You are on page 1of 15

Journal of Field Archaeology

ISSN: 0093-4690 (Print) 2042-4582 (Online) Journal homepage: https://www.tandfonline.com/loi/yjfa20

Lasers Without Lost Cities: Using Drone Lidar


to Capture Architectural Complexity at Kuelap,
Amazonas, Peru

Parker VanValkenburgh, K. C. Cushman, Luis Jaime Castillo Butters, Carol


Rojas Vega, Carson B. Roberts, Charles Kepler & James Kellner

To cite this article: Parker VanValkenburgh, K. C. Cushman, Luis Jaime Castillo Butters, Carol
Rojas Vega, Carson B. Roberts, Charles Kepler & James Kellner (2020) Lasers Without Lost Cities:
Using Drone Lidar to Capture Architectural Complexity at Kuelap, Amazonas, Peru, Journal of Field
Archaeology, 45:sup1, S75-S88, DOI: 10.1080/00934690.2020.1713287

To link to this article: https://doi.org/10.1080/00934690.2020.1713287

© 2020 The Author(s). Published by Informa


UK Limited, trading as Taylor & Francis
Group

Published online: 12 Feb 2020.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=yjfa20
JOURNAL OF FIELD ARCHAEOLOGY
2020, VOL. 45, NO. S1, S75–S88
https://doi.org/10.1080/00934690.2020.1713287

Lasers Without Lost Cities: Using Drone Lidar to Capture Architectural Complexity
at Kuelap, Amazonas, Peru
Parker VanValkenburgh a, K. C. Cushman a
, Luis Jaime Castillo Buttersb, Carol Rojas Vegac, Carson B. Robertsd,
Charles Keplerd and James Kellnera
a
Brown University; bPontificia Universidad Católica del Perú, Lima, Peru; cIndependent Scholar, Lima, Peru; dHeadwall Photonics, Bolton, MA, USA

ABSTRACT KEYWORDS
We report the results of drone lidar survey at a high-elevation archaeological site in the Chachapoyas Lidar; UAV; Drone;
region of Peruvian Amazonia. Unlike traditional airborne remote sensing, drone lidar produces very architecture; Peru;
high-density measurements at a wide range of scan angles by operating at low altitudes and slow Chachapoya; Kuelap
flight speeds. These measurements can resolve near vertical surfaces and novel dimensions of
variability in architectural datasets. We show in a case study at Kuelap that the number of detected
structures almost exactly matches the number reported from previous ground level surveys, and
we use these data to quantify the relative circularity and size frequency distribution of architectural
structures. We demonstrate variability in domestic architecture that was obscured in previous
models produced using low-resolution remote sensing. Spatial analysis of these attributes produces
new hypotheses about the site’s construction history and social organization.

Introduction
for mapping previously un-surveyed archaeological landscapes
Airborne lidar (light detection and ranging) has revolutionized and capturing new dimensions of variability within them. Dec-
the mapping of archaeological sites and landscapes over the last ades of excavations and pedestrian reconnaissance have
15 years. Lidar’s combination of automated, active, high den- demonstrated that the montaña has a deep history of human
sity, and long-distance signal emission has proven to be increas- occupation and that late Prehispanic population sizes were, in
ingly effective at generating high resolution ground models of many areas, substantially larger than those found in the region
archaeological sites located in a wide variety of field conditions, today (Bonavia 1981; Church 1996; Church and Von Hagen
including areas of dense vegetation. Flown on aerial platforms, 2008; Clasby 2014; Guengerich 2015; Le Moine and Raymond
lidar sensors are capable of producing datasets that are “big” in 1987; Raymond 1992; Schjellerup 2003). As population
two senses: 1) they have higher levels of spatial resolution (and declined after the Spanish invasion due to colonial violence, dis-
file sizes) than satellite remote sensing datasets and even image- ease, and migration, many settlements, terracing systems, and
based models; and 2) they achieve wider geographic coverage grazing lands were abandoned, and tropical forest vegetation
than is often possible through terrestrial laser scanning (TLS). recolonized landscapes that had formerly been the sites of
Analytically, lidar datasets (even those with limited geographic intensive economic exploitation and habitation. As a result,
footprints) may illustrate one of the key characteristics of big many sections of the ceja de selva now contain extensive
aerial data—the revelation of new dimensions of variability anthropogenic features that are obscured by dense overgrowth,
(and, sometimes, unexpected insights) in mapped phenomena, making them quite difficult to map on an individual basis and
by virtue of their resolution, scale, and/or ability to capture pre- impeding the progress of systematic pedestrian survey.
viously unmapped features. More than simply presenting practical challenges for
To be sure, some airborne lidar surveys have merely served archaeological fieldwork, landcover and topography in the
to confirm existing interpretations of archaeological land- eastern slopes have also impeded scientific understanding of
scapes rather than generating novel insights about them. regional history, obscuring the place of the montaña in
Yet a great number of surveys have indeed enabled scholars regional social developments, including those linking the
to create extensive ground models of unmapped sites and Andes and the Amazon. For much of their postcolonial his-
yielded transformational insights into the scale and character tory, the eastern slopes have been treated as a marginal region
of urban and agricultural landscapes, particularly in forested caught between the “green hell” or “counterfeit paradise” of
environments (Canuto et al. 2018; Chase et al. 2011; Chase the lowland Amazon (Meggers 1971; Goodland and Irwin
et al. 2012; Chase et al. 2014; Fisher and Leisz 2013; Fisher 1975) and the supposedly more developed highland and
et al. 2016; Garrison, Houston, and Alcover Firpi 2019; coastal regions of the Andes (Church 1996; Church and
Golden et al. 2016; Hesse 2010; Masini, Coluzzi, and Lasapo- Von Hagen 2008; Church and Guengerich 2017). Even as
nara 2011; Opitz et al. 2015). scientific investigations at sites in the ceja de selva have
The eastern slopes of the Peruvian Andes (also known as the been carried out since the mid 20th century (e.g., Reichlen
montaña or ceja de selva—the “eyebrow of the jungle”) are a and Reichlen 1950; Ruiz Estrada 1972, 1979; Thompson
region in which airborne lidar holds a great deal of promise 1976), explorers and the tourist industry have continued to

CONTACT Parker VanValkenburgh parker_vanvalkenburgh@brown.edu Department of Anthropology, Brown University, 128 Hope St., Providence, RI 02912,
USA
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
S76 P. VANVALKENBURGH ET AL.

portray the area as mysterious and impenetrable (Savoy among some archaeologists working in the region that air-
1970). Archaeologists working up and down the eastern borne lidar has limited applications in the eastern slopes of
edge of the cordillera have chipped away at these myths, illus- the Andes.
trating the diversity of montaña archaeology and socio-cul- Our experience employing drone lidar for localized map-
tural formations (Church 1996; Guengerich 2014; Narváez ping at Kuelap suggests that more satisfying results can be
Vargas 2013; Schjellerup 1997; Wilkinson 2013). In parallel, achieved through modifications to instrumentation and
research in the lowland Amazon has demonstrated that that flight parameters. As with any application of imaging tech-
region was far more populated in Precolumbian times than nology, successful data collection depends on careful pairings
once presumed (Arroyo-Kalin 2010; Clement et al. 2015; of hardware and collection parameters with project areas and
Heckenberger et al. 2008; Heckenberger 2005; Levis et al. research questions. The configurations of drone lidar that we
2017; Neves et al. 2003; Stahl 2002) and that some inno- employ here offer one solution to mapping relatively small
vations and cultural traditions later shared across extensive areas of forest and underlying terrain in great detail. We illus-
areas of the Andes may have arisen in Amazonia (Valdez trate its potential and limitations through data collection at
2008; Valdez 2007; Clasby 2014). Together, these develop- the site of Kuelap, located in the middle Utcubamba River
ments have reframed the importance of the eastern slopes Valley of the Peruvian department of Amazonas, an area typi-
to continental archaeology and seeded the development of a cally affiliated with the Chachapoya.
robust regional research community.
Yet landcover and topography continue to present meth-
odological challenges for research in the ceja de selva. Because Background
of the difficulty of conducting conventional pedestrian sur-
Kuelap and the Chachapoya
vey, regional histories of the eastern slopes of the Peruvian
Andes continue to rely on site-based research and chronolo- Alongside the sections of Cuzco’s Urubamba valley made
gies, and substantial disagreements remain concerning major famous by the archaeological site of Machu Picchu, the Cha-
demographic, social, and political developments. On the level chapoyas region is perhaps the sector of the Peruvian mon-
of individual sites, vegetation cover impedes the use of many taña that is best known to archaeologists, if not also the
analog and digital technologies for recording settlement general public. Described in early Spanish colonial chronicles
structure and architectural variation. Tree cover obscures as a province of the Inka empire, Chachapoyas became an
GNSS signals and makes detailed work with optical instru- administrative unit of the Spanish Viceroyalty following its
ments such as theodolites and total stations so time-consum- occupation in the 1540’s C.E. These same sources frequently
ing that individual structures often cannot be recorded in credit the people of this province with holding off an Inka
great detail. As a result, site maps necessarily represent struc- invasion in the 1450’s C.E., before succumbing to a second
tures as somewhat simplified polygons and lines, in a process campaign around 1470 C.E., earning them a reputation for
of distillation and projection similar to that carried out by ferocity and military prowess—albeit one that may reflect
archaeologists in another forest-covered region, lowland Inka imperial propaganda more than lived experience
Mesoamerica (Hutson 2012). In such contexts, the simplifica- (Toyne and Narváez 2014).
tion of architectural form is a practical necessity that still For generations, archaeologists have also used the word
allows for a great deal of insight into the organization of indi- Chachapoya to describe a complex of supposedly common
vidual settlements. However, it can also conceal meaningful cultural traits found among Late Intermediate Period (ca.
variation. As previous researchers have shown, attention to 1100–1450 C.E.) archaeological sites, to the east of the
subtle differences in architectural style and execution may Marañón River ranging from beyond the furthest southern
reveal the work of individual builders and outline the con- border of what is now Rio Abiseo National Park, in the
tours of local traditions, enabling archaeologists to “people” department of La Libertad, to as far north as the confluence
the past and to build more fine-grained understanding of of the Utcubamba and Marañón Rivers (Figure 1). Early
social life and site chronology (Tringham 1991; Guengerich models presumed that these traits—which include hilltop
2014; Halperin and Schwartz 2016). settlements comprised of circular houses, sometimes with
Lidar, both terrestrial and airborne, provides means for large platform bases and exterior mosaic friezes, as well as
improving the capture of subtle variation in architectural certain decorative motifs in pottery and textiles—outlined a
form, across large areas, in challenging field conditions. cultural unity that underlay the apparent regional political
Because of its proven ability to create ground models in den- coherence presented in the chronicles (Church 1996: Church
sely forested contexts more efficiently than TLS, airborne and Guengerich 2017). However, historical research suggests
lidar will form an important component of solutions for that the political and ethnic landscapes of the eastern slopes
both regional and local mapping projects on the eastern of Amazonas, La Libertad, and San Martin departments
slopes of the Andes (Khan, Aragão, and Iriarte 2017). How- were quite fragmented (Espinoza Soriano 1967; Ruiz Estrada
ever, among Andeanists, there has been some skepticism of 2017), and recent archaeological research has highlighted
lidar’s potential contributions to mapping sites in the mon- additional diversity in material culture, casting doubt on
taña. In 2016, a consortium called the Peru Lidar Project the unity of Chachapoya culture, even as some elements of
(www.facebook.com/mplidar/) conducted limited airborne shared traditions remain evident (Crandall 2017; McCray
lidar survey over Machu Picchu Archaeological Park and 2017; Church and Guengerich 2017).
Río Abiseo National Park, both located in heavily forested The archaeological site of Kuelap, centered on a limestone
regions of the selva alta. Subcanopy point densities from ridge at 3000 meters above sea level on the western flank of
these flights proved to be too low to resolve major archaeolo- the Utcubamba River Valley, has been almost inseparable
gical features, in part due to the density of understory veg- from the identity of the Chachapoya as an archaeological cul-
etation. These limited results have led to a perception ture since the times of its initial description in the 19th and
JOURNAL OF FIELD ARCHAEOLOGY S77

Figure 1. Map of Kuelap’s location within Peru and the department of Amazonas; locations mentioned in text highlighted.

early 20th century. Comprised of remains spread over as archaeologists for generations (Bandelier 1907; Langlois
many as 460 hectares and dating from at least the Middle 1940; Reichlen and Reichlen 1950; Ruiz Estrada 1972; Bradley
Horizon (800–1100 C.E.) through the Late Intermediate 2005; Narváez Vargas 2013; Toyne and Narváez 2011).
Period (1100–14570 C.E.), Late Horizon / Inka (1470–1535 While tourist literature has taken to promoting Kuelap as a
C.E.), Spanish Colonial (1535–1821 C.E.), and Republican “new Machu Picchu” and frequently depicted the site as being
(1821 C.E.–present) periods, Kuelap is nevertheless often a mysterious settlement in the jungle, the terrain around it is
equated with the so-called “fortress” or citadel, a walled settle- in fact quite diverse, consisting of a patchwork of tropical for-
ment that lies atop the limestone mass of Cerro la Barreta est, scrubland, and agricultural fields. During state-sponsored
(Church and Von Hagen 2008, 914). Its monumental exterior archaeological and conservation work in the walled settle-
walls, which consist of worked limestone blocks stacked ment between 2003 and 2011, tree cover was substantially
between 10 to 20 meters above the ground surface and reduced for the purpose of architectural conservation and
extending some 1500 meters around the ridgetop (a measure- consolidation. Still, remaining tree cover poses significant
ment that includes walls flanking the site’s three main challenges for capturing the site surface using total stations,
entrances), have attracted the interest of explorers and high resolution terrestrial GNSS and aerial photography.
S78 P. VANVALKENBURGH ET AL.

Detailed mapping was conducted by Alfredo Narváez Vargas closely aligned with a vein of exposed bedrock within the
in 1986–1987, but published versions of that map and more upper town, the other built directly atop circular structures
recent cartography of the site are primarily focused on repre- in the south. These features make it clear that Kuelap was
senting the relative number and position of structures and do also actively incorporated into the Inka empire. Isolated
not capture fine-grained variation in individual constructions finds of post-conquest material culture point to continued
(e.g., Narváez Vargas 2013, 122–123; Toyne and Narváez activity following the Spanish invasion of the Utcubamba val-
2014, 345). ley (Narváez Vargas 2013, 151, 155–156). As we discuss
Kuelap is surely the most extensively studied of all sites below, variation in structural morphology recorded in ground
located in the Utcubamba river valley, if not the Chachapoyas models of the site frame additional hypotheses about its con-
region as a whole (Guengerich 2015, 367), but a number of struction sequence, which might be tested with further exca-
essential questions remain about its history and construction. vations and dating.
The site’s architecture abounds in defensive affordances: the Previous excavations in the citadel have already generated
citadel is built atop a ridgetop with a cliff on one side, is sur- insights into the character and pacing of Kuelap’s construc-
rounded by massive walls, and has highly restricted access via tion. Behind the exterior walls, archaeologists have revealed
three narrow, stepped entryways that are themselves flanked an enormous mass of rubble and earthen fill (Ruiz Estrada
by high stone walls. In the northwestern sector of the citadel, 2009, 39–45), but the exposed bedrock within the upper
the “upper town” (pueblo alto) is further separated from the town indicates that not all of the space seemingly “behind”
rest of the settlement by walls 5–10 meters high, with two the exterior walls is filled with undifferentiated rubble. Rather,
narrow access points quite similar to those that penetrate like many Precolumbian monuments built in the Peruvian
its exterior walls. Other evidence suggests that Kuelap wit- highlands and coast, the walled settlement is a modified
nessed at least occasional moments of conflict during its his- mountaintop whose architects incorporated underlying rock
tory. At the far northern portion of the walled settlement, formations into its construction. Still, the immensity of the
adjacent to the D-shaped torreón, excavations uncovered a fill and its stone facing, as well as “clearly defined segments”
cache of over 2500 sling stones, the most common form of within the fill itself—which recall the columned segments of
projectile weapon used in the Precolumbian Andes (Narváez adobe bricks in coastal monuments such as the Huaca de la
Vargas 2013, 138; Church and Von Hagen 2008, 916). Toyne Luna (Hastings and Moseley 1975; Shimada and Cavallaro
and Narváez (2014, 351) also present evidence of a “mas- 1985)—suggest to Narváez (2013, 140) that it must have
sacre” in which 94.3% of the observable collection of 66 skulls been built over centuries rather than in one single flurry.
of individuals buried in the site’s plataforma circular (circular Excavations sponsored by Peru’s Plan COPESCO between
platform) showed evidence of perimortem blunt force 2016 and 2018 have also revealed evidence of at least one pre-
trauma. Differences in chronological scenarios discussed by vious construction stage near the site’s main entrance, in the
Narváez Vargas (2013) and Toyne and Narváez (2014) form of an earlier worked stone wall, nested more than ten
suggest that dating of this deposit to the Late Intermediate, meters within the outermost one (Jorge Chiguala, personal
Inka, or even Spanish colonial periods is not yet definitive. communication 2017). Coupled with Ruiz’s (1972, 2009,
Yet both Bradley (2005; n.d.) and Narváez Vargas (2013) 15) inference that the upper town is the oldest portion of
also suggest that the classification of the site as a “fortress” the site (because excavations within it produced the oldest
is overly simplistic, noting the remains of highly ritualized ceramics), these patterns make it possible that the Kuelap
activities in many sectors of the site, including great numbers citadel grew in stages from west to east and perhaps north
of human burials associated with exterior walls. Excavations to south. We return to this theory in the discussion and con-
in the idiosyncratic structure known as the tintero (ink clusion section below.
well), a solid mass of stone and fill with a bulb-shaped shaft Further debate centers on sociopolitical organization at
down its middle, have produced numerous secondary Kuelap during the Late Intermediate Period. Many early
human burials, carved Spondylus shell figures and exotic interpretations took the site’s monumental walls as evidence
styles of pottery. Narváez interprets these remains as evidence that it must have served as a center of concentrated political
that the structure functioned as the primary ritual locus of the power (Gil 1938; Brush 1977). Narváez (2013, 131) argues
walled settlement, and he has accordingly renamed it the tem- that the exceptionality of Kuelap’s architecture and occu-
plo mayor (main temple). pation makes it likely that the site served as the “political capi-
Site chronology is also a matter of debate. Based on stylis- tal of the Chachapoyas kingdom” whose “power was capable
tic similarities between some pottery found in lower levels of of mobilizing the collective labor of diverse affiliated nations
excavations, Arturo Ruiz Estrada (1972) has suggested that under its rule … [and] of organizing the political control of a
Kuelap was occupied since at least the Middle Horizon vast territory whose limits reached the Marañón river in the
(800–1100 C.E.)—an hypothesis echoed by Narváez (1987, west and the Huallaga River in the east” (my translation).
2013). The site’s dates of foundation have reportedly been In contrast, Church and Von Hagen (2008), Guengerich
pushed back even further via radiocarbon dating of samples (2015) and Church and Guengerich (2017) note that while
from the templo mayor (Narváez Vargas 2013, 150), but the the site’s monumental walls are indeed exceptional, Kuelap’s
results remain unpublished. More fine-grained dating of size does not place it atop the kinds of settlement hierarchies
structures and occupations (particularly those located near that are typical of highly centralized political systems. Indeed,
the surface) has yet to be conducted, though Narváez the number of structures at the site (and likely, population
(2013) suggests that many of these structures may have numbers) are not exceptional when compared to other
been constructed following the Inka invasion. Inka imperial large sites in Amazonas and La Libertad, such as Caserones
pottery and prominent Inka constructions conventionally (Schjellerup 1997), Purun Llaqta de Cheto (Ruiz Estrada
called kallankas (long, multi-doored feasting halls) are indeed 2004), and Olán / La Jalca (Lerche 1986). Moreover, archae-
located near the southern and northern limits of the site, one ological excavations at the site have yet to identify unequal
JOURNAL OF FIELD ARCHAEOLOGY S79

distributions of high-quality foods or prestige goods in dom- in some conditions. Notably, the fact that large aircraft need
estic structures, suggesting relatively little in the way of mate- to fly at high altitudes over ground surfaces and forest cano-
rialized class distinctions. pies (both for safety reasons and to avoid disturbing some fea-
tures they seek to image, such as vegetation) limits both their
positional accuracy and their ability to resolve smaller objects
Lidar and Drones
and complex three-dimensional surfaces. Because GPS errors
Lidar is a relatively old technology that has undergone signifi- accumulate as the distance between target and antenna
cant advances in recent years. Both laser scanning and the increases, higher flight altitudes result in worse horizontal
term “lidar” (sometimes also rendered as “LiDAR,” placement of laser pulses (Kellner et al. 2019, 6). Moreover,
“LIDAR,” or “Lidar”) have existed since the 1960’s, but the they also produce larger laser footprints due to beam diver-
technology’s use in civilian mapping applications was limited gence, and scan angles (that is, the angles of emission of the
before precision GPS and IMUs (inertial measurement units) laser pulse with respect to nadir) are restricted to increasingly
became available in the 1990’s. The development of higher narrow ranges (Kellner et al. 2019, 6; Figure 2). Terrestrial
rates of pulse emission and more accurate positioning sys- laser scanning (TLS) offers the advantage of much higher
tems has allowed sensors mounted on crewed aircraft to col- point densities and smaller beam footprints, but can be
lect high-resolution spatial datasets over expansive quite tedious to deploy because TLS instruments need to be
geographic areas. Initial deployment of airborne lidar for moved incrementally in order to ensure scanning of objects
environmental applications produced some datasets useful from multiple angles. Moreover, TLS may perform poorly
for archaeologists, but the high cost of sensors themselves, for scanning the tops of high features such as walls or trees
as well as the associated costs of deployment (including (Calders et al. 2017; Disney et al. 2018).
flight time and fuel) greatly limited archaeologically-specific Two recent advances—commercial UAVs capable of car-
lidar missions until the 2000’s. In the UK and Ireland, the ear- rying substantial payloads and lighter, higher resolution
liest archaeological lidar missions focused on mapping rela- lidar sensors—offer major advantages for some archaeologi-
tively small sites and landscapes with little vegetation cover cal mapping applications. Relatively few published archaeolo-
(e.g., Shell and Roughley 2004; Devereux et al. 2005; Bewley, gical studies have employed drone lidar, but existing literature
Crutchley, and Shell 2005). While measurement rates (33,000 has reported some enhancements vs. sensors mounted on
pulses per second achieved by the likes of the Optech ALTM crewed aircraft (Khan, Aragão, and Iriarte 2017; Risbøl and
3033 sensor used by Shell and Roughley 2004) and point den- Gustavsen 2018). The smaller size of UAVs in comparison
sities (between one point per .5 and 2 square meters Shell and
Roughley 2004; Bewley, Crutchley, and Shell 2005) offered
significant improvement over many previous datasets used
for the production of digital elevation models, they still
proved too low when flown on relatively fast-moving light
aircraft to allow for effective penetration of forest cover.
The applications of aerial laser scanning in archaeology
expanded following the development of sensors with higher
scan speeds and measurement rates, allowing for significantly
higher point densities and measurements of ground surfaces
below forest cover in some conditions. Airborne lidar gained
new visibility in Americanist archaeology following the pub-
lication of large-scale surveys by Chase and colleagues (2011;
2012) at the Maya settlement of Caracol, Belize and Fisher
and Leisz (2013) at Angamuco, Mexico. In both circum-
stances, resulting terrain models revealed scales and densities
of constructions and anthropogenic landscape modifications
that greatly exceeded earlier estimates based on pedestrian
survey. Evans et al. (2013) achieved similar results at Angkor,
Cambodia, demonstrating that the number of structures in
the settlement and the area of modified landscapes around
it was far greater than previously estimated. These studies
have sparked a veritable explosion of interest in the use of
lidar for mapping archaeological sites in forested terrain,
which has only grown with the publication of additional
impactful research (Chase et al. 2014; Garrison, Houston,
and Alcover Firpi 2019; Canuto et al. 2018).
Among Americanist archaeologists, this family of appli-
cations—which for the purpose of shorthand, we refer to
here as “big tropical forest lidar”—has become almost synon-
ymous with aerial laser scanning. But while sensors with long
ranges, flown on crewed aircraft, have proven to be transfor-
mative tools for mapping sites under forest cover in relatively
flat terrain, they are merely one of an expanding array of Figure 2. Comparison of the effects of sensor parameters and flight altitudes on
instrumental configurations, and they face notable challenges scan angles.
S80 P. VANVALKENBURGH ET AL.

to crewed aircraft allows them to fly closer to the surfaces of a Visible/Near Infrared (VNIR) hyperspectral line scan ima-
archaeological sites while following national laws and safety ger and a data acquisition and control computer. During
guidelines. UAVs can also fly at slower speeds than crewed research flights, hyperspectral data were also collected, but
aircraft. Deployed lower altitudes, in slower moving craft, analysis of that data is still ongoing and we do not discuss
UAV-mounted lidar sensors are capable of collecting higher results in this paper. In addition to capturing the spectral
density datasets with smaller beam footprints and improved image data, the Hyperspec Nano stored raw lidar return
horizontal beam placement. Moreover, as Risbøl and Gustav- data and real-time data from the payload GPS-IMU, an
sen (2018) note, lower flights also enable sensors to capture Applanix APX-15. Following collection, raw lidar data were
objects at a greater range of scan angles, improving imaging sent from the sensor to the Nano as a stream of packets
of more vertical surfaces and complex objects—a character- encoded in the User Datagram Protocol and high frequency
istic that should prove particularly valuable for the documen- GPS-IMU data were downloaded directly from the GPS/
tation of archaeological features in the montaña, such as high IMU unit. The GPS data were post-processed using simul-
walls, cliff tombs, stone reliefs, and settlements located on taneously acquired data from a Trimble GNSS Smart Target
very steep slopes (see Figure 2). Base Station in Applanix POSPac UAV software to generate
Using drones to fly lidar sensors can also pose several chal- a Smoothed Best Estimate of Trajectory (SBET) GPS-IMU
lenges for archaeological applications. Battery-powered file and a geolocated point cloud.
UAVs—particularly, multi-copters—have much shorter The aircraft used for these measurements was a DJI
maximum flight times than fueled aircraft, which limits the Matrice 600 Pro, outfitted with high altitude propellers. The
total amount of data that can be collected in a single flight sensor package was mounted on a three-axis actively stabil-
and also places practical restrictions on the locations in ized DJI Ronin gimbal, with the GPS antenna mounted on
which drone lidar can be deployed. Even in countries where top of the aircraft. The lidar sensor, a Velodyne Puck, was
it is legal to fly battery-powered UAVs beyond line-of-sight, mounted so that its axis of rotation was parallel to the direc-
aircraft must be launched relatively close to the sites one tion of motion of the aircraft, so that approximately half the
wishes to capture, to avoid using up battery power simply time the lasers were directed toward the ground and could
flying to and from the desired points of data collection. Find- potentially generate returns. The Puck has 16 lasers, each
ing open areas that provide clear sight lines during flight can paired with a detector, producing six nanosecond pulses at
be particularly challenging in heavily forested terrain. More- 903 nanometers. In the configuration used here, the laser/
over, the large lithium ion batteries that are necessary for detector pairs rotated at approximately 10 Hz and generated
powering motors capable of lifting heavier craft and sensors approximately 300,000 pulses per second. The lidar system
have significant transport restrictions. Fuel-powered UAVs was set to collect the strongest returns from each pulse. Col-
and fixed-wing UAVs are quite promising for improving lecting dual returns (strongest and last) would have provided
flight times and avoiding some of these challenges (Khan, better penetration of the canopy, and potentially more points
Aragão, and Iriarte 2017; Kellner et al. 2019), but even the lar- on the ground below trees.
gest commercially available drones have ranges that are sig- Survey flights were flown under automatic control, using
nificantly shorter than crewed aircraft. Furthermore, the UgCS flight control program. Flight patterns pro-
restrictions on the import, export, and deployment of larger grammed to cover the survey area in parallel flight paths
drones often make it challenging to use them outside of the with approximately 40% overlap between the fields of view
countries where they are built and stored. of the hyperspectral imager. The overlap of the lidar data
Our experience using drone lidar reflects some of these between flight lines was significantly larger. Flights were pro-
limitations. Import restrictions on our first-choice drone—a grammed to be at an altitude of approximately 50 m above
large, fuel-powered UAV—led us to pursue a second option ground level, following a 30m-resolution DEM generated by
involving use of a mass-produced, battery-powered multi- the Shuttle Radar Topography Mission (SRTM). The raw
copter. Internal shipping restrictions then limited the number lidar data, along with the SBET GPS/IMU data were pro-
of batteries to which we had access, which impacted how cessed with Headwall Photonics’ LiDAR Tools program,
much data we could collect in one day. At elevations between which produced point clouds in LAS format.
2800 and 3000 meters, we experienced reduced flight times
(on average, 10 minutes of data collection time per flight)
Data Processing
even while employing propellers designed for use at high
elevations, which limited our ability to fly aircraft at greater Both automated and manual processing methods were used
distances from our launch points. Despite these challenges, to produce a model of the ground and architectural structures
our results demonstrate the transformational potential of at Kuelap using free and open-source tools. First, all lidar
drone lidar for mapping some archaeological sites in forested returns were automatically classified as ground or non-
terrain. ground using the lidR package in R (version 3.5.1) (R Core
Team 2013; Rousell and Auty 2019). To enable processing
on a desktop workstation (64-bit with 12 GB RAM) while
Methods minimizing edge effects on classification, lidar data were re-
tiled to 100 m by 100 m subsets with 5 m overlap prior to
Instrumentation
ground classification, resulting in 37 separate tiles. The
Our instrumentation for this project consisted of a hyper- ground classification algorithm in lidR uses progressive mor-
spectral imager, a lidar sensor, and a GPS-IMU (Global Posi- phological filtering to identify and remove non-ground
tioning System and Internal Measurement Unit) flown on a objects using height thresholds for neighboring points in
commercial hexacopter. The core of the data acquisition sys- increasingly large windows (Zhang et al. 2003). For this appli-
tem was the Headwall Hyperspec Nano, which contains both cation, ground classification parameters were chosen to
JOURNAL OF FIELD ARCHAEOLOGY S81

Figure 3. 40 × 40 m tile representing classification procedures. Ground depicted with elevation color ramp. Non-ground points identified through automatic classifi-
cation shown in black. Non-ground points identified through manual classification shown in white.

remove as much vegetation as possible while retaining all too little separated from the site’s archaeological features to
architectural structures as “ground” points. To achieve this allow for them to be effectively edited out of the point
goal, the ground classification was performed using a single cloud. However, they did not significantly obscure structures
threshold height taller than the architectural structures in of archaeological interest.
the landscape (5.5 m) and using a sequence of window sizes After manual editing, a surface mesh was made of the
(0.6, 0.8, and 1.0 m) that produced consistently good results, walled settlement. To create the surface mesh, all lidar points
as determined by visual inspection on subsets of the land- from ground and architectural structures were loaded into
scape. This automated ground classification procedure was CloudCompare. A height raster of the area was created
successful in removing tall vegetation, but in order to retain using the “Rasterize” projection tool, selecting the maximum
architectural structures of interest tree trunks and other veg- point height in each grid cell and interpolating the value of
etation shorter than 5.5 m were not correctly classified as any empty cells. The raster was made using a 0.1 m grid
non-ground. Additionally, some points were erroneously size, then the raster grid was exported as a mesh. The surface
classified as “non-ground” along the tall outer walls of Kuelap mesh was smoothed using the “Smooth (Laplacian)” mesh
(Figure 3). function with 10 iterations and a smoothing factor of 0.1.
Automatically-classified lidar data were further edited To enable visualization and further measurement of architec-
manually to remove tree trunks, short vegetation, and struc- tural structures, the scene was given a uniform color (grey),
tures not of archeological interest (e.g., wooden handrails set to top view, and shaded using the “EDL Shader” display
along tourist routes within the site, and wooden architectural function.
supports added by conservators to support stone structures at Following mesh generation, the resulting, georeferenced
risk of collapse). To facilitate manual inspection, the lidar hillshade was imported into Imagej version 1.8.0, an open
point cloud was re-tiled to 40 m by 40 m subsets with no source image processing program that runs Java (Rasband
overlap, resulting in a total of 77 tiles. All manual editing 1997). Initially developed for research with imagery produced
was completed using CloudCompare, an open-source plat- in microbiological research, Imagej provides a suite of tools
form for visualizing and processing point clouds (CloudCom- that allow for efficient creation of two-dimensional vector fea-
pare 2019). For each tile, the automated ground classification tures and their morphological analysis. Outlines of each
was loaded as a scalar field and the point cloud was split into building foundation were carefully traced using the Freehand
ground or architectural structure (scalar field = 2) and non- Selection Tool in Imagej, to create a total of 421 individual
ground vegetation (scalar field = 1) using the “Filter by polygons (see Figure 4). The Measure tool in Imagej was
value” scalar field function. Then, manual edits to the auto- then used to generate a series of measurements for each
mated classification values were made using the segment traced feature—area, perimeter, centroid coordinates (X and
and merge features. This method led to great success in gen- Y), aspect ratio (major axis / minor axis), circularity
erating a ground model of the site, though some recent archi- (4*area/perimeter^2, such that a perfect circle has value of
tectural features added to the site—particularly, low wooden 1 and an increasingly elongated shapes approach a value of
walkways located quite close to the ground surface—were 0), and roundness (4pi*area/(pi*major_axis^2, which is also
S82 P. VANVALKENBURGH ET AL.

Figure 4. Features created based on structures visible in lidar ground model of


Kuelap, traced using Imagej.

the inverse of the aspect ratio). These measurements were


then exported as a table and loaded into ArcMap 10.6,
where coordinate values of feature centroids were re-pro-
jected to create a point feature class. The measurements
table from Imagej was then re-joined to this point feature
class, and spatial analyses including hotspot analysis (Getis- Figure 5. Hillshade produced from ground model of Kuelap, with major features
Ord Gi*) were performed on feature attributes. Incremental labeled.
Spatial Autocorrelation did not return any peak values, and
Z-scores were above 1.95 for 11.32 m to 69.92 m sampling At the level of individual walls and buildings, the model’s
distances. Therefore, Getis-Ord Gi* was run using an interval resolution is variable. The survey produced dense point cover-
of 20 m, just over twice the mean diameter (7.3 m) of circular age from the exterior walls of most circular structures, but not
structures in the sample of measured buildings, ensuring that all internal wall surfaces were captured in great detail (see
all features had at least one neighbor but avoiding oversam- Figure 3). This pattern holds across both the forested and
pling that could have resulted from using a larger interval unforested areas of the site, suggesting that the transects we
in an area with close spacing among structures. flew across did not provide the coverage necessary to fully
scan interior surfaces. This trend is unsurprising, as many of
the structures incline inwards, obscuring them from aerial
Results view. Moreover, point density does not appear to be high
enough to resolve fine differences in masonry and individual
Evaluating the Bare Earth Model
building blocks on either the site’s monumental walls or the
Following automated and manual classification, just over walls of smaller domestic structures. For scanning internal
129,897,000 points remained in the point cloud used to pro- portions of domestic structures and resolving individual
duce our bare earth model, at a mean density of approxi- stones, terrestrial laser scanning would provide improved cov-
mately 2000 points per square meter. While the absence of erage, as would drone lidar with flight plans focused on
multiple returns resulted in reduced point density in some increased point density (Kellner et al. 2019). Point densities
of the most forested portions of the site, comparison with on the vertical surface of the site’s monumental external
previous maps of Kuelap suggests that we were able to cap- walls are generally quite high, but they are lower than on the
ture all previously documented surface structures on the top of the citadel, with patchy coverage on the wall’s southeast-
summit of the walled settlement, though we expect that ern portion. Here, point density could no doubt be improved
additional structures remain to be defined beneath the cur- through data collected from lower flights around the exterior
rently exposed surface (Figure 5). The total number of struc- walls, an option we avoided in the field due to safety concerns.
tures recorded (421) is nearly identical to the estimate of 420 Finally, modern features present on the site surface also
generated by previous research (Ruiz Estrada 2009; Narváez impacted data collection. During our survey, the main
Vargas 2013), but the level of detail provided on structural entrance to the citadel was covered with a plastic tarp that
form is much greater. While we concentrate here only on absorbed laser pulses, resulting in a large blank space in the
the formal dimensions of structures, we should also note bare earth model. In addition, while we removed stairways
that the model also captures additional walls that define and handrails from the point cloud during the manual edition
internal patios and divisions. process, low wooden trackways that guide tourist foot traffic
JOURNAL OF FIELD ARCHAEOLOGY S83

through the site are still clearly visible, as are numerous piles of Two clusters of structures have floor areas that are signifi-
stones accumulated during recent conservation work. cantly smaller. The first is a group of eight in the middle por-
tion of the lower town, near the western external wall. They
do not appear to be arranged around a common open
Data Analysis
space. The second is the group atop the raised circular plat-
The high spatial resolution of our model provides a means of form at the citadel’s southern tip, as well as three adjacent
examining previously unstudied aspects of architectural lay- structures. The largest structure on the circular platform,
out and construction at Kuelap, beginning with precise esti- which Narváez (2013) suggests may have been the house of
mates of floor area. Among the 421 structures we mapped an elite family because of its size, is not particularly remark-
on the summit of the citadel, median floor area is 40.53 able for its floor area (51.53 square meters) and it forms part
square meters, with a standard deviation of 16.18 square of the cluster of lower floor area values.
meters and three outliers—the large kallanka in the upper Spatial patterns in building floor areas are mirrored in
town (237.88 sq m), the tintero/templo mayor (158.29 sq m) building perimeters, which is strongly correlated with floor
and a single large circular structure 40 m south of entrance area. Both building aspect ratio and roundness measures
three (126.31 sq m) (see Figure 6). With the outliers removed, (which are the inverse of one another) distinguish the two
SD drops to 11.06 m. Each of the three large structures are kallankas from the site’s circular structures, but neither set
located in distinct sectors of the site, rather than being clus- of values clusters in a spatially significant way. More pro-
tered in a single district. Indeed, relatively little statistically nounced patterns are evident among circularity indices—
significant spatial clustering of either low or high floor area which, indeed, are the attribute whose values are the single
values is present in the dataset. most spatially clustered of the entire dataset. In figure 7,
The appearance of the tintero and the larger of the two kal- both “hot” (highly circular buildings—i.e., those with high
lankas as hotspots seems to reflect the fact that they have rela- circularity indices) and “cold” (structures with exceptionally
tively few neighbors at the 20 m sampling interval—a factor low circularity) appear in several areas of the site. In the
that, at least in the case of the kallanka, is related to its history southern portion of the lower town, just south of entrances
as an Inka structure around which previously existing struc- one and two, a circularity hotspot contains a total of 44 struc-
tures were destroyed in order to make room for an imperial tures. Near the geographic center of the walled site, two adja-
installation placed atop a ritually important element of the cent cold spots (containing 9 and 14 structures, respectively)
landscape (exposed bedrock). Elsewhere, two areas with nearly form a band across the width of the elevated plateau. A
high-confidence clusters of larger floor area values are a further cold spot is located in the highest portion of the site,
group of nine structures arranged around an open space in the circular center of the upper town. Due to its lack of
just to the south of the third entrance, among which is one neighbors and low circularity value (0.562), the kallanka in
of the three largest structures in the site. The other is another the upper town again appears as a singular cold spot. Finally,
set of nine structures that also contain an open space and abut an additional cluster is visible just to the north of entrance
the eastern exterior wall. number three. Notably, none of the clusters in the floor

Figure 6. At left, graduated symbols representing variation in building floor area at Kuelap. (symbols not proportional—larger icon = greater floor area); At right, hot
spot analysis of building size.
S84 P. VANVALKENBURGH ET AL.

Figure 7. At left, graduated symbols representing variation in building circularity values at Kuelap. (Symbols are not proportional—smaller icon = more circular
building). At right, hot spot analysis of building circularity.

area data and the circularity indices significantly overlap with tend to be slightly larger than houses with more coarse
one another. stone walls. There, masonry type is not correlated with the
In addition to data generated by the Measure tool in Ima- remains of higher quality foods or portable material culture,
gej, we should also note that the dataset captures qualitative but still appears convincingly related to social capital. Thus,
differences in structures that are the result of conservation while housing in typical Late Intermediate Period Chachapoya
and archaeological excavation—i.e., the “sharpness” of walls settlements does not demonstrate the kinds of spatial size seg-
apparent among structures in the far southern portion of regation characteristic of hereditary monarchies or modern
the model, near the main entrance, along the central tourist industrial societies, structure size may be interpreted as at
corridor, and in a few isolated areas of the upper town and least partially reflective of social status (Blanton 1994; Sivert-
northern sector of the lower town. Clusters among circularity sen 1963; Beardsley, Hall, and Ward 1959). At Kuelap, four
values crosscut areas of differential conservation, suggesting clusters of 8–11 houses with modestly larger and smaller
that these patterns are not the product of recent conservation floor areas may reflect status shared among co-residential
efforts or reconstruction but rather that they reflect under- groups, which at least in some cases appear to have shared
lying variability in structural form. common patios. Notably, median floor area among structures
at Kuelap appears to be nearly double that at Monte Viudo,
where mean sizes are reportedly between 17 and 22 sq m
Discussion
(Guengerich 2014: figure 12), perhaps reflecting substantial
These two trends—modest clustering of floor area among differences in household sizes between the two settlements
small groups of structures and larger clusters of structures (Naroll 1962; LeBlanc 1970; Kramer 1980).
with similar circularity measures can be considered in light We further interpret large clusters of buildings with simi-
of other trends in architectural data at Kuelap to suggest sev- lar circularity indices as reflecting construction by distinct
eral interpretations regarding the site’s organization and con- communities of builders, as well as perhaps major differences
struction sequence. Each should be further tested through in site chronology. As previous authors have noted, morpho-
excavation and systematic dating efforts. logical features of houses often reveal highly localized and
The absence of sizable clusters of particularly large or small idiosyncratic histories of construction (Blanton 1994; Guen-
houses at Kuelap is a trend previously noted for other Late gerich 2014). The regularity of architectural form among
Intermediate Period settlements in the region (Guengerich Late Intermediate Period Chachapoya settlements (the circu-
2014)—one that differs markedly from more marked lar house, with conical roof, sleeping platform, and central
patterns of size clustering in sites in the central and southern hearth) suggests a normative construction of the home—a
Peruvian Andes (Arkush 2017; Arkush 2011; De Marrais structure of feeling, if not also certain more explicit norms,
2002). At Monte Viudo, Guengerich finds a statistically sig- that guided the work of individual builders. Yet each moment
nificant correlation between house size and masonry type, in of construction articulated these standards in distinct ways,
which houses with some forms of finely-worked masonry reflecting distinct pathways of learning and communities of
JOURNAL OF FIELD ARCHAEOLOGY S85

Figure 8. Boxplots showing distribution of building floor area and circularity of structures at Kuelap, calculated from features created in Imagej based on lidar ground
model.

practice (Wenger 1999; Minar and Crown 2001; Crown 2007; Spatial analysis of these attributes has, in turn, enabled us to
Roddick and Stahl 2016). If, as Guengerich (2014) convin- produce new hypotheses about the history and organization
cingly argues, Chachapoya structures were built not under of the site. Our results illustrate the potential of drone lidar
the watch of strong authorities that imposed standards on for enhancing topographic modeling on what Howey and col-
architects, but rather by groups of builders who introduced leagues (this volume) call the “safe” side of the spatio-temporal
idiosyncrasies into the masonry and layouts of the structures chart of remote sensing technologies, at high levels of spatial
they erected, we should expect the architectural construction resolution. In our case study, it is spatial resolution and file
to vary in time and space due to both drift and innovation. size itself, rather than geographic or temporal scale, that charac-
Under these conditions, we might also expect that subtle differ- terize a collection of “big” data. And just as with the more
ences in architecture would index chronological differences. geographically extensive datasets examined in other papers
Building circularity is one aspect of architectural style we within this themed issue, the true potential of big data lies
might reasonably expect to reflect distinct performances of not in scale itself but in the application of critical analysis to
architectural norms, reflecting different techniques for datasets. Used in a conventional manner, higher resolution
measuring the dimensions of structures, as well as different ground models based on drone lidar data risk reifying the site
standards and expectations of symmetry. At Kuelap, our as a unit of archaeological analysis, under the guise of greater
data reveal large clusters of buildings with similar circularity precision and accuracy (see McCoy, this issue). But examined
values—two of them in east-to-west swaths extending across critically, models based on lidar data can also reveal meaningful
the site, which may reflect its southward and eastward growth patterns in archaeological landscapes that were previously
over time, as suggested by Ruiz Estrada (2009). We suggest undocumented.
that, as new sectors were added to the platform, its exterior While our experience illustrates the interpretive potential of
walls may have been expanded out and circular structures high-resolution datasets of archaeological landscapes, it also
placed atop the platform. In this scenario, each new construc- outlines several of the challenges and limitations of working
tion stage would have been carried out by distinct groups of with lidar data collected from UAV platforms—particularly
builders. In the upper town, where Ruiz’s excavations found the instrumental configuration we employed. In the coming
what appears to be the earliest pottery at the site, exposed years, innovation (and lowering costs) for autonomous vehicles
bedrock suggests that relatively little fill is present. The circu- and optical sensors, driven in part by the rapid development of
lar arrangement of structures at the southern end of the lower technologies for driverless vehicles, will no doubt provide
town, at the highest point of the site, greatly resembles some additional and more affordable options for instrumentation.
smaller hilltop settlements in the middle Utcubamba drainage But certain challenges will remain. Among the varieties of
that we have documented through pedestrian survey (Rojas “big” aerial data, lidar is unique because of its potential to
Vega and VanValkenburgh 2017). Here, we find a substantial combine very high spatial (and in the case of lower cost,
group of houses with low circularity values clustered together, UAV-mounted platforms, high temporal) resolution with
in a manner quite similar to tiered mountaintop settlements extensive geographic coverage. As a result, however, lidar data-
such as Crestón, to the west of the modern town of Chocta- sets can be of sizes that are challenging to process with the com-
mal (Rojas Vega and VanValkenburgh 2017). We suggest putational resources available to archaeologists. In the near
that Kuelap may have begun as just such a settlement—a future, the tradeoff between spatial resolution and prohibitively
small hilltop village that gradually expanded into a walled large file sizes will mean that archaeological projects employing
necropolis whose monumental stone walls, packed with bur- lidar will have to carefully consider their research questions
ials, emulated the form of cliff faces where regional traditions when choosing instrumentation and data collection par-
had long placed their dead (see Bradley, n.d.). We advocate ameters. In the more distant future, additional options for auto-
further stratigraphic excavations to evaluate this proposed mated processing and classification of data may reduce the size
site chronology and history of construction. of these tradeoffs, making it possible for us to handle lidar data-
sets with resolutions of thousands of points per square meter
covering hundreds or thousands of square kilometers.
Conclusion
Our high-resolution ground model of Kuelap, generated from
drone lidar data, has revealed dimensions of variability in dom- Acknowledgments
estic architecture at the site that were obscured in previous Funding for this research was provided through a grant from the
models produced using lower-resolution instrumentation. National Geographic Committee on Research and Exploration (HJ-
S86 P. VANVALKENBURGH ET AL.

044R-17). We would like to thank Jorge Trajo Ramos, Manuel Malaver, Blanton, R. E. 1994. Houses and Households: A Comparative Study.
Jorge Chiguala, Jaime Jimenez, José la Torre, Alejandro Cuipal, Alcira New York: Springer Science & Business Media.
Chavez, Llony Cuipal, Ema Perea Rios, David Blair, Henry Johnson, Bonavia, D. 1981. “Tello y La Arqueologia de La Ceja de Selva.” Historica
and the Institute at Brown for Environment and Society. Cushman’s 5 (2): 149–158.
research has also been supported by a Presidential Fellowship from Bradley, R. 2005. “The Architecture of Kuelap.” Ph.D. Diss., Columbia
Brown University and a Graduate Research Fellowship from the University.
National Science Foundation. Bradley, R. n.d. “Reconsidering the Notion of Fortaleza Kuelap.” https://
www.academia.edu/10678131/Reconsidering_the_Notion_of_Fortal
eza_Kuelap.
Disclosure Statement Brush, S. B. 1977. Mountain, Field, and Family: The Economy and
Human Ecology of an Andean Valley. Philadelphia: University of
No potential conflict of interest was reported by the author. Pennsylvania Press.
Calders, K., M. I. Disney, J. Armston, A. Burt, B. Brede, N. Origo, J. Muir,
and J. Nightengale. 2017. “Evaluation of the Range Accuracy and the
Notes on Contributors Radiometric Calibration of Multiple Terrestrial Laser Scanning
Parker VanValkenburgh (Ph.D. 2012, Harvard University) is Assistant Instruments for Data Interoperability.” IEEE Transactions on
Professor of Anthropology at Brown University, head of the Brown Digi- Geoscience and Remote Sensing 55 (5): 2716–2724.
tal Archaeology Library and an elected fellow of the Institute at Brown Canuto, M. A., F. Estrada-Belli, T. G. Garrison, S. Houston, M. J. Acuña,
for Environment and Society. Among other research projects, he co- M. Kováč, D. Marken, P. Nondédéo, L. Auld-Thomas, and C.
directs the Paisajes Arqueológicos de Chachapoyas project with Carol Castanet. 2018. “Ancient Lowland Maya Complexity as Revealed by
Rojas Vega and is and is co-editor, with Steven Wernke, of GeoPACHA: Airborne Laser Scanning of Northern Guatemala.” Science 361
Geospatial Platform for Andean History, Culture, and Archaeology. (6409): eaau0137.
Chase, A. F., D. Z. Chase, J. J. Awe, J. F. Weishampel, G. Iannone, H.
K. C. Cushman received her Ph.D. in Ecology and Evolutionary Biology Moyes, J. Yaeger, and M. K. Brown. 2014. “The Use of LiDAR in
at Brown University and the Institute at Brown for Environment and Understanding the Ancient Maya Landscape: Caracol and Western
Society and is postdoctoral fellow at the Smithsonian Tropical Research Belize.” Advances in Archaeological Practice 2 (3): 208–221.
Institute Chase, A. F., D. Z. Chase, C. T. Fisher, S. J. Leisz, and J. F. Weishampel.
Luis Jaime Castillo (Ph.D. 2012, University of California, Los Angeles) is 2012. “Geospatial Revolution and Remote Sensing LiDAR in
Professor of Archaeology at the Pontificia Universidad Catolica del Peru, Mesoamerican Archaeology.” Proceedings of the National Academy
directs the San Jose de Moro Archaeological Program and the Digital of Sciences 109 (32): 12916–12921.
Archive of Cultural Heritage. He is Foreign Associate member of the Chase, A. F., D. Z. Chase, J. F. Weishampel, J. B. Drake, R. L. Shrestha, K.
National Academy of Sciences and a National Geographic Society C. Slatton, J. J. Awe, and W. E. Carter. 2011. “Airborne LiDAR,
Explorer and has served as Vice Minister of Cultural Patrimony and Archaeology, and the Ancient Maya Landscape at Caracol, Belize.”
Minister of Culture in Peru. Journal of Archaeological Science 38 (2): 387–398.
Church, W. B. 1996. “Prehistoric Cultural Development and
Carol Rojas Vega is an archaeological researcher based in Lima Peru and
Interregional Interaction in the Tropical Montane Forests of Peru
co-director of the Paisajes Arqueológicos de Chachapoyas Project, as
Volume.” Ph.D. Diss., Yale University.
well as other archaeological projects in Peru.
Church, W. B., and A. Von Hagen. 2008. “Chachapoyas: Cultural
Carson B. Roberts is a Senior Applications Engineer at Headwall Photo- Development at an Andean Cloud Forest Crossroads.” In The
nics, Inc. His work concentrates on remote sensing and hyperspectral Handbook of South American Archaeology, edited by H. Silverman,
and LiDAR data analysis. and W. H. Isbell, 903–926. New York: Springer.
Charles Kepler is s a Manufacturing Engineer at Headwall Photonics, Church, W. B., and A. Guengerich. 2017. “La (Re)Construcción de
Inc. He is chief pilot for Headwall’s UAV operations, and supervises Chachapoyas a Través de La Historia e Histografía.” Boletín de
the construction and testing of UAV systems. Arqueología PUCP 23: 5–38.
Clasby, R. 2014. “Early Ceremonial Architecture in the Ceja de Selva
James Kellner (Ph.D. 2008, University of Georgia) is the Peggy and (800-100 BC): A Case Study from Huayurco, Jaén Region, Peru.”
Henry D. Sharpe Jr. Assistant Professor of Environmental Studies, and Ph.D. Diss., Yale University.
an Assistant Professor of Ecology and Evolutionary Biology at Brown Clement, C. R., W. M. Denevan, M. J. Heckenberger, A. B. Junqueira, E.
University. He is a member of the Science Team of the NASA Global G. Neves, W. G. Teixera, and W. I. Woods. 2015. “The Domestication
Ecosystem Dynamics Investigation, which placed a waveform lidar sen- of Amazonia before European Conquest.” Proceedings of the Royal
sor on the International Space Station. Society B: Biological Sciences 282 (1812): 20150813.
CloudCompare. 2019. GPL Software. Ver 2.10.2 http://www.
cloudcompare.org/.
ORCID Crandall, J. M. 2017. “El Desarrollo Espacial de Las Comunidades
Parker VanValkenburgh http://orcid.org/0000-0002-4289-6170 Chachapoyas Bajo La Dominación Colonial Inka y Española.”
K. C. Cushman http://orcid.org/0000-0002-3464-1151 Boletín de Arqueología PUCP 23: 283–312.
Crown, P. L. 2007. “Life Histories of Pots and Potters: Situating the
Individual in Archaeology.” American Antiquity 72 (4): 677–690.
References De Marrais, E. 2002. “The Architecture and Organization of Xauxa
Settlements.” In Empire and Domestic Economy, edited by C. A.
Arkush, E. 2017. “The End of Ayawiri: Abandonment at an Andean Hastorf and T. N. D’Altroy, 115–153. New York: Springer.
Hillfort Town of the Late Intermediate Period.” Journal of Field Devereux, B. J., G. S. Amable, P. Crow, and A. D. Clif. 2005. “The
Archaeology 42 (3): 241–257. Potential of Airborne Lidar for Detection of Archaeological
Arkush, E. 2011. Hillforts of the Ancient Andes: Colla Warfare, Society, Features under Woodland Canopies.” Antiquity 79 (305): 648–660.
and Landscape. Gainesville, FL: University Press of Florida. Disney, M. I., M. B. Vicari, A. Burt, K. Calders, S. L. Lewis, P. Raumonen,
Arroyo-Kalin, M. 2010. “The Amazonian Formative: Crop and P. Wilkes. 2018. “Weighing Trees with Lasers: Advances,
Domestication and Anthropogenic Soils.” Diversity 2 (4): 473–504. Challenges and Opportunities.” Interface Focus 8 (2): 20170048.
Bandelier, A. F. 1907. The Indians and Aboriginal Ruins Near Espinoza Soriano, W. 1967. “Los Señoríos Étnicos de Chachapoyas y La
Chachapoyas in Northern Peru. New York: Historical Records and Alianza Hispano-Chacha.” Revista Histórica 30: 224–333.
Studies. Evans, D. H., R. J. Fletcher, C. Pottier, J.-B. Chevance, D. Soutif, B. S.
Beardsley, R. K., W. J. Hall, and R. E. Ward. 1959. Village Japan. Tan, S. Im, D. Ea, T. Tin, S. Kim, et al. 2013. “Uncovering
Chicago: University of Chicago Press. Archaeological Landscapes at Angkor Using Lidar.” Proceedings of
Bewley, R. H., S. P. Crutchley, and C. A. Shell. 2005. “New Light on an the National Academy of Sciences 110 (31): 12595–12600.
Ancient Landscape: Lidar Survey in the Stonehenge World Heritage Fisher, C. T., and S. J. Leisz. 2013. ““New Perspectives on Purépecha
Site.” Antiquity 79 (305): 636–647. Urbanism Through the Use of LiDAR at the Site of Angamuco,
JOURNAL OF FIELD ARCHAEOLOGY S87

Mexico.” In Mapping Archaeological Landscapes from Space, edited Masini, N., R. Coluzzi, and R. Lasaponara. 2011. ““On the Airborne Lidar
by D. C. Comer and M. J. Harrower, 199–210. New York: Springer. Contribution in Archaeology: From Site Identification to Landscape
Fisher, C. T., J. C. Fernández-Diaz, A. S. Cohen, O. N. Cruz, A. M. Investigation.” In Laser Scanning, Theory and Applications, edited by
Gonzáles, S. J. Leisz, F. Pezzuitti, R. Shrestha, and W. Carter. 2016. C. C. Wang, 263–290. Rijeka, Croatia: IntechOpen.
“Identifying Ancient Settlement Patterns through LiDAR in the Meggers, B. J. 1971. Amazonia: Man and Culture in a Counterfeit
Mosquitia Region of Honduras.” PloS One 11 (8): e0159890. Paradise. Chicago: Aldine-Atherton, Inc.
Garrison, T. G., S. Houston, and O. Alcover Firpi. 2019. “Recentering the McCray, B. 2017. “Límites e Interfaces de Chachapoyas En La Región
Rural: Lidar and Articulated Landscapes among the Maya.” Journal of Rodríguez De Mendoza.” Boletín de Arqueología PUCP 23: 187–206.
Anthropological Archaeology 53: 133–146. Minar, C. J., and P. L. Crown. 2001. “Learning and Craft Production: An
Gil, N. 1938. “Dos Pueblos Prehistóricos Kuelapenses: Kacta y Chipuric.” Introduction.” Journal of Anthropological Research 57 (4): 369–380.
Boletin de La Sociedad Geográfica (Lima) LV (2–3): 132–139. Narváez Vargas, A. 1987. “Kuélap: Una Ciudad Fortificada en los Andes
Golden, C., T. Murtha, B. Cook, D. S. Shaffer, W. Schroder, E. J. Hermitt, Nor-orientales de Amazonas, Peru.” In Arquitectura y Arqueologıa:
O. Alcover Firpi, and A. K. Scherer. 2016. “Reanalyzing Pasado y Futuro de la Construcción en el Peru, edited by V. Rangel,
Environmental Lidar Data for Archaeology: Mesoamerican 115–142. Chiclayo: Universidad de Chiclayo.
Applications and Implications.” Journal of Archaeological Science: Narváez Vargas, A. 2013. “Kuelap: Centro Del Poder Político y Religioso
Reports 9: 293–308. de Los Chachapoyas.” In Los Chachapoyas, edited by F. Kaufmann
Goodland, R. J. A., and H. S. Irwin. 1975. Amazon Jungle: Green Hell to Doig, 87–159. Lima: Banco de Credito.
Red Desert? An Ecological Discussion of the Environmental Impact of Naroll, R. 1962. “Floor Area and Settlement Population.” American
the Highway Construction Program in the Amazon Basin. Elsevier Antiquity 27 (2): 587–588.
Scientific Publishing Co: Amsterdam. Neves, E. G., J. B. Petersen, R. N. Bartone, and C. A. Da Silva. 2003.
Guengerich, A. 2014. “The Architect’s Signature: The Social Production “Historical and Socio-Cultural Origins of Amazonian Dark Earth.”
of a Residential Landscape at Monte Viudo, Chachapoyas, Peru.” In Amazonian Dark Earths: Origin, Properties, Management, edited
Journal of Anthropological Archaeology 34: 1–16. by J. Lehmann, 29–50. Boston: Kluwer Academic.
Guengerich, A. 2014. “Monte Viudo: Residential Architecture and the Opitz, R. S., K. Ryzewski, J. F. Cherry, and B. Moloney. 2015. “Using
Everyday Production of Space in a Chachapoya Community.” Airborne LiDAR Survey to Explore Historic-Era Archaeological
Ph.D. Diss., The University of Chicago. Landscapes of Montserrat in the Eastern Caribbean.” Journal of
Guengerich, A. 2015. “Settlement Organization and Architecture in Late Field Archaeology 40 (5): 523–541.
Intermediate Period Chachapoyas, Northeastern Peru.” Latin Rasband, W. S. 1997. ImageJ. Version 1.8. Bethesda, Maryland, USA.
American Antiquity 26 (3): 362–381. https://imagej.nih.gov/ij/.
Halperin, C., and L. Schwartz, eds. 2016. Vernacular Architecture in the Raymond, J. S. 1992. “Highland Colonization of the Peruvian Montaña
Pre-Columbian Americas. New York: Taylor & Francis. in Relation to the Political Economy of the Huari Empire.” Journal of
Hastings, C. M., and M. E. Moseley. 1975. “The Adobes of Huaca Del Sol the Steward Anthropological Society 20 (1–2): 17–36.
and Huaca de La Luna.” American Antiquity 40 (2): 196–203. Reichlen, H., and P. Reichlen. 1950. “Recherches Archéologiques Dans
Heckenberger, M. 2005. The Ecology of Power: Culture, Place, and Les Andes Du Haut Utcubamba: Deuxième Rapport de La Mission
Personhood in the Southern Amazon, A.D. 1000-2000. New York: Ethnologique Française Au Perou Septentrional.” Journal de La
Routledge. Société Des Américanistes 39: 219–246.
Heckenberger, M. J., J. C. Russell, C. Fausto, J. R. Toney, M. J. Schmidt, Risbøl, O. and L. Gustavsen. 2018. “LiDAR from Drones Employed
E. Pereira, B. Franchetto, and A. Kuikuro. 2008. “Pre-Columbian for Mapping Archaeology: Potential, Benefits and Challenges.”
Urbanism, Anthropogenic Landscapes, and the Future of the Archaeological Prospection 25 (4): 329–338.
Amazon.” Science 321 (5893): 1214–1217. Roddick, A. P., and A. B. Stahl, eds. 2016. Knowledge in Motion:
Hesse, R. 2010. “LiDAR-Derived Local Relief Models–a New Tool for Constellations of Learning Across Time and Place. Tucson:
Archaeological Prospection.” Archaeological Prospection 17 (2): 67–72. University of Arizona Press.
Hutson, S. R. 2012. “‘Unavoidable Imperfections’: Historical Contexts Rojas Vega, C., and P. VanValkenburgh. 2017. Informe de la Primera
for Representing Ruined Maya Buildings.” In Past Presented: Temporada del Proyecto Paisajes Arqueológicos de Choctamal,
Archaeological Illustration and the Ancient Americas, edited by J. Chachapoyas. Report Submitted to the Puruvian National Ministry
Pillsbury, 282–316. Washington, D.C.: Dumbarton Oaks Research of Culture.
Library and Collection. Rousell, J.-R., and D. Auty. 2019. lidR: Airborne LiDAR Data Manipulation
Kellner, J. R., J. Armston, M. Birrer, K. C. Cushman, L. Duncanson, C. and Visualization for Forestry Applications. R package version 2.1.2.
Eck, C. Falleger, B. Imbach, K. Král, and M. Krŭček. 2019. “New Ruiz Estrada, A. 1972. “La Alfareria de Cuelap: Tradición y Cambio.”
Opportunities for Forest Remote Sensing through Ultra-High- B.A. Thesis, Universidad Nacional Mayor de San Marcos, Lima.
Density Drone Lidar.” Surveys in Geophysics 40 (4): 959–977. Ruiz Estrada, A. 1979. “Los Monumentos Arqueológicos de
Khan, S., L. Aragão, and J. Iriarte. 2017. “A UAV–Lidar System to Map Leimebamba.” Boletín de Lima 42: 69–82.
Amazonian Rainforest and Its Ancient Landscape Transformations.” Ruiz Estrada, A. 2009. La Alfareria de Cuelap: Tradición y Cambio. Lima:
International Journal of Remote Sensing 38 (8–10): 2313–2330. Avqi Ediciones.
Kramer, C. 1980. ““Estimating Prehistoric Populations: An Ruiz Estrada, A. 2004. “Purumllacta: Un Centro Administrativo Incaico
Ethnoarchaeological Approach.” L’Archèologie de L’Iraq: En Chachapoyas.” Investigaciones Sociales 8 (13): 73–84.
Perspectives et Limites de l’Interprétation Antropologique Des Ruiz Estrada, A. 2017. “Deslindes Étnicos en la Historia de Amazonas,
Documents.” In Colloques Internationaux du CNRS 580, 315–334. Peru.” Boletín de Arqueología PUCP 23: 41–56.
Paris: Editions de Centre National de la Recherche Scientifique. Savoy, G. 1970. Antisuyo: the Search for the Lost Cities of the Amazon.
Le Moine, G., and J. S. Raymond. 1987. “Leishmaniasis and Inca New York: Simon and Schuster.
Settlement in the Peruvian Jungle.” Journal of Historical Geography Schjellerup, I. R. 1997. Incas and Spaniards in the Conquest of the
13 (2): 113–129. Chachapoyas: Archaeological and Ethnohistorical Research in the
Langlois, L. 1940. “Utcubamba: Investigaciones Arqueológicas En Este North-Eastern Andes of Peru. GOTARC, series B, Gothenburg
Valle Del Departmento de Amazonas (Perú).” Revista Del Museo Archaeological Theses, 7, Göthenburg, Sweden: Göteborg University.
Nacional 9 (1): 33–72. Schjellerup, I. R. 2003. Forgotten Valleys: Past and Present in the
LeBlanc, S. 1970. “An Addition to Naroll’s Suggested Floor Area and Utilization of Resources in the Ceja de Selva, Peru. Aarhus,
Settlement Population Relationship.” American Antiquity 36 (2): Denmark: Aarhus Universitetsforlag.
210–211. Shimada, I., and R. Cavallaro. 1985. “Monumental Adobe Architecture
Lerche, P. 1986. Häuptlingstum Jalca: Bevölkerung Und Ressourcen Bei of the Late Prehispanic Northern North Coast of Peru.” Journal de
Den Vorspanischen Chachapoya Peru. Berlin: D. Reimer. La Société Des Américanistes 71 (1): 41–78.
Levis, C., F. R. C. Costa, F. Bongers, M. Peña-Claros, C. R. Clement, A. R. Shell, C. A., and C. F. Roughley. 2004. “Exploring the Loughcrew
Junqueira, E. G. Neves, E. K. Tamanaha, F. O. G. Figuereido, and R. P. Landscape: A New Approachwith Airborne LIDAR.” Archaeology
Salomão. 2017. “Persistent Effects of Pre-Columbian Plant Ireland 18 (2): 20–3.
Domestication on Amazonian Forest Composition.” Science 355 Sivertsen, D. 1963. When Caste Barriers Fall: A Study of Social and Economic
(6328): 925–931. Change in a South Indian Village. New York: Humanities Press.
S88 P. VANVALKENBURGH ET AL.

Stahl, P. W. 2002. “Paradigms in Paradise: Revising Standard Women and Prehistory, edited by M. Conkey and J. Gero, 93–131.
Amazonian Prehistory.” The Review of Archaeology 23 (2): 39–51. Cambridge, MA: Blackwell.
Team, R. Core. 2013. R: A Language and Environment for Statistical Valdez, F. 2007. “Un Formativo Insospechado En La Ceja de Selva: El
Computing. Complejo Cultural Mayo Chinchipe.” In II Congreso Ecuatoriano
Thompson, D. 1976. Prehistory of the Uchucmarca Valley in the North de Antropología y Arqueología, edited by F. S. García, 549–576.
Highlands of Peru. Proceedings of the 41st International Congress of Quito: Abya Yala.
Americanists. 99–106. Valdez, F. 2008. “Inter-Zonal Relationships in Ecuador.” In The
Toyne, J. M., and A. Narváez. 2011. Patrones de Trauma Craneales Handbook of South American Archaeology, edited by H. Silverman,
Perimortem de Una Matanza de La Ocupacion Tardia En La and W. H. Isbell, 865–888. New York: Springer.
Fortaleza de Kuelap, Chachapoyas. Paleopathology Association in Wenger, E. 1999. Communities of Practice: Learning, Meaning, and
South America, Lima, Peru. Identity. Cambridge, UK: Cambridge University Press.
Toyne, J. M., and A. Narváez. 2014. “The Fall of Kuélap: Wilkinson, D. A. 2013. “Politics, Infrastructure and Non-Human
Bioarchaeological Analysis of Death and Destruction on the Eastern Subjects: The Inka Occupation of the Amaybamba Cloud Forests.”
Slopes of the Andes.” In Embattled Bodies, Embattled Places, edited Ph.D. Diss. Columbia University.
by A. K. Scherer and J. Verano, 341–364. Washington D.C.: Zhang, K., S.-C. Chen, D. Whitman, M.-L. Shyu, J. Yan, and C. Zhang.
Dumbarton Oaks Research Library and Collection. 2003. “A Progressive Morphological Filter for Removing Nonground
Tringham, R. 1991. “Households with Faces: The Challenge of Gender in Measurements from Airborne LIDAR Data.” IEEE Transactions on
Prehistoric Architectural Remains.” In Engendering Archaeology: Geoscience and Remote Sensing 41 (4): 872–882.

You might also like