The May 2003 Extended Tornado Outbreak

1
The May 2003 Extended Tornado Outbreak

• Tom Hamill (CDC), Russ Schneider (SPC), Harold
  Brooks (NSSL), Greg Forbes (Weather Channel),
  Howie Bluestein (Oklahoma), Mike Steinberg
  (AccuWx), Daniel Melendez (NOAA HQ), and Randy
  Dole (CDC)

2
May 2003 Extended Outbreak Map
3
Extended Outbreak Statistics

• 41 fatalities, 642 injuries
• 829 million dollars property damage. 2300 homes
  and businesses destroyed, and 11,200 damaged
• 26 states
• Most tornadoes ever in a week (334)
• Strong tornadoes (F2) occurred in an unbroken
  sequence of nine straight days
• May 2003 more tornadoes than any previous month
  (559) , 361 during 9-day extended outbreak

4
Outline

• Review outbreak
• Refresher weather conditions necessary to
  support tornado-spawning supercells
• Were numerical forecasts and SPC forecasts
  accurate from short to long lead times?
• How unusual was this outbreak?

5
Tornado Damage, Kansas City, 4 May 2003
Source Larry Gonnello, KCFD
6
Northmoor, MO (near KC). Source AP
7
For better, or for worse
Northmoor, MO (near KC). Source AP
8
Source AP
9
Girard, KS
3 people dead near McCune KS 1 dead in Liberal,
KS
Girard, KS
10
Source AP
Stockton damage. Source OSHA
3 dead in Stockton, MO town nearly destroyed
11
Pierce City, MO 5 dead, old downtown destroyed
Pierce City Armory. Photo AP
Source Jimmy Sexton, AP
12
Tennessee Tornadoes occurred after dark, 03Z
09Z 5 May 2003 Two killed in Jackson, TN and
nine in the town of Denmark, 12 miles southwest
of Jackson. (Grazulis).
Searching Relatives of Lee McLaughlin, aged
seven, killed by a tornado in Denmark, Tennessee,
search for possessions. Photograph Mark
Humphrey, AP
Carolyn Tucker looks over her wrecked home and
car in Madison, a suburb of Nashville, Tennessee.
Photograph Larry McCormack, AP/The Tennessean
13
 
Source NWS Storm Prediction Center
14
May 2003 Extended Outbreak Map
15
Climatological Tornado Threat in Early May
P(tornado within 40 km)
Source Harold Brooks, NSSL, based on 1980-2002
data
16
Environmental Conditions Necessary to Support
Supercells
Source AccuWeather
17
Rotating Thunderstorms
Courtesy of COMET
18
(No Transcript)
19
Large-scale weather patterns related to
outbreak? MJO before event.
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
Average analyzed (true) wind and instability
conditions from 0000 UTC 4 May 2003 0000 UTC
11 May 2003
41
NCEP GFS ensemble-mean wind and instability,0000
UTC 28 April 2003 initial time, valid 0000 UTC 4
May 2003 0000 UTC 11 May 2003 (6-13 day fcst)
42
NOAA Forecast Products
Hazards assessment issued 30 April
Severe Wx Forecast Issued 12Z 2 May valid 4 May
Tornado Probability Forecast Issued 1600 UTC 04
May 2003 for 1630 UTC 04 May to 1200 UTC 05 May
2003
43
Severe Prediction Center Watches/Warnings
Tornado and Severe Watch Boxes Tracks
Tornado Warnings
44
May 3, 2003
45
May 4, 2003
46
2245 UTC 4 May 2003
SGF
47
May 5, 2003
48
May 6, 2003
40 homes destroyed in Pulaski County, Illinois,
causing one death and about 40 injuries. A man
died when the chimney of his house collapsed on
him while he was protecting his young son. In
Massick County, a 65-year-old woman was found
dead in a ravine, 100 yards from the site of her
trailer. Her car was thrown 200 feet. The
wreckage of the trailer was wrapped around two
tree trunks. (Grazulis)
49
(No Transcript)
50
(No Transcript)
51
May 7, 2003
52
May 8, 2003
The city of Moore was struck by a major tornado
for the 4th time in 5 years. This one produced
100 million damage and more than 100 injuries,
but no deaths. (Grazulis)
53
(No Transcript)
54
Moore, OK, 8 May 2003
Photo by Roger M. Bondy
Moore, OK, 3 May 1999
Satellite photo by Space Imaging
55
May 9, 2003
A night-time tornado family swept across the
western and northern suburbs of Oklahoma City,
hitting Bethany and Yukon, with a few injuries.
An 80-year old man hit his head on a door jamb
while running for shelter. The house wasn't hit,
but he died in the hospital the following day.
This event should not be counted as a killer
tornado, but some lists may call it that. The
man's father died in the Leedey, Oklahoma,
tornado of May 31st, 1947. (Grazulis)
56
(No Transcript)
57
May 10, 2003
At about 430 AM, an F2 tornado damaged or
destroyed at least 40 homes west and northwest of
Harrodsburg, Kentucky. One death in mobile home.
Another woman was found downstream after an 18
hour search. She had apparently drowned.
(Grazulis)
58
May 11, 2003
59
How rare was this extended outbreak?Problem
tornado observations not consistent over time.
Source Grazulis /NWS
60
Comparison using 9-day F2 tornado counts
74 Super-outbreak
Palm Sunday 65
61
Palm Sunday 1965
256 deaths major impetus for national weather
radar network
Midway, Elkhart County, Indiana
62
1974 Super-outbreak
Xenia OH, photo by Fred Stewart
63
Summary of comparable outbreaks
 
 Summary of four long-sequence outbreaks and two
short sequence outbreaks. Parenthetical values
after year are number of days with at least one
strong tornado out of total days in sequence.
64
Other outbreaks significant in other ways,
e.g.,Tri-State Tornado, 1925
740 deaths (689 of these due to the Tri-State
tornado)
65
Regions of past extended outbreaks
66
How rare? Evaluate using records of past F2
tornadoes
1 day, gt26 F2 tornadoes
Raw data from Grazulis (1993) for period
1921-1995.
67
How rare was it to have tornado-favorable
conditions on 9 straight days?
Step 1 Develop model for predicting tornado
probability from analyzed shear and LI Step 2
Plot time series of the fraction of blue box with
elevated risk of tornadoes.
68
Developing an objective model of tornado
probability from shear and lifted index
P (Tornado LI, shear) 1.0 1.0 / (1.0 exp
(b0 b1shear b2LI ) )
69
Reliability
Data from April-May 1979 to 2002 using grid boxes
inside blue box in previous figure.
70

• Notes
• No other string
• of days with such large
• area covered by high
• risk as May 2003.
• Looks like theres
• a trend with time. Is
• there a trend in strong
• tornado frequency?
• (3) What happened during
• some of the other peaks?

Andover, KS
?
71
Is there a problem with the probability model?
Few tornadoes in 2002 under large risk area.
72
Heres the data for the day in April 2001 with
the large fraction of areal coverage.
73
Skew-T at Springfield, Missouri, in middle of
high risk area
Looks like a sounding that could support
tornadoes.it just didnt
74
Conclusions

• Outbreaks of this magnitude happen every 10-100
  years.
• Forecasts for this outbreak appeared pretty good
  large-scale pattern for outbreak seen a week in
  advance, 5 day forecasts consistently hinted at
  tornado-favorable patterns.
• Outbreak lasted so long because of persistent
  trough in intermountain west and no respites due
  to intrusions of Canadian high pressure
• Fewer deaths, due to (1) better forecasts
  awareness, (2) fewer F4-F5 tornadoes.
• Acknowledgments Cathy Smith, Xue Wei, Jeff
  Whitaker, Andres Roubicek for help generating
  figures and extracting data.