Flash Flooding Across the Southern Appalachians:
An Abbreviated Climatology and Discussion
M.S. Thesis Defense
Anthony D. Phillips
Department of Geography
Ball State University





Introduction
Previous Research
Research Focus
Data and Methodology
Results
– Statistical Analyses
– Climatological Findings
– Discussions with local WFOs
 Brief Summary & Future Work
 Questions
April 13, 2012
A. Phillips — M.S. Thesis Defense
2
 The NWS defines a flash flood as:
– “Within six hours (often within one hour) of a causative event such as
intense rain, dam break, or ice jam formation, one or more of the
following occurs:
• River or stream flows out of banks and is a threat to life or property.
• Person or vehicle swept away by flowing water from runoff that inundates
adjacent grounds.
• A maintained county or state road closed by high water.
• Six inches of fast-flowing water over a road or bridge. This includes low
water crossings in a heavy rain event that is more than localized (i.e., radar
and observer reports indicate flooding in nearby locations) and poses a
threat to life or property.
• Dam break or ice jam causes dangerous out-of-bank stream flows or
inundates normally dry areas creating a hazard to life or property.
• Any amount of water in contact with, flowing into, or causing damage to a
residence or public building as a result of above ground runoff from adjacent
areas.
• …”
April 13, 2012
A. Phillips — M.S. Thesis Defense
3
 Residents of the southern mountains are at greater risk
due to:
– Steep, complex terrain
– Rapid accumulation of precipitation
– Competition with mountain streams for roads, bridges, housing,
etc.
– Mountaintop removal
April 13, 2012
A. Phillips — M.S. Thesis Defense
4
April 13, 2012
A. Phillips — M.S. Thesis Defense
5
 Several precipitation and flash flood climatologies have
been developed over the past few decades.
– Each defined flash flooding differently
– Varying spatial and temporal scales
 Three main climatologies focused on:
– Maddox et al. (1979)
– Gaffin and Hotz (2000)
– Stonefield and Jackson (2009)
April 13, 2012
A. Phillips — M.S. Thesis Defense
6
 Maddox et al. (1979) examined 151 flash floods that
occurred between 1973 and 1977.
– Reports from across the entire United States
– Researched individual storm event reports for mentions of flash
flooding.
– Examined mesoscale and synoptic scale situations during flash
floods
 No specific criteria existed regarding what constituted a
flash flood.
– Reports often lacked information about timing, duration, and
amount of precipitation.
 Maddox firmly called for a national database to be
developed.
– Lead to the development of the database Storm Data.
April 13, 2012
A. Phillips — M.S. Thesis Defense
7
 Maddox et al. (1979) discovered that 86 percent of flash
floods in their study occurred during the warm season
(April to September).
– 25% occurred in the month of July.
 They also found that a large
segment of events occurred
during the nighttime hours
between 0600 and 1800 local
time.
Temporal frequency of events per month (from
Maddox et al., 1979).
April 13, 2012
A. Phillips — M.S. Thesis Defense
8
 Maddox et al. (1979) identified several characteristics
common to almost all flash floods in their research:
– Heavy rains were produced by convective storms.
– Surface dew point temperatures were very high.
– Large moisture contents were present through a deep
tropospheric layer.
– Convective storms and/or cells repeatedly formed and moved
over the same area.
– Vertical wind shear was weak to moderate through the cloud
depth.
– Storms often occurred during nighttime hours.
April 13, 2012
A. Phillips — M.S. Thesis Defense
9
 Gaffin and Hotz (2000) examined flash floods over a 39year period from 1960 to 1998.
– Study was limited to the extent of the Morristown, TN WFO
CWA.
– Utilized records from Storm Data.
– Synoptic patterns were also examined
 Research ignored any type of urban flooding
– Can be caused by even slight amounts of precipitation
 Divided multi-county flash flood events into single events
per county
April 13, 2012
A. Phillips — M.S. Thesis Defense
10
 Gaffin and Hotz (2000) recognized a large percentage of
flash flood reports occurred across counties with high
population densities
– Result of urban development along creeks and streams.
 Counties with steep terrain had higher frequencies of
flash flooding
– Most destructive events
occurred across these
areas
Number of flash flood reports by county across the WFO Morristown
CWA between 1960 and 1998 (from Gaffin and Hotz, 2000).
April 13, 2012
A. Phillips — M.S. Thesis Defense
11
 As with Maddox et al. (1979) , Gaffin and Hotz (2000)
found that the month of July had the highest frequency of
flash floods, followed by March and May, respectively.
 Most events occurred during the afternoon and evening
hours.
Flash flood events per month across the Morristown,
TN CWA (from Gaffin and Hotz, 2000).
April 13, 2012
Flash flood events per time of day across the
Morristown, TN CWA (from Gaffin and Hotz, 2000).
A. Phillips — M.S. Thesis Defense
12
 Most recent flash flood climatology across the region is
from Stonefield and Jackson (2009)
– Very similar to research by Gaffin and Hotz (2000).
– Events were compiled from the Storm Data database
 They developed a severity classification scheme
– Flood Severity Index (FSI)
– Based on the impact (i.e., monetary damage) of individual flash
flood events
– Flash floods were ranked from FS1 (nuisance) to FS5
(catastrophic)
April 13, 2012
A. Phillips — M.S. Thesis Defense
13
 766 flash flood events were analyzed
– County distributions
– Annual, monthly, and hourly frequenices
 Discovered that 75 percent of events occurred during the warm
season
– 25 percent occurred during June, followed by January
 Large percentage of events
occurred during afternoon
and evening hours between
1200 and 2100 local time
 83 percent of flash floods
were categorized as FS1 or
FS2 events
 Identified geographic regions
prone to flash flooding
April 13, 2012
Regions of synoptic-scale significant flash flood events
(from Stonefield and Jackson, 2009).
A. Phillips — M.S. Thesis Defense
15
 Created a climatology of flash flooding across the
southern Appalachian Mountains using verifiable reports
from the NCDC Storm Data database.
 Analyzed the spatial and temporal extent of flash floods
across the region.
 Examined the influences of regional topography and
social characteristics on the spatial distribution of flash
floods.
 Gathered information from local NWS WFOs about flash
flood issues they are familiar with; specifically locations
prone to flash flooding and meteorological conditions
associated with heavy precipitation across the higher
terrain.
April 13, 2012
A. Phillips — M.S. Thesis Defense
16
 Defined the Southern Appalachians as:
– Mountainous areas south of Mason-Dixon Line (~39.7° N)
– Locations within the USGS Appalachian Highlands physiographic
division; namely the Appalachian Plateau, Valley and Ridge, and Blue
Ridge provinces

Data obtained from the National Climatic Data Center
– Storm Data from 1950 to present (minus 6 months)
 Focus on events after the Modernization and Associated
Restructuring (MAR) of the NWS in the mid-1990’s
– Storm reports and verification
– Abbreviated climatology: January 1, 1996 to December 31, 2010
April 13, 2012
A. Phillips — M.S. Thesis Defense
17
April 13, 2012
A. Phillips — M.S. Thesis Defense
18
April 13, 2012
A. Phillips — M.S. Thesis Defense
19
 Number of events per county
 Number of events per square
kilometer
 Number of events per county
population
 Mean number of events per year
 Fatalities per county
 Fatalities per county population
 Injuries per county
 Injuries per county population
 Total damage incurred
 Total property damage incurred
 Total crop damage incurred
April 13, 2012
 Number of events per county
per year
 Number of events per county
per month
 Number of FS1, FS2, FS3,
FS4, and FS5 events per
county
 Fatality explanations
 Hot spot analysis (Getis-Ord
Gi*)
 Cluster and Outlier Analysis
(Anselin Local Moran’s I)
 Directional Distribution
(Standard Deviational Ellipse)
 Point density
A. Phillips — M.S. Thesis Defense
20
April 13, 2012
A. Phillips — M.S. Thesis Defense
21
 Originally it was hypothesized that those counties with a
higher frequency of flash floods also had higher mean
percent slope values.
 By comparing only the average slope to the number of
flash floods per county, the findings suggest that this is
not the case.
April 13, 2012
A. Phillips — M.S. Thesis Defense
22
 It was discovered that there is a positive relationship
between the number of events per square kilometer and
county population per square kilometer
– Either a result of more people witnessing the flash flood or more
events actually occur over urbanized areas
– From NWS discussions, most likely due to urban landscape
April 13, 2012
A. Phillips — M.S. Thesis Defense
23
April 13, 2012
A. Phillips — M.S. Thesis Defense
24
 Poisson distribution was calculated to determine both the
observed and predicted probability of a given number of
flash floods occurring within a fixed area of 64.75 km2
(25 mi2).
April 13, 2012
A. Phillips — M.S. Thesis Defense
25
 Frequency of events per year
 Noticeable variations, especially
between drought (’99, ‘07) and
non-drought years (‘03)
April 13, 2012
A. Phillips — M.S. Thesis Defense
26
 Number of events that occurred during each month
 Frequency of events increases dramatically starting in
May
– Substantial decrease in October
April 13, 2012
A. Phillips — M.S. Thesis Defense
27
April 13, 2012
A. Phillips — M.S. Thesis Defense
28
 Events per time of day divided into 1-hr increments
 Number of events increases during afternoon/early night
hours
April 13, 2012
A. Phillips — M.S. Thesis Defense
29


Events per flash
day compared
to events
Significant
flood events
per dayper
andhour
per hour
Clustering
of events
between
and September
& 13:00
LST
Same
temporal
clustering,
butApril
significant
events occur
later and
into 23:00
the night
April 13, 2012
A. Phillips — M.S. Thesis Defense
30
April 13, 2012
A. Phillips — M.S. Thesis Defense
31
 Considerable number of fatalities due to children and
young adults entering flood waters
April 13, 2012
A. Phillips — M.S. Thesis Defense
32
 Emails were sent to each WFO that serves the southern
Appalachians asking for their thoughts on conditions favorable for
flash flooding as well as any locations prone to events.
 Eight of the thirteen WFOs responded with their insights
 Their responses were compared to our derived climatology to
determine if it was valid, some commonalities include:
– Flash floods occur with greater frequency over urbanized areas and
terrain highly influences the development and evolution of heavy
precipitation storms, especially those that cause significant damage.
– Nearly 75 percent of flash floods occur during the warm season from
April to September.
– June is typically the most active month for flash floods.
– Most flash floods are nuisance or minor events, resulting in little to no
damage.
– Significant flash floods generally occur across rural counties.
April 13, 2012
A. Phillips — M.S. Thesis Defense
33






4,938 unique flash flood reports from 1996 to 2010
71 fatalities and 64 injuries
As expected, greater number of events during warm season months and
during afternoon/overnight hours
Correlation between number of events per km2 and population per km2
Probability of a flash flood occurring within 64.75 km2 (25 mi2): 0.36
Future work should include:
– Expansion to include eastern U.S.
– Further analysis of the synoptic and
mesoscale environments associated
with flash flooding
April 13, 2012
A. Phillips — M.S. Thesis Defense
34
 Committee Chair: David A. Call, Ph.D.
 Committee Members: Jill S. M. Coleman, Ph.D.
Petra A. Zimmermann, Ph.D.
Special thanks:
 Stuart Hinson, NCDC, Asheville, NC
 Various NWS Forecast Offices
 BSU Department of Geography
 Graduate Students
April 13, 2012
A. Phillips — M.S. Thesis Defense
35
Any questions or comments?
For more information:
Anthony Phillips
http://www.wx4sno.com
wx4sno@vt.edu