Post on 01-Jun-2020
Archaeological aerial thermography: a case study at the Chaco-eraBlue J community, New Mexico
Jesse Casana a,*, John Kantner b, Adam Wiewel a, Jackson Cothren c
aDepartment of Anthropology, University of Arkansas, Main 330, Fayetteville, AR 72701, United StatesbUniversity of North Florida, 1 UNF Dr., Jacksonville, FL 32224, United StatescDepartment of Geosciences and Center for Advanced Spatial Technologies, University of Arkansas, JBHT 320, Fayetteville, AR 72701, United States
a r t i c l e i n f o
Article history:
Received 18 September 2013
Received in revised form
17 February 2014
Accepted 20 February 2014
Keywords:
Thermal imaging
LWIR
UAV
Photogrammetry
Archaeo-geophysics
Ancestral Puebloans
a b s t r a c t
Despite a long history of studies that demonstrate the potential of aerial thermography to reveal surface
and subsurface cultural features, technological and cost barriers have prevented the widespread appli-
cation of thermal imaging in archaeology. This paper presents a method for collection of high-resolution
thermal imagery using an unmanned aerial vehicle (UAV), as well as a means to efficiently process and
orthorectify imagery using photogrammetric software. To test the method, aerial surveys were con-
ducted at the Chaco-period Blue J community in northwestern New Mexico. Results enable the size and
organization of most habitation sites to be readily mapped, and also reveal previously undocumented
architectural features. Our easily replicable methodology produces data that rivals traditional archaeo-
logical geophysics in terms of feature visibility, but which can be collected very rapidly, over large areas,
with minimal cost and processing requirements.
� 2014 Elsevier Ltd. All rights reserved.
1. Introduction
Archaeologists have recognized since the 1970s that aerial im-
ages which record thermal infrared wavelengths of light (7.5e
13 mm) could be a powerful means for recognizing both surface and
subsurface cultural remains (e.g., Tabbagh, 1977; Fourteau and
Tabbagh, 1979; Berlin et al., 1977), yet technological and cost bar-
riers have largely prevented the widespread application of ther-
mography in archaeological contexts. This paper presents interim
results of a National Endowment for the Humanities-funded proj-
ect to develop techniques for collection of thermal imagery using a
UAV, to designworkflows for efficient photogrammetric processing
of imagery, and to assess the effectiveness of the technology at
archaeological sites located in a variety of different environments.
Herein we present a case study that provides an illustration of our
methodology, conducted at a Chaco-period archaeological com-
munity known as Blue J, located in northwestern New Mexico
about 70 km south of Chaco Canyon (Fig. 1). Results demonstrate
the enormous potential of aerial thermography in archaeology,
revealing the organization of occupation at Blue J and aiding in
the discovery of many previously undocumented subsurface
architectural remains, as well as providing a blueprint for similar
investigations elsewhere.
2. Archaeological thermography
The principles behind archaeological thermographic imaging
are relatively simple: due to their composition, density and mois-
ture content, materials on and below the ground surface absorb,
emit, transmit, and reflect thermal infrared radiation at different
rates. Experimental data illustrates the complexity of the variables
that affect the potential visibility of archaeological features in a
thermal image (Scollar et al., 1990: 593e611). However, in general,
such features should be resolvable if when compared to the back-
ground matrix they have sufficient contrast in thermal inertia, a
product of a material’s thermal conductivity and its volumetric heat
capacity (Perisset and Tabbagh, 1981), and if the image is acquired
at a time in the diurnal cycle when such differences are pro-
nounced. Several studies have demonstrated that thermography
can detect features at or near the ground surface such as pits,
ditches and field boundaries as well as buried architectural features
up to a half meter below the ground (e.g., Perisset and Tabbagh,
1981; Lunden, 1985; Bellerby et al., 1990).
Despite its potential, archaeological applications of the tech-
nology are scarce, largely because few archaeologists had access to* Corresponding author. Tel.: þ1 479 595 9569.
E-mail address: jcasana@uark.edu (J. Casana).
Contents lists available at ScienceDirect
Journal of Archaeological Science
journal homepage: http : / /www.elsevier .com/locate/ jas
http://dx.doi.org/10.1016/j.jas.2014.02.015
0305-4403/� 2014 Elsevier Ltd. All rights reserved.
Journal of Archaeological Science 45 (2014) 207e219
the highly specialized radiometers used by researchers in the 1970s
(Scollar et al., 1990: 614e18), while thermal images acquired from
civilian satellites such as Landsat and ASTER are of too low spatial
resolution to reveal most archaeological features. Even the aircraft-
acquired TIMS (Thermal Infrared Multispectral Scanner) imagery
that Sever and colleagues famously utilized to document ancient
roadways in the Chaco Canyon area (Sever and Wagner, 1991) and
in Costa Rica (Sheets and Sever, 1991) was only 5 m resolution.
While a handful of other more recent archaeological investigations
have used higher-resolution, aircraft-acquired thermal imagery
(e.g., Challis et al., 2009), the exorbitant cost of imagery acquisition,
requiring tasked flights of specialized aircraft equipped with
advanced sensors, puts the technology well beyond the reach of
most archaeologists.
Several recent studies have instead explored the archaeological
potential of much higher resolution thermal images collected with
conventional handheld cameras. Ben-Dor et al. (2001) use imagery
collected from a helicopter to detect subsurface stone architecture
in Israel, while Buck et al. (2003) use thermal images collected by a
camera mounted on a platform to reveal differing concentrations of
surface artifacts. A few avocational archaeologists have similarly
collected images of archaeological sites using handheld thermal
cameras attached to kites, revealing many subsurface remains
(Wells, 2011). One of the most systematic studies was undertaken
at late prehistoric mound center in Mississippi, where thermal
imagery acquired with a handheld camera suspended from a heli-
um blimp was used to successfully document subsurface house
remains as well as the diurnal variation in their thermal values
(Haley et al., 2002; Giardino and Haley, 2006). Kvamme (2006,
2008) was the first to deploy such an approach on a site-wide
scale, flying a handheld thermal camera aboard a manned pow-
ered parachute (see, Hailey, 2005), at Fort Riley, Kansas and Double
Ditch, North Dakota. His results are inspiring in their possibilities as
they reveal much of what can be seen in more traditional
geophysics, as well as distinct features not recorded by other
methods. However, implementing Kvamme’s approach would be
very challenging, requiring a microlight aircraft, an experienced
pilot, ideal weather conditions, and extensive work to rectify and
mosaic the images into a usable map.
In recent years, several technologies have emerged that make
possible the cost-effective acquisition of high-resolution thermal
data at the scale of sites and landscapes. Firstly, unmanned aerial
vehicles (UAVs) are rapidly improving in sophistication and reli-
ability, enabling archaeologists to collect aerial imagery from spe-
cific altitudes with precise camera orientation at any time of day
and in a range of weather conditions. The possibilities for UAVs to
aid in archaeological research have recently generated much in-
terest, particularly for documenting sites, monuments and exca-
vations (e.g., Verhoeven and Docter, 2013; Hill, 2013; Chiabrando
et al., 2011; Seitz and Altenbach, 2011; Sauerbier and Eisenbeiss,
2010; Eisenbeiss and Zhang, 2006). At the same time, advances in
photogrammetric image processing software packages, such as
AgiSoft’s PhotoScan and Microsoft’s Image Composite Editor (ICE),
now make quite straightforward the once laborious process of
mosaicking, orthorectifying and georeferencing images, a fact
which is rapidly transforming archaeological documentation more
broadly (e.g., De Reu et al., 2013; Verhoeven et al., 2012a, 2012b;
Eisenbeiss and Sauerbier, 2012; Verhoeven, 2011; Koutsoudis
et al., 2007). Finally, commercially available thermal cameras have
been significantly improved upon over the past several years, with
increased spatial resolution (from 320 � 240 to 640 � 512 pixels)
and thermal sensitivity (from 0.2 to <0.05 K), while also being
reduced substantially in both size and cost. Through a case study at
the Chaco-period Blue J Community, New Mexico, research
Fig. 1. Locator map showing the Blue J Community and Chaco Canyon in northwestern New Mexico, USA.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219208
presented herein integrates these various technologies and dem-
onstrates an efficient methodology for their deployment in
archaeological investigations.
3. The Blue J community
Located in northwestern New Mexico (Fig. 1), Blue J is an
Ancestral Puebloan community that was first identified by ar-
chaeologists in the 1970s, although official records do not appear
until the 1980s and 1990s through cultural resource management
reconnaissance efforts (Kantner, 2010). These projects, however,
only examined a few of the households located on the edges of the
larger prehistoric community. It was not until Kantner’s in-
vestigations in the 2000s under the auspices of Georgia State
University and the School for Advanced Research that the full
extent of the visible surface occupation was identified and the
community fully recorded (Kantner, 2013).
The Blue J Community consists of nearly 60 Ancestral Puebloan
households loosely clustered over approximately two square kilo-
meters around what was once a large spring (Fig. 2A). The majority
of the households were built along a sandstone escarpment, lead-
ing to considerable alluvial, colluvial, and aeolean deposits over the
community (Fig. 2B). Each household, which is recorded as a
separate site, typically includes a masonry roomblock of 3e15
rooms that faces a flat plaza area to the east or southeast. Beyond
this, a shallow and usually expansive midden area is typically
found. This layout is known in the American Southwest as a
“Prudden unit,” named after the archaeologist who first discussed
the ubiquity of this layout across the early Puebloan world
(Prudden, 1903). In principle, Prudden unit households should also
feature subterranean structures in the plaza that are usually
interpreted as having served a combination of ceremonial and
domestic functions. In the Blue J Community, however, households
were buried by up to a meter of eroded sandstone and wind-blown
silt, making identification of subsurface deposits difficult without
excavation (Kantner, 2013). In fact, the only two subterranean
structures identified by Kantner were found through a program of
subsurface test excavations. Similarly, although the roomblocks
could usually be identified on the surface, their full extent was only
revealed through excavation; testing of several roomblocks found
that approximately 30% of each roomblockwas too deeply buried to
be identified on the surface.
The issue of extensive deposition over the Blue J Community is
relevantbecauseof the sites’potential importance forunderstanding
Fig. 2. (A) Oblique aerial photograph showing part of the Blue J Community and adjacent sandstone cliffs; (B) Site LA 170607 as it appears on the surface today.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 209
regional dynamics of the Ancestral Puebloan occupation. In-
vestigations in the community show that it was occupied from at
least A.D. 600 to as late as A.D.1150, but the greatest intensity of use
was in the eleventh century (Kantner, 2013). This is during the
fluorescence of an influential ceremonial center in Chaco Canyon,
during which the vast majority of Puebloan communities across the
northern Southwest constructed two forms of monumental archi-
tecture known as “great kivas” and “great houses.” Great kivas are
large, circular subterranean structures with diameters of 10 m or
more that have a deep history in the Puebloan Southwest long pre-
dating the Chaco period, but which became more formalized in
Chaco Canyon and surrounding areas during the eleventh century.
Great houses, in contrast, are monumental surface structures that
seem to have been inspired by archetypes constructed in Chaco
Canyonbeginningperhaps in theninth century butfluorescing in the
late AD 1000s. The presence of both great houses and great kivas in
communities across the Puebloan Southwest are thought to reflect
the degree of adherence to the ceremonial tradition based in Chaco
Canyon (Kantner and Kintigh, 2006; Van Dyke, 2003, 2007).
Kantner’s investigations in the Blue J Community have located
neither a great house nor a great kiva, making it unusual among its
neighbors. In fact, these architectural forms have been identified in
every other Chaco-era community in the immediate region, even
though few have been the subject of intensive archaeological
investigation (Kantner, 2010). Whether the absence of these fea-
tures is the result of deposition over the community, or whether
these structures were never constructed there, is important for
understanding when, how, and why communities like Blue J
aligned themselves with the ceremonial center of Chaco Canyon
(e.g., Kantner and Vaughn, 2012; Lekson, 2009; Reed, 2011).
Although some targeted magnetometry and resistivity work has
been conducted in the Blue J Community, with mixed results, its
considerable size calls for a remote sensing approach that can
effectively and efficiently cover a large area.
4. Instrumentation and methods
4.1. Thermal camera
The thermal camera we use is built around FLIR’s Tau II LWIR
uncooled camera core. The camera records light in the 7.5e13 mm
range, with a sensitivity of 0.05 K at f/1.0, effective at ambient
temperatures from �40 to 160 �C. The camera has a resolution of
640 � 512 pixels, with a pixel pitch of 17 mm, and records 8-bit
video images at a frame rate of 30 Hz. While the resolution of the
camera is very low by comparison to color light cameras, it is five
times better than the best cameras on the market only a few years
ago and among the highest resolution thermal cameras available
commercially. To maximize the camera’s field of view, and thereby
reduce the altitude at which the camera would need to be
positioned during aerial surveys, we use a relatively short focal
length, the 19 mm lens option. The Tau II camera core, designed
more for laboratory than field applications, was attached to a power
source and SD card recorder, and fitted inside an aluminum case for
aerial deployment (Fig. 3).
4.2. Unmanned aerial vehicle
UAVs offer key advantages over traditional forms of archaeo-
logical aerial imaging (cf. Verhoeven, 2009), particularly in their
ability to cover large areas at a fixed altitude and speed under a
wide range of wind and weather conditions. For aerial thermog-
raphy, these capabilities are essential, because imagery acquisition
is highly time-sensitive. To maximize archaeological visibility,
thermal imagery has to be collected at a time of day when the
contrast in thermal inertia is highest between the archaeological
target and the background soil and ground cover. Moreover, the
entirety of the survey area must be covered with as little temporal
variability as possible in order to reduce drift in thermal values
across a single mosaicked scene. Additionally, the relatively stable
and slow moving platform of a copter-type UAV is advantageous
because the pixels of the Tau II camera heat and cool continuously
as the IR signal is integrated. With an unstable platform, IR frames
are comparatively more difficult to mosaic as the imagery is
subject to artifacts related to the camera’s shutter type (FLIR,
2013).
In this project we utilize a CineStar 8 (Fig. 3), one of the
preferred UAVs for aerial cinematography and other similar appli-
cations. The eight-rotor “octocopter” can lift around 2 kg (4.4 lbs) of
payload, with cameras mounted on an independently operated
gimbal suspended below the UAV. The gimbal is capable of a full
360� of motion, enabling cameras to be pointed in a pre-
determined direction or at a specific point regardless of the mo-
tion of the UAV (Freefly Systems, 2014). It also has vibration control
and stabilization devices that greatly reduce image blur. The flight
and navigation controls for the CineStar are produced by industry
leader MikroKopter, and are well suited to the needs of surveying.
The custom software enables users to download a Google Maps
image of the survey area, and then to plan flight paths with up to 35
GPS-guided waypoints (Fig. 4). Users can therefore specify the
precise location and speed of the UAV, as well as the camera
orientation throughout the survey. A telemetry device enables the
craft to be monitored while in flight on a laptop ground control
station, and all flight data is recorded on an SD card aboard the
craft. Finally, the fact that the craft is sold as a kit means it can be
easily customized to suit users’ needs, and can also be readily
repaired in the event of a crash or other equipment failure.
With these advantages, there are still many issues with the
system, particularly for novice operators, as a perusal of the dis-
cussion boards dedicated to UAVs will quickly demonstrate. While
Fig. 3. The CineStar8 in flight with thermal camera aimed downward (left) and a closeup of the FLIR Tau II camera core inside our custom-built case.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219210
the CineStar 8 can have a flight time of up to 24min if outfittedwith
a GoPro camera and flown in windless conditions, using a larger
camera in real world conditions flight times are considerably less,
probably around 15 min (although one would not want to find out
the limit during a survey, so some additional buffer time is always
required). Thus, flight time becomes the major limiting factor in the
size of survey areas. The engineering of many components is also
far from robust, and in our experiments with the UAV, we found it
prone to frequent failures for a multiplicity of reasons, including
faulty electronics, wires of inadequate gage, vibration loosening of
propellers, and many other issues. Nonetheless, considering the
low cost and flexibility of the CineStar 8, it suits the needs of
archaeological aerial thermography reasonably well.
4.3. Imagery acquisition
As discussed above, one of the main strengths of the UAV over
other aerial photographic platforms such as kites and balloons is
the control it gives users in planning surveys precisely. While the
relatively short focal length of our thermal camera (19 mm) in-
creases the area of the ground that can be covered in one image, its
small sensor size (8.70 � 10.88 mm) restricts the camera’s field of
Fig. 4. (A) The MikroKopter-Tool software used to program and pilot the CineStar8, with waypoints, flight path, and telemetry display as used at the Blue J Community, and (B)
actual flight path of 7:18am flight as recorded by the onboard GPS, overlaid on a Google Earth image.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 211
view to only 26 � 32�. Combined with its rather low resolution of
only 512 � 640 pixels, we are faced with a trade-off between the
ground area of an image and the resolution of the resultant data,
determined by the altitude of the camera. While it would be
theoretically possible to simply collect a larger number of high
resolution images from a lower altitude, the short flight time of the
CineStar 8 at around 15 min, makes this impractical. Thus to
maximize the areal coverage of a survey, which is typically the goal,
one can either increase the altitude of the UAV, at the expense of
image resolution, or increase the speed of transects, at the risk of
blurring images during flight. For the surveys at the Blue J site, we
selected a compromise of 70 m elevation for the UAV, yielding 6e
7 cm ground sampling distance for thermal images (comparable to
high-resolution geophysical datasets) and an image footprint of
approximately 32 � 40 m. Survey speed was set at 5 m per second
(11.2 mph) as our past experiments suggest that at 70 m, flying
faster than this will result in an excessive number of blurred
images.
In order to ensure that images can be effectively processed in
PhotoScan or other photogrammetric software packages (see Sec-
tion 4.4) a high percentage of image overlap is required, generally
70e80%. In order to reduce the number of transects in a survey, we
orient the camera perpendicular to the flight path. The relatively
high frame rate of the video feed (30 Hz) means that wewill collect
many images per meter along each transect, enabling us to extract
images with 80% or greater overlap. To align transects with one
another, we plan on approximately 40% cross lapping images. At the
altitude and orientation described above, this requires transects to
be spaced approximately 20 m apart.
At the Blue J community, we conducted surveys with eight east-
west oriented transects of 300 m length, spaced at 20 m, enabling
us to image an area of approximately 220 � 360 m in thermal
infrared. Color imagery was also collected using a Nikon D6000
camera, with its timer set to take photos at 1 s intervals. The color
image helps to interpret results of the thermal data, producing
images of 2e3 cm ground sampling distance, and is also used to
generate a high resolution digital surface model. The wider field of
view of the camera, with focal length of its variable zoom lens set to
18 mm, enables a larger area of the ground (280 � 420 m) to be
imaged using the same flight parameters.
To aid in georeferencing of image mosaics, we placed eight
ground control targets across the survey area, and using a total sta-
tion, recorded their locations within the coordinate system used in
previous investigations at the site. Targets were made of aluminum
sheeting because metals have very low thermal emissivity, and thus
tend to appear very dark on thermal imagery. Our targets measured
60 � 15 cm, with the expectation that they would appear as 2 � 10
pixel targets in thermal data.While all of the targets are visible in the
thermal data, most tend to blur in the imagery, appearing more as
dots than Xs, and so in future surveys we plan to increase their size,
perhaps to 1.2m in length. A larger number of targets could improve
theaccuracyof thefinal images, althoughour results usingonlyeight
targets were still very good (see Section 4.4).
4.4. Image processing
We rely primarily on Agisoft PhotoScan to produce both color
and thermal ortho-imagery. For each thermal flight, we first extract
still images from the video feeds recorded by the Tau II camera.
Although the camera records at a high frame rate (30 Hz), we found
that only one frame per second is necessary for sufficient overlap
along flight transects. For color photographs, we selected approx-
imately every fourth image between the first and final waypoints,
avoiding inclusion of either blurred or excessively oblique images.
PhotoScan is one of many modern photogrammetric applica-
tions which integrate relatively new feature extraction and
matching capabilities with traditional aero-triangulation based on
the self-calibrating bundle adjustment (Mikolajczyk and Schmid,
2004; Triggs et al., 2000). As such it is capable of creating accu-
rate orthoimagery and digital surface models from overlapping
two-dimensional imagery with manual point measurement only
required for ground control identification (Verhoeven, 2011;
Verhoeven et al., 2012a; De Reu et al., 2013). Briefly put, after
adding all photographs from a flight to PhotoScan, the software
automatically aligns the images by identifying corresponding fea-
tures shared among photographs (Agisoft, 2013). The software es-
timates the locations of these points, from which a sparse point
cloud is derived, and more importantly, camera locations are
approximated and internal camera parameters are estimated
(Fig. 5). Using the estimated camera positions and the actual pho-
tographs, PhotoScan then calculates depth maps from which a
dense point cloud and meshed model are generated. We used the
highest accuracy setting to align photographs and the height field-
smoothmethod to build geometry. Initially, we produced amedium
qualitymodel, enabling us to use a guided approach by then placing
our eight ground control points. After refining marker placement in
each photograph and adding ground control point coordinates,
PhotoScan optimizes the alignment to build a more accurate and
georeferenced mesh. We used the height field-smooth method to
reconstruct geometry to produce an ultrahigh quality mesh from
Fig. 5. Thermal image mosaic from the 7:18am flight illustrating the estimated camera locations along the flight path as calculated in Photoscan.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219212
which we generate color and thermal orthophoto mosaics, as well
as a digital surface model.
Microsoft ICE, a freely available panoramic image stitching tool,
offers an alternative to PhotoScan, and a particularly good method
for processing thermal data because ICE is capable of producing a
seamless high-resolution panorama directly from a video feed.
While not offering the flexibility to change settings and parameters
like Photoscan, ICE requires dramatically less processing time and
memory demands, meaning one does not need a high-end com-
puter to process imagery. The software also seems to select images
from the video feed that have good contrast and clarity, thereby
producing image mosaics on which it can be somewhat easier to
recognize archaeological features. After producing image mosaics
in ICE, we then simply used our ground control points to geore-
ference the mosaic in ArcGIS. To be clear, ICE does not produce an
orthophoto and in areas of high topographic relief, the image
stitching applicationwill produce excessive errors. For small scenes
over relatively flat terrain, we found results to be visually pleasing
and somewhat easier to interpret than the resampled orthophoto
produced by PhotoScan.
5. Results
We flew five surveys at the Blue J community, collecting four
thermal datasets and one color photographic series. Each flight
took approximately 11 min from takeoff to touchdown. The four
thermal flights were begun respectively at 9:58pm, 5:18am,
6:18am and 7:18am, while the color photographic flight was begun
at 5:45am, just prior to sunup (Table 1). We had intended to collect
imagery at peak afternoon heat as well, but on the days of the
surveys there were frequent wind gusts of up to 30 mph during the
afternoon and early evening, making flights impossible until
around 9pm. While the CineStar 8 can fly in far more diverse wind
conditions than other aerial photographic platforms, strong, gusty
winds make flying challenging even for experienced pilots. On the
days in question (June 20e21, 2013), temperature variability was
high, as is common in deserts, reaching 32 �C (90 �F) on the af-
ternoon of June 20 and a low of 8 �C (46 �F) in the early morning
hours of June 21. Such temperature oscillations may improve the
likelihood that archaeological features will appear because the
ground will likewise experience more extreme heating and cooling
over a diurnal cycle.
As outlined in Section 4.4, color and thermal imagery was
orthorectified, and georeferenced mosaics of each flight were
produced. Each thermal orthophoto mosaic was composed of
approximately 350 still frames while 118 color images were used to
create a color orthophoto mosaic (Fig. 6) and digital surface model
of the site (Fig. 7). Alignment of the thermal images was based on at
least 200,000 matched features across the scene. Interior orienta-
tion of the camera followed Brown’s model (Triggs et al., 2000)
with only the focal length, principal point, pixel aspect ratio and
three radial distortion parameters treated as unknowns. In all cases,
the resulting estimated parameters (19.80 mm calibrated focal
length, 3% longer in the camera x direction) did not deviate
significantly from our initial estimates. Four of the surveyed ground
points were used in the alignment while the other four were
withheld and used as check points (Fig. 7). The alignment (or aero-
triangulation) resulted in control point RMSE errors of approxi-
mately 6 cm horizontal and 3 cm vertical. With so few control
points this is expected and so we rely on the check point errors to
indicate the quality of the alignment. Check point errors at each of
the four points for the 6:18am flight are shown in (Table 2). Hori-
zontal and vertical errors at the check points are larger than at the
control points e again, expected e but still only about the size of a
pixel in the final orthophotos.
In all thermal images, the modern road, an unimproved dirt
track, appears very clearly at the top of the image, as does the tarp
used to launch the UAV (in order to minimize dust particles from
getting inside the motors) (Fig. 6). Our two vehicles and team
members also appear at the upper left of the images. Because the
UAV tends to swing as it adjusts to new waypoints, we collected
more oblique images at transect ends than we had anticipated. The
best image for archaeological visibility is the 5:18am flight (Figs. 6
and 8A), which reveals many surface and subsurface features. The
ambient warming of the vegetation at the site by 6:18am results in
increased noise (Fig. 8D), while by the time the 7:18am image,
intense raking sunlight covered the entire site, producing an image
that primarily records how surface features differentially reflect
thermal radiation (Figs. 7 and 8E). While subsurface features are far
less apparent in the 7:18am image, the fact that the relief of the
ground surface is so clearly visible makes archaeological features
with topographic expression quite evident. Experimental data from
the 1970s show that in some cases the ability of early morning
thermal imagery to reveal micro-relief can be used to extensively
map archaeological remains (Scollar et al., 1990: 623e8).
In all nighttime images, modern vegetation creates a very strong
signature, as thewater in trees and plants tends to retain heat much
more readily than the dry desert soils. In the 9:58pm image, there is
more vegetation noise than in the pre-dawn images, as even grasses
and small shrubs produce high thermal infrared values (Fig. 8F). By
the following morning, the larger trees and woody shrubs still
appear as high value spots, while most of the grasses and smaller
plants had cooled, reducing noise and increasing the likelihood that
archaeological features will appear (Fig. 8A). Similarly, the complex
geology of the cliffs and talus slopes at the northern edge of the
survey produce very strong thermal noise (Fig. 6), with extreme
high and low values, and thus make detection of potential archae-
ological remains unlikely. In both evening and pre-dawn images
there are numerous low value, round features measuring 1e3 m in
diameter, some of which could be ancient storage pits, which are
commonly found on Puebloan sites in the region (Kantner, 2004).
However, comparison with the color imagery shows that most of
these features correspond with patches clear of vegetation, and
ground truthing reveals that most of these features are antmounds.
In the 7:18am image, many of these features have a small mound at
their center, confirming this interpretation (Fig. 8E).
After images were georeferenced, we plotted the locations of all
the archaeological sites and features that were previously identified
through surface survey at Blue J in order to evaluate whether and
how these features appear on the thermal imagery. Careful analysis
of the thermal image datasets in comparison with the color image
enables us to quickly eliminate those features in the imagery that
are caused by modern, surface landscape features from those that
likely represent subsurface archaeological remains. Section 5 out-
lines key findings and their interpretation.
6. Discussion
In the 5:18am image, most of the previously identified sites at
Blue J can be recognized, and in many cases it provides a detailed
Table 1
Summary of flight times, temperature, and weather conditions during aerial surveys
at Blue J on June 20e21, 2013.
Time Camera �C/�F Conditions
9:58pm FLIR Tau II 22/72 Dark, occasional wind gusts
5:18am FLIR Tau II 10/50 Dark, no wind
5:45am Nikon D6000 11/52 Light on horizon, no wind
6:18am FLIR Tau II 13/55 First sunlight on site, light wind
7:18am FLIR Tau II 16/61 Strong raking sunlight, light wind
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 213
view of subsurface remains, including the size and orientation of
roomblocks, individual walls, and potential subterranean features
(Fig. 6). Field investigations show that many of these sites are re-
mains of stone-built architectural roomblocks, often visible at the
surface todayas piles of rubble and artifact scatters (Fig. 2B; Kantner,
2013). These features generally appear on the early morning ther-
mal imagery as high value clusters, presumably because the stone
walls and collapsed masonry retain heat longer than surrounding
desert soils. The deepest middens, such as the one just east of LA
170609, also appear as high value patches, probably because they
similarly contain dense concentrations of stone and artifacts.
Significantly, atmanyof these sitesweare also able to distinguish
subsurface architectural remains. One of the best examples comes
from LA 170609 (Fig. 8). First identified during survey operations in
2004, this domestic roomblock and its associated plaza andmidden
areas were tested and completely backfilled in 2006 (Fig. 8B).
Excavation of the roomblock was limited to uncovering the tops of
the buried walls in order to determine masonry type, room sizes,
and structure orientation. While the survey estimated roomblock
size at five rooms based on the surface rubble, excavations revealed
at least six and probably eight buried rooms and, most notably, a
lengthymasonrywall surrounding the plaza area to the southeast of
the roomblock. This feature was not visible on the surface, but can
readily be seen as a high-value arc in the thermal data that may, in
fact, extend farther north than the test excavations identified
(Fig. 8A). A dark circular area indicating lower thermal values just
Fig. 6. Thermal imagery of Blue J from the 5:18am flight (top) compared to color imagery (bottom). Locations of features discussed in the text are indicated. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.)
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219214
inside the plazawall might indicate a subsurface structure such as a
kiva, but this area was not tested during the 2006 operations. The
subtle topographic expression of architectural features at LA 170609
is also evident in the 7:18am thermal image, which reveals the
micro-relief of the site quite well (Fig. 8E). The previous partial
excavation of LA 170609 may somewhat improve the clarity with
which it appears on thermal imagery, although architectural details
of numerous other roomblocks that have never been excavated can
also be resolved quite clearly, as described below.
At nearby LA 170607, just 30 m to the southwest, surface survey
recorded two possible clusters of buildings, based on the presence
of stones and artifacts, but the paucity of material prevented cer-
tainty. On the thermal data, these two building clusters appear
clearly, and remains of subsurface rectilinear walls can also be
traced (Fig. 9A). The clearest of these room blocks are the same size
and shape as the architecture excavated at LA 170609, providing
some confidence in our interpretation.
At LA 170608 just south of the modern road, surface manifes-
tations were especially difficult to discern during survey operations
due to dense vegetation growing over the site (Fig. 9B). Thermal
data, in contrast, provide a clearer indication of the roomblock’s
layout because by comparing thermal and color imagery we can
Fig. 7. Thermal imagery from the 7:18am flight revealing micro-relief (top) compared to a digital surface model created in Photoscan using color photographs (bottom). Locations of
ground control points (dots) and check points (crosses) are indicated. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of
this article.)
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 215
easily eliminate high value features caused by vegetation. Analysis
reveals what is likely a long arc of rectilinear rooms flanking a
central plaza area and a smaller architectural feature, perhaps an
enclosure wall like the one that identified at LA 170609 (Fig. 8), to
the east. Similarly, at LA 170613 (Fig. 10A), thermal data reveal a
rectilinear roomblock facing east towards a small plaza or courtyard
area and a smaller arc of rubble to the east that may also be an
enclosing wall. Several other similar clusters of architectural
Fig. 8. Roomblock LA 170609 as it appears in the (A) 5:18am thermal image and (B) architectural plan produced by test excavations. Compare with the (C) color image, and thermal
images from (D) 6:18am, (E) 7:18am, and (F) 9:58pm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Table 2
Error summaries at four check points used to indicate the quality of alignment for the 5:18am thermal images.
Label X (cm) Y (cm) Z (cm) Total (cm) # images on which
this point appears
Reprojection error
(pixels)
Point 1 �12.00 6.26 4.54 14.31 10 0.15
Point 2 �7.88 0.21 29.48 30.51 15 0.11
Point 7 �4.14 6.39 11.41 13.71 13 0.28
Point 8 11.53 �1.63 13.00 17.46 13 0.12
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219216
remains are clearly visible in the thermal data, but none of them are
evident in color imagery, highlighting the potential of our approach
for identification and mapping of such features at regional scales.
As mentioned previously, identification of subsurface features
such as domestic pithouses and kivas has been difficult in the Blue J
Community. Thermal imagery, however, reveals several features
that may prove to be such structures. One such case discussed
earlier is the possible pit structure in the plaza area of LA 170609
(Fig. 8). Another example is at LA 170612, where surface survey
discovered a low-density scatter of artifacts but no indication of
stone architecture (Fig. 10B). Unlike the high-value clusters that
typify the buried masonry roomblocks, a 10 m diameter, circular
low-value feature with a large number of plants growing inside it is
visible at LA 170612. Its size and configuration indicates it could be
a buried great kiva, a feature expected but not yet identified in the
Blue J Community. It could also be some other type of subsurface
feature like a large pit house. Recent ground-truthing at LA 170612
found nothing on the surface to conclusively indicate the presence
of a subsurface feature, nor did it identify another reason for the
low-value circular signal. Certainly, if a deep kiva structure or large
pit house had been filled by aeolian and slope wash sediments, the
differences in composition and porosity could reasonably be ex-
pected to create a low-value (i.e., cool) thermal anomaly. Of course,
subsurface testing will be needed to determine what the feature is,
but it offers an example of themany features that might be revealed
by future thermal imaging.
7. Conclusions
The aerial thermographic imaging presented here produced
valuable insights regarding our study site, the Blue J Community,
and show both the potential of the methodology for archaeological
investigations at sites in similar environments as well as for rapid
archaeo-geophysical data collection over very large areas. At Blue J
itself, the thermal imagery we collected reveals nearly all archae-
ological features that had been previously recorded through
pedestrian survey and subsurface testing, suggesting that in the
future, thermography could be used as a means to identify sites
ahead of field survey. The data also reveal many features that were
not evident on the ground surface and which merit additional
investigation. This includes the possible northern extension of the
enclosing wall at LA 170609 that testing may have overlooked
(Fig. 8), the roomblock layout at LA 170613 that dense vegetation
otherwise obscured (Fig. 9B), and the low-value circular features at
LA 170609 and LA 170612 that may be buried pit structures or kivas
(Figs. 8 and 10B). Future fieldwork at the Blue J Community will be
able to target these areas for additional investigation.
Previous studies have already demonstrated thermography can
be a powerful complement to conventional archaeological
geophysics, by revealing a distinct set of physical phenomena (e.g.,
Kvamme, 2006, 2008). Moreover, investigators working at sites in
the American Southwest that are similar to Blue J, where soils are
very dry and architecture is built of stones with little or no
Fig. 9. Sites LA 170607 (A) and LA 170608 (B) in 5:18am thermal image (left), interpretive plan (center), and color image (right). (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 217
magnetic signature, have often found that both electrical resistivity
and magnetic gradiometry yield poor results (e.g., Ernenwein,
2008). The extremely rocky, topographically complex surface at
Blue J, combined with the dense, thorny vegetation would likewise
make ground penetrating radar survey extremely challenging. Our
aerial thermographic method thus offers a key tool to complement
other more conventional archaeological geophysics, and an
approach that is of particular value to investigations in the Amer-
ican Southwest as well as at sites in other similar environments
worldwide.
Finally, the method for UAV-based thermography outlined in
this paper offers a means to collect and process thermal imagery
over very large areas extremely rapidly, which is perhaps its
greatest advantage. For example, the technique could be a powerful
method for investigations of large sites where conventional
geophysics might require many months of fieldwork, as well as in
situations where more ephemeral remains are scattered over
extensive areas. It could also provide a valuable complement to
regional archaeological surveys, by revealing the likely location of
archaeological features across large areas. While it would likely be
difficult to acquire archaeologically useful thermal data in situa-
tionswhere cultural remains are covered by dense vegetation or are
deeply buried, the methods we have described herein have broad
applications in many contexts worldwide. Moreover, the ease with
which imagery can nowbe acquired opens upmany newavenues of
future research. Investigators could, for example, exploit diurnal
and seasonal differences in the appearance of archaeological fea-
tures to enhance their visibility or determine their depth. Similarly,
the differential thermal properties of vegetation could be used to
indirectly detect archaeological features, as could the subtle topo-
graphic relief that is recorded in early morning thermal imagery
(Fig. 7). Likely improvements in the reliability and simplicity of
UAVs will make acquisition of thermal imagery increasingly easy
and inexpensive, facilitating continued research into its many
possibilities and ultimately making aerial thermography a standard
stage in archaeological investigations.
Acknowledgements
Research presented herein was supported by a grant from the
National Endowment for the Humanities through the Digital Hu-
manities Start-Up program (HD-5159012). Previous archaeological
investigations at Blue J, including excavations at LA 170609, were
supported by grants from the National Science Foundation
(#0237579 and #0715996). Instruments, hardware and software
used in this project were made available through the University of
Arkansas’ Center for Advanced Spatial Technologies (CAST). Special
thanks goes to CAST staff members who provided invaluable
consultation and assistance with various aspects of this project,
especially Bill Johnson, Adam Barnes, and Katie Simon. Ken
Fig. 10. Sites LA 170613 (A) and LA 170612 (B) in 5:18am thermal image (left), interpretive plan (center), and color image (right). The low-value circular anomaly evident at LA
170612 may indicate the location of subterranean pit house or kiva. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219218
Kvamme also offered useful suggestions regarding data acquisition
and processing, and the development of our methodologies
benefited from collaboration with Kevin Fisher and Andrew Clark.
Finally, we owe enormous thanks to the student volunteers who
piloted the UAV during all flights undertaken as part of this project.
References
Agisoft, 2013. Agisoft PhotoScan User Manual: Professional Edition, Version 1.0.0.Agisoft LLC.
Bellerby, T.J., Noel, M., Branigan, K., 1990. A thermal method for archaeologicalprospection: preliminary investigations. Archaeometry 32, 191e203.
Berlin, G.L., Ambler, J.R., Hevly, R.H., Schaber, G.G., 1977. Identification of a Sinaguaagricultural field by aerial thermography, soil chemistry, pollen/plant analysis,and archaeology. Am. Antiq. 42, 588e600.
Ben-Dor, E., Kochavi, M., Vinizki, L., Shionim, M., Portugali, J., 2001. Detection ofburied ancient walls using airborne thermal video radiometry. Int. J. RemoteSens. 22, 3689e3702.
Buck, P.E., Sabol, D.E., Gillespie, A.R., 2003. Sub-pixel artifact detection using remotesensing. J. Archaeol. Sci. 30 (8), 973e989.
Challis, K., Kincey, M., Howard, A., 2009. Airborne remote sensing of valley floorgeoarchaeology using Daedalus ATM and CASI. Archaeol. Prospect. 16, 17e33.
Chiabrando, F., Nexb, F., Piatti, D., Rinaudo, F., 2011. UAV and RPV systems forphotogrammetric surveys in archaeological areas: two tests in the Piedmontregion (Italy). J. Archaeol. Sci. 38, 697e710.
De Reu, J., Plets, G., Verhoeven, G., De Smedt, P., Bats, M., Cherretté, B., DeMaeyer, W., Deconynck, J., Herremans, D., Laloo, P., Van Meirvenne, M., DeClercq, W., 2013. Towards a three-dimensional cost-effective registration of thearchaeological heritage. J. Archaeol. Sci. 40, 1108e1121.
Eisenbeiss, H., Zhang, Li, 2006. Comparison of DSMs generated from mini UAVimagery and terrestrial laser scanner in a cultural heritage application. IAPRS 36(5), 90e96.
Eisenbeiss, H., Sauerbier, M., 2012. Investigation of UAV systems and flight modesfor photogrammetric applications. Photogramm. Rec. 26 (136), 400e421.
Ernenwein, E.G., 2008. A geophysical view of Pueblo Escondido: implications for thepithouse to Pueblo transition in the Jornada Mogollon. Bull. Tex. Archaeol. Soc.79, 125e145.
FLIR, 2013. Is the Tau/Quark a Rolling Shutter Camera or a Framing Camera?. http://www.flir.com/cvs/cores/knowledgebase/index.cfm?CFTREEITEMKEY¼337&view¼35773 (accessed 02.02.14.).
Fourteau, A.M., Tabbagh, A., 1979. Parcellaire fossile et prospection thermique, re-sultants des recherches à Lion en Beauce (Loiret). Rev. d’Archéom. 3, 115e123.
Freefly Systems, 2014. Cinestar 3-axis Gimbal. http://www.freeflysystems.com/products/cinestar-3-axis-gimbal.php (accessed 02.02.14.).
Giardino, M., Haley, B., 2006. Airborne remote sensing and geospatial analysis. In:Johnson, J. (Ed.), Remote Sensing in Archaeology: an Explicitly North AmericanPerspective. University of Alabama Press, Tuscaloosa, pp. 47e77.
Hailey, T.I., 2005. The powered parachute as an archaeological aerial reconnaissancevehicle. Archaeol. Prospect. 12, 69e78.
Haley, B., Johnson, J., Stallings, R., 2002. The Utility of Low Cost Thermal Sensors inArchaeological Research. Report Prepared for the Office of Naval Research NASAGrant NAG5-7671. Center for Archaeological Research, University of Mississippi.
Hill, A.C., 2013. UAVs at Marj Rabba, Israel: low-cost high-tech tools for aerialphotography and photogrammetry. SAA Archaeol. Rec. 13 (1), 25e29.
Kantner, J., 2004. The Ancient Puebloan Southwest. Cambridge University Press,Cambridge.
Kantner, J., 2010. Casamero in the Chaco world. In: Mathien, J.F. (Ed.), The CasameroCommunity in the Red Mesa Valley of Northwestern New Mexico. Archaeo-logical Society of New Mexico, Albuquerque, pp. 189e207. Special PublicationNo. 4.
Kantner, J., 2013. Archaeological Investigation of the Blue J Ancestral PuebloanCommunity Near Prewitt, New Mexico. Report on file with the Bureau of LandManagement. Farmington Office, New Mexico.
Kantner, J., Kintigh, K., 2006. The Chaco world. In: Lekson, S.H. (Ed.), The Archae-ology of Chaco Canyon: an Eleventh-Century Pueblo Regional Center. School ofAmerican Research Press, Santa Fe, NM, pp. 153e188.
Kantner, J., Vaughn, K.J., 2012. Pilgrimage as costly signal: religiously motivatedcooperation in Chaco and Nasca. J. Anthropol. Archaeol. 31 (1), 66e82.
Koutsoudis, A., Arnaoutoglou, F., Chamzas, C., 2007. On 3D reconstruction of the oldcity of Xanthi. A minimum budget approach to virtual touring based onphotogrammetry. J. Cult. Herit. 8 (1), 26e31.
Kvamme, K.L., 2006. Integrating multi-dimensional geophysical data. Archaeol.Prospect. 13, 57e72.
Kvamme, K.L., 2008. Archaeological prospecting at the double ditch state historicsite, North Dakota, USA. Archaeol. Prospect. 15, 62e79.
Lekson, S.H., 2009. A History of the Ancient Southwest. SAR Press, Santa Fe, NM.Lunden, B., 1985. Aerial thermography: a remote sensing technique applied to
detection of buried archaeological Remains at a site in Dalecarlia. Sweden.Geogr. Anna. Phys. Geogr. 67 (1/2), 161e166.
Mikolajczyk, K., Schmid, C., 2004. Scale and affine invariant interest point detectors.Int. J. Comput. Vis. 60 (1), 63e86.
Perisset, M.C., Tabbagh, A., 1981. Interpretation of thermal prospection on bare soils.Archaeometry 23 (2), 169e187.
Prudden, T.M., 1903. The prehistoric ruins of the San Juan watershed in Utah, Ari-zona, Colorado, and New Mexico. Am. Anthropol. 5 (2), 224e288.
Reed, P.F., 2011. Chacoan immigration or emulation of the Chacoan system? Theemergence of Aztec, Salmon, and other great house communities in the middleSan Juan. Kiva 77, 119e138.
Sauerbier, M., Eisenbeiss, H., 2010. UAVS for the documentation of archaeologicalexcavations. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 38 (5), 526e531.In: http://www.isprs.org/proceedings/XXXVIII/part5/papers/214.pdf.
Scollar, I., Tabbagh, A., Hesse, A., Herzog, A., 1990. Archaeological Prospecting andRemote Sensing. Cambridge University Press, Cambridge.
Seitz, C., Altenbach, H., 2011. Project ARCHEYE: the quadrocopter as the archaeol-ogist’s eye. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 38 (1), 1e6.
Sever, T.L., Wagner, D.W., 1991. Analysis of prehistoric roadways in Chaco Canyonusing remotely sensed data. In: Trombold, C.D. (Ed.), Ancient Road Networksand Settlement Hierarchies in the New World. Cambridge University Press,Cambridge, pp. 42e52.
Sheets, P., Sever, T.L., 1991. Prehistoric footpaths in Costa Rica: transportation andcommunication in a tropical rainforest. In: Trombold, C.D. (Ed.), Ancient RoadNetworks and Settlement Hierarchies in the New World. Cambridge UniversityPress, Cambridge, pp. 53e65.
Tabbagh, A., 1977. Sur la determination du moment de mesure favorable et l’in-terprétation des resultants en prospection thermique archéologique. Ann.Géophys. 33, 243e254.
Triggs, B., Mclauchlan, P., Hartley, R., Fitzgibbon, A., 2000. Bundle Adjustmente a Modern Synthesis. Vision Algorithms: Theory and Practice, LNCS,pp. 298e375.
Van Dyke, R.M., 2003. Bounding Chaco: great house architectural variability acrosstime and space. Kiva 69, 117e139.
Van Dyke, R.M., 2007. Great kivas in time, space, and society. In: Lekson, S.H. (Ed.),The Architecture of Chaco Canyon, New Mexico. University of Utah Press, SaltLake City, pp. 93e126.
Verhoeven, G., 2011. Taking computer vision aloft e archaeological three dimen-sional reconstructions from aerial photographs with photoscan. Archaeol.Prospect. 18, 67e73.
Verhoeven, G., 2009. Providingan archaeological bird’s-eye view: an overall pictureof ground-based means to execute low-altitude aerial photography (LAAP) inarchaeology. Archaeol. Prospect. 16, 233e249.
Verhoeven, G., Doneus, M., Briese, C., Vermeulen, F., 2012a. Mapping by matching: acomputer vision-based approach to fast and accurate georeferencing ofarchaeological aerial photographs. J. Archaeol. Sci. 39, 2060e2070.
Verhoeven, G., Taelman, D., Vermeulen, F., 2012b. Computer vision-based ortho-photo mapping of complex archaeological sites: the ancient quarry of Pitaranha(Portugal-Spain). Archaeometry 54, 1114e1129.
Verhoeven, G., Docter, R., May 29th e June 2nd 2013. The amphitheatre of Car-nuntum e towards a complete 3D model using airborne structure from motionand dense image matching. In: Proceedings of the 10th International Confer-ence on Archaeological Prospection, pp. 438e440. Vienna, Austria.
Wells, J., November 2011. Kite aerial thermography. Int. Soc. Archaeol. Prospect.News 29, 9e10.
J. Casana et al. / Journal of Archaeological Science 45 (2014) 207e219 219