Demands and expectations at SMHI on the European Reanalysis for observations and climate Per Und én Tomas Landelius SMHI
EURO4M WP1 observational datasets WP2 reanalysis and evaluation WP 3 evaluation WP 4 management and coordination WP 2.1 4D-VAR developments and radar precipitation, a year of data – MO WP 2.2 3D-VAR downscaling, most of 20 years, 25 km SMHI WP 2.3 MESAN and SAFRAN downscaling, 12-4 km, MF and SMHI
What SMHI expects from EURO4M Dynamical downscaling of ERA data using HIRLAM 3D/4D-VAR – Consistent data set for ~ 20 years or more Access to observations additional to the ECMWF archive 2D high resolution downscaling of HIRLAM – Using these auxiliary observations – Driven by consistent HIRLAM model fields – Consistent data set ~ 20 years or more
Expect project members to share data
User requirements at SMHI onwards Every 3:rd hour 5 km (0.05°) Parameters for: Evaluation of climate change models Atmospheric environment models Oceanographical models Wind energy studies Hydrological models Surface radiation models Observation monitoring and replacement
SMHI KOAKK 40 years for QC of observations for climate National archives of climate data have discrepancies Need to be re-checked – Corrections when necessary and possible EERA-40 ? >SMHI KOAKK 125 km -> HIRLAM reanalysis 22 km ->11 km? N Europe ? MESAN downscaling – at 11 km
Data-assimilation system, model and analysis, unchanged through the period Analysis product quality improves in time Observing systems including SST/ ICE improves: Better quality, more data types, higher time frequency Reanalysis philosophy
Intermittent data assimilation 06 UTC 12 UTC 18 UTC (06 UTC(12 UTC3 h) (18 UTC tid
ITN 4/ Dimensional Variational Data Assimilation Iterative fitting of a Forecast trajectory to observations Over a time window of 6 hours
SMHI expertise and resources HIRLAM 3D and 4D-VAR Observation handling Re-analysis – ERA expertise – DAMOCLES coupled HIRLAM/HIROMB reanalysis Surface parameterisation Cloud parameterisation Radar and satellite data and algorithms HARMONIE (ALADIN) models and data assimilation 3D(4D) MESAN 2D-analysis OI with anisotrophic structure functions Observation processing including radar, precip, satellite and road stations Long operational experience ERA-MESAN
FoUp redov HIRLAM ALADIN High Resolution Limited Area Modelling Aire Limitee Adaption Dynamique InterNationale
Improved 2D reanalysis for Europe ERA-40 as first guess 1980 – , 06, 12, and 18 UTC 11 km (0.1°) ERA-MESAN
Workpackage 2.2 ERA-Interim downscaling 25 km ENSEMBLES area ? ECMWF observations conv AMVs? HIRLAM 3D-VAR 25 km Jk ((large-scale mix)) > HIRLAM 3D-VAR 11 km EU area MESAN downscaling 11 km T2m, Td, uv, prec, clouds 3D-VAR developments Jk MESAN/SAFRAN developments Snow/ orography etc Advanced features VARAN type structure functions Coupled surf-upper air 3D ? Validation KNMI/MO
11-4 km km ERA-40 / ERA Interim ECMWF HIRLAM MESAN 22 km
Signatur HIRLAM - Large Scale Mixing, LSM Reruns from ECMWF analysis, updating first guess Instead: Include ECMWF information in assimilation! Related work done with ALADIN at Météo-France
Signatur Vorticity, model state Short forecast, ECMWF Constrain Vorticity Begin as simple as possible: - Vorticity only - Univariate NMC statistics from ECMWF forecasts, interpolated to HIRLAM RCR geometry
EURO4M downscaling with HIRLAM 4D-Var possible areas (Per Kållberg, Per Dahlgren – SMHI) HIRLAM rotated lat-long coordinates S.P. at -35º/20º three resolutions: 0.2º, 0.15º and 0.1º
294*260 = points 0.2º*0.2º (27/-31/-24.7/27.5)
an example one day 4D-Var 0.15º*0.15º LBC from ERA_Interim 1 January Z 4D-Var and +12h fcst ~215 System Billing units on C1A one cycle took ~40 minutes (run on daytime Nov 17)
306x306 points 0.2 x 0.2 °
another example one 4D-Var 0.2º*0.2º – 2 outer loops LBC from ERA_Interim +12 h forecasts Analysis 20 mins Forecast 5 mins => 25 mins per cycle – Possible to run 1D / 2h – Or 12 days / day (but depending on queues etc) Special Project at ECMWF? Will apply....
65 levels inst of 60 och 10 m lowest mod lv inst of 30
Development of the MESAN 2D analysis Anisotropic structure functionsParameterized downscaling