The EnKF Analyses and Forecasts of The EnKF Analyses and Forecasts of the 8 May 2003 Oklahoma City Tornadic the 8 May 2003 Oklahoma City Tornadic Supercell Storm Supercell Storm By Nusrat Yussouf 1,2 Edward Mansell 2 , Louis Wicker 2 , Dustan Wheatley 1,2 , David Dowell 3 , Michael Coniglio 2 and David Stensrud 2 1. Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, OK. 2. NOAA/National Severe Storms Laboratory , Norman, OK. 3. NOAA/ESRL/Global Systems Laboratory, Boulder, CO.
19
Embed
The EnKF Analyses and Forecasts of the 8 May 2003 Oklahoma City Tornadic Supercell Storm
The EnKF Analyses and Forecasts of the 8 May 2003 Oklahoma City Tornadic Supercell Storm. By Nusrat Yussouf 1,2 Edward Mansell 2 , Louis Wicker 2 , Dustan Wheatley 1,2 , David Dowell 3 , Michael Coniglio 2 and David Stensrud 2. - PowerPoint PPT Presentation
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
The EnKF Analyses and Forecasts of the The EnKF Analyses and Forecasts of the 8 May 2003 Oklahoma City Tornadic 8 May 2003 Oklahoma City Tornadic
Supercell StormSupercell Storm
By
Nusrat Yussouf1,2
Edward Mansell2, Louis Wicker2 , Dustan Wheatley1,2, David Dowell3,
Michael Coniglio2 and David Stensrud2
1. Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, OK. 2. NOAA/National Severe Storms Laboratory , Norman, OK. 3. NOAA/ESRL/Global Systems Laboratory, Boulder, CO.
MotivationMotivation Most storm-scale NWP modeling studies assume horizontally
homogenous environmental conditions Much easier to obtain a high-quality analysis of supercell storm than a
accurate forecast Stensrud and Gao (2010): Substantial improvement in storm
forecast accuracy when using realistic inhomogeneous mesoscale environment
This work focuses on ensemble data assimilation experiments of tornadic supercell within full mesoscale complexity
In support of Warn on Forecast - a numerical model-based probabilistic convective-scale analysis and forecast system to support warning operations within NOAA
A very short-range probabilistic forecasts of tornadic supercell storms
The 8 May 2003 Oklahoma City Tornadic The 8 May 2003 Oklahoma City Tornadic SupercellSupercell
KOUN Radar Observations at 22:10 UTC
NWS Damage Path of OKC Tornado
HPC Synoptic Scale Surface Analyses at 18:00 UTC
Hu and Xue (2007)
Mesoscale EnsembleMesoscale Ensemble
• WRF-ARW v3.2.1 Mesoscale data assimilation on CONUS domain 18-km horizontal grid spacing; 51 vertical levels Mean initial and boundary conditions from GFS final analysis
• 45 member mesoscale ensemble IC/BC perturbations from WRF-Var (Torn et al. 2006) Physics Options: - Cumulus: Kain-Fritsch - PBL: MYJ - Microphysics: Thompson - Shortwave Radiation: Dudhia - Longwave Radiation:RRTM - Land Surface: Noah
• Ensemble Adjusted Kalman Filter (EAKF) approach from the Data Assimilation Research Testbed (DART)
Mesoscale Data Assimilation Mesoscale Data Assimilation Observations assimilated: - Altimeter setting (p) - Temperature (T) - Dewpoint (Td) - Horizontal winds (u and v) Observation platforms: - METAR, Radiosonde, Maritime and Automated Aircraft from MADIS Adaptive prior inflation & localization (1600 obs) Localization half width: 287/4 km for horizontal/vertical Filter configuration adapted from Glen Romine’s system at NCAR
Timeline of mesoscale data assimilation experiment: - Continous cycling for 3 days
- Every 6 hour DA: 18 UTC May 5 – 12 UTC May 8 - Every 1 hour DA: 13 UTC May 8 - 0 UTC May 9
• A 45 member storm-scale ensemble One-way nested down from mesoscale ensemble analyses at 21Z, May 8 2-km horizontal grid spacing , 450 x 360 km wide, 50 vertical levels KTLX WSR-88D radar doppler velocity (Vr) and reflectivity (dBZ) Radar data objectively analyzed to 4-km grid using OPAWS Both adaptive inflation and additive noise to maintain spread Adaptive localization (2000 obs) Observation errors: Vr = 2 m s-1, Z = 5 dBZ Localization half-width: 12/6 km for horizontal/vertical
• Timeline of storm-scale data assimilation experiment: - One hour DA every 3 minutes: 21 UTC – 22 UTC, May 8 - One hour ensemble forecast: 22 UTC - 23 UTC, May 8
Storm-Scale Data AssimilationStorm-Scale Data Assimilation
Storm-Scale Data Assimilation Storm-Scale Data Assimilation Experimental DesignExperimental Design
Three ensemble DA experiments using different bulk microphysics schemes:
- Thompson 1.5 moment (Thompson et al. 2004, 2008) Mixing ratio: Qc, Qi, Qs, Qr and Qg Number Concentrations: ice (Ni) and rain(Nr) - NSSL Variable Density Double Moment (NVD-DM; Mansell et al. 2010) Mixing ratio: Qc, Qi, Qs, Qr, Qg and Qh Number Concentrations: Nc, Nr, Ni, Ns, Ng and Nh - NSSL Fixed Density Single Moment (NFD-SM; Gilmore et al. 2004) Mixing ratio: Qc, Qi, Qs, Qr and Qg
Remaining physics options are identical to mesoscale ensemble
Observation-Space Diagnostics: rmsi and total ensemble spreadObservation-Space Diagnostics: rmsi and total ensemble spread
Vr statics are calculated at all observed values over the entire domain
Z statistics are calculated where observed Z > 10 dBZ
Ensemble spread for reflectivity is consistently smaller than the rmsi
Radial velocity ensemble spread is comparable to rmsi
Reflectivity rmsi from Thompson is relatively smaller during the later assimilation period
rmsi and spread are similar in magnitude for the 3 microphysics Scheme experiments for Vr
Observation-Space Diagnostics: Consistency ratioObservation-Space Diagnostics: Consistency ratio
Consistency ratio = (ens. variance + obs-error variance)
/ (mean-squared innovation)
Reflectivity consistency ratio is well below 1.0
Analyses at 2200 UTC at 1 km AGLAnalyses at 2200 UTC at 1 km AGL
Thompson NFD-SM NVD-DM
mesocyclone
Vorticity contours from 0.001 to 0.01 at 0.001 s-1
mesocyclonemesocyclone
KTLX Reflectivity Obs.
Member 12 Member 14Member 31
U-V Winds vector (m/s)
The areal extent and the reflectivity distribution in the forward flank region is closer to the observation in Thompson and NVD-DM scheme compared to NFD-SM.
Thompson NFD-SM NVD-DM KTLX Reflectivity Obs
15 m
in F
cst
at 2
215
UT
C45
min
Fcs
t at
224
5 U
TC
Reflectivity Forecast at 1 km AGLReflectivity Forecast at 1 km AGL
Member 12 Member 14Member 31
Member 12 Member 14Member 31
Ensemble Mean Coldpool Analyses and ForecastEnsemble Mean Coldpool Analyses and Forecast
1-hr Forecast Probability of Vorticity 1-hr Forecast Probability of Vorticity (2145-2245 UTC) after 45-min assimilation(2145-2245 UTC) after 45-min assimilation
≥ 0
.00
15
s-1 a
t 1
50
m A
GL
≥ 0
.00
3 s
-1 a
t 1
km
AG
L
Thompson NFD-SM NVD-DM
% Probability
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
45-min Forecast Probability of Vorticity 45-min Forecast Probability of Vorticity (2200-2245 UTC) after 1-hr assimilation(2200-2245 UTC) after 1-hr assimilation
≥ 0
.00
15
s-1 a
t 1
50
m A
GL
≥ 0
.00
3 s
-1 a
t 1
km
AG
L
Thompson NFD-SM NVD-DM
% Probability
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Observed damage track and times
~22:06
~22:38
Summary and Future workSummary and Future work
The results show promise for short-range, ensemble-based, storm-scale tornadic supercell forecasts initialized from EnKF analyses
The reflectivity structure of the supercell storm using a DM scheme compare better to the observations than that using a SM scheme
Storm-scale ensemble system can predict the track of the strongest rotation with some accuracy in 0-1 hour time frame
Future work:
Vary the microphysical parameters across the ensemble to improve spread
Use of higher resolution grid of 1 km or less
AcknowledgementAcknowledgement
Glen Romine and Nancy Collins for help with DARTKevin Manross for providing the edited radar data
Additional SlidesAdditional Slides
Reflectivity Analyses Reflectivity Analyses at 2200 UTCat 2200 UTC
Reflectivity ForecastsReflectivity Forecastsat 2230 UTCat 2230 UTC