Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Analyzing NEXRAD doppler radar images to assess nightlydispersal patterns and population trends in Brazilian free-tailedbats (Tadarida brasiliensis)Jason W. Horn1,* and Thomas H. Kunz*�Center for Ecology and Conservation Biology, Department of Biology, Boston University, Boston, MA 02215 USA
Synopsis Operators of early weather-surveillance radars often observed echoes on their displays that did not behave
like weather pattern, including expanding ring-like shapes they called angels. These echoes were caused by high-flying
insects, migrating birds, and large colonies of bats emerging from roosts to feed. Modern weather-surveillance radar
stations in the United States (NEXt-generation RADar or NEXRAD) provide detailed images that clearly show evening
bat emergences from large colonies. These images can be used to investigate the flight behavior of groups of bats and
population trends in large colonies of Brazilian free-tailed bats (Tadarida brasiliensis) in south-central Texas which are
clearly imaged by local NEXRAD radar stations. In this study, we used radar reflectivity data from the New Braunfels,
Texas NEXRAD station to examine relative colony size, direction of movement, speed of dispersion, and altitude
gradients of bats from these colonies following evening emergence. Base reflectivity clear-air-mode Level-II images were
geo-referenced and compiled in a GIS along with locations of colonies and features on the landscape. Temporal sequences
of images were filtered for the activity of bats, and from this, the relative size of bat colonies, and the speed and heading
of bat emergences were calculated. Our results indicate cyclical changes in colony size from year to year and that initial
headings taken by bats during emergence flights are highly directional. We found that NEXRAD data can be an effective
tool for monitoring the nightly behavior and seasonal changes in these large colonies. Understanding the distribution of
a large regional bat population on a landscape scale has important implications for agricultural pest management and
conservation efforts.
Introduction
The ability to observe free-ranging animals interact-
ing within their environment is essential for under-
standing their behavior and ecology. Increasingly,
studies have broadened the scope of ecological
research from local processes and interactions to
changes at the level of the landscape by employing
remote sensing methods. One such approach is the
use of radar (Radio Detection and Ranging) systems
to study behaviors and patterns of movement that
occur at altitudes and distances not visible to human
observers.
Operators of early radars were often challenged
with interpreting echoes in images that they believed
were of empty airspace (Lack and Varley 1945;
McKay 1945; Buss 1946; Plank 1956). Researchers
have since successfully employed radar to describe
the distribution and movement of smoke, dust,
birds, bats, and insects. The timing, altitude, speed,
and density of nightly and seasonal flights of birds
has been studied extensively by examining echoes
created by groups of animals aloft in ‘‘clear’’ (no
precipitation) air (Schnell 1965; Cooper et al. 1991;
Cooper and Ritchie 1995; Bruderer 1997; Gauthreaux
1998; Alerstam and Gudmundsson 1999; Klaassen
and Biebach 2000; Burger 2001; Williams et al. 2001;
Dinevich et al. 2003). Entomologists have found
that radar echoes can also be attributable to insects
flying at altitudes higher than previously thought
(Crawford 1949; Glover et al. 1966; Richter et al.
1973; Wolf et al. 1995). Williams et al. (1973) con-
firmed by helicopter that the reflectivity on airport
radars in Texas was caused by groups of Brazilian
free-tailed bats (Tadarida brasiliensis) dispersing
from caves to forage.
Early long-range weather-surveillance radars
(WSR-57D) also detected biological phenomena.
Operators noted shapes and movements on their
displays during clear-sky conditions, such as expand-
ing ring-like shapes and long finger-like forma-
tions that did not correspond to weather events.
From the symposium ‘‘Aeroecology: Probing and Modeling the Aerosphere—The Next Frontier’’ presented at the annual meeting of the Societyfor Integrative and Comparative Biology, January 2–6, 2008, San Antonio, Texas.1E-mail: [email protected]
24
Integrative and Comparative Biology, volume 48, number 1, pp. 24–39
doi:10.1093/icb/icn051
Advanced Access publication June 18, 2008
� The Author 2008. Published by Oxford University Press on behalf of the Society for Integrative and Comparative Biology. All rights reserved.
For permissions please email: [email protected].
The current United States National Weather Service
weather-surveillance radars known as WSR-88D or
NEXRAD (NEXt-Generation RADar) have been used
extensively for studying bird migration on a larger
geographic scale (Gauthreaux and Belser 1992;
Gauthreaux and Belser 1998; Russell and Gauthreaux
1998; Diehl and Larkin 2005). Nocturnal and diurnal
exodus flights (large-scale departure of groups of
migrating birds from a roost), overwater migrations,
arrival times, and selection of roost sites can be
examined using images produced from the first 0.58elevation scan (base reflectivity). Flights of large
groups that occur along migration routes can be
monitored, and thus long-term studies to examine
changes in the intensity and scope of migration are
possible. Insect migrations are also detectable in
NEXRAD images (Westbrook and Isard 1999).
Westbrook et al. (1998) were able to confirm that
NEXRAD echoes were due to migrating corn earworm
moths (Helicoverpa zea) by using additional 3 cm radar
that revealed the orientation and size of the targets.
Biologists and meteorologists have also noted
that the timing, location, and seasonal changes in
some clear-air NEXRAD images are consistent with
bat activity. The first report of bats appearing in
NEXRAD images described a unique dispersal
pattern of reflectivity from the location of a known
large colony located in a cave in Oklahoma (Ruthi
1994). NEXRAD radial-velocity images indicated that
bats were dispersing at 25–32 km/h away from the
cave and that large numbers of bats were flying well
above 2000 m above ground level (AGL). Similar
emergences have been observed to originate and then
disperse from most large maternity colonies of
Brazilian free-tailed bats (T. brasiliensis) in Texas
and Oklahoma (McCracken and Westbrook 2002;
Cleveland et al. 2006).
An important and frequently posed question
about NEXRAD images and bats is whether it is
possible to distinguish echoes that are due to bats
from those produced by weather, insects, birds, and
other airborne objects. Reflectivity that is caused by
Brazilian free-tailed bats during evening emergence
from roosts is visually identifiable in sequences of
NEXRAD images because its appearance and move-
ment are correlated with known flight behavior
and life history patterns. Areas of intense reflectivity
that expand and disperse are viewable in NEXRAD
images 30–40 min after the onset of evening emer-
gence from roosts. These expanding shapes are
located directly over the known geographic position
of several large, well-known colonies. These emer-
gence patterns do not appear in winter when bats are
not present, and appear strongest during June and
July when these colonies are at their peak numbers.
Bird migrations and morning departures from roosts
occur at times of night and the early morning that
generally do not overlap with the timing of evening
bat emergence (Diehl and Larkin 2005). Insect con-
centrations appear morphologically distinct on
NEXRAD (Westbrook et al. 1998) and do not origi-
nate from bat colony locations. However, NEXRAD
reflectivity may be a mix of birds, bats, insects,
dust, and anomalous reflections from objects on the
ground. Few attempts have been made to quantify
the underlying flight behavior and population fluc-
tuations of bats by attempting to isolate reflectivity
from bats from other sources based on the timing,
geographic location, and known emergence behavior
of bats.
The present study has two primary goals: (1) to
establish methodology for systematically quantifying
magnitude, movement, and dispersal patterns of
groups of Brazilian free-tailed bats from large
numbers of NEXRAD reflectivity images and (2) to
demonstrate that these data can be used to inves-
tigate population changes and feeding ecology of
Brazilian free-tailed bats in Texas.
Methods
Species studied and study area
The Brazilian free-tailed bat is a semi-tropical species
that typically migrates annually between Mexico and
the south-central United States. It roosts in build-
ings, mines, and bridges, but is best known for the
enormous colonies that it forms annually during the
summer in Texas, Oklahoma, and New Mexico.
These colonies are comprised mostly of females, 96%
of which are pregnant or lactating (Davis et al. 1962;
McCracken and Gustin 1991). These females each
give birth to a single pup, on average, around 10
June (McCracken and Gustin 1991). Young Brazilian
free-tailed bats begin to forage when they are �6
weeks of age (Kunz and Robson 1995), during which
time, mothers return to the roost cave nightly to
nurse their pups. Historical accounts describe the
seasonal maximum population size of these colonies
in the tens of millions (Davis et al. 1962).
McCracken’s (2003) review of available reports
estimated the total population of Brazilian free-
tailed bats in the southwestern United States during
this peak to be 150 million individuals. Brazilian
free-tailed bats are also noteworthy for their high-
altitude and long-distance nightly foraging flights
(Davis et al. 1962; McCracken et al. this sympo-
sium). As Brazilian free-tailed bats emerge from
caves to forage, they typically form a closely-packed
Analyzing NEXRADDoppler radar images 25
stream which rises several hundred meters above
ground before breaking up into smaller groups that
disappear from view. Williams et al. (1973) found
that these bats dispersed from caves at an average
maximum speed of 53 km/h and reached an average
maximum altitude of 2500 m AGL. Brazilian free-
tailed bats radio-tracked from Carlsbad Caverns in
New Mexico have flown as far as 56 km to reach
foraging areas (Best and Geluso 2003) and echoloca-
tion and terminal feeding buzzes have been recorded
at 300 m AGL (Griffin and Thompson 1982;
McCracken et al. 2008).
For our analysis, we selected four colonies inhab-
iting caves and two colonies inhabiting bridges
from among the many colonies that are located in
the Balcones Escarpment area between Austin and
San Antonio, Texas, where the Edwards plateau
lies to the northwest and agricultural areas to the
south and east (Fig. 8). The two colonies inhabiting
Congress Avenue Bridge and McNeil Bridge are
located near Austin Texas. Congress Avenue Bridge is
located just south of the downtown area and McNeil
Bridge is located 25 km to the northeast along
Interstate 35. Two colonies inhabiting Frio Cave
and Ney Cave are located 65 and 115 km west of
San Antonio, respectively. Two other colonies that
inhabit Davis Cave and Eckert James River Cave are
located in the foothills 83 and 156 km northwest of
Austin, respectively. Agricultural landscapes com-
prised largely of cotton, corn, and beets lie northeast
of Austin and south of San Antonio in the ‘‘Winter
Garden’’ area of Texas, where bats are known to feed
on the adult forms of crop pests such as the corn
earworm moth (H. zea) (Lee and McCracken 2005;
Cleveland et al. 2006). The study area contains a
total of 54 known colonies of Brazilian free-tailed
bats; 14 in caves, 38 in bridges, and 2 in sinkholes.
These colonies also contain small numbers of the
cave bat, Myotis velifer, and perhaps other species
(Ritzi 1999). Bracken Cave, widely regarded as
housing the largest colony in the study area, report-
edly contains upwards of 20 million individuals at its
peak population (Davis et al. 1962; McCracken 2003;
but see Betke et al. 2008). The geographic area where
these colonies are found, and the range over which
they are thought to forage (50 km from the roost), is
100,238 km2. The NEXRAD station (KEWX) in New
Braunfels, Texas, is centrally located in this area.
Sampling NEXRAD data
To quantify the nightly emergence behavior of
Brazilian free-tailed bats, we examined archived
NEXRAD data available from the NOAA National
Climatic Data Center (NCDC, http://hurricane.ncdc.
noaa.gov/pls/plhas/has.dsselect). For our analysis,
we only used ‘‘clear-air’’-mode images which are
produced by NEXRAD stations when there is no
local precipitation, and which are more sensitive to
the presence of bats. We selected days for which
clear-air mode data were available while attempting
to achieve as close to an even distribution as possible
for each period of time we examined.
To examine intra-annual seasonal changes in
relative colony size at each of the six sites, we
assembled a data set that included 27 days beginning
January 12, 2005 and ending December 20, 2005. We
collected 904 separate images of bat emergence, with
each day containing between 33 and 34 images taken
at 10 min intervals. We also tested for directional
trends in initial headings taken by bats following
evening emergence using a sub-set of 19 days of
this data set. To examine long-term inter-annual
changes in relative colony size, we obtained images
of emergence from the six colonies spanning an
11-year period (1995–2005). To improve our confi-
dence in estimating colony size for any given year,
we used images from the 5–6 week period following
parturition, when mothers are nursing their devel-
oping pups and are unlikely to relocate to new
roosts. For each year, we selected two days between
June 20 and July 8, from which we used the mean
value as an estimate of the most stable number of
adults in each colony.
Spatially explicit image processing
We estimated relative colony size and initial heading
from radar images of emerging bats by creating a
software-based NEXRAD data analysis system. The
system consists primarily of a custom-written soft-
ware framework and applications using the scripting
language php 5 (http://www.php.net) that extends
the open-source GRASS GIS (Geographic Resources
Analysis and Support System, http://www.geog.
uni-hannover.de/grass/index.php). Together, the appli-
cations make possible geo-referencing, geographic
processing and spatial analysis, data storage and
statistical analysis, and visualization and animation
of NEXRAD data. Native NEXRAD data ‘‘bins’’, each
representing the amount of reflectivity in a wedge-
shaped volume of air, are first converted into a grid
pattern using the NCDC Java NEXRAD Exporter
version 1.0.12 (http://www.ncdc.noaa.gov/oa/radar/
jnx/batch.html). The NEXRAD exporter geo-
references the cells as a grid of latitudinal and
longitudinal coordinates based on a World Geodetic
System spheroid (WGS84) model of the earth.
26 J.W. Horn and T. H. Kunz
Latitude and longitude coordinate grids must be pro-
jected onto a flat surface (a map projection) before
measurement of area, distance, and scale can be
calculated without distortion errors. We used a
custom-written application to project large numbers
of NEXRAD images onto the Texas State Plane
Mapping System (TSMS), a Lambert conformal
conic projection based on a GRS80 spheroid model
of the earth and the 1983 North American Map
Datum (NAD83) that minimizes distortion of
distance and area in maps of Texas. The resulting
geo-referenced NEXRAD layers are raster grids of
1 km2 cells that are 1072 km wide (east-west) and
994 km high (north-south) and are centered at the
New Braunfels, Texas NEXRAD station. NEXRAD
reflectivity is measured in units of decibels Z on
a logarithmic scale. Z is a ratio of the amount of
RADAR energy that is emitted by the station to the
strength of the reflected echo received for a given
volume of airspace. The range (including zero) of
values found in typical Level-II NEXRAD images that
include ground clutter, insects, birds, and bats is
from �25 to 25 dBZ. Object density cannot be
inferred by comparing or summing dBZ values from
volumes of airspace, as dBZ values are on a log10 scale.
To allow comparison of reflectivities in our analysis,
we transformed the dBZ values for each pixel into
Z-values; the ratio of the emitted signal to the strength
of the return echo. In a typical clear-air-mode image
containing bats, these Z-values range from 1 to 10,000,
with 0 indicating no reflectivity. Because NEXRAD
Z-values are linearly proportional to the number of
birds aloft in a volume of space (Gauthreaux and
Belser 1998, 1999; Black and Donaldson 1999), we used
Z-values as estimates of the relative number of bats
aloft in a volume of airspace.
Modeling the emergence of bats
To identify activity of bats in NEXRAD images,
it was necessary to build a model of bat behavior
during emergence. We developed a set of qualitative
and quantitative criteria, based on the foraging
ecology of Brazilian free-tailed bats and the physics
of the NEXRAD beam propagation, for determining
which areas (or patches) within an image contain
reflectivity due to bats. (1) Emergences of free-tailed
bats contain large dispersing groups of individuals.
We chose a threshold-sized area that was large
enough to confidently identify bats, but small
enough to filter unwanted NEXRAD reflectivity.
Therefore, we define only patches of reflectivity
covering at least 35 km2 (35 pixels) to be a group of
bats. (2) Emergences originate from specific roosting
locations and bats disperse outward from these
locations. The first patch of reflectivity from an
emergence must occur no further that 15 km distant
from the location of the colony. (3) Emergences of
bats on NEXRAD manifest as developing, dispersing
patterns. Therefore, reflectivities should at first
appear weak, then become centralized and stronger,
then become weaker and ultimately disappear. (4)
The patch must conform to one of three common
shapes: a characteristic expanding ring centered over
the colony, a partial expanding ring or curved wave
front, or a funnel shape, expanding outward as
distance from the colony increases.
Identifying bat activity
Quantifying the reflectivity in NEXRAD images attrib-
utable to emergence of Brazilian free-tailed bats
requires two steps. First, areas or patches of reflec-
tivity in the images that are due to bats aloft must be
identified and retained while all other reflectivity is
removed. Second, patches of reflectivity in successive
images must be tracked through time to attribute them
to a single emergence event. In the first step, a series of
nine image-analysis procedures are performed in
sequence (Fig. 1). In Procedure 1, an area is selected
near the colony in which bats are expected to be found
(see the next section on ‘‘Tracking movements of bats’’
for a detailed explanation of how these areas were
calculated). In Procedure 2, a contrast filter is applied
that generates a sigmoidal response (from 0 to 1) for all
the pixel values in the masked image using the
following formula
response ¼1
1 þ e�12rx,yþ6where rx,y ¼
rx,y � rmin
�r þ ð�r � rminÞ
where rx,y is the reflectivity value of any pixel at
position x,y in the image, rmin is the minimum
reflectivity value, rmax is the maximum value, and r is
the mean reflectivity value in the image.
This has the effect of lowering lower-than-mean
values and raising higher-than-mean values, condens-
ing patches of reflectivity and separating them from
one another. Procedure 3 masks this image using
the mask from Procedure 1. Procedure 4 removes the
values below the mean value in the Procedure-3
image. Procedure 5 uses the Procedure-4 image as a
mask for the original data, selecting only the pixels
from the original image that match the contrast-
filtered image. Procedure 6 reintroduces original
pixels into the image that are above a threshold value
by using a dilation procedure. This approach retains
below-mean reflectivities that are spatially associated
with likely bat patches, while removing the same
values elsewhere in the image where they may
Analyzing NEXRADDoppler radar images 27
represent clutter or noise. The threshold value is
based on the mean value of the pixels in Procedure
1, but is modulated slightly by the pixel density (the
proportion of non-zero pixels). This is an important
aspect of this procedure because it is common in
NEXRAD sequences for pixel density to increase
as the night sky becomes saturated with dispersed
bats, leading to images where there are a few high
concentrations of bats surrounded by a sea of lower
reflectivities. Thus, the threshold value is raised
slightly as pixel density increases. The threshold is
calculated as follows, where r is reflectivity.
thresholda ¼ �r þ ð2 density� 1Þ � 0:5rstddev
Procedure 7 separates patches of bats that are con-
nected by only a few pixels. This is a critical step
A B
C D
E F
Fig. 1 The main steps in the processing of NEXRAD images to isolate reflectivity generated by dispersing Brazilian free-tailed bats.
Values from a mask (B) of the original image (A) are used to apply a contrast function that removes low Z-values (C). The image is
then threshold (D) to remove lower values. Neighborhood kernel operators are applied to remove small bridges interconnecting
larger patches (E). Finally, a recursive algorithm identifies distinct patches in the image (F).
28 J. W. Horn and T. H. Kunz
because emergence is a dispersal phenomenon and as
emergence progresses, patches of bats from different
source colonies may begin to overlap, leading to over
counting. To separate patches, a series of 25-pixel
neighborhood kernels are passed over each value in
the image, and the center value is either retained or
removed, based on the value of its neighbors in the
kernel. These functions have the effect of breaking
small bridges 1 or 2 pixels wide that connect two
larger patches. An additional kernel function uses
two alternate 9-pixel patterns and effectively removes
any diagonally oriented small bridges connecting
two larger patches. Procedure 8 fills any small holes
with original pixels. The final Procedure (9) uses
a recursive connected-components algorithm that
assigns an ID number to each of the remaining
patches.
Tracking movements of bats
To attribute multiple patches of reflectivity in
successive images to a single bat emergence event,
we applied a tracking algorithm (the Tracker) to
the filtered images (Fig. 2). To accomplish this, the
Tracker first determines the position of each patch
by calculating its value-weighted spatial mean (Dacey
1962). Thus, the center of the patch is determined
both by the geometric shape of the patch, and by the
distribution of values within the patch. The spatial
mean coordinates of a raster image are given by
the following where ri is the reflectivity at location
xi or yi.
�x ¼rixi
riand �y ¼
riyi
ri
The Tracker’s algorithm is based on the criteria
given previously and proceeds as follows: (1) The
Tracker searches for the first patch of reflectivity
within a circular zone of 15 km radius (707 km2)
around the colony’s geographic location. Once a
patch is found in this zone, its center is recorded as
the first point of a new branch that will become a
line of successive points that describes the move-
ment of the center of the patch through time. In
subsequent images, the Tracker searches for patches
whose centers fall within an elliptical model of the
probability of the position of that patch. One focus
of the ellipse is placed at the endpoint of a branch.
The position of the second focus is determined by
the average distance moved by the patch in the
previous two images (T�2 and T�1) and the angle
of the vector from the center of the T�2 patch to the
T�1 patch. The ellipse is then drawn using the
formula given below where e is the eccentricity of
the ellipse, a is 1/2 the major axis length, and r is
a radius from the focus to a point on the ellipse
at angle �.
r ¼a 1 � e2� �
1 þ e cos �
If the center of the present patch falls within this
ellipse, and the endpoint was recorded in the T�1
image, the new patch is added to the endpoint’s
Fig. 2 A sequence of filtered NEXRAD image patches that
indicate dispersing Brazilian free-tailed bats are tracked from
Frio Cave on May 19, 2005. Each is a snapshot of every other
step in the tracking progression. The left panel of each image
shows the unprocessed data, while the right shows processed
data, and the progress of the Tracker. Crosses mark the location
of the spatial mean of the patch. Circles and ellipses mark the
areas that were used as criteria to accept the patches tracked
in that image. The colony location is marked in magenta. The first
wave of the emergence is tracked in yellow, the second in red.
As the emergence progresses, the Tracker effectively records
only the reflectivity that originates from Frio Cave.
Analyzing NEXRADDoppler radar images 29
branch. If not, the Tracker again determines if this
new patch is within the circular zone around the
colony. If the patch is in this area, again, it is
recorded as the first point of a new branch. If the
center of the patch does not fall into any of these
areas, the patch is assumed to be unassociated with
the emergence from the colony being examined and
is discarded. Subsequently, all recorded patches are
measured and the following values are recorded to
a database: (1) number of pixels, (2) sum of all
the reflectivity, (3) mean reflectivity, (4) standard
deviation, (5) coordinates of the spatial mean,
(6) weighted standard distance (spatial standard
deviation), and (7) distance to, and heading from,
the previous point in the branch. The output of the
combined filtering and tracking steps is a group of
patches which, when considered together, represent
the total of all of the reflectivity generated by a single
colony during an emergence period.
Results
Our filtering and tracking system identified and
recorded 642 patches of bats in 548 images. Each
emergence event contained between 1 and 21 patches
(x¼ 7.3� SD¼ 4.6). Both weak (low Z-value) and
strong (high Z-value) areas of bat reflectivity were
identified and tracked. The Tracker successfully
recorded one or more branches of bat patches
when emergence proceeded in a series of several
waves. This observation correlates with observations
of emergence at roost sites, which, when at peak
colony size, typically proceeds in 2–4 waves separated
by pauses of 15–30 min (Frank et al. 2003; Betke
et al. 2007). The number of branches recorded per
emergence does not necessarily indicate the number
of emergence waves, however, as the Tracker occa-
sionally broke a single wave into one or more
branches. Tracking performance was, by design,
conservative, and some weak patches that were
visually identifiable as part of an emergence branch
were not counted. Only in very rare cases did the
filtering and tracking system select areas of reflectiv-
ity that did not appear related to bats.
The seasonal variation in the total reflectivity con-
formed to the pattern of activity typically observed
on the ground at these colonies (Fig. 3). No reflec-
tivity was recorded in January. Reflectivity from
bats was first observed on February 21 at Congress
Avenue Bridge. Total reflectivity increased at the
Frio Cave, Ney Cave, Congress Avenue Bridge, and
McNeil Bridge colonies throughout March and April.
Strength varied in May, June, and July, but decreased
during the first week of June. Reflectivity decreased
markedly in August. Sporadic reflectivity was
recorded in September and October, with some
colonies showing little or no reflectivity (Davis Cave
and Eckert James River Cave), and others showing a
marked increase (McNeil Bridge). No reflectivity was
recorded in December.
During the period from April 1, to August 1,
2005, Frio Cave produced the highest total reflectiv-
ity per night (1.2 MZ). Ney Cave, Congress Avenue
Bridge, and McNeil Bridge produced similar
amounts of reflectivity (712, 947, 628 kZ, respec-
tively). Davis and Eckert James River caves produced
much smaller amounts of reflectivity (14 and
127 kZ), respectively.
The patterns of nightly reflectivity measured at
Frio Cave and Ney Cave are highly correlated with
the expected seasonal migration of Brazilian free-
tailed bats. In early May, a large increase in numbers
of bats typically occurs at these caves, and there is a
corresponding increase in nightly reflectivity. Toward
the end of the summer maternity period, there is a
drop in reflectivity as pups are weaned and adult
females begin to move among different roosts at the
onset of migration. We also observed a correspond-
ing decrease in reflectivity in the last week of July.
The pattern at Frio Cave, Ney Cave, and Congress
Avenue Bridge may also correspond to the timing
of the onset of parturition at these colonies, which
typically occurs between the first and second week
of June. Both Frio Cave and Ney Cave experienced a
significant drop in reflectivity during this time.
The pattern of reflectivity differed significantly
between bridge colonies and cave colonies. Bridge
colonies both showed their first significant increases
in reflectivity in early March, two months before
increases occurred at cave colonies. We observed a
peak of activity at the Congress Avenue and McNeil
bridge colonies on April 15 and April 3, respectively,
while during the same period, almost no reflectivity
was recorded for both Frio Cave and Ney Cave.
Also, the bridge colonies showed higher reflectivity
through the end of October, whereas cave colonies
showed little to no reflectivity after August 1 (Fig. 4).
These data suggest that bats first occupy bridge
colonies when they arrive in Texas following spring
migration, and then later expand into caves in
late spring and early summer. A reverse pattern
of reflectivity is evident in the autumn prior to
migration, when bats appear to move from caves to
bridges.
Mean and total reflectivity both varied annually
(Fig. 5). The highest mean nightly reflectivity (mean
of the total cumulative Z for the two sampled nights
for that year) occurred in 2000, and the lowest in
30 J.W. Horn and T. H. Kunz
2002. An increasing trend in reflectivity was observed
from 1995 through 2000, followed by a precipitous
decrease to the lowest recorded measurements in
2001, 2002, and 2003. From 2002 through 2005, an
increase in annual reflectivity was observed, reaching
the pre-2001 levels. We found no evidence of a
general downward trend in total reflectivity from the
six colonies sampled over an 11-year period.
Nightly emergences were observed on 17 out of
the 19 nights we evaluated in 2005. Using Rayleigh
tests (a statistical measure of directionality) we
found the distributions of headings of patches to
be significantly directional (P50.05) for each of the
six colonies. Mean heading varied from 158 (NNW)
to 1288 (SW). Typically, individual colonies moved
in a predominant direction during nightly emer-
gence. Bat colonies at Frio Cave and Ney Cave
produced emergence reflectivity that typically moved
in a southeasterly direction (x¼ 111.98, � SD¼ 25.78and x¼ 99.28, � SD¼ 58.28). Bats from the McNeil
Bridge colony typically moved toward the northeast
(x¼ 54.68, � SD¼ 49.78). Bats from the Congress
Avenue colony typically dispersed in a northeasterly
direction, but also exhibited several emergences
distinctly oriented toward the southwest as well
(x¼ 109.38, � SD¼ 70.78). Bats from Eckert James
River Cave and Davis Cave both exhibited mean
directionality toward the northeast, but this obser-
vation may not be significant owing to the small
number of times they were observed (n¼ 5, n¼ 3,
respectively).
Significant differences existed between the mean
vectors for all cave colonies versus all bridge colonies
(Fig. 6). The mean vector for emergence from
bridge colonies was directional toward the southeast
(x¼ 116.98, � SD¼ 43.38, Rayleigh test¼ 0.75,
P50.001) and that for cave colonies was oriented
to the northeast, (x¼ 76.38, � SD¼ 66.08, Rayleigh
test¼ 0.51, P50.001). Moreover, the mean vectors
for colony emergences from caves and bridges were
significantly different from each other (Watson
t-test¼ 0.44, P50.05).
We also tested for the differences in headings
between cave colonies and bridge colonies during the
date
Mea
n R
efle
ctiv
ity (
Z)
Feb Apr Jun Aug Octdate
Feb Apr Jun Aug Oct
dateFeb Apr Jun Aug Oct
dateFeb Apr Jun Aug Oct
dateFeb Apr Jun Aug Oct
dateFeb Apr Jun Aug Oct
020
000
5000
0
Mea
n R
efle
ctiv
ity (
Z)
020
000
5000
0
Mea
n R
efle
ctiv
ity (
Z)
020
000
5000
0
Mea
n R
efle
ctiv
ity (
Z)
020
000
5000
0
Mea
n R
efle
ctiv
ity (
Z)
020
000
5000
0
Mea
n R
efle
ctiv
ity (
Z)
020
000
5000
0
Frio
A B
C D
E F
Ney
(c) Davis James River
Congress McNeil
Fig. 3 Seasonal patterns in the level of NEXRAD radar reflectivity recorded for each of six colonies of Brazilian free-tailed bat
colonies in south-central Texas. Cave colonies and bridge colonies reflect a similar population increase beginning in early May (A, B,
E, and F), but bridge colonies (E and F) show radar reflectivity in March and April that is absent from caves (A–D) and may
represent formation of colonies for courtship and mating.
Analyzing NEXRADDoppler radar images 31
spring mating/migration period (before May 15),
the summer maternity and lactation period (May
15 to August 15), and the fall migration period
(after August 15). Emergence headings from bridges
in spring, summer, and autumn and from caves in
spring and summer were significantly directional.
Emergences from caves in autumn were not signif-
icantly directional, most likely owing to the small
number of emergences observed during this period
(n¼ 3). During spring and summer, there were
significant differences in the mean vector of emer-
gence between cave- and bridge-colonies. Thus, we
observed no effect of season on the directionality of
emergences from caves or bridges. We continued this
analysis to the level of the individual colony, but
again found no effect of season on the distribution
of headings.
We analyzed the mean heading of emergences for
all six colonies combined over the same inter-annual
period (1995–2004) for which we found large varia-
tion in total reflectivity. Eight of the ten years
showed significant directionality in emergence head-
ing. Mean heading varied between years, but was
typically in an easterly direction. Emergences were
significantly directional in years 1995 through 2001,
and in 2004, but were not significantly directional for
years 2002 and 2003. Significance in the Rayleigh test
was correlated with the inter-annual pattern of total
reflectivity. As total reflectivity increases, the variance
in direction of emergence decreases. During the
2001–2003 period, when reflectivity was at its lowest
point, direction of emergence was most variable.
To test the effect of wind heading on observed
headings of emerging bats, we compared the
mean surface wind heading measured during the
Mean Total Annual Reflectivity 1995–2005
Date
Mea
n R
efle
ctiv
ity (
Z)
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
020
000
050
000
080
000
01
100
000
140
000
0 all coloniescave coloniesbridge colonies
Fig. 5 Total NEXRAD reflectivity for six colonies of Brazilian free-tailed bat colonies over 11 years in south-central Texas.
Each measurement is the average of two days of reflectivity measure during the June 20–July 8 period, a time when females
are lactating and populations are expected to be stable with the maximum number of adults.
Frio Ney Davis Jms Rvr Congress McNeil
May–August 2005A
B
Colony
Frio Ney Davis Jms Rvr Congress McNeil
Colony
Ref
lect
ivity
(Z
)
0e+
004e
+05
8e+
05
September–October 2005
Ref
lect
ivity
(Z
)
0e+
004e
+05
8e+
05
Fig. 4 Differences in the amount of NEXRAD reflectivity
from dispersing Brazilian free-tailed bats between cave colonies
(in grey) and bridge colonies (in black) during (A) May–August
2005 versus (B) August–November 2005. Bridge colonies show
activity in the fall when there is little to none at cave colonies.
32 J.W. Horn and T. H. Kunz
emergence period (17:00–23:00 CDT) to the mean
emergence heading of all colonies. Mean wind head-
ing and mean heading of emerging bats were signi-
ficantly different in 2005 for all colonies (Watson¼
0.449, P50.05, Fig. 7). A circular–circular regression
(Sarma and Jammalamadaka 1993) showed that wind
heading was not a significant predictor of emergence
headings of bats (k¼ 1.70, px¼ 0.65, py¼ 0.10).
When considering only Frio Cave and Ney Cave
and wind measurements taken from the KSAT
station (San Antonio), mean heading of emerging
bats was close to mean wind heading (mean wind
heading¼ 136.78� SD 25.58, mean heading emerging
bats¼ 116.98� SD¼ 43.38), but the two were still
significantly different (Watson¼ 0.221, P50.05,
Fig. 7). Again, a circular–circular regression showed
no effect of wind on emergence direction (px¼ 0.89,
py¼ 0.66). The difference between the mean wind
heading at the Austin station (KAUS) and the
heading of the emergences at the two bridges,
Congress Avenue and McNeil, was larger than that
for the two caves (mean wind heading¼ 75.88� SD
47.78, mean heading of emergences¼ 158.78�SD¼ 62.18, Fig. 7). Again, the mean headings were
significantly different (Watson¼ 0.531, P50.05), and
there was no effect of surface wind on direction of
emergence (k¼ 2.69, px¼ 0.42, py¼ 0.07). Speed of
dispersal was measured as the mean distance traveled
by patches of reflectivity during an emergence,
divided by the time between images (10 min).
Mean speed of dispersal of colonies for 2005 was
27.4 km/h� SD 6.14. We observed no discernible
pattern in annual variation in speed of dispersal
from all colonies, from bridges alone, or from caves
alone. Wind speed at the time of emergence
(14.12 km/h� SD 5.07) was also not a predictor of
emergence speed, nor did wind speed predict the
total distance traveled by patches of bat reflectivity.
We estimated total distance traveled during dispersal
by calculating the distance of the furthest point in
each tracked path of an emerging wave of bats. Total
mean distance of dispersal of the center of patches of
reflectivity was 15.1 km. Patches of reflectivity from
bats emerging from Frio Cave traveled the furthest
(18.6 km), whereas patches from Congress Avenue
Bridge traveled the shortest distance (10 km).
Discussion
A primary question raised by this study is whether
patterns in NEXRAD reflectivity accurately reflect
changes in sizes of colonies and nightly dispersal of
Brazilian free-tailed bats. Taken together, the absence
of organized reflectivity near known colonies in
November, December, and January, the seasonal
peak of total Z occurring when these colonies
are known to be largest, and the synchronous drop
in Z in August when migration begins, are strong
indicators that NEXRAD images contain valuable
information about nightly behavior and colony
dynamics.
An important finding is that Brazilian free-tailed
bats appear to populate bridge roosts earlier in
the season, and again later in the season, than cave
roosts. Cavernous limestone caves and modern
precast concrete bridges are clearly very different
types of roosting habitat for bats. The environment of
highway or urban bridges includes noise, vibration,
constant air pollution from vehicle emissions, and
the increased potential for disturbance by humans.
90
180
270
0.05
0.10
0.15
0.20
0.25 n=21
p < 0.001
Caves
90
180
270
0.05
0.10
0.15 n=34
p < 0.001
Bridges
Fig. 6 Circular histograms showing that mean heading of dispersing Brazilian free-tailed bats during nightly emergence is both
highly directional, and significantly different between cave colonies and bridge colonies. The single black indicator on the circle
indicates the mean emergence heading for the number of emergence events (n) shown. The standard deviation is marked in grey.
Reflectivity from Frio Cave and Ney Cave typically moves in an east-southeasterly direction. Reflectivity from bridges more often
moves toward the east or northeast and may be a bimodal pattern.
Analyzing NEXRADDoppler radar images 33
There may be a number of benefits from roosting
in bridges that would explain the attractiveness of
these sites to Brazilian free-tailed bats in spite of
these potential drawbacks. One possibility is that bats
gain an energetic advantage by using crevices in
bridges during cooler weather in spring and fall (L.C.
Allen-Hristova, personal communication). Brazilian
free-tailed bats depend on group living for energetic
benefits. Maternity roosts in caves are warm, which
benefits individual females by reducing gestation
times and reducing energetically costly re-warming
after bouts of torpor induced by lower temperatures
(Racey and Swift 1981; Kunz and Robson 1995).
Crevices and expansion joints in precast concrete
bridges in which Brazilian free-tailed bats typically
roost have lower thermal inertia than does the
rock mass of cave walls and ceilings, and thus may
have shorter temperature cycles. For a small group
of bats, or even large groups of bats on cooler
days, these spaces may warm more quickly, and
thus bats may benefit energetically from the
warmth of group huddling (Herreid 1963, 1967;
Vickery and Millar 1984; Kurta 1985). Temperatures
recorded inside bridge crevices in the study area
(L.C. Allen-Hristova, personal communication) indi-
cate that temperatures in bridge roosts are indeed
90
180
270
0.05
0.10
0.15 n=34
p < 0.001
Emergence
90
180
270 0.05
0.100.150.200.250.30 n=34
p < 0.001
Wind
All colonies
90
180
270
270
0.05
0.10
0.15
0.20 n=29
p < 0.001
Emergence
90
0.050.100.150.200.250.300.35 n=29
p < 0.001
Wind
Austin Wind and Bridges
90
180
270
0.050.100.150.200.25 n=21
p < 0.001
Emergence
90
180
270
0.10.20.30.40.5 n=21
p < 0.001
Wind
San Antonio Wind and Caves
180
Fig. 7 Headings of surface winds measured at nearby weather stations do not predict headings of nightly emergences of Brazilian
free-tailed bats from colonies of Brazilian free-tailed bats in south-central Texas.
34 J.W. Horn and T. H. Kunz
warmer and more stable than those in caves. Before
precast concrete bridges were available, Brazilian
free-tailed bats may have used small crevices in
buildings and other structures as stopover roosts
during migration or during the mating season.
Concrete bridges with expansion joints may repre-
sent a significant thermal advantage to bats over such
spaces.
Another possibility is that bridges represent a
reduction in competition for roosting space.
Davis (1962) reported that many limestone caves in
Texas were unoccupied by bats, but those that were
occupied were concentrated in the Balcones escarp-
ment (our study area), and shared characteristics,
such as large, domed ceilings which trap heat
generated by roosting bats. Suitable roosts may be
limited in caves, which are experiencing increasing
intrusion and recreational use by humans, while
new bridges are increasing in number. Occupancy
of bridges may relieve overcrowding in the limited
number of suitable roosting space in caves. This
factor may be combined with the thermal advantages
of bridges, such that when all available thermally
‘‘good’’ roosting spaces in caves are occupied during
the maternity season, bats may roost in bridges as an
alternative.
Finally, the extended activity period in early spring
suggests that bridge colonies may also provide ideal
sites for courtship and mating. Keeley and Keeley
(2004) reported that a bridge within the study area
(Williamson County, Texas) is used for courtship
and mating during late March and early April. There
may be some intrinsic characteristics of bridges
that make them favorable sites for mating, such as
the ability of males to move easily among groups
of females roosting in crevices (Keeley and Keeley
2004). Combining radar monitoring with detailed
ground observations might shed light on any plastic-
ity in this gregarious mating behavior.
It has been suggested or reported many times that
there may be a decline in the size of many large,
well-known Brazilian free-tailed bat colonies
(Cockrum 1969; Mohr 1973; Altenbach et al. 1979;
McCracken 1986, 1989, 2003; Clark 2001; Betke et al.
2008). Use of pesticides, increased intrusion by
humans into caves, and shifting use of roost and
feeding areas by bats have been variously suggested
as the cause. We found no evidence of a net decline
in the reflectivity from colonies that we monitored.
We did, however, find a nonrandom pattern of
annual fluctuation in total reflectivity that suggests a
cyclical fluctuation in colony sizes over an 11-year
period. The 5-year steady increase in reflectivity from
1995 to 2000, as well as the marked decrease in
2001–2003 may be caused by several factors. One
possibility is that climatological events, such as
droughts, floods, or storms may affect the timing
of seasonal movements, foraging success, reproduc-
tive rates, and survival of bats, leading to differential
population of colonies. Events that occur in over-
wintering areas in Mexico also may result in popu-
lation changes that are measurable downstream over
the range of northward migration. For example,
a warmer spring may allow individuals to conserve
energy and continue moving further north to large
caves in Oklahoma, rather than stop in Texas,
resulting in smaller sizes of colonies in Texas than
in cooler years. Moreover, severe meteorological
events, such as tropical storms may further influence
patterns of migration and movement. The decrease
in reflectivity that we observed in 2002 is coincident
with a damaging flood that occurred that year in our
study area. Although our measurements were taken
before the rains began, the floods were caused by
a large circulating tropical storm system fed by
moisture from the Gulf of Mexico, known to occur
in cyclic patterns. Storm-prone years could have
effects in overwintering areas in Mexico, or along
migration routes that result in smaller sizes of
colonies. An El Nino-Southern Oscillation (ENSO)
occurred in 2002–2003, and could have contributed
to the pattern we observed. If this hypothesis were
correct, reductions in reflectivity in one region might
be offset by increases in other regions. This question
could be addressed by examining NEXRAD images
from several stations along the migration path of
Brazilian free-tailed bats.
Another important finding of this study is that
nightly dispersal of Brazilian free-tailed bats is highly
directional. Surface wind headings at the time of
emergence do not appear to affect the initial direc-
tion in which bats set out to forage, nor does surface
wind speed predict either the total distance traveled
or the speed of emergence. The absence of an
influence from wind suggests that some other
factor(s) are important in determining the highly
directional behavior exhibited by Brazilian free-tailed
bats. This is reasonable from an energetic perspec-
tive. If bats used winds to their advantage to con-
serve energy en route to foraging areas, they would
likely pay the opposite penalty on the reverse trip,
expending additional energy to fly against the wind
when returning to their roosts. However, Westbrook
et al. (1995) showed that there is an altitudinal profile
of wind speeds in the south-central United States,
with slower speeds near ground, peak velocities of
9–15 m/s between 400 and 600 m above ground,
and slower speeds again with increasing altitude.
Analyzing NEXRADDoppler radar images 35
Thus, surface wind speed may have a weak relation-
ship to dispersal phenomena because it does not
correlate with winds at higher altitudes where bats
are dispersing. Moreover, Brazilian free-tailed bats
may minimize or overcome the energetic disadvan-
tage of having to fly into prevailing winds by flying
at different altitudes on outgoing and return trips.
A standing hypothesis is that Brazilian free-tailed
bats increase foraging success by pursuing masses of
emerging, dispersing, or migrating insects at high
altitudes (Cleveland et al. 2006). Our observation
that nightly emergence is highly directional generally
supports this hypothesis. That nearby colonies share
common directionality, as in the case with the bridge
colonies near Austin and Frio Cave and Ney Cave in
the San Antonio/Winter Garden area, suggests that
there may be a common local food resource in those
areas. It is not surprising that emergences from
Ney Cave and Frio Cave tend to move toward the
southeast, as this is the general direction of dense
areas of corn and cotton agriculture where pest
insects are highly abundant (Fig. 8). For the two
bridge colonies, an initial northeasterly heading may
be explained by similarly dense plantings of field
crops in the regions immediately to the north and
east. Lee and McCracken (2005) reported that the
daily composition of the diet that consisted of adult
corn earworms and fall armyworms in these colonies
was correlated with the emergence of noctuid moths
from crop fields. Bats may intersect high densities
of insects within altitudinal bands where insects
are taking advantage of winds for dispersal. If so,
bats may improve their foraging success by flying
over, but not descending entirely to, large areas of
field crops from which insects often emerge in large
number (Westbrook 2008). We also noted in our
images, several instances of ‘‘radar fine lines’’ which
have been attributed to insect densities in the
atmosphere (Russell and Wilson 1997). Such fine
lines moved across the study area from southeast to
northwest, directionally against the prevailing winds.
We did not observe direct convergence of emerging
patches of bats with fine lines, but if such
interactions occur, they may occur when bats have
dispersed to the point that they are indistinguishable
from background reflectivity and thus are unlikely to
be detected in NEXRAD images. Future studies could
use images from a combination of vertical profiling
Fig. 8 Mean heading of emergences in March, April, and May 2005, overlaid on the geographic study area in the foothills and
agricultural plains near San Antonio and Austin, Texas. Cave colonies are marked as circles, bridge colonies as triangles, and agricultural
areas as yellow patches. Headings of bat emergences may be related to the location of agricultural areas that may be dense with
potential insects prey for Brazilian free-tailed bats.
36 J.W. Horn and T. H. Kunz
radar (T.A. Kelly, personal communication) and
NEXRAD to test for evidence of spatial and temporal
convergence of bats and insects.
Limitations and future work
Although our results show there is strong evidence
that NEXRAD reflectivity measured during emer-
gence of bats reflects real events observed on the
ground, there are several factors that can introduce
error. NEXRAD images themselves contain inherent
variability. Sensitivity of the radar decreases with
distance from the station. Although this is compen-
sated for to some extent in the Level-II data that
we used, patches of bats (particularly low-intensity
patches) are less likely to be detected when further
from the station. This is partly the reason for the
weaker returns from colonies at Eckert James River
Cave and Davis Cave. Another reason is that the
elevation of the NEXRAD beam means that the
altitude band both increases in width and height
as distance from the station increases. Therefore,
patches of bats appear weaker or stronger at various
distances from the NEXRAD station. We used a
tracking algorithm, rather than counting reflectivity
based on an arbitrary distance from the station
or arbitrary strength. This means that we do not
count reflectivity unless it originates from a colony
location, which gives us a high degree of confidence
that the reflectivity counted was due to bats.
Timing and duration of precipitation were
other important variables we considered. We used
only clear-air-mode data that is generated by the
NEXRAD station during times when there is no
reflectivity caused by precipitation. Whenever possi-
ble, we selected dates that did not follow long
periods of precipitation. Bat emergences may be
larger, denser, or have altered timing after many days
of precipitation because females may have conse-
quently experienced low foraging success, and may
change their emergence behavior to compensate.
Another source of error is the filtering and tracking
algorithms themselves. Although our analytical
approach successfully isolates and measures bat reflec-
tivity, there are some scenarios that are problematic,
such as when two patches of emerging bats merge.
Applying additional ecologically-based rules to the
tracking algorithm will help limit this source of
variation.
The behavior of bats in the aerosphere is unknown
once they climb to higher altitudes following emer-
gence. Bats likely disperse in a variety of patterns
and change their rate of climb as they pass through
the volume of aerosphere being sampled by the
radar beam. This causes them to appear in a variable
number of images, leading to over- or under-
estimation by our sampling method. Unfortunately,
we cannot hold constant the behavior of bats while
adjusting the performance of our detection or track-
ing algorithms. Thus, the accuracy of the filtering
and tracking system we used cannot be easily
evaluated. Vertically integrating all of the reflectivity
in all elevation scans into a three-dimensional data
set for analysis (rather than just base reflectivity 0.58scan as we did) should provide a more complete
understanding of the spatial and temporal patterns of
nightly emergence behavior of Brazilian free-tailed
bats.
Finally, our measurements of colony size are based
on variation in relative strength and geographic
distribution of reflectivity. To compare actual
numbers of bats present during evening emergences,
a relationship must be developed between radar
reflectivity and the actual density of bats in the
aerosphere. One possible approach is to correlate
censuses of colonies taken on the ground with radar
observations to create a calibration curve as has
been done for migrating birds with NEXRAD data
(Gauthreaux and Belser 1998, 1999). The number of
bats present in each volume of the aerosphere can
then be estimated from the reflectivity value for
that volume, and total counts can be estimated
by integrating this over all the reflectivity associated
with the event in question. Ongoing research that is
generating accurate ground-based censuses of emer-
ging bats using high-speed thermal infrared imaging
(Betke et al. 2008; Hristov et al. 2008) should make
this possible in the near future.
Analysis of NEXRAD reflectivity can be a powerful
and flexible remote sensing tool for studying nightly
dispersal and population changes in Brazilian free-
tailed bats. Our approach has potential for uncover-
ing population trends, for detecting changes in the
timing of migration, and for characterizing differ-
ential use of roosts by bats. The large NEXRAD
repository provides opportunities to use both current
and historical data to understand long-term popu-
lation trends and to examine how anthropogenic
factors may be affecting this species. Thus, this
approach also holds promise for assessing con-
servation efforts and for improving understanding
of the foraging ecology of this wide-ranging bat
species.
Acknowledgments
We wish to thank Ray Dezzani of the University
of Idaho Department of Geography for guidance
Analyzing NEXRADDoppler radar images 37
with spatial analysis and statistical techniques and
Margrit Betke of the Boston University Department
of Computer Science for her expertise in computer
vision analysis. We would also like to thank the
following individuals for invaluable field assistance
and logistical support: Steve and Bill Rafferty,
Tommy Reardon, and Clinton Schulze for assistance
at the Eckert James River Cave; Lynn and Tex
Barnett and Thomas (Bub), and Marge Keese for
assistance at Ney Cave, Pat Morton at Texas Parks
and Wildlife; and Jim Kennedy from Bat Conserva-
tion International. For contributing additional coor-
dinates of bat colonies, we would like to thank Brian
Keeley of Bat Conservation International and Jerry
Fant of the Texas Speleological Survey. We also
thank Theresa Labriola for assistance with analysis.
Finally, we wish to thank the Society for Integrative
and Comparative Biology for waiving our registration
fees and for their support in hosting the symposium.
This research was funded in part by grants from
NSF: DBI-9808396 (to T.H.K. and C.J. Cleveland)
and EIA-ITR 0326483(to T.H.K., M. Betke, G.F.
McCracken, J.K. Westbrook, and P.W. Morton). We
also wish to thank the Air Force Office of Scientific
Research, through a grant to Boston University
(FA9550-7-1-0449 to T.H.K.) for providing partial
travel support to participate in this symposium.
References
Alerstam T, Gudmundsson GA. 1999. Migration patterns
of tundra birds: tracking radar observations along the
northeast passage. Arctic 52:346–71.
Altenbach JS. Geluso KN, Wilson DE. 1979. Population size
of Tadarida brasiliensis at Carlsbad Caverns in 1973. In:
Genoways HH, Baker RJ, editors. Biological investigations
in the Guadalupe Mountains National Park, Texas.
National Park Service Proceedings and Transactions,
Series No. 4. Washington (DC): National Park Service.
p. 341–8.
Best TL, Geluso KN. 2003. Summer foraging range of Mexican
free-tailed bats (Tadarida brasiliensis mexicana) from
Carlsbad Cavern, New Mexico. Southwest Nat 48:590–6.
Betke M, Hirsh D, Bagchi A, Hristov NI, Makris NC, Kunz
TH. 2007. Tracking large variable numbers of objects in
clutter. Proceedings of IEEE Computer Society Conference
on Computer Vision and Pattern Recognition.
Minneapolis, MN. p. 8.
Betke M, et al. 2008. Thermal imaging reveals significantly
smaller Brazilian free-tailed bat colonies than previously
estimated. J Mammal 89:18–24.
Black JE, Donaldson NR. 1999. Comments on ‘‘display of
bird movements on the WSR-88D: patterns and quantifica-
tion’’. Weather Forecast 14:1039–40.
Bruderer B. 1997. The study of bird migration by radar 2.
Major achievements. Naturwissenschaften 84:45–54.
Burger AE. 2001. Using radar to estimate populations and
assess habitat associations of marbled murrelets. J Wildl
Manage 65:696–715.
Buss I. 1946. Bird detection by radar. Auk 63:315–8.
Clark DR. 2001. DDT and the decline of free-tailed bats
(Tadarida brasiliensis) at Carlsbad Cavern, New Mexico.
Arch Environ Con Tox 40:537–43.
Cleveland CJ, et al. 2006. Economic value of the pest control
service provided by Brazilian free-tailed bats in south-
central Texas. Front Ecol Environ 5:238–43.
Cockrum EL. 1969. Migration in the guano bat. Miscellaneous
Publications, Museum of Natural History, University of
Kansas. Lawrence, KS. 51:303–36.
Cooper BA, Day RH, Ritchie RJ, Cranor CL. 1991. An
improved marine radar system for studies of bird
migration. J Field Ornithol 62:367–77.
Cooper BA, Ritchie RJ. 1995. The altitude of bird migration
in east-central Alaska: a radar and visual study. J Field
Ornithol 66:590–608.
Crawford AB. 1949. Radar reflections in the lower atmo-
sphere. P IRE 37:404–5.
Davis RB, Herreid CF II, Short HL. 1962. Mexican free-tailed
bats in Texas. Ecol Monogr 32:311–46.
Dacey MF. 1962. Analysis of central place and point pattern
analysis by a nearest neighbor method. Lund Studies in
Geography, B. Human Geog 24:55–75.
Diehl RH, Larkin RP. 2005. Introduction to the WSR-88D
(NEXRAD) for Ornithological Research. In: Ralph CJ,
Rich TD, editors. Bird Conservation Implementation and
Integration in the Americas: Proceedings of the Third
International Partners in Flight Conf, March 20–24, 2002,
Asilomar, CA. Gen Tech Rep PSW-GTR-191. U.S. Dept of
Agr, Forest Service, Pacific Southwest Research Station,
Albany, CA. p. 876–88.
Dinevich L, Matsyura A, Leshem Y. 2003. Temporal charac-
teristics of night bird migration above Central Israel – a
radar study. Acta Ornithol 38:103–10.
Frank JD, Kunz TH, Horn JW, Cleveland CJ, Petronio C.
2003. Advanced infrared detection and image
processing for automated bat censusing. Infrared
Technology and Applications XXIX. Proceedings of SPIE
5074:261–71.
Gauthreaux SA. 1992. The use of weather radar to monitor
long-term patterns of trans-Gulf migration in spring.
In: Hagan JM, Johnston DW, editors. Ecology and
conservation of neotropical migrant landbirds.
Washington (DC): Smithsonian Institution Press. p. 96–100.
Gauthreaux SA, Belser CG. 1998. Displays of bird movements
on the WSR-88D: patterns and quantification. Weather
Forecast 13:453–64.
Gauthreaux SA, Belser CG. 1999. Reply to ‘‘Comments on
display of bird migration on the WSR-88D: patterns and
quantification’’. Weather and Forecast 14:1041–2.
Glover KM, Hardy KR, Konrad TG, Sullivan WN,
Michaels AS. 1966. Radar observations of insects in free
flight. Science 154:967–72.
38 J.W. Horn and T. H. Kunz
Griffin DR, Thompson D. 1982. High altitude echolocation of
insects by bats. Behav Ecol Sociobiol 10:303–6.
Herreid CF II. 1963. Temperature regulation of Mexican
free-tailed bats in cave habitats. J Mammal 44:560–73.
Herreid CF, Kessel B. 1967. Thermal conductance in birds
and mammals. Comp Biochem Physiol 21:405–14.
Hristov NI, Betke M, Kunz TH. 2008. Applications of
thermal infrared imaging for research in aeroecology.
Integr Comp Biol. doi:10.1093/icb/icn053.
Keeley ATH, Keeley BW. 2004. The mating system of
Tadarida brasiliensis (Chiroptera: Molossidae) in a large
highway bridge colony. J Mammal 85:113–9.
Klaassen M, Biebach H. 2000. Flight altitude of trans-Sahara
migrants in autumn: a comparison of radar observations
with predictions from meteorological conditions and water
and energy balance models. J Avian Biol 31:47–55.
Kunz TH, Arnett EB, Cooper BA, Erickson WIP, Larkin RP,
Mabee T, Morrison ML, Strickland JD, Szewczak JM. 2007.
Assessing impacts of wind energy development on noctur-
nally active birds and bats: a guidance document. J Wildl
Manage 71:2449–86.
Kunz TH, Robson SK. 1995. Postnatal-growth and develop-
ment in the Mexican free-tailed bat (Tadarida brasiliensis):
birth size, growth-rates, and age estimation. J Mammal
76:769–83.
Kurta A. 1985. External insulation available to a non-nesting
mammal, the little brown bat (Myotis lucifugus). Comp
Biochem Phy A 82:413–20.
Lack D, Varley GC. 1945. Detection of birds by radar. Nature
156:446.
Lee YF, McCracken GF. 2005. Dietary variation of Brazilian
free-tailed bats links to migratory populations of pest
insects. J Mammal 86:67–76.
McCracken GF. 1986. Why are we losing our Mexican
free-tailed bats? Bats 3:1–2.
McCracken GF. 1989. Cave conservation: special problems
with bats. Bull Natl Speleol Soc 51:47–51.
McCracken GF. 2003. Estimates of population sizes in
summer colonies of Brazilian free-tailed bats (Tadarida
brasiliensis). In: O’Shea TJ, Bogan MA, editors. Monitoring
trends in bat populations of the United States and
Territories: problems and prospects. Washington (DC):
U.S. Geological Survey, Biological Resources Discipline,
Information and Technology Report, USGS/BRD/ITR-
2003-003. p. 21–30.
McCracken GF, Gustin MK. 1991. Nursing behavior in Mexican
free-tailed bat maternity colonies. Ethology 89:305–21.
McCracken GF, Westbrook JK. 2002. Scientists discover that
high-flying mammals are bad news for bugs. Natl Geogr
201:114–23.
McCracken GF, Gillam EH, Westbrook JK, Lee Y, Jensen ML,
Balsley BB. 2008. Brazilian free-tailed bats (Tadarida
brasiliensis: Molossidae, Chiroptera) at high altitude: links
to migratory insect populations. doi:10.1093/icb/icn033.
McKay HAC. 1945. Detection of birds by radar. Nature
156:446.
Mohr CE. 1973. The status of threatened species of cave-
dwelling bats. Bulletin of the National Speleological
SocietyBull Natl Speleol Soc 34:33–47.
Plank VG. 1956. Geophysical research paper No. 52. Air Force
Cambridge Research Center. Bedford, MA.
Racey PA, Swift SM. 1981. Variations in gestation length in a
colony of pipistrelle bats (Pipistrellus pipistrellus) from year
to year. J Reprod Fertil 61:123–9.
Richter JH, Jensen DR, Noonkester VR, Kreasky JB,
Stimmann MW, Wolf WW. 1973. Remote radar sensing:
atmospheric structure and insects. Science 180:1176–8.
Ritzi CM. 1999. Utilization of cliff swallow (Petrochelidon
pyrrhonata) nests in west Texas by cave myotis (Myotis
velifer). Southwest Nat 44:414–5.
Russell KR, Gauthreaux SA. 1998. Use of weather radar to
characterize movements of roosting purple martins.
Wildlife Society Bulletin 26:16.
Russell RW, Wilson JW. 1997. Radar-observed ‘‘fine lines’’ in
the optically clear boundary layer: reflectivity contributions
from aerial plankton and its predators. Bound-Lay
Meteorol 82:235–62.
Ruthi L. 1994. Observation of bat emergence from Reed
Bat Cave (HR-004) with the WSR-88D RADAR. Okla
Underground 17:54–6.
Sarma Y, Jammalamadaka S. 1993. Circular regression.
Statistical science and data analysis. Proc 3rd Pacific Area
Stat Conf. VSP: Utrecht, Netherlands. p. 109–28.
Schnell G. 1965. Recording the flight speed of birds by
Doppler radar. Living Bird 4:79–87.
Vickery WL, Millar JS. 1984. The energetics of huddling by
endotherms. Oikos 43:88–93.
Westbrook JK. 2008. Noctuid migration in Texas within the
nocturnal aeroecological boundary layer. Integr Comp Biol.
doi:10.1093/icb/icn040.
Westbrook JK, Esquivel JF, Lopez JD, Jones GD, Wolf WW,
Raulston JR. 1998. Validation of bollworm (Lepidoptera:
Noctuidae) migration across south-central Texas in 1994-
1996. Southwest Entomol 23:209–19.
Westbrook JK, Isard SA. 1999. Atmospheric scales of biotic
dispersal. Agr Forest Meteorol 97:263–74.
Westbrook JK, Raulston JR, Wolf WW, Pair SD, Eyster RS,
Lingren PD. 1995. Field observations and simulations
of atmospheric transport of noctuids from northeastern
Mexico and the south-central US. Southwest Entomol
Suppl 18:25–44.
Williams T, Ireland L, Williams J. 1973. High altitude flights
of the free-tailed bat, Tadarida brasiliensis, observed with
radar. J Mammal 54:807–21.
Williams TC, Williams JM, Williams PG, Stokstad P. 2001.
Bird migration through a mountain pass studied with
high resolution radar, ceilometers, and census. Auk
118:389–403.
Wolf WW, Westbrook JK, Raulston JR, Pair SD, Lingren PD.
1995. Radar observations of orientation of noctuids
migrating from corn fields in the lower Rio-Grande
Valley. Southwest Entomol 18:45–61.
Analyzing NEXRADDoppler radar images 39