Title: Mesocyclone Climatology of the Southern Great Plains using the NSSL Mesocyclone Detection Algorithm
1Mesocyclone Climatology of the Southern Great
Plains using the NSSL Mesocyclone Detection
Algorithm
Kevin M. McGrath School of Meteorology,
University of Oklahoma http//mesocyclone.ou.edu
2Objectives
- Produce a climatology of mesocyclones in the
Southern Great Plains using the NSSL Mesocyclone
Detection Algorithm (MDA) - Improve the quality of the data set by developing
methods to identify and automatically remove
spurious detections - Perform post-processing analysis and generate
statistics of output - Geographic variability in mesocyclone occurrences
- Atmospheric features influencing detections
3Motivation
- Large percentage of mesocyclones are associated
with severe weather - JDOP (1976 1978) determined that 50 of the
observed mesocyclones were tornadic 80 were
severe and/or tornadic (Burgess et al. 1979) - Burgess and Lemon (1991) reported that 31 of the
observed mesocyclones were tornadic 94 were
severe and/or tornadic - A recent study indicated as low as 25.5 of
mesocyclones are tornadic (Trapp et al. 2002) - Drop in percentage likely due to increase in
detection of very weak circulations - Difficultly in constructing a climatology of
mesocyclones using in situ observations - Build on Pittsburgh, PA study (Mitchell et al.
2000) - Assist in training and guidance
4Radar Data
- Level II data acquired from multiple Southern
Plains radars under the auspices of the
Collaborative Radar Acquisition Field Test
(CRAFT) project
- Convective cases from March 2000 through June
2002 from the initial set of six radars (KAMA,
KFWS, KINX, KLBB, KSRX, and KTLX) have been
processed using the MDA
- KDDC, KFDR, KICT, and KVNX added to CRAFT feed in
spring of 2002. Subsequent data were processed
during this study data sets too small to include
in results.
5Mesocyclone Detection Algorithm
- Attempts to detect all storm-scale vortices (1-10
km diameter) and then diagnose them to determine
if they are significant - Process
- Find shear segments and construct 2D vortices
- Vertically associate 2D vortices ? 3D features
- Classify and diagnose
- Adaptable parameters used were those suggested by
the NSSL - Data processed in cyclonic mode
- Detections defined on a per-volume-scan basis
no time correlation attempted - Adapted to process archived data
6Near Storm Environment Data
- RUC NSE data incorporated into suite of
algorithms when used in real-time. The velocity
dealiasing algorithm uses upper-level wind
information and the first guess storm cell motion
are estimated for storm cell identification and
tracking.
- NSE data was not available for this project FSL
sounding data used instead - NSSL software utilized to estimate 253K and 273K
heights - Tests indicate MDA not particularly sensitive to
this data
7Initial Results
(May 29 30, 2001)
8Challenges
- The relatively high number of weak detections
tend to obscure the stronger detections - Spurious detections (often of strong rank)
caused, in part, by - Incorrectly dealiased velocity data (especially
near boundaries parallel to beam) - Beam broadening and cone-of-silence
- Ground clutter and beam blockage
- Sidelobe contamination
- Anomalous propagation
9Filtering Techniques 1 Initial Filter
Raw MDA output often contains large number of
spurious and weak detections ? noise
- Initial Filter Criteria
- Located within 5 km of the radar (3.0)
- Located at the maximum unambiguous velocity range
(first trip ring, commonly located at 147 km)
(0.40) - Weak in intensity (Mesocyclone Strength Rank 0)
(88.7) - Detected in clear air mode (VCP 31 or 32) (4.0)
- 8.8 of raw detections retained
10Filtering Techniques 1 Initial Filter
(KAMA, 20010502 15Z 20010504 0Z)
Raw MDA detections (N 7,866)
KAMA
KAMA
MDA detections remaining after application of
initial filter (N 948, 12.1 retained)
11Results Post Initial Filtering
12Storm Cell Identification and Tracking (SCIT)
Algorithm
- Designed to identify, characterize, track, and
forecast the short-term movement of storm cells
identified in three dimensions (Johnson et al.
1998) - Maximum reflectivity must be ? 30 dBz
- Bases detections solely on reflectivity data
- Thresholds used in this study were those
suggested by the NSSL and are commonly used in an
operational setting
13Filtering Techniques 2 SCIT Filter
- Attempts to correlate mesocyclone detections with
the associated storm cells as defined by the SCIT
algorithm - Filter searches for storm cell centroids within a
user defined circular window centered on each
mesocyclone detection centroid during the same
volume scan
- Detections that do not have a storm cell centroid
simultaneously within the search window were
labeled as false - Similar techniques were employed by Thomas Jones
14Determining Filter Radius
Correlation of Mesocyclone Low-level Rot. Vel.
And SCIT Derived Storm Cell VIL as a Function of
Separation Distance Between Centroids
Percent of Mesocyclone Detections Retained as a
Function of SCIT Filter Search Radius
15Example of SCIT Filtering
(KAMA, 20010502 15Z 20010504 0Z)
MDA detections, post-initial filtering. Note
region of high ranking, false detections. (N
948)
Mesocyclone track
KAMA
KAMA
MDA detections remaining after passage through
the SCIT filter (10 km circular window). Meso
track now much more noticeable. (N 544, 6.9
of original)
16Results Post SCIT Filtering
17True Detections Using a 10 km search window
KINX
KTLX
KAMA
KSRX
KLBB
KFWS
18Density of True Detections Using a 10 km search
window
KINX
KTLX
KAMA
KSRX
KLBB
KFWS
19Density of True Detections Using a 10 km search
window (bin 77 km2)
KAMA
KTLX
KINX
75
90
75
60
90
75
KLBB
KFWS
KSRX
20False Detections Using a SCIT Filter 10 km
search window
KINX
KTLX
KAMA
KSRX
KLBB
KFWS
21Density of False Detections Using a 10 km
search window
KINX
KTLX
KAMA
KSRX
KLBB
KFWS
22Density of False Detections Using a 10 km
search window (bin 77 km2)
KAMA
KTLX
KINX
90
90
90
75
90
75
KLBB
KFWS
KSRX
23KAMA Beam Blockage
24Concentration of False Detections NNW of KTLX
25Equal Area Range Distribution 0 229 km (bin
5,542 km2)
26Equal Area Range Distribution 0 146 km (bin
2,239 km2)
27Time of Detection
KTLX
KAMA
KLBB
KFWS
28Time of Detection
KTLX
KINX
KFWS
KSRX
29Mesocyclone Attributes
30Mesocyclone Strength as a Function of Time
True detections
31Mesocyclone Strength as a Function of Time
False detections
32Mesocyclone Strength as a Function of Equal Area
Range Real detections
33Mesocyclone Strength as a Function of Equal Area
Range False detections
34Mesocyclone Strength Distribution
Meso Strength Distribution of True SCIT
Filtered Detections
Meso Strength Distribution of False SCIT
Filtered Detections
35Mesocyclone Base vs. Range
36May 5 6, 2002 Mesocyclones
37May 5 6, 2002 Mesoanticyclones
38June 12 14, 2002 Severe Squall Line
39Conclusions
- A large percentage of MDA detections are false
in the sense that they are not mesocyclones - The quality of a mesocyclone detection data set
can be significantly improved using simple
filtering techniques - Spatial distribution varies geographically
- Temporal offset between real and false
detections - MDA produces more true detections with VCP 11
- Possible radar calibration issues identifiable
40Recommendations
- Develop filtering techniques that require less
human interaction, i.e., location of first trip
ring - Algorithm should process data in cyclonic and
anticyclonic modes - Importance of utilizing ground clutter
suppression, updated quarterly - Multi-radar algorithms
- Operate in VCP 11
- Combine MDA and TDA (VDDA Vortex Detection and
Diagnosis Algorithm) - Examine seasonal variations in mesocyclone
climatology - Remove detections w/ range ? 10 km and/or MSR ?
3/5
41Acknowledgements
- Don Burgess
- Karen Cooper
- Greg Stumpf
Kelvin Droegemeier Jason Levit Kevin Thomas
Courtney Garrison and staff Thomas Jones John
Snow and staff Andy White
Friends and Family NOAA Warning
Decision Training Branch Oklahoma NASA
Space Grant Consortium