Impact of ensemble perturbations provided by convectivescale ensemble
Impact of ensemble perturbations provided by convective-scale ensemble data assimilation in the COSMO-DE model Florian Harnisch 1, Christian Keil 2 1 Hans-Ertel-Centre 2 Meteorologisches for Weather Research, Data Assimilation, LMU München, Germany Institut, LMU München, Germany Special thanks to Hendrik Reich & Andreas Rhodin, DWD ISDA 2014, Feb 24 – 28, Munich 1
KENDA-COSMO Kilometer-Scale Ensemble Data Assimilation (KENDA) → Lokal Ensemble Transform Kalman Filter (LETKF) (Hunt el al. 2007) applied for the COSMO-DE model ensemble of COSMO-DE first-guess forecasts + set of observations → ensemble of analyses → ensemble of high-resolution initial conditions to directly(? ) initialise ensemble forecasts ISDA 2014, Feb 24 – 28, Munich 2
KENDA-COSMO: Inflation è LETKF: background error covariance matrix Pb is estimated from ensemble forecasts xb Problem: not all sources of forecast error are sampled in Pb → sampling errors due to limited ensemble size & model error → estimate of Pb will systematically underestimate variances Solution: Inflation of estimate of Pb to enhance the variance (1) multiplicative covariance inflation (adaptive / fixed) (2) relaxation-to-prior-perturbations / relaxation-to-prior-spread = (Zhang et al. 2004) ISDA 2014, Feb 24 – 28, Munich (Whitaker and Hamill, 2012) 3
Setup of experiments (1) 15 UTC 10 June - 00 UTC 12 June 2012: → 21 -h fc at 00 UTC 11 / 12 June (2) 06 UTC 18 June – 12 UTC 19 June 2012: → 21 -h fc at 12 UTC 18 June KENDA: - 3 -hourly LETKF data assimilation of conventional data - 3 -hourly analysis ensemble with 20 ensemble members - 20 member ECMWF EPS lateral boundary conditions (16 km) - No physics parametrization perturbations (PPP) - Multiplicative adaptive covariance inflation KENDAppp: including 10 physics parametrization perturbations (PPP) KENDArtpp: relaxation-to-prior-perturbation inflation (α = 0. 75 ) KENDArtps: relaxation-to-prior-spread inflation (α= 0. 95 ) KENDArtps 40: 40 ensemble members / relaxation-to-prior-spread ISDA 2014, Feb 24 – 28, Munich 4
KENDA covariance inflation, 12 UTC 11 June 2012 First-guess ensemble spread U-Wind (m s-1) Radar derived precipitation (mm/h) Analysis ensemble spread U-Wind (m s-1) Observation used in the LETKF data assimilation ISDA 2014, Feb 24 – 28, Munich
KENDA relaxation-to-prior-pert, 12 UTC 11 June 2012 First-guess ensemble spread U-Wind (m s-1) Radar derived precipitation (mm/h) Analysis ensemble spread U-Wind (m s-1) Observation used in the LETKF data assimilation ISDA 2014, Feb 24 – 28, Munich
Departure statistics for KENDA experiment N Obs Radiosonde temperature KENDArtps KENDA è Accuracy of the analysis ensemble mean (solid) compared to the first-guess (+3 h) ensemble mean (dashed) → relaxation method inflation ensemble = better accuracy ISDA 2014, Feb 24 – 28, Munich 7
Departure statistics for KENDA experiment N Obs Radiosonde temperature KENDArtps 40 è Accuracy of the analysis ensemble mean (solid) compared to the first-guess (+3 h) ensemble mean (dashed) → larger ensemble = better accuracy ISDA 2014, Feb 24 – 28, Munich 8
Ensemble mean error and ensemble spread +3 h forecast of U-Wind: KENDAppp KENDArtpp § Average over 11 cycles § Verification against COSMO-DE analysis § PPP increase the spread error § Relaxation methods lead to the largest spread (RMSE~SPREAD) ISDA 2014, Feb 24 – 28, Munich 9
Ensemble rank histogram Verified against COSMO-DE analysis (similar results OPER KENDArtps rank frequency KENDAppp frequency +3 h forecasts of 10 m wind speed KENDA against observations) ISDA 2014, Feb 24 – 28, Munich 10
Ensemble dispersion è Normalized variance difference (NVD): KENDA / OPER KENDA / KENDAppp KENDA / KENDArtps 1 -h prec KENDA / KENDArtps 40 1 -h prec NVD 1 -h prec var(eps 1) - var(eps 2) var(eps 1) + var(eps 2) average all cycles forecast steps (h) ISDA 2014, Feb 24 – 28, Munich 11
BSS: 3 -h ensemble forecasts of precipitation KENDAppp KENDArtps 40 OPER 06 UTC 18 June – 12 UTC 19 June 2012 KENDArtps OPER BSS 15 UTC 10 June – 00 UTC 12 June 2012 thresholds (mm / 3 h) è Brier Skill Score = [resolution – reliability] / uncertainty è Hard to beat COSMO-DE-EPS on up to 3 -h hours: LHN in analysis è Impact of model physics perturbations, inflation method and ensemble size ISDA 2014, Feb 24 – 28, Munich 12
BSS: 21 -h ensemble forecasts of precipitation 3 -21 h forecasts averaged over Germany KENDAppp KENDArtps OPER BSS KENDAppp KENDArtps OPER 00 UTC 11 June 2012 thresholds (mm / 3 h) 00 UTC 12 June 2012 thresholds (mm / 3 h) è Brier Skill Score = [resolution – reliability] / uncertainty è Accounting for model errors with PPP shows positive impact è Large impact of inflation procedure ISDA 2014, Feb 24 – 28, Munich 13
Summary è KENDA-COSMO ensemble of analyses → Consistent ICs for ensemble forecasts → ICPs are present at all scales / all levels from the beginning è Necessary to use inflation methods to account for unrepresented error sources: large impact of different methods è Physic parameter perturbations can only partially account for model error è Ensemble size matters (initialize 20 member FC from 40 member AN? ) ISDA 2014, Feb 24 – 28, Munich 14
- Slides: 14