![Page 1: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/1.jpg)
Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data
L. Wicker1, C. Potvin1,2, T. Thompson1,2, D. Stensrud1, and P. Heinselman1 1NOAA National Severe Storms Lab, Norman OK
2CIMMS, University of Oklahoma, Norman OK
![Page 2: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/2.jpg)
20Z 20Z
22Z HRRR 5 HR FCST
![Page 3: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/3.jpg)
National Weather Radar Testbed
Multifunction Phased Array Radar (MPAR)
• 9.4 cm single-face phased array antenna
• Forms beam electronically using 4,352 transmit/receive elements
• Completes volumetric scans (of 90° sector) 5-7x faster than WSR-88D
• Beam width broadens from 1.5° at bore-site to 2.1°at a 45° angle
! See Forsyth et. al. 2005, Zrnic et.
al. 2007, Heinselman et. al. 2008
![Page 4: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/4.jpg)
Truth
PAR Analysis
WSR-88D Analysis
Benefits from MPAR data
after 15 min of
data assimilation
(OSSE) !
Yussouf and Stensrud
(MWR, 2010)
![Page 5: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/5.jpg)
Truth
PAR Analysis
WSR-88D Analysis
Benefits from MPAR data
after 15 min of
data assimilation
(OSSE) !
Yussouf and Stensrud
(MWR, 2010)
![Page 6: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/6.jpg)
Q: Using real MPAR data, will storm-scale NWP forecasts be improved using rapid-scan radar
data? !
Use the MPAR data set from the 24 May 2011 Radar at 20:38 UTC
Tornado B1: 20:38-20:46 UTC
Tornado B2: 20:50-22:35 UTC
![Page 7: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/7.jpg)
Technical Stuff• Model and ensemble:
• NCOMMAS cloud model, 160 km x 160 km 20 km (moving domain)
• 2 km horizontal resolution with 51 vertical levels (dz = 250 m below 2.5 km)
• Model initialized using 1 hour forecast sounding from operational RUC model
• Complex microphysics (2-moment scheme, 4-ice class [ice, snow, graup, hail])
• 48 member ensemble
• Initial perturbations added to the wind profile in each ensemble member
• DA System: 4D-LETKF (T. Thompson et al. 2014 QJRMS)
• 5 min cycles, +/- 2.5 min temporal window centered on time
• 9/4.5 km localization cutoff, no temporal localization used
• Inflation Methodology: Zhang (2004) relaxation to prior perturbation
• MPAR radar data thinned to 4 km grid on conical surfaces (Dowell et al 2009)
![Page 8: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/8.jpg)
Validation Challenge• Problem: How to verify these forecasts?
• use observed severe weather information • forecaster/warning perspective
• Avoid a “tuned” result! • One Experiment ==
• two temporal frequencies of radar data • 1 minute volumes (actual MPAR data collection) • 5 minute volumes (approximating the WSR-88D)
• 6 forecasts generated for experiment • after 10,15,20,30, and 40 minutes of data assimilation • forecasts to same time
• More than dozen experiments run
![Page 9: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/9.jpg)
Examine Two Questions…• Do 1 minute volumes spin up faster?
• Short assimilation cycles (10-20 min) • Will 5x more data make up for this relative to the conventional
radar data? • Do 1 minute volumes continue to add value later?
• For longer assimilation cycles (30 min) • Will 5x more data continue to improve the forecasts?
• One Experiment shown • Initialization at 19:40 with reflectivity-based perturbations • Ensemble “cooks” for 20 minutes, radar DA begins at 20:00 • Show 30 minute forecasts from 20:10, 20:20, 20:30
![Page 10: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/10.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
![Page 11: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/11.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
Sig rotation
![Page 12: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/12.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
Sig rotation Obs 40 dbz
![Page 13: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/13.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
Sig rotation Obs 40 dbz20% Prob
![Page 14: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/14.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
Sig rotation Obs 40 dbz20% Prob
Ens Mean 40 dbz
![Page 15: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/15.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
1 km Reflectivity Forecast valid @ 20:30 UTC
88D 3
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
![Page 16: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/16.jpg)
0-1.5 km Mean Layer Rotation Forecast 20:10-20:50 UTC
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
1 km Reflectivity Forecast valid @ 20:30 UTC
MPAR 11
Volumes
After 10 min of radar DA: Does ensemble using 1 minute data spin-up quicker?
![Page 17: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/17.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
0-1.5 km Mean Layer Rotation Forecast 20:20-20:50 UTC
1 km Reflectivity Forecast valid @ 20:40 UTC
After 20 min of radar DA
88D 5
Volumes
![Page 18: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/18.jpg)
0-1.5 km Mean Layer Rotation Forecast 20:20-20:50 UTC
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
1 km Reflectivity Forecast valid @ 20:40 UTC
After 20 min of radar DA
MPAR 21
Volumes
![Page 19: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/19.jpg)
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
88D 7
Volumes
After 30 min of radar DA: Do 1 minute volumes still improve forecast?
0-1.5 km Mean Layer Rotation Forecast 20:30-21:00 UTC
1 km Reflectivity Forecast valid @ 20:50 UTC
![Page 20: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/20.jpg)
0-1.5 km Mean Layer Rotation Forecast 20:30-21:00 UTC
Shaded regions: Prob(dBZ > 40) Solid blue line: Observed 40 dBZ.
Thick dashed line: Mean ENS 40 dBZ Thin dashed line: Prob(40 dBZ) > 20%
Shaded regions: Prob( > 7.5x10-3 s-1)
Blue dots: members where > 1.5x10-2 s-1
1 km Reflectivity Forecast valid @ 20:50 UTC
MPAR 31
Volumes
After 30 min of radar DA: Do 1 minute volumes still improve forecast?
![Page 21: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/21.jpg)
Summary• Results are similar to Yussouf & Stensrud’s OSSE
• 1 minute volumes spin storm up faster
• 1 minute volumes continue to improve forecasts even after 30 minutes
• Other experiments (not shown) are consistent with these results (different superob density, different initialization periods, KTLX radar DA for 20 min before 20:00, etc.)
• Assimilation of 1 minute data synchronously (e.g., every minute) is very sensitive to data density and other assimilation parameters. Inconsistent results are found.
• Storm-scale NWP will likely greatly benefit from a next-generation radar network having rapid-scan capability.
![Page 22: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/22.jpg)
4D Asynchronous LETKF 1) run 5 min forecast 2) assimilate radar data 3) run next 5 min forecast
Each Volume = 0.5, 0.9, 1.3, 1.8, 2.4, 3.1, 4.0, 5.1, 6.4, 8.0 degrees
VolumeVolume
1 min volumes
Volume Volume VolumeVolume
20:03 20:04 20:05 20:06 20:07 20:08Time (UTC):
Each Volume (10 tilts) distributed over 5 min (~VCP-212)
0.5, 0.9 degrees
5 min volumes
3.1, 4.0 degrees
5.1, 6.4, 8.0 degrees
0.5, 0.9 degrees
1.3, 1.8, 2.4 degrees
5.1, 6.4, 8.0 degrees
-2.5 min +2.5 min5 min DA Window
20:03 20:04 20:05 20:06 20:07 20:08Time (UTC):
Radar Data
ONLY PAR data is used to examine impact from rapid-scan capability
1.3, 1.8, 2.4 degrees
![Page 23: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/23.jpg)
Background Mesoscale Environment• Assumption: In tornado outbreaks the warm sector is more homogeneous
than in other situations? (else you would not have an outbreak....)
• homogeneous background environment (HBE) is sufficient for basic radar DA research
• HBE reduces computational cost considerably (enables many more tests)
![Page 24: Improved Convective Scale Prediction from the Assimilation ......Improved Convective Scale Prediction from the Assimilation of Rapid-Scan Phased Array Radar Data L. Wicker1, C. Potvin1,2,](https://reader030.vdocument.in/reader030/viewer/2022040213/5e9e3d014bd8591fda432874/html5/thumbnails/24.jpg)
Data Assimilation & Ensemble Forecasts
4D Asynchronous LETKF 1) run 5 min forecast 2) assimilate radar data 3) run next 5 min forecast
20:00
Time (UTC)
20:1020:20
20:3020:40
21:10
10 min DA
15 min DA
20 min DA
30 min DA
40 min DA
60 min Forecast
55 min Forecast
50 min Forecast
40 min Forecast
30 min Forecast
19:40
Init
End
Control Run
90 min Forecast
6 forecasts Control Run: Shows improvement from radar DA
“Cook”