Content uploaded by M. Parise
Author content
All content in this area was uploaded by M. Parise on Feb 07, 2019
Content may be subject to copyright.
Sinkhole distribution in a rapidly developing urban environment: Hillsborough
County, Tampa Bay area, Florida
R. Brinkmann
a,
⁎, M. Parise
b
,D.Dye
a
a
Department of Geography, University of South Florida, Tampa, USA
b
National Research Council (IRPI), Bari, Italy
article info abstract
Article history:
Accepted 14 November 2007
Available online 26 February 2008
Sinkhole formation in Florida is a common event. The Florida karst plain is significantly altered by human
development and sinkholes cause considerable property damage throughout much of the state. We present in
this paper a morphometric analysis of karst depressions in the Tampa Bay area, and the relation with the
known distribution of sinkholes. We selected the Tampa Bay area because it is particularly susceptible to the
evolution of karst depressions in relation with development of the built-up environment. Karst depressions
were mapped from the 1:24,000 USGS topographic maps and a morphometric analysis was performed by
using parameters such as shape, circularity index, perimeter, area, length, width, and orientation. Maps
showing the distribution of depression density, and the sectors with greatest areas of karst depression were
produced using a GIS. These results were compared with data compiled from the database of sinkhole
occurrences in Florida maintained by the Florida Geological Survey. Our analysis demonstrates that the
distribution of new sinkhole occurrences differs from the distribution of existing sinkholes, indicating that
there are processes acting today that are influencing karst landscape formation that are different from those
acting in the past.
© 2008 Elsevier B.V. All rights reserved.
Keywords:
Sinkholes
Hazard
Karst
Planning
Management
1. Introduction
Karst landforms occur throughout the world where soluble rock is
present at or near the surface (Sweeting 1973; White 1988;Ford &
Williams, 1989). There are a number of landforms that are associated
with karst landscape including caverns, karst depressions, disappear-
ing streams, and springs. The landscape associated with karst is
unique in that much of the drainage occurs underground in conduit
systems that dissolved and formed over long periods of time and in
that the landscape is unstable due to the irregular, but common,
formation of karst depressions when the land surface collapses
(Cooley, 2002; Klimchouk, 2002; Waltham, 2002). There are many
regions of the world where karst landforms are the dominant
expression of the landscape including southern China, many areas
on the Mediterranean fringe, portions of the Carpathian Mountains,
and many areas of the United States, as in Florida where karst
depressions dot the landscape of much of the state (Williams, 1966;
Day, 1976; Hansel, 1980; Troester et al., 1984; Hung et al., 2002).
The geology of Florida is conducive to karst landscape formation.
Karst depressions in central Florida occur largely as a result of cover
subsidence processes (White, 1970; Schmidt, 1997; Tihansky, 1999;
Fig. 1). The Cenozoic limestone in the state occurs in horizontal strata
overlain by Quaternary marine deposits of varying thickness
(Upchurch and Randazzo, 1997). The two units are separated by the
Miocene/Pliocene Hawthorne Group that consists largely of clay-rich
sediment. Active depressions most commonly formwhere the surface
cover is thin, although karst depressions can occur almost anywhere
in the state due to the nearly ubiquitous presence of limestone in the
subsurface. When they form, sand ravels into void spaces in the
limestone with time to create a weakness within the sand cover. Often,
the sand slowly filters into the subsurface and depression formation is
a long process. Sometimes, the sand cover collapses suddenly to create
a cover collapse depression (Tharp, 1999; Waltham et al., 2005).
Although there are many terms used for these types of depressions
around the world, each of these karst depressions (those that form
from cover collapse and cover subsidence processes) is known in
Florida as a sinkhole. Thus, we use the term sinkhole for these
features, although the term ‘doline’or ‘karst depression’may be more
appropriate. Many of the karst depressions we identified at Tampa
have been partly or totally modified by anthropogenic activities:
development of roads and houses quite often have changed the
landform shape, so that at present one or more sides of the depression
appear to be linear, and follow, for instance, a road. In other cases,
water retention ponds have been constructed using the natural
depression, which is now occupied totally or in part by water.
It is our goal to provide an analysis of karst depressions in the City
of Tampa in order to better understand how karst landscape develops
in west central Florida and in order to measure the morphometric
Engineering Geology 99 (2008) 169–184
⁎Corresponding author.
E-mail addresses: rbrinkmn@cas.usf.edu (R. Brinkmann), m.parise@ba.irpi.cnr.it
(M. Parise).
0013-7952/$ –see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.enggeo.2007.11.020
Contents lists available at ScienceDirect
Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo
characteristics of the sinkholes within the context of a city. While
there are certain natural boundaries that would be more appropriate
delimiters of the study area, the city is a convenient scale when one
considers that cities have distinct rules and regulations regarding
sinkholes, development, and land use policy. Thus, understanding the
characteristics of sinkholes at the city scale is important to evaluating
the hazards associated within the city and some of the development
limitation that exist within the region.
2. Study area
The Florida karst is known to be very flat, with broad, shallow
depressions; this was clearly shown by Troester et al. (1984) through
comparison of the sinkhole frequency–depth distributions for six
karst regions. They obtained (by counting all the depression within the
karst area and dividing by the area) a depression density of 7.94 km
2
for Florida, the highest among the investigated regions.
Sinkholes in Florida impact many people. They are both an amenity
for real estate developers who wish to develop the fast-growing state
by building around aesthetically-pleasing water features, and a
problem for home owners who find themselves in financial difficulty
due to the damage sinkholes cause. The state of Florida mandates
sinkhole insurance for all homeowners insurance policies. Payouts for
insurance claims have been increasing in the last decades as reports of
sinkholes in the region have increased (Eastman et al., 1995; Maroney
et al., 2005).
Early research on Florida karst focused on reviewing the distribu-
tion of karst in the state and how it formed (Lane, 1986). Later on, karst
research in the state blossomed, largely because so many people were
impacted by karst. Perhaps the ground-breaking moment occurred
when the great Winter Park sinkhole formed in 1981 near Orlando.
There, a massive sinkhole about 100-m wide and 27-m deep formed
over the course of a few days, while destroying many local businesses
and some transportation routes. This event spawned a huge public
interest in the formation of sinkholes in Florida and the risk they pose
to the population.
Thus, over the last few decades, much of the karst research that
occurred in the state was applied in nature and many studies focused
on engineering aspects of sinkholes such as how best to stabilize
sinkholes (Sowers, 1984), how to detect sinkholes (Horwitz and Smith,
2003), and how human activity influences sinkhole formation
(Sinclair, 1982; Bengtsson, 1987; Randazzo and Smith, 2003). Over
that time period, the Florida Sinkhole Research Center collected a
database of sinkhole events (Florida Department of Environmental
Protection, 2006) and new laws were enacted for insurance of
property and protection of sinkhole features (Florida Legislature,
2006).
Recently, there has been growing interest on understanding what
the distribution of sinkhole events and existing karst depressions tell
us about landscape formation and how humans are interacting with
the landscape. For example, Tihansky (1999) demonstrated that
sinkhole distribution in west-central Florida is largely associated
with the thickness of the surface cover over bedrock. It is important to
stress that while there is a distinct natural cause to where sinkholes
are located, Wilson (2004) and Brinkmann et al. (2007) found that
many older (fossil) sinkholes are destroyed in the development
process in cities.
2.1. The city of Tampa
Tampa is a city of 303,447 residents located on the western edge of
the Florida Peninsula (Fig. 2). It is bisected by the meandering
Hillsborough River that drains into the Tampa Bay estuary in the
downtown area of the city. There are numerous sinkholes and
sinkhole lakes throughout the city, as well as many springs. Within
these latter, Sulphur Springs is a high-discharge spring that releases
1.25 m
3
/s of water into the Hillsborough River north of downtown
Tampa. The landscape of the city, like much of west-central Florida, is
dominated by karst features that developed in the covered karst
terrain. The coastal fringes on Tampa Bay are erosional and do not
contain any marine features except for the presence of a drowned
mangrove coastline. In many areas of the city, the coastline has been
Fig. 1. Typical cross section of sinkhole formation in the Tampa area. The cross section consists of Quaternary marine sands at the surface, a Miocene/Pliocene aged confining clay unit
known as the Hawthorne Group, and Eocene Ocala Limestone bedrock. Sediment ravels slowly into void space into the limestone to create a subdued sinkhole feature at the surface.
170 R. Brinkmann et al. / Engineering Geology 99 (2008) 169–18 4
replaced by a sea wall. Tampa is a relatively new city. It developed
largely in the 20th century, first around tourism, cigar manufacturing,
and phosphate mining. With the advent of air conditioning, the
region's economy diversified. Currently, the Tampa region has one of
the most successful and diversified economies in the southeastern
United States. The population of the Tampa region doubles approxi-
mately every twenty years due to the many economic opportunities
that exist in the area. One of the key areas of economic growth is real
estate development. This industry has an interest in karst science due
to the possible sinkhole damage to homes. While some of the karst
depressions were certainly modified through the development
process, the mapping of sinkholes in the city should still provide
considerable information on the characteristics of karst depressions in
the region. At this regard, a recent study that interpreted sinkhole
distribution within the highly urbanized areas of Pinellas County, in
western Tampa, was very illuminating (Wilson, 2004).
3. Methods
There are two databases that we used to analyze the characteristics
of sinkholes in Tampa. The first one is the sinkhole database of the
Florida Sinkhole Research Institute, currently maintained by the
Florida Geological Survey (FGS), and the second one is a database of
topographic sinkholes that we created for this project. The first
database will be referred to as the FGS database and the second
database will be referred to as the topographic database. Each
database has its limitations, but together they provide an interesting
picture of the characteristics of sinkholes in the city. Previous work on
Fig. 2. The location of sinkholes mapped by the Florida Sinkhole Research Institute and the Florida Geological Survey. One area of clustered sinkholes (circled) exists north and south
of the Hillsborough River.
171R. Brinkmann et al. / Engineering Geology 99 (2008) 169–184
sinkholes in the urbanized Tampa Bay area has demonstrated the
problems of using air photos or satellite imagery in correctly
identifying sinkholes, and particularly the difficulty of discriminating
differences between natural and anthropogenic depressions. Thus, we
chose to use topographic maps as the best possible way to map
sinkholes in the region quickly and accurately.
3.1. FGS database
The FGS database consists solely of sinkholes that formed between
approximately 1960 and the present (Fig. 2). The database was
managed by the Florida Sinkhole Institute from the 1980s until the
early 1990s. At that time, the responsibility for managing the database
was given to the FGS. The Florida Sinkhole Institute collected reports
of sinkhole occurrences dating back to 1960. To do this, they relied on
historical records, to which contemporaneous sinkhole events were
added. Included in the database are spectacular sinkholes, such as the
Winter Park event, and subtler sinkholes that might cause small
foundation cracks. Many of the sinkhole events are not necessarily
large topographic features. Instead, they were included in the
database because some type of property damage occurred, regardless
of the sinkhole size. Thus, in many ways, the database has its roots in
Fig. 3. A. The distribution of topographic sinkholes. Note the grouping of sinkholes in three areas: A) a cluster of larger sinkholes in Carrolwood area in north-western Tampa; B) the
area south of the Hillsborough River in the Sulphur Springs area of eastern Tampa: C) the New Tampa area in the northern area of the city. B. In this detail from the map in the
Carrolwood Area within cluster A, one can see that the sinkholes vary by size, shape, and orientation. Nevertheless, some patterns are observable.
172 R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
engineering and urban planning and not in geology. The Florida
Geological Survey continues this approach today and collects records
of both noteworthy and subtle sinkholes.
In order to assess the distribution of the sinkholes within the city of
Tampa, the database was downloaded from the FGS website (Florida
Department of Environmental Protection, 2006), it was queried to
select only those sinkholes located within the city limits (28 total), and
the data were analyzed within a Geographic Information System (GIS).
3.2. Topographic database
The topographic database was constructed by digitizing karst
depressions and lakes located within several 1:24,000 USGS topo-
graphic maps (originally published in 1984, 1987, and 1995) of the City
of Tampa that were compiled within a GIS spatial database (Fig. 3A).
Mapping karst features at this scale has undoubtedly some limitations,
due to the 1.5 m (5 ft) contour interval and to the impossibility of
representing small features. Notwithstanding these limits, topo-
graphic maps have often been used, even at smaller scale, in karst
morphometry studies, since they provide the main source of
information on landscape features over large areas, and the database
extracted from the maps are particularly suitable to statistical analysis.
Even within a city, as is the case for this study, many depressions can
still be identified at the 1:24,000 scale. In addition, there are
limitations with the accuracy of what one can digitize from a
topographic map. In this study, we digitized depressions and lakes
using the standard topographic map symbols as a guide. A total of 293
sinkholes were measured in this way. A detailed example of the
dataset used in the study is shown in Fig. 3B.
4. Morphometry and statistics using the topographic database
The importance of morphometric analysis in karst areas has been
well highlighted in the last decades, starting from the works by La
Valle (1968),Williams (1966, 1971, 1972) and Day (1976, 1978). More
recently, the use of GIS has made rapid and spatially distributed
analysis in different karst regions of the world possible (Hung et al.,
2002; Denizman, 2003; Florea, 2005; Gao et al., 2005). Aim of the
morphometric approach is to describe karst landforms and to look for
any relationships between their distribution and density with other
relevant factors (Wopat, 1974; White and White, 1979), from geology
to hydrology, topography, and, last but not least, the anthropogenic
environment. In a low topography environment as Florida, the
morphometric analysis of karst landforms may result particularly useful
to identify relationships between some of the factors above that are not
directly or easily appreciated due to the overall flatness of the land.
Each depression was catalogued in the Topographic Database and
assigned the following attributes: size in m
2
, perimeter in m, longest
axis length in meters, orientation of longest axis in degrees, and the
circularity index. Depth was not measured because most sinkholes fall
within one or two contour intervals of the map due to the subtle
Fig. 3 (continued ).
Table 1
Measurements calculated from the topographic database
Variable Definition
Size in m
3
This variable is a measurement of the area of the
depression. It is calculated based on the perimeter
of the lake or the depression contour.
Perimeter in m This variable is defined as the length of the
perimeter of the lake or the depression contour.
Longest axis length in m This variable is defined as the length of the
longest axis in the polygon created by mapping the
perimeter of the lake or the depression contour.
Orientation of the longest
axis length in degrees
This variable is defined as the angle, measured
clockwise from north, of the longest axis in the
polygon created by mapping the perimeter
of the lake or the depression contour.
Circularity Index This variable is defined as the ratio between the
measured area of the depression and the area of a
circle with the same perimeter as the depression.
173R. Brinkmann et al. / Engineering Geology 99 (2008) 169–18 4
nature of the features in Florida and the relative indistinctness of the
topographic features.
Once the database was completed, general statistics were
computed on each of the variables (Table 1). Range, mean, and
standard deviation were obtained once the data were standardized
and outliers of very large sinkhole lakes removed. In addition, a
stepwise cluster analysis was completed to determine if any patterns
were evident within the datasets of each of the variables.
4.1. Area
The statistical summary of sinkhole area is shown in Table 2. The
average area of sinkholes in the city is 18,423 m
2
with a standard
deviation of 39,416 m
2
. The smallest detectable sinkhole is only
457 m
2
while the largest is over half a square kilometer in area. The
data distribution of the area is shown in Fig. 4. The data clearly have a
non-normal distribution in that sinkhole area in the city is log
distributed. There are many small sinkholes and fewer large sinkholes,
which is a relatively common phenomenon with landforms of this
type. Even when the largest of these sinkholes are removed, the data
are still non-normal (Fig. 4B). When normalized (Fig. 4C), the average
sinkhole area is 7447 m
2
and the standard deviation is 3.7 m
2
(Table 2).
Fig. 4. The mathematical distribution of sinkhole area: A) note the skewed nature of the distribution; B) sinkhole area distribution with extreme outliers removed; C) distribution of
log transformed sinkhole area.
Table 2
Basic statistics of sinkhole area
n=293 Area (m
2
) Area (m
2
)
(with outlier removed)
Area (m
2
)
(normalized)
Range 457–524,681 457–194,296 457–524,683
Mean 18,423 16,698 7447
SD 39,416 25,985 3.7
SD is for standard deviation
174 R. Brinkmann et al. / Engineering Geology 99 (2008) 169–18 4
A closer look at the distribution reveals that the vast majority of
sinkholes are less than 50,000 m
2
(Fig. 4A). In addition, it is clear that
there is one exceptionallylarge sinkhole that is rather unusual and not
like the others.
A two-step cluster analysis was performed on the area parameter
and three groups were identified (Table 3) and defined by size. The
largest sinkholes were defined as having an area greater than
97,000 m
2
. There are only six sinkholes in this group. In contrast,
the vast majority of the sinkholes (n=240) are less than 27,000 m
2
in
size and are classified as small. A medium-sized sinkhole group
consisting of 47 sinkholes falls between these two size classes. The
distribution of these three sinkhole groups was mapped (Fig. 5). Four
of the largest sinkholes were found south of the Hillsborough River in
the east Tampa sinkhole group. This is the same area where sinkhole
activity is present today. Two of the other large sinkholes are
associated with lakes in the northwest Tampa sinkhole group. There
is no particular distribution evident for small and mid-sized sinkholes.
They are found scattered in all sinkhole groups and diffusely
throughout the city. However, it is important to note that the north
Tampa sinkhole group contains no large sinkholes.
Table 3
Classification of sinkholes by size
Cluster type Number of sinkholes Size range (m
2
)
Small sinkholes 240 0–26,999
Medium sized sinkholes 47 27,000–96,999
Large sinkholes 6 N97,000
Fig. 5. Using a two-step cluster analysis, the area of sinkholes can be broken down into three distinct sizes (see Table 3).
175R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
4.2. Perimeter
Sinkhole perimeter is defined as the length of the outline of the
sinkhole area. Perimeter is often related to area, although highly
crenulated sinkholes can have a long perimeter, but a relatively small
area. The general statistics of sinkhole perimeter is listed on Table 4.
The sinkhole perimeter ranges from 81 m to 4499 m and the average
sinkhole perimeter is 493 m. The standard deviation of the population
is 467 m. A graph of sinkhole perimeter is shown in Fig. 6A. It is clear
that the population is not normally distributed and that, like area,
there are extreme outliers. When these outliers are removed, the
population is still not normally distributed (Fig. 6B). Thus, a log
transformation was performed to normalize the data (Fig. 6C). When
this is completed, the average of the perimeter is 369 m with a
standard deviation of 2.1 m.
The two-step cluster analysis allowed to identify two groups of
sinkholes (Table 5): those with a perimeter less than 860 m and those
with a perimeter greater than 860 m. There are many more short
perimeter sinkholes (n=255) than there are long perimeter sinkholes
(n=38). These two groups were mapped to discern any patterns
within the sinkhole regions (Fig. 7). There are long perimeter
sinkholes in each of the three sinkhole cluster areas, but very few
Table 4
Basic statistics of sinkhole perimeter
n=293 Perimeter
(m)
Perimeter (m) (with outlier
removed)
Perimeter (m)
(normalized)
Range 81–4499 81–2517 81–4499
Mean 493 479 369
SD 467 404 2.1
SD is for standard deviation
Fig. 6. The mathematical distribution of sinkhole perimeter: A) note the skewed nature of the distribution; B) population of sinkhole perimeter with outliers removed; C) when
transformed, sinkhole perimeter has a normal distribution.
Table 5
Two-step clustering approach for sinkhole perimeter
Cluster type Number of sinkholes Perimeter (m)
Small perimeter 255 81–860
Large perimeter 38 861–4500
176 R. Brinkmann et al. / Engineering Geology 99 (2008) 169–18 4
long perimeter sinkholes outside of these groups. Small perimeter
sinkholes are most common within the three groups, but are also
found scattered throughout the city.
4.3. Diameter
Sinkhole diameter is defined as the longest line that can be drawn
through the sinkhole outline without intersecting the perimeter. The
basic statistics associated with sinkhole diameter in the study area are
shown in Table 6. The diameter range is 31–1144 m with an average
diameter of 183 m. The standard deviation of the population is 153 m.
The mathematical distribution of the sample suite is shown in Fig. 8A.
Clearly the data are skewed. When log normalized (Fig. 8B), the
average diameter length is 140 m and the standard deviation is 2 m.
Based on diameter, two distinct groups of sinkholes were defined by
means of a two-step cluster analysis (Table 7): those sinkholes with
diameters over 250 m (n=58) and those with diameters less than
250 m (n=225). When these two groups are mapped (Fig. 9), very few
patterns emerge.
Fig. 7. The distribution of the two groups of sinkhole perimeter using a two-step clustering analysis.
Table 6
Basic statistics for sinkhole diameter
n=293 Diameter (m) Diameter (m) (normalized)
Range 31–114 4 31 –114 4
Mean 183 140
SD 153 2.0
SD is for standard deviation
177R. Brinkmann et al. / Engineering Geology 99 (2008) 169–184
4.4. Circularity
The most typical shape of a depression in karst is circular. In this case,
it may be simplyexpressed by area,perimeter,and diameter of the circle.
A way to discriminate between circular and non circular depression
is the use of the circularity index. The circularity of a sinkhole is mea-
sured by comparing the diameter and area of a sinkhole with the area
of a circle of the same diameter of the sinkhole. A circularity index of
1 is a perfect circle. Sinkholes have a circularity index of less than 1. In
the study area, the range of circularityvalues ranges from 0.153 to 0.954
(Table 8). The mean circularity is 0.715 and the standard deviation
is 0.173. The mathematical distribution of the circularity index may be
seen in Fig. 10. It is evident that the data are normally distributed,
although they are skewed to the right, which means that the population
is more circular than not circular. This makes sense when one considers
that the natural form of sinkholes is circular. However, it must be noted
that multiple coalescing sinkholes have a circularity index that is much
less than 1. Coalescing sinkhole landscapes are older than landscapes
with individual circular depressions.
The sinkhole circularity figures were divided into three clusters
using a two-step clustering analysis (Table 9): circular (n=143),
somewhat circular (n=89), and not circular (n=61). It is interesting
that the cluster with the largest number of sinkholes is the circular
group. There are not many depressions in the region that are irregular
or non-circular, which indicates that the landscape is a relatively
young karst landscape. This is not particularly surprising since much
of the landscape is less than 30 m above sea level and was inundated
in the Quaternary. Areas at higherelevations outside of Tampa contain
much more complex and less circular karst depressions. The three
circularity classes were mapped (Fig. 11) and distinct patterns emerge.
The lowest circularity indexes are located in the east Tampa and north
Tampa sinkhole groups. Very few of these irregular karst depressions
are found outside of these two areas. Instead, the northwest sinkhole
group and the areas outside of the three groups contain mainly
circular sinkholes.
4.5. Orientation
The orientation, or azimuth, of a sinkhole is calculated as the
orientation of the long axis from which the sinkhole diameter is
calculated. The orientation of the sinkhole population ranges from 1 to
179°, with a mean orientation of 91°, and a standard deviation of 56°.
The population distribution graph in Fig. 12A indicates that the data
are bimodal. Interestingly, the bimodal distribution is such that there
is a distinct separation at 90° and each bimodal peak has a distinct
normal pattern to its distribution (Fig. 12B and C). The peaks of each of
the curves are at approximately 30° and 140°.
The orientation data were examined using a two-step clustering
process. Two distinct clusters were identified with the sample
population essentially split between the two groups. As clearly
indicated in the bimodal distribution of the sample, one group
contained sinkholes that were oriented between 1 and 89° and the
other contained sinkholes that were oriented between 90 and 180°.
The two peaks at 30° and 140° and the distinct clustering of the
sinkhole orientations indicates distinct joint control in the orientation
of the long axis of the sinkholes. The two clusters were mapped
(Fig. 13) and no distinct pattern developed that indicated that some
areas were more likely to develop one particular orientation over
another. Instead, the distribution suggests that the mechanisms and
controls for the development of sinkhole orientation act uniformly
across the study area. These results are interesting in that they verify
the works done by Littlefield (1988) and Upchurch and Littlefield
(1988) who suggested that the distribution of sinkholes in the region
was influenced by location of the bedrock lineaments.
5. Results and discussion
The results section is divided into two sections: a description and
analysis of sinkholes that formed in the area since 1960 using the FGS
sinkhole database and a summary of the locational attributes of fossil
topographic sinkholes using the topographic sinkhole database.
5.1. Sinkholes that formed since 1960 —the FGS database
Using the Florida Sinkhole Research Center's database on sink-
hole events, all the sinkholes in the Tampa Metropolitan Statistical
Fig. 8. The mathematical distribution of sinkhole diameter: A) note the skewed nature of the distribution; B) when transformed, sinkhole diameter has a normal distribution.
Table 7
Two-step clustering approach for sinkhole diameter
Cluster type Number of sinkholes Diameter range (m)
Long diameter 58 251–1144
Short diameter 225 31–250
178 R. Brinkmann et al. / Engineering Geology 99 (2008) 169–18 4
area since 1960 were mapped (Fig. 2). A total of twenty-eight
sinkholes were reported within the city between 1960 and 2006.
This is a quite low rate of formation (only 0.6 sinkholes per year).
There are often media reports of sinkholes in the Tampa area and
there are numerous sinkhole insurance claims that occur in the
region as well. One of the main reasons of incompleteness of the FGS
database is that many people are not reporting sinkhole events,
probably because of the implication for property values. Thus,
individuals are likely conducting property repairs or filling in voids
without making official reports. It must be noted that reporting of
sinkhole events to the FGS is voluntary. Thus, given the negative
impact that sinkholes can have on property values, there are few
incentives to reporting sinkholes. The rate of 0.6 sinkholes/year is
certainly a modest rate. However, the rate does demonstrate the
significance of sinkhole development on the landscape evolution in
the region.
The location of the sinkholes mapped using the FGS database
provides clues as to the more active areas of karst landscape
Fig. 9. Distribution of the two classes of sinkhole diameter based on a two-step cluster analysis.
Table 8
Basic statistics for the circularity index
n=293 Circularity
Range 0.153–0.954
Mean 0.715
SD 0.173
SD is for standard deviation
179R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
development, and thus the most likely hazard to residents of the City
of Tampa. The majority of the sinkholes are located in one cluster in
east Tampa north or south of the Hillsborough River. This is an area of
the community that developed between 1940 and 1970. Many of the
springs in the city are located in this area as well. The fact that the
Hillsborough River roughly bisects this sinkhole cluster suggests that
the river was captured within a karst valley. Only a few sinkholes
occurred on the Interbay Peninsula that separates Tampa Bay from Old
Tampa Bay. This is an older portion of the city and it is also the most
densely populated. Campbell (1984) notes that the sands that cover
the bedrock in the region are thin near the central portion of the
county. Interestingly, there are very few sinkholes that were mapped
in the northern portion of the city where sand cover is thickest. Here,
only one sinkhole was identified. However, it must be noted that this is
a newly developed portion of the city. It was largely rural until it was
annexed by the City of Tampa in the 1990s. Much of the development
occurred here between the mid 1990s and the present. There have
been recent media reports of sinkhole occurrences in this area. Thus, it
is quite likely that sinkholes were forming in this area, but they were
not noticed until recently.
5.2. Fossil sinkholes —the topographic database
There were 293 karst depressions mapped within the City of
Tampa for this project (Fig. 3). Interestingly, in contrast to the FGS
database, the vast majority of them are located within the undeve-
loped or newly developed portions of the city. Older portions of the
city, particularly near downtown and the Interbay Peninsula are
largely devoid of sinkholes. However, there is one area of the city, east-
central Tampa, where the location of topographic sinkholes is
conterminous with the location where the greatest cluster of
sinkholes was found using the FGS database.
Sinkholes occur within three distinct clusters in Tampa. The
clusters are defined as those areas where the density of sinkholes is
greatest on the landscape. While it might appear that some of the
clusters may in fact be conterminous, there are distinct topographic
differences between the three clustered regions that help to delineate
the areas. The first cluster is located within the northwestern area of
the city in the Carrolwood area (Cluster A in Fig. 3) which developed
between the 1940s and the 1980s. Here, the sinkholes are large and
typically in the form of lakes. These sinkholes have been preserved
through the development process and serve as amenities for high-end
housing developments. Most of these lakes are used for a variety of
recreational activities, including fishing, waterskiing, and swimming.
The second cluster of sinkholes (Cluster C in Fig. 3) is located in the
northern portion of the city in an area known as New Tampa, an area
that has been undergoing development since the 1990s. Here, the
sinkholes occur in three main forms: lakes, wetlands, and dry
depressions. As in Carrolwood, the lakes are typically used as
amenities for housing developments. This locally occurs even for
wetlands and dry depressions, although they can be destroyed with
development increase. Thus, the presence of perennial lakes is an
important factor in the preservation of sinkhole forms in urbanized
areas of Florida. The third area of clustering is the area of east Tampa
south of the Hillsborough River (Cluster B in Fig. 3) where sinkholes
are likely to form as indicated by the FGS database.
The northwestern cluster is bounded on the south by the no rthern
edge of the Interbay Peninsula, on the west by lowlands that extend
north from the northern fringe of Tampa Bay and to the east by the
lowlands of the Hillsborough River. The northern edge of the
northwestern cluster abuts the New Tampa sinkhole cluster, which
is much wider than the northwestern cluster. In the New Tampa area,
the eastern fringe of the sinkhole cluster is the Green Swamp
lowland. The New Tampa cluster extends beyond the northern and
western edges of the City of Tampa into a broad undulating sinkhole
plain.
The east Tampa cluster is associated with a complex karst
environment where there are springs and where sinkholes are known
to occur with some regularity. This area is bisected by the Hillsborough
River, which has likely been captured as a result of karst collapse.
6. Hazards implications
The results of the analysis of the FGS sinkhole database indicate
that sinkholes, while able to form anywhere within the study area, are
most likely to form in the east Tampa portion of the city that
developed in the middle of the 20th century. There, official reports of
sinkholes suggest that at least one will occur every two years,
although the reporting rate is quite low and it is likely that the real rate
is somewhat higher. The sinkhole hazardin the rest of the city is lower,
although it is difficult to assess the hazard to the north Tampa area
because there are few official records of sinkhole formation in this
newly forming part of the city. Nevertheless, media reports of home
damage from sinkhole activity in the region have been reported.
6.1. Implications for the understanding of the Karst evolution in Tampa
An analysis of the topographic information in conjunction with
information available from the Florida sinkhole database provides
evidence for the understanding of karst landscape evolution in this
part of Florida. It is clear that the processes that act to form sinkholes
on the landscape are not acting at the same rate at all times. There are
three clusters of sinkholes within the city limits. Each will be
discussed below. In addition, the areas outside of the clusters provide
interesting information about the development of the landscape.
Cluster A: The northwest TampaCluster. Thissinkhole group consists
of a groupof large, round sinkholes on the northwest fringeof the city. As
shown in Fig. 3, this cluster extends northwards out of Tampa for
approximately 10 km. This sinkhole groupis distinctly different from the
others in that the depressions are larger, rounded, and not associated
Fig. 10. Distribution of the circularity index. The data are skewed to the right (more
circular than not).
Table 9
Two-step clustering approach for circularity index of sinkholes
Cluster type Number of sinkholes Circularity Index
Circular 143 1–0.772
Semi-circular 89 0.771–0.564
Not circular 61 0.563–0.152
180 R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
with many smaller sinkholes. The distinct morphometric characteristics
suggests that there are distinct reasons for the formation of these
sinkholes that is different from the other groups.
Cluster B: The East Tampa Cluster. This cluster consists of a group of
sinkholes that occur south of the Hillsborough River in the eastern
portion of the city. This cluster partly corresponds to the cluster of
active sinkhole activity that was identified using the FGS database.
This is one of the more interesting karst areas in the region. There are
springs, solution valleys, and many sinkholes, some of which are quite
large. The fact that sinkholes are active in this area suggests that there
are distinct geologic conditions that enhance sinkhole formation here.
The sinkholes consist of circular to non-circular forms.
Cluster C: The North Tampa Cluster. This cluster is associated with a
low sinkhole plain on the northern fringes of the city. There are few
reports of sinkholes forming in this area between 1960 and the present,
largelybecause there werefew people living in this area up until thelast
decade.It will be interestingto monitor this area to determine if sinkhole
reports increase in the future now that the area has developed. The
mapped topographic landforms are small and consist of circular and
non-circular depressions. Like the east Tampa area, this region has a
more mature sinkhole landscape than other areas of Tampa.
Outside of the Clusters, the remainder of the city of Tampa has few
sinkholes and very little evidence of active sinkhole formation. Indeed,
there are areas within Tampa where sinkhole formation, although not
unknown, is quite rare. Many of these areas aren ear Tampa Bay and are at
very low elevations. In addition, some of them have very little cover above
the bedrock, so the raveling of sand, which causes so many problems for
home owners, is unusual. There is another explanation why some of these
areas and areas away from the coast do not have many topographic
sinkholes. Urbanization destroys many subtle topographic features.
Fig. 11. The distribution of the three groups of sinkholes defined by circularity index.
181R. Brinkmann et al. / Engineering Geology 99 (2008) 169–184
Evidence from earlier work in the Tampa Bay area indicate that over half
of all sinkholes are lost through development in Florida.
7. Conclusions
This studydemonstratesthe spatial characteristicsof sinkholes in the
City of Tampa using two different databases. While there are limitations
to mapping using each dataset, some conclusions can be made: 1)
Sinkholes are not distributed regularly across the landscape. There are
three main clusters that can be seen in the datasets. The distribution
provides evidence that there are different geological factors that
influence the pattern. 2) Sinkholes are usually more circular than not.
This indicates that the sinkholes in theregion are relatively young. Older
karst depressions may be filled by sediment deposited during the last
sea level high. 3) Sinkholes are occurring at a rapid rate. The historic FGS
sinkhole database provides clear evidence that sinkhole formation rates
are not geologically slow. Indeed, over theshort span of theexistence of
the database, clear patterns emerge. Finally, it is important to mention
that mapping of sinkholes always represents a very difficult task: they
are often a subtle phenomenon, can be easily modified or even canceled
by human activities, and range in size from very small to very large.
Nevertheless, it is worth producingan effort to defineand map sinkholes
using existing as well as properly built databases to better understand
the geologic history and the hazard associated with living within an
active karst environment. The outcomes from this type of study can
provide thebasis for further and more detailed analysis and, at the same
time, for land use management in densely populated karst areas. The
study also demonstrates the need to continue efforts at mapping new
sinkhole events within the active karst plain of Florida. Future research
should focus on the main geologic factors that seem to influence the
distributional variation found in this research.
Acknowledgments
The Authors express their gratitude to Philip LaMoreaux and to two
anonymous reviewers for the usefulcomments on the first version of the
Fig. 12. Frequency distribution of orientation: A) note the bimodal distribution of the data; B) frequency distribution of the azimuth of the low angle cluster; C) frequency distribution
of the azimuth of the high angle cluster.
182 R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
article. We also benefitted from further suggestions from the Engineer-
ing Geology's Editors-in-Chief Giovanni Crosta and Roy Shlemon.
Support for Mario Parise's work at the University of Florida was provided
by a grant from the National Research Council of Italy (Short-Term
Mobility Program –Year 2005, Research Project: “Assessment of the
hazard related to karst sinkholes in urban and peri-urban areas”).
References
Bengtsson, T.O., 1987. The hydrologic effects from intense ground-water pumpagein
east-central Hillsborough County, Florida. In: Beck, B.G., Wilson, W.L. (Eds.), Karst
hydrogeology: engineering and environmental applications, pp. 109–114.
Brinkmann, R., Wilson, K., Elko, N., Seale, L.D., Florea, L., Vacher, L., 2007. Sinkhole
distribution based on pre-development mapping in urbanized Pinellas County,
Florida, USA. In: Parise, M., Gunn, J. (Eds.), Natural and Anthropogenic Hazards in
Karst Areas: Recognition, Analysis and Mitigation. Geol. Soc. London, spec. paper,
vol. 279, pp. 5–11.
Campbell, 1984. Geology of Hillsborough County, Florida. Open File Report, vol. 6.
Florida Geological Survey (17 pp.).
Cooley, T., 2002. Geological and geotechnical context of cover collapse and subsidence
in mid-continent US clay-mantled karst. Environ. Geol. 42, 469–475.
Day, M., 1976. The morphology and hydrology of some Jamaican karst depressions.
Earth Surf. Proc. 1, 111–129.
Day, M., 1978. Morphology and distribution of residual limestone hills (mogotes) in the
karst of northern Puerto Rico. Bull. Geol. Soc. Am. 89, 426–432.
Denizman, C., 2003. Morphometric and spatial distribution parameters of karstic
depressions, lower Suwannee River Basin, Florida. J. Cave Karst Stud. 65 (1), 29–35.
Eastman, K.L., Butler, A.M., Lilly, C., 1995. The effect of mandating sinkhole coverage in
Florida homeowners insurance policies. CPCU J. 9, 165–176.
Florea, L., 2005. Using state-wide GIS data to identify the coincidence between
sinkholes and geologic structure. J. Cave Karst Stud. 67 (2), 120–124.
Florida Department of Environmental Protection, 2006. Sinkhole database. http://www.
dep.state.fl.us/geology/gisdatamaps/sinkhole_database.htm.
Fig. 13. The two sinkhole orientation groups identified using a two-step cluster analysis.
183R. Brinkmann et al. / Engineering Geology 99 (2008) 169–184
Florida Legislature, 2006. Statute 627.706. http:/ /www.leg.state.fl.us/Statutes/ind ex.
cfm?App_mode=Display_Statute&Search_String=&URL=Ch0627/SEC706.
HTM&Title=-N2005-NCh0627-NSection%20706#0627.706.
Ford, D.C., Williams, P.W., 1989. Karst Geomorphology and Hydrology. Unwin Hyman,
London (601 pp.).
Gao, Y., Alexander Jr., E.C., Barnes, R.J., 2005. Karst database implementation in
Minnesota: analysis of sinkhole distribution. Environ. Geol. 47, 1083–1098.
Hansel, A.K., 1980. Sinkhole form as an indicator of process in karst landscape evolution.
Ph.D. Thesis, University of Illinois, Urbana-Champaign.
Horwitz, M.H., Smith, T.J., 2003. Characteristics of stable and reactivated in-filled paleo-
karst features in west-central Florida. Proc. 9th Multidisc. Conf. on Sinkholes and
the Engineering and Environmental Impact of Karst, Huntsville, Alabama. ASCE
publ., vol. 122, pp. 50–60.
Hung,L.Q.,Dinh,N.Q.,Batelaan,O.,Tam,V.T.,Lagrou,D.,2002.RemotesensingandGIS-based
analysis of cave development in the Suoimuoi catchment (Son La –NW Vietnam).
J. Cave Karst Stud. 64 (1), 23–33.
Klimchouk, A., 2002. Subsidence hazard in different types of karst: evolutionary and
speleogenetic approach. Int. J. Speleol. 31 (1/4), 5–18.
Lane, E., 1986. Karst in Florida. sp. publ., vol. 29. Florida Geological Survey, Tallahassee,
Florida.
La Valle, P., 1968. Karst depression morphology in south-central Kentucky. Geogr. Ann.
50 A, 94–108.
Littlefield, J.R., 1988. Evaluation of data for sinkhole-development risk models:
west-central Florida. Uni versity of South Florida Geology Master's Thesis. 75 p.
Maroney, P.F., Cole, C.R., Corbett, R.B., Dumm, R.E., Eastman, K.L., Gatzlaff, K.M.,
McCollough, K.A., 2005. Final report: insurance study of sinkholes. State of Florida
Report (130 pp.).
Randazzo, A.F., Smith, D.L., 2003. Subsidence-induced foundation failures in
Florida's karst terrain. Proc. 9th Multidisc. Conf. on Sinkholes and the
Engineering and Environmental Impact of Karst, Huntsville, Alabama. ASCE
publ., vol. 122, pp. 82–94.
Schmidt, W., 1997. Geomorphology and physiography of Florida. In: Randazzo, A.F.,
Jones, D.S. (Eds.), The geology of Florida. University Press of Florida, pp.1–12.
Sinclair, W.C., 1982. Sinkhole development resulting from groundwater development
in the Tampa area, Florida. U.S. Geol. Survey Water Resources Investigations Report
81–50, p. 19.
Sowers, G.F., 1984. Correction and protection in limestone terrain. Proc. 1st Multidisc.
Conf. on Sinkholes, Orlando, 15–17 October, pp. 373–378.
Sweeting, M.M., 1973. Karst landforms. Columbia University Press, New York. (362 pp.).
Tharp, T.M., 1999. Mechanics of upward propagation of cover-collapse sinkholes.
Eng. Geol. 52 (1–2), 23–33.
Tihansky, A.B., 1999. Sinkholes, West-Central Florida. In: Galloway, D., Jones, D.R.,
Ingebritsen, S.E. (Eds.), Land subsidence in the United States. U.S.G.S. Circular,
vol. 1182, pp. 121–140.
Troester, J.W., White, E.L., White, W.B., 1984. A comparison of sinkhole depth
frequency distributions in temperate and tropical karst regions. In: Beck, B.F.
(Ed.), Sinkholes: their geology, engineering, and environmental impact. Balkema,
Rotterdam, pp. 65–73.
Upchurch, S.B., Littlefield, J.R., 1988. Evaluation of data for sinkhole-development risk
models. Environ. Geol. 12, 135–140.
Upchurch, S.B., Randazzo, A.F., 1997. Environmental geology of Florida. In: Randazzo, A.
F., Jones, D.S. (Eds.), The geology of Florida. University Press ofFlorida, pp. 217–249.
Waltham, T., 2002. The engineering classification of karst with respect to the role and
influence of caves. Int. J. Speleol. 31 (1/4), 19–35.
Waltham, T., Bell, F., Culshaw, M., 2005. Sinkholes and subsidence. Praxis Publ.
White, W.A., 1970. The geomorphology of the Florida peninsula. Geol. Bull., vol. 51. State
of Florida (164 pp.).
White, W.B., 1988. Geomorphology and hydrology of karst terrains. Oxford University
Press.
White, E.L., White, W.B.,1979. Quantitative morphology of landforms in carbonate rock
basins in the Appalachian Highlands. Bull. Geol. Soc. Am. 90, 385–396.
Williams, P.W.,1966. Morphometric analysis of temperate karst landforms. Ir. Speleol. 1,
23–31.
Williams, P.W., 1971. Illustrating morphometric analysis of karst with examples from
New Guinea. Zeit. für Geomorph. 15, 40–61.
Williams, P.W., 1972. Morphometric analysis of polygonal karst in New Guinea. Bull.
Geol. Soc. Am. 83, 761–796.
Wilson, K.V., 2004. Modification of karst depressions by urbanization in Pinellas County,
Florida. University of South Florida Masters Thesis, 68 pp.
Wopat, M.A., 1974. The karst of southeastern Minnesota and methods for statistical
analysis in polymodal two-dimensional orientation data. M.A. thesis, University of
Wisconsin, Madison.
184 R. Brinkmann et al. / Engineering Geology 99 (20 08) 169–184
