WWOSC 2014 Assimilation of 3D radar reflectivity with an Ensemble Kalman Filter on a convection- permitting scale WWOSC 2014 Theresa Bick 1,2,* Silke Trömel 1,2 , Kathrin Wapler 1,3 , Clemens Simmer 2 1 Hans-Ertel-Centre for Weather Research, Atmospheric Dynamics and Predictability Branch 2 Meteorological Institute, University of Bonn, Germany 3 German Meteorological Service, Offenbach, Germany * [email protected]Special thanks to K. Stephan, Y. Zeng, R. Potthast, H. Reich, H. Lange August 18, 2014 1
26
Embed
WWOSC 2014 Assimilation of 3D radar reflectivity with an Ensemble Kalman Filter on a convection-permitting scale WWOSC 2014 Theresa Bick 1,2,* Silke Trömel.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
WWOSC 2014
Assimilation of 3D radar reflectivity with an Ensemble Kalman Filter on a convection-permitting scale
WWOSC 2014
Theresa Bick1,2,*
Silke Trömel1,2, Kathrin Wapler1,3, Clemens Simmer2
1Hans-Ertel-Centre for Weather Research, Atmospheric Dynamics and Predictability Branch2Meteorological Institute, University of Bonn, Germany3German Meteorological Service, Offenbach, Germany*[email protected]
Special thanks to K. Stephan, Y. Zeng, R. Potthast, H. Reich, H. Lange
August 18, 2014
1
WWOSC 2014
Motivation
Why radar data assimilation?
Highly resolved in space and time, dense coverage
3D information of convective systems
Improve short-term model forecasts of high impact weather events
Why ensembles?
No Ad/TL model or linearization necessary
Flow-dependent covariances
2
WWOSC 2014
COSMO-DE
Convection-permitting numerical weather prediction model
Δx = 2.8km, 50 vertical layers
Domain size: ~ 1200 km x 1300 km
KENDA
(Km-scale ensemble data assimilation):
Local Ensemble Transform Kalman Filter (LETKF) for COSMO-DE (Reich et al, 2011, following Hunt et al, 2007)
3
Radar forward operator
Observation
Radial velocity vr
Reflectivity Ze
Radar grid: range, azimuth + elevation
WWOSC 2014 4
COSMO-DE
Temperature T
Wind components U, V, W
Mixing Ratios QR, QS, QG, ..
COSMO-DE model grid
Radar forward operator derives pseudo radar volume scan from COSMO-DE model output (Zeng 2013)
No-reflectivity
Assimilate clear air information
Constrain all values to 5dBZ
Huge amount of data:
Superobbing: reduce observation density (cf. talk Y. Zeng)
Relaxation to prior spread: maintain ensemble spread after analysis (Harnisch and Keil 2014, submitted to MWR following Whitaker and Hamill, 2012)
Analysis does not deteriorate unobserved variables
During forecast….
Cells produced by analysis survive for several hours
FSS indicates clear benefit of radar reflectivity assimilation on small scales, slight disadvantage on larger scales
RMSE of unobserved variables evolves similarly to control run
WWOSC 2014 15
Outlook
Verification of more case studies
Analysis only yields improvement when ensemble spread is large
Additive noise to allow for non-linear development
Combination of radar reflectivity with other observation types:
Radial velocity
Polarimetric moments, inference on mixing ratios (need for 2 moment schemes?)
Cloud information to predict convective initiation
WWOSC 2014 16
WWOSC 2014
Assimilation of 3D radar reflectivity with an Ensemble Kalman Filter on a convection-permitting scale
WWOSC 2014
Theresa Bick1,2,*
Silke Trömel1,2, Kathrin Wapler1,3, Clemens Simmer2
1Hans-Ertel-Centre for Weather Research, Atmospheric Dynamics and Predictability Branch2Meteorological Institute, University of Bonn, Germany3German Meteorological Service, Offenbach, Germany*[email protected]
Special thanks to K. Stephan, Y. Zeng, R. Potthast, H. Reich, H. Lange
August 18, 2014
17
FSS for different scales. Threshold 10dBZ
WWOSC 2014 18
16UTC (1h fc) 21UTC (6h fc)
RUC CNTRL
FSS: forecast window. Threshold 10dBZ
WWOSC 2014 19
10km radius 20km radius 200km radius
RUC CNTRL
FSS for single grid points: 16UTC, scale = 200km
WWOSC 2014 20
FSS for single grid points: 21UTC, scale = 200km
WWOSC 2014 21
Ensemble spread after analysis
WWOSC 2014 22
How to treat errors in inital conditions? First trial.
Experiment RUC:
Reflectivity, no-reflectivity
Update every 15 minutes
Δx = 2.8km
WWOSC 2014 23
Experiment CNTRL:
Synop, temp, airep
Hourly update
Case study: June 6th 2011, radar station Essen
2h cycling of a single cell over North-Rhine Westphalia