1. Analysis and Prediction Thrust - IP1 Report
2007 Site Visit
Presented by Ming Xue and Jerry Brotzge

2. Outline Detection algorithms in MC&C
Nowcasting and integration into MC&C
Near-realtime NWP DA and forecasts
IP1 cases and morphology studies

3. MC&C Feature Repository
Detection Algorithms
KSAO
KCYR
KRSP
KLWE
KTLX
KFDR
ldm2netcdf
Reflectivity
Velocity
KVNX
w2merger_refl
w2circ
w2mdalite
w2merger_vel
NEXRAD Feature Table
casaIngestor
Currently running
Velocity-data
Reflectivity-data
Dual Pol-data
DebrisSig
Debris Sig Table
AzShear
MDALite Table
WindField
Shear feature
VortDiv
LLSD Table
Multi-Doppler Table
NetRad Circulation Table
MR_feature
Merged Reflectivity
SCIT-CASA
Reflectivity Threshold
Algorithms (not including velocity-based detection) in the shaped area are running in MC&C currently.
Cell Features
Storm Cell Table
MC&C Feature Repository

4. SCIT (Storm Cell ID and Tracking)-CASA Running in MC&C - Identifies local maxima in reflectivity data
Identifies local maxima in reflectivity data
Feature must match size criteria (~2.5km2)
If not large enough, will include next lower reflectivity threshold and re-analyze
Limit to the number of threshold lowerings allowed
Uses thresholds of 5 dBZ increments between 20 and 60 dBZ
Reports all features
Only features whose min dBZ values are >40 dBZ are passed to the MC&C
Colored areas are identified as local reflectivity maxima (or storm centroids)
0043 UTC, 5/9/2007

5. Local Max Feature (SCIT-CASA)
Pros
Works well with strong isolated echoes
Not fooled by ground clutter or 2nd trip echoes
Variable threshold limits and MC&C reporting limits
Cons
Won’t detect initial echoes if they are too small/weak
Allows large regions where reflectivity varies by small values
Might not identify storms if only 2 deg elevation surveillance scan data is available

6. Reflectivity Threshold (RT) and SCIT-CASA
RT: Operates on single radar data.
Relativistic 2d model, no hard dBZ thresholding.
Detects reflectivity echoes as they first appear
Complement SCIT
Designed to steer radars to important reflectivity areas, not to classify storms.
Operates on single radar data, in polar coordinate system.
Relativistic 2d model, identifies areas of significance in relation to others contained in the scan.
dBZ thresholds are met in areas of a minimum 1 Km radius that are greater than 2 standard deviations above the mean of the data file. No hard dBZ thresholding
Crafted specifically to work within MC&C architecture.
Detects reflectivity echoes as they first appear unless areas of substantially higher reflectivity already exist in the domain.
During widespread, strong, but non-stratiform storms (e.g., squall lines), RT focuses in only on strongest areas, which is not ideal to forecasters. A fix is being tested.
5/27/2007
Click for movie

7. Linear Least Squares Derivative (LLSD) Azimuthal Shear
1 deg elev
LLSD Shear Velocity Reflectivity
2 deg elev
Improved shear estimates (compared to MDA)
Less subject to noise
Calculates both rotational shear and divergence
Filter kernel must be matched to size of feature being detected
Requires feature detector for automated use in the MC&C

8. Detection Plan for Spring 2008
Data Quality Control
Remove residual clutter/noise
Use dual-pol data for identification of non-met targets – RhoHV is being used.
Testing and adjustments with/for IP1 data
Thresholding and identification of bad detections with MDAlite
Modify LLSD to run at multiple scales to identify both mesocyclones and tornados
May also need testing with larger samples of simulated data, produced by CASA radar emulator
Activate velocity algorithms (MDAlite and LLSD) and WSR-88D detection in MC&C and evaluate their impact
Run velocity algorithms in realtime in spring 2008
Future Work
Will be running in realtime in spring 2008.

9. Nowcasting Systems
Growth-Decay Storm Tracker (GDST) (Wolfson et al 1999, MIT Lincoln Laboratory)
Multiscale Storm Identification and Forecast (MSIF) in NSSL WDSS-II system (Lakshmanan et al JAR)
Thunderstorm Identification, Tracking, Analysis and Nowcasting (TITAN) (Dixon & Wiener 1993, NCAR)
Dynamic and Adaptive Radar Tracking of Storms (DARTS): a spectral based real time nowcasting system achieves robust global estimation of the spatially non-uniform motion fields
Designed to take advantage of CASA high spatial and temporal resolution data
runs very fast
  author =       {Lakshmanan et al JAR), V. and Rabin, R. and DeBrunner, V.},   title =        {Multiscale Storm Identification and Forecast},   journal =      {J. Atm. Res.},   year =         2003,   volume =       67,   pages =        { },   publisher =    {Elsevier},   address =      {Shannon, Ireland},   month =        {July}

10. Example: DART v.s. GDST with IP1 Data Aug 15 – Aug 16, 2006
5 min forecast DARTS
5 min forecast GDST
Observed
GDST
DARTS
DARTS
GDST
GDST
DARTS
Critical Success Index
Probability of Detection
False Alarm Rate

11. Nowcasting Goal, Status and Plan
Goal: To produce realtime reflectivity-based forecasts up to 60 min, and use 1-5 min forecast in MC&C
Status: Real time experiments had been done by streaming archived CASA data.
DARTS nowcasting performance is similar to GDST but uses significantly less computation time.
Preliminary results show improvement in performance in DARTS with better spatial / temporal resolution.
Hour-long nowcasting will be implemented in real time for Spring 2008 season, for use/evaluation by forecasters.
Possible use of nowcasting product in MC&C in spring 2008.
will evaluate impact carefully.

12. 2007 Spring NWP Experiments
Goals: Examine, in near realtime, the impact of IP1 data on convective storm forecast; collect data and provide a baseline benchmark for post-real time case studies.
Four 6-hour 1-km resolution forecasts daily, assimilating
1. Both 4 CASA IP1 and 6 WSR-88D radars
2. WSR-88D radars only
3. CASA IP1 radars only
4. No radar data
Using 280 processors on OU OSCER supercompuer (~4 hours for each forecast)
Forecast products posted at

13. Forecast Configurations
22UTC
00 UTC
04 UTC
Example 40-min Assimilation & 5.5-hour forecast
Dx = 1 km
600x540x20 km
Start time determined daily by the time when convection initially develops within or when an existing system enters the IP1 network.
Run only on days with echoes in IP1 network.
Actual runs are done in quasi-real time. Takes up to 18 h to complete all four forecasts.

14. Assimilation Strategy
2210
2220
2230 04
2200 00
IAU 01 02 03
5.5-hour Forecast
40-min Assimilation
2150
Using ADAS and its cloud analysis package
Reflectivity data were used, but not radial velocity (parallelization of software not ready)
10-min assimilation cycles over 40 min window, with Incremental Analysis Update (IAU)
Prediction using ARPS
Special handling and QC of IP1 sector data (data, form, and scanning new)
ETA 12-km analysis and forecasts used as analysis background and BC

15. 2007 Spring Cases Date Time Weather Summary 1 10 April 23 Z
Cluster of cells, one rotating with funnel 2
13 April 21
N-S Oriented Squall Line Moving W-E 3
17 April 19
Synoptic Low, TRW in Okla, Severe TRW in N-Tx 4
24 April 16
Squall Line developing & moving ESE 5
27 April 18
Weak Convection in IP1 ahead of cold front 6
1 May
Broken line cells developing in IP1 & moving SE 7
3 May
Cells develop in IP1, move E, form line in E Okla 8
7 May 05
Squall line forms in IP1, additional cells in IP1 later 9
8 May 07
E-W line convection in IP1
10a 23
Cells form ahead of squall line in IP1, rotation after 00z
10b
9 May 01
Additional start time during the 1st rotating cells within IP1 11
11 May 22
E-W line of TRW moving S, skirts E side of IP1 12
15 May 14
Weak rain showers on E-W cold front 13
24 May
Nocturnal Squall line moving from NW to SE
26 May 20
Clusters of moderate-to-heavy cells moving from SSW to NNW 15
30 May 10
Strong nocturnal squall line with high winds moving from NW to SE
1 June 09
Decaying nocturnal squall line moving from W to E across network 17
Cells develop within IP1, severe warnings within network
3 June
Squall line with severe reports in IP1, declines after about 13 UTC

16. Verification. 11:00pm (0400 Z) May 8-9, 2007 A series of low-level
© Patrick Marsh
11:00pm (0400 Z)
May 8-9, 2007
A series of low-level
circulations.
NWS Tornado
Warnings: 7:16pm,
7:39pm, 8:29pm
9:54pm
8:30pm (0130Z)
© KSWO TV
7:21pm (0021Z)

17. Impact of IP1 Data on Low-level Analysis
0050 UTC
9 May 2007
Analyzed
Hydrometeors
Converted to Reflectivity
Hgt = ~250 m AGL
WSR-88D Only
NetRad & WSR-88D

18. N-S Vertical Cross-Section
5km
5km
WSR-88D Only
WSR-88D & NetRad
0050 UTC 9 May 2007

19. 24-min Prediction
~ 600 m AGL
KCYR Indicated Circulation
Vorticity centers
WSR-88D Only
WSR-88D & IP1
Positive impact on low-level vorticity prediction with ADAS and even without Vr data

20. 04 UTC May 09 Forecast Animation
Used both IP1 and 88D data
Link to movie – put a picture here
3 h 10 min forecast
Click on figure for movie

21. Near-realtime NWP Accomplishments
Completed MPI of ADAS with cloud analysis
Pre-processing of CASA radar data
Data selection scripts
Data I/O and Quality Control (ref & vel)
Remapping and qr assignment
Reflectivities are well-handled in model
Automated assimilation system operational
Showed preliminary positive impact of including IP1 data
Provide a set of baseline forecasts for post-realtime case studies, using better DA methods

22. Near Future Plan for NWP
Quantify the impact of IP1 data on spring 2007 forecasts
Test additional benefit of Vr data from 2007 cases
Parallelize ARPS 3DVAR and optimize other programs and prepare for spring 2008 realtime experiments
Perform detailed case studies and refine assimilation capabilities
Secure computational resources to run 88D+IP1 case in real time for use by forecasters in spring of 2008
Plan on comparing realtime NWP and nowcasting in spring 2008

23. Evaluate higher temporal/spatial resolution.
KFDR
02:16:10 UTC
KCYR
02:16:39 UTC
© Val Castor, News9
Higher Resolution Data
Nov 30, 2006

24. Evaluate higher temporal/spatial resolution.
12:36:42 UTC
12:37:30 UTC
© Val Castor, News9
Higher Resolution Data
April 24, 2007

25. KLWE at 1.0 deg elevation, 21km range (~200m AGL)
Evaluate ability to scan adaptively.
May 8-9, 2007
KLWE at 1.0 deg elevation, 21km range (~200m AGL)
7:26pm
7:28pm
7:32pm
7:34pm
7:36pm
7:38pm
7:39pm
© KSWO TV
7:23pm
7:28pm
7:34pm
7:39pm
7:40pm

26. Evaluate collaborative scanning. May 9, 2007
KCYR
© Val Castor, News9
KLWE
Higher Resolution Data
01:16:28UTC

27. Verification. 2324 UTC 14.0° 11.0° 9.0° April 10, 2007 7.0°
3.0°
5.0°
7.0°
9.0°
2.0°
11.0°
14.0°
© Val Castor, News9
April 10, 2007
Cluster of isolated
severe storms.
Photo taken near
Elgin, OK.

28. Verification. 11:00pm May 8, 2007 Series of low-level circulations.
© Patrick Marsh
11:00pm
May 8, 2007
Series of low-level
circulations.
NWS Tornado
Warnings: 7:16pm,
7:39pm, 8:29pm
9:54pm
8:30pm
© KSWO TV
7:21pm