed (m/s) ALARO E11 E05 E05(7.3) RH (%) Scores for verification against observations February 2011 RMS/BIAS as function of forecast range ALARO Surface Pressure (hPa) E11 E05 T2m (K) E05(7.3) 10 m Wind speed (m/s) Scores for verification against observations February 2011 T2m BIAS (K) averaged over forecast lengths ALARO E05 E05(7.3) 00 UTC 12 12 UTC UTC Scores for verification against observations February 2011 T (K) RMS/BIAS averaged over forecasts length as function of vertical level Wind speed (m/s) ALARO E11 E05 E05(7.3) RH (%) Monitoring of satellite data and VarBC 06 h update 24 h update coverage 00 UTC 06 UTC 12 UTC A comparison of two off-line soil analysis schemes for assimilation of screen level observations (Mahfof et al., 2009) OI-equations 1/(2π) 1/(2) (ztiner in cactus.F90) Table of coefficients Conclusions Scores for verification against observations I month parallel exp, January 2010 RMS/BIAS T2M (K) as function o forecast range ALARO (1/2π) ALARO 1/2 T2m BIAS/RMS (K) averaged over forecast lengths ALARO (1/2π) ALARO 1/2 Experimental set-up Two parallel exp. for July & August 2009 and January & Febr. 2010: • 6 h intermittent data assimilation cycle • 3 h intermittent data assimilation cycle Lateral boundary conditions from 6 to 9 h old ECMWF forecasts and observations from ECMWF MARS archive Observation time window: • HH +/- 3 h for 6 h itermittent DA cycle • HH +/- 1.5 h for 3 h intermittent DA cycle (no modifications of error statistics or IDFI settings applied when modifying from 6 h to 3 h cycle) At 00 and 12 UTC 30 h forecasts were launched 20 Aug 2009 12 UTC sum6h mslpfc (black, hPa) sum6h-sum3h psdiff (red, conint 1 hPa) +6 h +18 h 20 August 2009 09 UTC (RUC 3h) sum3h mslpfc(black,hPa) 3h analysis incr.(red,conint 0.1hPa) Scores for verification against observations January, 2010, first 14 days (experiment for January and July ongoing) 4D-Var 3D-Var 14 Day RMS/BIAS time series 500 Temperature (K) 500 hPa Wind Speed (m/s) Timings 96 processor on SMHI Linux Clustre 3D-Var:~925 s 4D-Var: ~5500 s 30 h forecast: ~5200 s RADAR radial wind DA experiments Recent HARMONIE data assimilation developments • Technical problems with initialisation of snow in SURFEX solved (Trygve Aspelien). • Technical problems with syncronisation of snow between FA file and LFI file solved (Trygve Aspelien). • Data assimilation of with empty pools in ODB is enabled (Sami Saarinen) arp/obs_preprocs/readoba.F90 & odb/cma2odb/shuffle_odb.F90, export BASETIME=YYYYMMHHDD. • Spectral mixing of large scale information from first lateral boundary file enabled through LSMIXBC option (first step towards Jk-large scale constraint) (Ole Vignes). (xmix(m,n,l)=wbcxbc(m,n,l)+(1-wbc)xown(m,n,l)) Conclusions and Future Plans • • • • • • • • Verification scores indicate that the qualiy of SMHI ALARO upper air forecasts are at least as good as the quality of SMHI HIRLAM upper air forecasts. SMHI ALARO 10 m wind speeds too low and winter time 2mtemperatures too warm. SMHI system will be updated to next HARMONIE version (36h1.4) with recent developments for handling ODB empty pools, improved handling of snow and option for LSBCMIX (to be tested in SMHI system). Further experiments with modified surface data assimilation in OImain to be carried out, utilising updated version. AMSU-B, radar radial winds and later on ground based GPS to be introduced and evaluated (follow work by others on IASI brightness temperatures and radar reflectivities). Sensitivity studies to VarBC settings and wide extension zone. Potential application of EKF for surface data assimilation, Flake lake model, RUC and 4D-Var are in the longer term plans. Follow DMI work on varying structure functions. Close co-operation with met.no towards common operational HARMONIE system in 2014. Proposed domains for operational SMHI-met.no system in 2014 ~5.5 km hor res (1212*1360 gp), ~65 vert levels ~2.5 km hor res (1134*1720 gp), ~65-90 vert levels