Occasion What do we gain from the regional reanalysis products - a perspektive from multivariate analysis of PBL Profiles Ronny Petrik , Beate Geyer, Burkhardt Rockel (Helmholtz-Zentrum Geesthacht) 17.7.2018 / Bonn Many thanks to the data providers at CESAR, at Uni Hamburg (Ingo Lange, Felix Ament) and at MOL Lindenberg (Udo Rummel, Frank Beyrich) Many thanks to comments from UKMetOffice (Jemma Davie)
2
3
Regional hindcasts and reanalyses • regional reanalysis from the UERRA Projekt: SMHI (Harmonie) and UK-Met Office (UKMO) REA6/UERRA* • COSMO-REA6 provided by University of Bonn (Bollmeyer et al., 2015) • Realistic hindcasts of HZG using Large-Scale Constraint and pure regional downscaling oF-CCLM oF-CCLM COSMO-REA6 UERRA-SMHI UERRA-UKMO Helmholtz-Zentrum Deutscher Wetter- SMHI UKMet-Office Geesthacht dienst / Uni. Bonn Large-Scale constraint Continuous nudging OI / 3D-VAR 4D-VAR / pures Downscaling / OI MERRA2 / ERA inter ERA inter ERA inter - 0.06 ° 0.06 ° 0.11 ° 0.11 ° 4
Impact of assimilation / nudging techniques (Low Christian, October 2013) Contineous Nudging Large-Scale const. Pure downscaling MIKLIP_DECREG Large-Scale const (HD) 1.8 3.8 5.8 7.8 9.8 11.8 13.8 15.8 Sting jet dynamics: high level winds transported downwards within strong downdrafts (radio sounding Emden: 160 km/h at 500 m above ground) 5
October 2013 – Performance of 10m wind MAE, Benelux Cont. Nudg. (REA6) Pure down. (oF-CCLM) Large-scale HD (oF-CCLM) Large-scale (oF-CCLM) → most positive impact by large scale constraints, further improvement due to HD and assimilation (contineuous nudging) 6
Evaluation of lower boundary layer • Long-term data of wind or temperature or humidity only for some locations • not everywhere a mast correction is available (e.g. MOL) • masts at Hamburg, MOL, FINO1 and Cabauw considered here Standorte von Grenzschichtmasten (@googlemaps) @MOL (DWD) @Lange (UHH) 7
HH-Billwerder, Daily cycle of wind speed (2007) Large-Scales, oF-CCLM Cont. Nudg., REA6 • Daytime increase at near ground levels • Daytime decrease at elevated levels 10m-data not combined … … with 3D-data 8
Falkenberg, Daily cycle of wind speed (2007) Cont. Nudg., REA6 Pure Downscaling, oF-CCLM Large Scales, oF-CCLM 9
Lindenberg wind profile – error measures (2007) Cont. Nudg. (REA6) UE-SMHI UE-UKMO Pure Down. (oF-CCLM) MAE, wind speed Large-Scale (oF-CCLM) Cont. Nudg. (REA6) UE-SMHI Large-Scale (oF-CCLM) UE-UKMO Pure Down. (oF-CCLM) 10
HH-Billwerder, daily cycle of stability Number of hours with specific stability classes (2006-2007) Stratification (10m to 110m): very stable (dT/dz > 0.0 K/m), stable (-0.009 < dT/dz < 0.0 K/m), neutral (-0.011<dT/dz<0.009 K/m), unstable (dT/dz < -0.011 K/m) 11
HH-Billwerder, T-Gradient + wind profile (2006-2007) lnstabil Stabil Cont. Nudg. (REA6) measurement UE-SMHI UE-UKMO Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 12
HH-Billwerder, very stable stratification Sehr stabil Cont. Nudg. (REA6) measurement UE-SMHI UE-UKMO Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 13
HH-Billwerder, very stable stratification Hamburg Falkenberg Cabauw Cont. Nudg. (REA6) measurement UE-SMHI UE-UKMO Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 14
HH-Billwerder, distribution T-gradient at ~100m Falkenberg Cabauw Cont. Nudg. (REA6) Measurement (obsolete) UE-SMHI UE-UKMO Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 15
Turbulence intensity, HH-Billwerder (2006) HH-Billwerder Large Scales, oF-Merra Pure Down. (oF-CCLM) Large-Scales (oF-CCLM)
Temporal consistency, HH-Billwerder (2006) Cont. Nudg., REA6 UERRA-UKMO Large-Scales oF-CCLM Cont. Nudg. (REA6) UE-SMHI UE-UKMO Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 17
Resolving power up-ramps, Falkenberg (2006-2007) Cont. Nudg. (REA6) Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) UERRA-SMHI UERRA-UKMO 18
Conclusion 4D-Var/3D-Var Kontinuierl. Large-Scale Pures Down- (UKMO / SMHI) Nudging (REA-6) Constraint scaling 10m wind ++ + .. ++ o „PBL“ daily cycle o + + + „PBL“ profiles + + .. ++ + o „PBL“ temporal ++ (4D-Var); + .. ++ + - consistency. + (3D-Var) „PBL“ dT/dz ++ (SMHI, vs); + + + + (UKMO) „PBL“ turb. variab. + .. ++ ++ • Simulations using COSMO offer sharing of error characteristics: e.g. profiles during very strong stratitifcations -> best way to distinguish between model families / parameterizations • Runs without any assimilation / nudging clearly seperate from others, and runs with large-scale constraint are much closer to „ true “ reanalysis • Radiation errors within COSMO family call for postprocessing or resimulation 19
Season of maximum solar elevation angle (2006-2007) Cont. Nudg. (REA6) SARAH2 Aerocom v2 Aerosols (oF-CCLM-AC2) Pure Down. (oF-CCLM) Large-Scale (oF-CCLM) 20
Future • openFRED: database for open source energy system modelling • 1. step: fill the database with data from the COSMO-REA6 analysis (mostly best way of error correction, e.g. 10m wind) and with data from realistic hindcast (oF-CCLM) to fullfill the needs of users‘s • 2. step: dependent on application usage of ensemble spread for e.g. cloudy situations or temporally most-consistent member (UERRRA-UKMO) • hope to consider parameterizations/diagnostics and ideas for model development coming out of our investigation for the next reanalysis runs • hope to catch some data missing on the UERRA-archives Planetary boudary layers seen by most recent regional reanalysis (in prep.) Christopher Frank (tomorrow) 21
ronny.petrik@hzg.de Danke für Ihre Aufmerksamkeit Christian reaches FINO1 – very fast Measurement coastdatIII Cont. Nudg. (REA6) Large-Scale (oF-CCLM) Pure Down. (oF-CCLM) 22
Additionals 23
excurse “spectral nudging” • technically a nudging term is Practical point of view added in the „ spectral domain “ of specific prognostic variables • dependent on height (not altitude) • wave number truncation, i.e. „ free “ development of small scales) smaller scales Taken from Feser (2005) 24
HH-Billwerder vs. Falkenberg, Schichtungen Zum Vergleich: Lindenberg 25
Daily Cycle, LI-Temp-2006, BIAS
- SAFRAN surface analysis is reducing the error snow cover in the Alps and in France - Members of SURFEX-ENS show a uncertainty for snow cover in the alps - for regions at 3000 meters there is a clear underestimation of snow depth, but the lowest for SAFRAN - SURFEX guy mentioned that some measurements lead to problems for assimilation systems 27
Recommend
More recommend
