Content uploaded by Yaroslav Vasyunin
Author content
All content in this area was uploaded by Yaroslav Vasyunin on Sep 03, 2023
Content may be subject to copyright.
1
IDENTIFICATION OF ARCHEOLOGICAL SITES
IN THE PERUVIAN AMAZON
USING SATELLITE REMOTE SENSING
Ceslav Cieslar
1
and Yaroslav Vasyunin
2
The Amazon basin is still poorly studied, and new archaeological
discoveries will continue to arise. The authors selected the Andean zone
of the Manu National Park in Peru as a prominent research area. A
substantial archive of geospatial data is collected by integrating it into a
multi-user GIS. It includes a digital terrain model, high-resolution
aerospace imagery of optical and microwave ranges, and derivative
datasets. This paper introduces an original method to outline areas fit for
Andean people based on the geospatial analysis in a big data platform.
Thus, a settlement suitability map covering over 3000 km² is created and
assessed. Furthermore, for areas with high suitability scores, visual
interpretation of imagery reveals patterns and features that could
indicate archaeological sites. In total, the GIS analysis reveals six sites
that could contain human-modified terrain features. The authors also
attempt to relate these sites to recorded testimonies from witnesses who
encountered large ruins in the mountain rain forest.
Keywords: Archaeology; Remote sensing; GIS; Predictive modeling;
Suitability analysis; Inca; Google Earth Engine
1. INTRODUCTION
The archaeology of the Amazon is still poorly known. In particular, there is a scarce
theoretical development in the archaeological studies in the Peruvian Amazon. The little
relevance granted to this geographical area would be, firstly, due to the low visibility of the
archaeological sites, and, secondly, due to complicated conditions of life, transfer, stay, and
work [1].
The latest studies show the possibility that the average Amazonian indigenous person at
the end of the fifteenth century did not live in an isolated, autonomous village, but instead
was part of a regional polity or, at least, articulated with one in broad social networks
extended across the region [2]. In 1998 and 1999, two expeditions were that proved the
possibility of the river-based goods exchange between Incas and the cultures living in the
lower basin of the Amazon. Ancient people might have sailed boats made of totora, a special
kind of reed from Lake Titicaca [3]. It is also possible that ancient people inhabited the
Amazonian upland far from the main riverine routes. This hypothesis is based on the
1
Independent Scholar, Switzerland. Email: jellyczez@hotmail.com
2
Independent Scholar, Italy. Email: yaroslav.vasyunin@gmail.com
(Preprint)
2
discovery of hundreds of geoglyphs with an associated system of roads in the state of Acre,
Brazil [4].
A heavily forested, mountainous, and almost uninhabited territory in the Manu National
Park in Peru, between Cusco and Madre de Dios regions (Figure 1), could have potentially
acted as a barrier and corridor for human societies through time, considering its relevant
vicinity to the core of the Inca state. Moreover, two historic Jesuits’ manuscripts (Relazione
d’un Miracolo and Exsul Immeritus) [5], [6] and several recent testimonies [7]–[10] point to the
high-altitude selva, rain forest of southeast Peru, as to the place where a settlement (or
settlements) of some developed culture may exist.
To study inaccessible area covering approximately 1500 km2, it is essential to apply remote
sensing technology, widely used across many disciplines, including archaeology.
Remote sensing generally refers to
instrument-based techniques measuring
properties of objects at a distance, rather
than in situ, by utilizing electromagnetic
radiation, force fields, or acoustic energy
[11]. In this paper, we adhere to the term
Earth remote sensing (ERS) and limit its
meaning to observing the Earth's surface,
utilizing only electromagnetic signals
recorded on airborne and space-borne
platforms. A platform at the lowest flight
altitude, such as an aircraft or UAV,
achieves the highest spatial resolution but
provides only local information, limited
by cruising time and radius [12]. Modern
ERS satellites enable global
measurements with relatively high
resolution and are suitable to overcome
funding issues while mapping large areas
1
.
The history of successful applications of ERS for archaeology goes back more than a
century, and it proved to be an effective method for archaeological investigations, especially
in hardly accessible regions like tropical forests [13]. The principle behind archaeological
ERS, as explained in [14], is the following: ancient human transformations of the landscape
left subtle features called marks that are only visible when viewed from above. For example,
caused by subsurface archaeological remains, these marks can be observed as differences in
plant growth or variations [15] of micro-topographic relief visible by shadowing [16]. They
can be revealed as anomalies in a signal of different spectral ranges, recorded by an ERS
sensor, which finally allows detection of linked human-made features, buried and exposed.
ERS, applied in archaeology, can be divided into four distinct groups: photography,
multi/hyperspectral imaging, synthetic aperture radar (SAR), and light detection and ranging
(lidar) [13].
1
https://droneapps.co/price-wars-the-cost-of-drones-planes-and-satellites/
Figure 1. Location of the research area (red
mark). The dot pattern represents the Inca
Empire at its greatest extent ca. 1525.
3
Historically, photography was the first ERS method, implemented by passive film-based
cameras sensitive mainly to the visible part of the electromagnetic spectrum [17]. Historic
film photographs taken by airborne platforms and declassified spy satellites decades ago are
still used by archaeologists who study changes in land cover that have occurred over the past
century [18].
Multi and hyperspectral imaging involves the passive acquisition of electromagnetic
energy in different ranges simultaneously, from visible to far-infrared, which has a higher
potential to detect archaeological sites and monuments than photography [19]. It is especially
true for hyperspectral sensors with more than hundreds of spectral bands, which were
successfully used to indicate the presence of underground buried structures, mainly by
detection of crop marks in farmland [20]. Hyperspectral surveys are usually conducted by
airborne platforms because hyperspectral space-borne instruments are limited due to
technical constraints. Moreover, practical issues of hyperspectral sensor costs, data volumes,
and data processing mechanisms need to be addressed for operational use and favor
multispectral systems [21]. Multispectral sensors, especially space-borne, having fewer
spectral bands, became the most popular imaging technique in archaeology. They are applied
to detect and identify archaeological features, paleo-landscape analysis and reconstruction,
looting monitoring and assessment, urbanization mapping, risk monitoring and assessment,
and cultural heritage management and conservation [13].
Another emerging imaging technique is realized with SAR — an active type of ERS. In its
simplest form, SAR is a radar whose output product is a two-dimensional mapping of "radar
brightness" that resembles a black-and-white aerial photograph [22]. SAR can sense a target
under all-weather conditions at any time of day or night and even penetrate soil and
vegetation, depending on the wave frequency and target properties [23]. A significant amount
of studies demonstrated that SAR is superior at detecting buried features, soil marks, and
relief for archaeological purposes [24], [25], with notable implications for archaeology in
tropical forests, where SAR data was used to detect ancient settlements in densely vegetated
areas [26]–[28]. However, applications of SAR in mountainous areas are limited due to
inherent radiometric and geometric distortions in reconstructed images [29], [30].
Imaging lidar — the optical analog of SAR — transmits visible or near-infrared pulses to
the ground and takes backscattered echoes of radiation to reconstruct 3d topography [31].
The ability of lidar to penetrate overlying vegetation and forest canopies generated a
fundamental shift in Mesoamerican archaeology and transformed research in forested areas
worldwide [32]–[34]. However, lidar imaging is mainly airborne due to the practical
limitations of satellite platforms, which imposes restrictions on its application in inaccessible
distant regions or low-budget projects.
From the user point of view, access to ERS data and delivery systems, in particular to
space-borne data sets, became transparent and straightforward [35]. Nowadays,
governmental organizations, such as NASA
1
, Natural Resources Canada
2
, and the European
Space Agency
3
freely and openly distribute low-to-moderate-resolution ERS imagery covering
1
https://earthdata.nasa.gov
2
https://www.eodms-sgdot.nrcan-rncan.gc.ca/index_en.jsp
3
https://earth.esa.int/eogateway
4
the whole globe. To acquire high-resolution data, users refer to the commercial ERS market
[36], rich in data suppliers, such as Maxar
1
, Planet
2
, and Airbus Defense & Space
3
.
The combination of ERS data from multiple sensors and platforms, together with other
geospatial data and ground observations, should provide more precise insight and contextual
information about Earth's surface anomalies [37], [38]. It can be done in a Geographic
Information System (GIS), which is a "computer-based system to aid in the collection,
maintenance, storage, analysis, output, and distribution of spatial data and information." [39]
The early pioneers of archaeological GIS quickly recognized that it was more than a data
management tool. The analytic power of GIS to advance understandings of past landscapes
was demonstrated in several works, collected in [40]. Now GIS, ERS, and other digital
technologies became standard and well-routinized tools in archaeology [41], [42].
The use of GIS by archaeologists is typically grouped into three broad categories:
inventory, spatial analysis, and mapmaking [43]. Archaeological features can be extracted
from imagery using GIS-aided manual interpretation or automatic detection following image
enhancement. The spatial analysis used for prospecting of new archaeological sites has a long
and successful track record [43]–[46]. The advent of open-source software, including GIS,
triggered a vast research volume in different fields and disciplines [47]. Combined with ERS
data, which became an accessible and profitable commodity [48], it dramatically facilitates
archaeological explorations.
Petabytes of openly available big data from space-borne Earth observation missions led to
data processing and analysis issues. A robust and expandable cloud platform, Earth Engine
was designed by Google to tackle these issues — it empowers a broad audience that lacks the
technical capacity needed to utilize traditional supercomputers or large-scale cloud
computing resources [49]. Thus, the platform was successfully applied to the detection of
buried Neolithic tells and the investigation of the urban sprawl impact in the vicinity of
archaeological sites and monuments [50]. However, literature research in Scopus
4
performed
in June 2020 retrieves just three papers related to "Earth Engine" and "archaeology" keywords,
which makes it a prospective study direction.
In this work, we perform low-cost research based on non-invasive technology of Earth
remote sensing (ERS), an open-source geographic information system (GIS), and a big data
platform Google Earth Engine (GEE), to locate potential human-modified sites in a forested
mountainous region of Peru. We develop a cloud GIS infrastructure for multi-user work on
the project and integrate ERS imagery with various geospatial data sets, such as digital
elevation models (DEM) and historical maps. A novel predictive technique that allows
detecting areas suitable for ancient settlements is put into action based on GEE scripts, open
to the public. The most prominent areas are manually studied using the available imagery
and maps, which reveal marks that could indicate archaeological sites.
1
https://www.maxar.com
2
https://www.planet.com
3
https://www.airbus.com/space.html
4
Abstract and citation database of peer-reviewed journals: https://www.scopus.com/
5
2. MATERIALS AND METHODS
2.1. GIS Infrastructure
As the research team was scattered worldwide, this study's first step was creating a GIS
that enables simultaneous work on geospatial layers. Such implementations have already
been realized in [51], [52]. Our spatial infrastructure's core became the Relational Database
Service (RDS) platform from Amazon
1
, allowing cost-effective and straightforward
configuration, use, and scaling of relational databases in the cloud. The platform is equipped
with an open-source RDBMS PostgreSQL
2
with the PostGIS
3
extension to work with spatial
data. The overall architecture of the implemented GIS is provided in Figure 2.
The front end for interacting with geospatial information in the database is the desktop
open-source, cross-platform software QGIS
4
. To ensure security, connecting to the database
is limited to specific IP addresses and configured user roles, listed by ascending privileges:
Reader, Editor, and Administrator. All geospatial data used in the study are divided into
vector and raster categories. Vector layers are updated frequently and stored centrally in RDS
without duplication on local computers. However, a set of raster data is updated and
processed rarely, mainly at the beginning of the project. As a compromise, to reduce the
number of information transmitted via the Internet, it was decided to copy the same set of
raster data on each local machine. Thus, when a user launches QGIS and enters a password,
they see a map consisting of both local raster layers and remote vector layers.
1
https://aws.amazon.com/rds/
2
https://www.postgresql.org
3
https://postgis.net
4
https://qgis.org/
Figure 2. The architecture of the implemented multi-user GIS database for desktop
systems, which is based on open-source technologies.
6
It is worth mentioning that GIS is not an end in itself. It serves for the optimal
representation of the finished geospatial information embedded in it by developers and
researchers, who analyze and interpret this information based on their experience, intuition,
and assumptions.
2.2. Data Collection and Overview Study
The research area extended over ca. 8,000 km², lies where the southeastern Peruvian
Andes transitions to the rainforests of the Amazon basin, on the border between Cusco and
Madre de Dios regions. Western mountainous side, sierra, varies in altitudes from 2,000 m to
4,000 m, covered chiefly by low grass and shrubs. Eastern mountain selva goes down to 1000
m and has a very high density of woody vegetation — more than 75%
1
.
As in any project on the analysis of poorly studied, almost uncharted territories, work
begins with collecting all available territorial data, which are then ingested in GIS.
Thereupon, it becomes clear what methods of geospatial analysis apply to the data collected.
All the data sets collected by authors fit into two groups by their map detail level. The overview
level has moderate spatial resolution and covers the whole research region, while the detailed
level with high spatial resolution clarifies only specific spots.
It is worth noting the difficulty of obtaining cloud-free satellite images in the visible and
near-infrared bands for the territory of interest. In this regard, the authors developed an
algorithm for the Earth Engine platform that allows generating high-quality image mosaics
(Landsat-8 and Sentinel-2 image collections). The publicly available algorithm allows every
researcher to quickly download not only cloud-free satellite mosaics but a set of primary
geospatial data, such as land cover types and DEM, to any world region [54].
The subsequent processing of the acquired DEM with the QGIS geo-processing toolkit
allowed us to create derivative data sets: hydrographic network and morphometric terrain
parameters (slope and aspect).
2
A list of the collected and created (derived) datasets is given
in Table 1.
1
Percentage of horizontal ground in each 30-m pixel covered by woody vegetation greater than 5 meters in
height [53]
2
https://docs.qgis.org/3.16/en/docs/training_manual/processing/index.html
7
Table 1. Primary datasets used in the research.
RASTER DATASETS: REMOTE SENSING IMAGERY
Platform name
Platform
type1
Spectral
range2
GSD, m2
Cloud-free
coverage, km
Temporal range
Reference
Note
Landsat-8
SB
V-NIR-SWIR
15,0
8,000
2013-2019
landsat.gsfc.nasa.gov
cloudless mosaic created in GEE
WorldView-2
SB
V
1,0
7,000
2010-2014
www.bing.com/maps
high-resolution imagery base map from Microsoft
UAVSAR
AB
MW
10,0-20,0
7,000
2014
uavsar.jpl.nasa.gov
polarimetric L-band images in three polarization RGB
color overlay: HH, HV, VV
PlanetScope
SB
V-NIR
3,0
5,800
2018-2019
www.planet.com
RapidEye
SB
V-NIR
5,0
4,700
2012
www.planet.com
GeoEye-1
SB
V
0,5
95
2016-2017
www.maxar.com
RASTER DATASETS: OTHER TYPES
Product name
Type
GSD, m
Coverage
Reference
Note
Copernicus Global Land
Service: Land Cover 100m:
collection 3: epoch 2019: Globe
land cover
100,0
8,000
[55]
NASA SRTM3 SRTMGL1
terrain
30,0
8,000
[56]
Landsat Vegetation Continuous
Fields (VCF)
land cover
30,0
8,000
[53]
Slope
terrain
30,0
8,000
derived from DEM in QGIS
Aspect
terrain
30,0
8,000
derived from DEM in QGIS
Annual insolation
terrain
30,0
8,000
the direct incoming solar radiation, derived from DEM
with the Area Solar Radiation tool in ArcGIS Pro
VECTOR DATASETS
Product name
Vector data type
Note
Contour lines
linear
50 m vertical distance, derived from DEM in QGIS
Stream network
linear
derived from DEM with the Hydrology toolset of ArcGIS Pro - a set of vector features representing the linear hydrological network
Inca roads
linear
result of a joint interpretation of high-resolution satellite imagery and terrain datasets; contains mostly hypothetical trails
1
SB – space-borne; AB – air-borne
2
V – visible; (N/SW)IR – near/short-wave infrared; MW – microwave
8
2.3. Predicting Modelling
Mountain environments tend to constrain movements [57]. As a critical physiographic
feature of South America, the Andes acted as a barrier and corridor for human societies and
their displacements, enabling the interactions between costa (the Pacific coast), sierra, and
selva [58]–[60]. In mountainous areas, some authors used a combined GIS and remote sensing
approach to predict ancient human frequentations [61], [62]. We decided to extrapolate their
approach for our research area, identified in the Introduction.
We performed a preliminary study of well-known Inca ruins, such as Machu Picchu and
Vilcabamba (see the complete list in the Appendix A). Their coordinates were collected by
overpass turbo, a web-based data filtering tool for OpenStreetMap
1
with the following query:
[out:json][timeout:25];
(node["historic:civilization"~"inca",i];
way["historic:civilization"~"inca",i];
relation["historic:civilization"~"inca",i];);
out body; >; out skel qt;
As seen from Figure 3, they show relation to terrain morphology, namely slope that fits
the range approx. 0-20˚. To assure the derived conclusions, we assumed that the settling
patterns of small modern local communities did not change significantly. We retrieved 15738
samples of small Andean settlements from OpenStreetMap
2
. For each point, we created a 50-
meter-radius buffer and calculated terrain morphology values [54], which result is
represented in Figure 4.
1
https://overpass-turbo.eu/
2
https://www.openstreetmap.org/
Figure 3. Median slope angle for Inca archaeological sites, 59 samples.
9
Compared to other terrain-related factors taken into accounts, such as incoming solar
radiation and aspect, slope gives the most evident relationship with the distribution of
inhabited places. It is also easy to compute slope maps in an open-source QGIS instead of
computationally intensive solar radiation tools, e.g., available in commercial software
ArcGIS
1
.
Based on the obtained relationship between slope and populated places in the Andes, the
authors created a Slope Steepness/Settlement Suitability Map (SSM — Appendix B), which divides
the entire study area into five distinct categories by slope steepness: gentle (0-10˚), slightly
steep (10-20˚), steep (20-30˚), highly steep (30-60˚), and precipitous (60-90˚). The relation
between slope categories and suitability is the following:
• Gentle slope → Highly suitable
• Slightly steep → Suitable
• Steep → Poorly suitable
• Highly steep and Precipitous → Unsuitable
Thus, the efforts to collect and analyze high-resolution ERS data were focused only on the
territories mapped as highly suitable and suitable. It reduced the search area from 8000 km2
down to 2350 sq. km.
2.4. High-resolution Data Interpretation
Existing analytical tools for object detection associated with digital remote sensing data,
such as road or building detection, usually fail when targeting elusive archaeological traces.
[38]. Moreover, the possibility of locating new archaeological sites from aerial or satellite
1
https://desktop.arcgis.com/en/arcmap/latest/tools/spatial-analyst-toolbox/an-overview-of-the-solar-
radiation-tools.htm
Figure 4. Median slope angle for small modern Andean settlements, 15738 samples.
10
imagery is highly dependent on various parameters, such as image spatial resolution, the
extension of buried sites, ground characteristics, illumination conditions, and view geometry
[14].
A thorough visual interpretation of all available spatial data, specifically of high-
resolution, became the basis of our research approach. The criteria to identify potential
unknown archaeological sites were the position in the basin of the Nistron river, on terrain
of highly suitable and suitable classes, a water source presence, and the dominant position over
the surroundings. Undoubtedly, this is a subjective way, but it is the human eye that can
notice subtle marks in images, linking them with the surrounding context (Figure 5).
In GIS, the entire study area was divided into 1x1 km squares. We analyzed an area of
about 3000 km2, with a particular focus on the areas highlighted by the SSM — for which we
have acquired high-resolution commercial data. Different team members connected to the
database with their local GIS were marking "suspicious" image patterns as new vector
features, saving them in the database. Since the target's appearance was unknown, we tried
to discern any visual marks present in the images in the visible, near-infrared, and
microwave ranges. The bird's eye perspective and switching between different datasets
helped to get a holistic understanding of the observed territory. Upon completion of the
analysis of each square, it was marked as "studied." Of course, it was necessary to repeatedly
re-check the studied squares: each image interpreter has a different, unique experience, and
various factors, including simple fatigue, influence their performance. Therefore, mutual
control became is an essential part of the work.
Figure 5. One of the identified features — a subtle dark grid (3 hectares) formed by the play of
shadows in the forest canopy, which could correspond to an underlying human-made structures
that affect the distribution and growth of trees. Image from WorldView-2 satellite, acquired
2010-10-28. © 2017 DigitalGlobe, © 2017 HERE
11
3. RESULTS
Six areas of interest (AOI) in the basin of the Nistron river were identified as potential
candidates. Subtle image features were revealed inside these areas, which may refer to buried
or semi-buried human-made structures. Table 2 describes every AOI.
Table 2. Identified areas of interest.
N.
SSM view
Pseudo-color view
Approx.
area, ha
Description1
1
300
1050–1250 m MSL; linear and point features visible in
MS imagery.
2
80
1460–1800 m MSL; linear features in SAR and MS
optical imagery.
3
300
1360–1660 m MSL; abundance of linear and point
features visible in MS imagery.
4
100
2700–3170 m MSL: linear features visible in SAR and MS
imagery.
5
50
900–950 m MSL; absence of features in imagery, but
the position is advantageous over surroundings.
6
50
1150–1350 m MSL; linear features visible in SAR
imagery and a point feature visible in MS imagery.
1
MSL — mean sea level; SAR — synthetic aperture radar; MS — multi-spectral.
12
4. DISCUSSION
4.1. Limitations of the Proposed Approach
Based on the insights given in [26]–[28], it was initially presumed that microwave L-band
images (UAVSAR
1
) would penetrate through vegetation and expose bare terrain patterns.
However, this approach did not work in the study region due to physical limitations: L-band
waves do not penetrate through the thick foliage. Moreover, the image quality is heavily
affected by geometric distortions caused by rugged mountain terrain. The amount of image
noise made it hard to discern clear signals from the surface level. Thus, the authors switched
to visual and near-infrared image analysis.
The archaeological features, i.e., the ancient human transformations of the landscape, are
visible in imagery as subtle differences in vegetation growth or micro-topographic relief
variations. The dense forests in the research area significantly obstruct their identification,
and natural shapes can be misinterpreted as artificial. The authors expect to have false
positives in the recognized linear features.
With the lack of evidence on the existence of archaeological sites, the authors relied
heavily on geoinformation modeling. As the SRTM digital terrain model became the basis of
this study, it is worth mentioning that 30-m ground resolution does not allow the detailed
study of terrain morphometry. Thus, small-size archaeological sites cannot be discovered
with SRTM. The authors assumed that the potential sites would be large enough to be
expressed in this dataset.
4.2. Testimonies
Until today, the only known large ancient settlement in the basin of the Madre de Dios
river are the ruins of Mameria, discovered in the 1970s by Herbert Cartagena [8], [10, pp. 113–
116]. However, there are a few unverified testimonies describing the existence of a larger
settlement hidden in the selva of the Manu National Park. Such testimonies were reported by
Landa [7, pp. 56, 19, 51–53, 57], and Palkiewicz & Kaplanek [9, p. 374]. They narrow down the
research area to the Nistron (Maestron) river basin in the western Manu National Park.
The AOI 3 turned to be the most promising site due to its position between all the areas
described in the testimonies and the presence of several recognized image features. It has
favorable terrain conditions as it has a dominant position with mountains that limit its
western part. It should provide natural protection, the abundance of waterfalls and water
supply, as mentioned in several testimonies. See the visual presentation of the finding is in
the Figure 6 and different representations of the site in the Appendix C
1
https://uavsar.jpl.nasa.gov/
13
5. CONCLUSIONS
Six areas with potential archaeological sites were identified in the Manu National Park
1
based on the combined analysis of terrain data, remote sensing imagery, and other
cartographic information. On some of these sites, subtle image anomalies were recognized,
potentially indicating underlying human-made structures.
One of those areas, of approximately 2 km2 at 1360–1660 m above mean sea level, is
selected as the most prominent because it corresponds to the descriptions from Jesuits’
manuscripts (Relazione d’un Miracolo and Exsul Immeritus), recorded testimonies of a few
local natives, and has high suitability score according to the designed geospatial modeling.
Moreover, after the visual interpretation of high-resolution multi-spectral satellite images of
1
Manu Province in the Madre de Dios region, Peru
Figure 6. A potential location of a large site in the AOI 3 on the left side of the Nistron
river. Yellow boundaries highlight suitable flat areas: the “upper town” (about 20 ha, 1650
m above MSL) dominates the surroundings, protected from north-east by a 350-meter-tall
abrupt “wall” of a mountain range; the “lower town” (about 100 ha, 1300 m above MSL) has
a good water supply from descending streams — we suppose it could have been used for
agricultural needs. Underlying multi-spectral image from PlanetScope satellite, acquired
Sep 28, 2019, by courtesy of Planet Labs, Inc.
14
the area, several linear marks that could indicate potential archaeological features are
recognized in this area.
In addition, the complete geospatial database of the vast, poorly mapped Peruvian area,
created in this work, gives a unique holistic understanding of that environment. It is enabled
by a multi-user GIS that connects researchers from different places in the world. On the one
hand, the collected data brings value for future studies in the region, such as for the Manu
National Park administration to preserve biodiversity. On the other hand, the developed
open-source GIS workflow can be applied outside of this research area, and it is not limited
to only archaeology.
The authors are aware that the presented evidence is not yet conclusive and that more
additional data is required to be collected. The further investigation, either by fieldwork
activities or by an aerial lidar survey able to penetrate through thick vegetation, would be
crucial in the verification of the results of this work, in the understanding of the settlement
patterns on the eastern Andean side, and in the interaction between selva and sierra people.
6. ACKNOWLEDGMENTS
The authors made this study as their independent work, not affiliated with any
organization, and not receiving any external financial support. They thank the following
organizations for helping with the acquisition and processing of remote sensing data:
• Sputnik Lab, a Peruvian aerospace research laboratory of the National University of
Engineering, Lima, and personally Cristhian Jesus Neyra Kunkel;
• Geoproject Group, a Russian private entity from Moscow, and personally Viktor
Lavrov, Deputy Director;
• The Mission Control Center of the Academy of Engineering, the Peoples' Friendship
University of Russia, Moscow, and personally Vasily Lobanov, Deputy Director.
7. REFERENCES
[1] H. Tantaleán, “Un panorama de la teoría arqueológica en el Perú de comienzos del siglo
XXI,” Discursos sur, no. 5, pp. 201–243, Jul. 2020, doi: 10.15381/dds.v0i5.18150.
[2] M. Heckenberger and E. G. Neves, “Amazonian Archaeology,” Annu. Rev. Anthropol., vol.
38, no. 1, pp. 251–266, Oct. 2009, doi: 10.1146/annurev-anthro-091908-164310.
[3] J. Blashford-Snell and R. Snailham, East to the Amazon: In Search of Great Paititi and the
Trade Routes of the Ancients. London: John Murray, 2002.
[4] M. Pärssinen, W. Balée, A. Ranzi, and A. Barbosa, “The geoglyph sites of Acre, Brazil: 10
000-year-old land-use practices and climate change in Amazonia,” Antiquity, vol. 94, no.
378, pp. 1538–1556, Dec. 2020, doi: 10.15184/aqy.2020.208.
[5] “Relazione d’un miracolo... nel Regno di Paÿtiti a canto del Perù [Report of a miracle... in
the Kingdom of Paÿtiti near Peru],” Archivum Romanum Societatis Iesu, Perù, 19, ff. 38-
41, 1567/1625.
[6] L. Laurencich-Minelli, “Traduzione in italiano del quaderno Exsul Immeritus Blas Valera
Populo Suo [Italian translation of the notebook Exsul Immeritus Blas Valera Populo
Suo],” in Exsul immeritus blas valera populo suo e Historia et rudimenta linguae piruanorum:
Indios, gesuiti e spagnoli in due documenti segreti sul Perù del XVII secolo, CLUEB, 2005, pp.
307–340. doi: 10.1400/104366.
15
[7] C. N. Landa, Paititi en la bruma de la historia [Paititi in the mist of history]. Arequipa: Cuzzi,
1983.
[8] G. Deyermenjian, “Mameria: an Incan Site Complex in the High-Altitude Jungles of
Southeast Peru,” Athena Review, vol. 3, no. 4, pp. 80–88, 2003.
[9] J. Palkiewicz and A. Kaplanek, El Dorado. Polowanie Na Legende [El Dorado. Hunt For The
Legend]. Poznan: Wydawnictwo Zysk i S-ka, 2005.
[10] J. Edigo, Paititi - la ciudad perdida [Paititi - the lost city]. Lulu.com, 2017.
[11] V. V. Salomonson, “Remote sensing, historical perspective,” Encyclopedia of Remote
Sensing. Springer, New York, pp. 684–690, 2014. doi: 10.1007/978-0-387-36699-9_158.
[12] H. P. Röser and M. Von Schönermark, “Comparison of remote sensing experiments from
airborne and space platforms,” Acta Astronautica, vol. 39, no. 9–12, pp. 855–862, 1996,
doi: 10.1016/S0094-5765(97)00070-2.
[13] L. Luo et al., “Airborne and spaceborne remote sensing for archaeological and cultural
heritage applications: A review of the century (1907–2017),” Remote Sensing of
Environment, vol. 232, no. March, p. 111280, 2019, doi: 10.1016/j.rse.2019.111280.
[14] R. Lasaponara and N. Masini, “Remote Sensing in Archaeology: From Visual Data
Interpretation to Digital Data Manipulation,” in Satellite Remote Sensing: A New Tool for
Archaeology, vol. 16, no. 5, R. Lasaponara and N. Masini, Eds. Dordrecht: Springer, 2012,
pp. 3–16. doi: 10.1007/978-90-481-8801-7.
[15] R. Lasaponara and N. Masini, “Detection of archaeological crop marks by using satellite
QuickBird multispectral imagery,” Journal of Archaeological Science, vol. 34, no. 2, pp.
214–221, 2007, doi: 10.1016/j.jas.2006.04.014.
[16] D. Tapete, F. Cigna, and D. N. M. Donoghue, “‘Looting marks’ in space-borne SAR
imagery: Measuring rates of archaeological looting in Apamea (Syria) with TerraSAR-X
Staring Spotlight,” Remote Sensing of Environment, 2016, doi: 10.1016/j.rse.2016.02.055.
[17] J. N. Hampton, “Aerial Reconnaissance for Archaeology: Uses of the Photographic
Evidence,” The Photogrammetric Record, vol. 9, no. 50, pp. 265–272, 1977, doi:
10.1111/j.1477-9730.1977.tb00087.x.
[18] A. Agapiou, D. D. Alexakis, A. Sarris, and D. G. Hadjimitsis, “Colour to Greyscale Pixels:
Re-seeing Greyscale Archived Aerial Photographs and Declassified Satellite CORONA
Images Based on Image Fusion Techniques,” Archaeological Prospection, vol. 23, no. 4,
pp. 231–241, 2016, doi: 10.1002/arp.1536.
[19] A. Beck, “Archaeological applications of multi / hyper-spectral data – challenges and
potential,” in Remote Sensing for Archaeological Heritage Management. Occasional
Publication of the Aerial Archaeology Research Group No. 3, 2011, pp. 87–97.
[20] D. Cerra, A. Agapiou, R. M. Cavalli, and A. Sarris, “An objective assessment of
hyperspectral indicators for the detection of buried archaeological relics,” Remote
Sensing, vol. 10, no. 4, pp. 1–25, 2018, doi: 10.3390/rs10040500.
[21] M. Govender, K. Chetty, and H. Bulcock, “A review of hyperspectral remote sensing and
its application in vegetation and water resource studies,” Water SA, vol. 33, no. 2, pp.
145–151, 2007, doi: 10.4314/wsa.v33i2.49049.
[22] K. Raney, “Radar, Synthetic Aperture,” Encyclopedia of Remote Sensing. Springer, New
York, pp. 536–547, 2014. doi: 10.1007/978-0-387-36699-9.
[23] F. Wiig et al., “Mapping a subsurface water channel with X-band and C-band synthetic
aperture radar at the iron age archaeological site of ‘Uqdat al-Bakrah (Safah), Oman,”
Geosciences (Switzerland), vol. 8, no. 9, 2018, doi: 10.3390/geosciences8090334.
16
[24] D. Tapete and F. Cigna, “COSMO-SkyMed SAR for detection and monitoring of
archaeological and cultural heritage sites,” Remote Sensing, vol. 11, no. 11, 2019, doi:
10.3390/rs11111326.
[25] D. Tapete and F. Cigna, “Trends and perspectives of space-borne SAR remote sensing for
archaeological landscape and cultural heritage applications,” Journal of Archaeological
Science: Reports, vol. 14, pp. 716–726, 2017, doi: 10.1016/j.jasrep.2016.07.017.
[26] F. Yakam-Simen, E. Nezry, and J. Ewing, “Legendary lost city found in the Honduran
tropical forest using ERS-2 and JERS-1 SAR imagery,” in International Geoscience and
Remote Sensing Symposium (IGARSS), 1999, vol. 5, pp. 2578–2580. doi:
10.1109/igarss.1999.771582.
[27] W. Saturno, T. L. Sever, D. E. Irwin, B. F. Howell, and T. G. Garrison, “Putting Us on the
Map: Remote Sensing Investigation of the Ancient Maya Landscape,” in Remote Sensing
in Archaeology, 1st ed., J. Wiseman and F. El-Baz, Eds. New York: Springer, 2007, pp.
137–160. doi: 10.1007/0-387-44455-6_6.
[28] E. Moore, T. Freeman, and S. Hensley, “Spaceborne and Airborne Radar at Angkor:
Introducing New Technology to the Ancient Site,” in Remote Sensing in Archaeology, 1st
ed., J. Wiseman and F. El-Baz, Eds. New York: Springer, 2007, pp. 185–216. doi:
10.1007/0-387-44455-6_8.
[29] T. Bayer, R. Winter, and G. Schreier, “Terrain influences in SAR backscatter and attempts
to their correction,” IEEE Transactions on Geoscience and Remote Sensing, vol. 29, no. 3,
pp. 451–462, 1991, doi: 10.1109/36.79436.
[30] C. Jung and P. Rebillard, “Effects Of Slopes And Relief Two-dimensional Spatial
Frequencies On Spaceborne Sar Imagery,” in International Geoscience and Remote Sensing
Symposium, “Remote Sensing: Moving Toward the 21st Century,” 1988, vol. 3, pp. 1575–
1578. doi: 10.1109/IGARSS.1988.569526.
[31] D. M. Tratt, “Emerging Technologies, Lidar,” Encyclopedia of Remote Sensing. Springer,
New York, pp. 177–185, 2014. doi: 10.1007/978-0-387-36699-9.
[32] A. F. Chase, D. Z. Chase, C. T. Fisher, S. J. Leisz, and J. F. Weishampel, “Geospatial
revolution and remote sensing LiDAR in mesoamerican archaeology,” Proceedings of the
National Academy of Sciences of the United States of America, vol. 109, no. 32, pp. 12916–
12921, 2012, doi: 10.1073/pnas.1205198109.
[33] T. Inomata et al., “Monumental architecture at Aguada Fénix and the rise of Maya
civilization,” Nature, no. November 2019, 2020, doi: 10.1038/s41586-020-2343-4.
[34] A. Rowlands and A. Sarris, “Detection of exposed and subsurface archaeological remains
using multi-sensor remote sensing,” in Journal of Archaeological Science, vol. 34, no. 5,
2007, pp. 795–803. doi: 10.1016/j.jas.2006.06.018.
[35] R. Weaver, “Data Access,” Encyclopedia of Remote Sensing. Springer, New York, pp. 119–
121, 2014. doi: 10.1007/978-0-387-36699-9.
[36] W. Gail, “Commercial Remote Sensing,” Encyclopedia of Remote Sensing. Springer, New
York, pp. 78–83, 2014. doi: 10.1007/978-0-387-36699-9.
[37] D. Tapete, V. Banks, L. Jones, M. Kirkham, and D. Garton, “Contextualising
archaeological models with geological, airborne and terrestrial LiDAR data: The Ice Age
landscape in Farndon Fields, Nottinghamshire, UK,” Journal of Archaeological Science,
vol. 81, pp. 31–48, 2017, doi: 10.1016/j.jas.2017.03.007.
[38] K. Lambers, “Airborne and Spaceborne Remote Sensing and Digital Image Analysis in
Archaeology,” in Digital Geoarchaeology. Natural Science in Archaeology, C. Siart, M.
17
Forbriger, and O. Bubenzer, Eds. Cham: Springer, 2018, pp. 109–122. doi: 10.1007/978-3-
319-25316-9_7.
[39] P. Boldstad, GIS Funfamentals, A First Text on Geographic Information Systems, 5th ed. White
Bear Lake: XanEdu, 2016.
[40] K. M. S. Allen, S. W. Green, and E. B. W. Zubrow, Eds., Interpreting Space: GIS and
Archaeology. London: Taylor & Francis, 1990.
[41] M. C. L. Howey and M. Brouwer Burg, “Assessing the state of archaeological GIS research:
Unbinding analyses of past landscapes,” Journal of Archaeological Science, vol. 84, pp. 1–
9, 2017, doi: 10.1016/j.jas.2017.05.002.
[42] A. V. Postnov and E. G. Vergunov, “Подготовка специализированных ГИС для
археологии [Preparation of specialized GIS for archeology],” in Методика
археологических исследований Западной Сибири [Methodology of archaeological research
in Western Siberia], vol. 904, L. V. Tataurova, Ed. Omsk: Favorit, 2005, pp. 57–64.
[43] G. T. Edward, “Recent directions and future developments in geographic information
systems for historical archaeology,” Historical Archaeology, vol. 50, no. 3, pp. 24–49,
2016.
[44] B. Bobowski, “The history of digital methods in archeology (1951-1999) [Past Perfect
metod cyfrowych w archeologii (1951-1999)],” Echa Przeszlosci, vol. 22, no. 1, pp. 13–52,
2019, doi: 10.31648/ep.4827.
[45] M. D. McCoy and T. N. Ladefoged, “New developments in the use of spatial technology in
archaeology,” Journal of Archaeological Research, vol. 17, no. 3, pp. 263–295, 2009, doi:
10.1007/s10814-009-9030-1.
[46] F. Estrada-Belli and M. Koch, “Remote Sensing and GIS Analysis of a Maya City and Its
Landscape: Holmul, Guatemala,” in Remote Sensing in Archaeology, 1st ed., J. Wiseman
and F. El-Baz, Eds. New York: Springer, 2007, pp. 263–281. doi: 10.1007/0-387-44455-
6_11.
[47] G. von Krogh and S. Spaeth, “The open source software phenomenon: Characteristics
that promote research,” Journal of Strategic Information Systems, vol. 16, no. 3, pp. 236–
253, 2007, doi: 10.1016/j.jsis.2007.06.001.
[48] M. Craglia and K. Pogorzelska, “The Economic Value of Digital Earth,” in Manual of
Digital Earth, H. Guo, M. F. Goodchild, and A. Annoni, Eds. Singapore: Springer, 2020,
pp. 623–643. doi: 10.1007/978-981-32-9915-3_19.
[49] N. Gorelick, M. Hancher, M. Dixon, S. Ilyushchenko, D. Thau, and R. Moore, “Google
Earth Engine: Planetary-scale geospatial analysis for everyone,” Remote Sensing of
Environment, vol. 202, pp. 18–27, 2017, doi: 10.1016/j.rse.2017.06.031.
[50] A. Agapiou, “Remote sensing heritage in a petabyte-scale: satellite data and heritage
Earth Engine© applications,” International Journal of Digital Earth, vol. 10, no. 1, pp. 85–
102, 2017, doi: 10.1080/17538947.2016.1250829.
[51] Z. Magyari-Sáska, “Developing and implementing multiuser, fully relational gis database
for desktop systems using open source technologies,” Geographia Technica, vol. 10, no.
2, pp. 59–65, 2015.
[52] D. (eHealth A. Sonoiki, “An Open Source Approach to Multi-user Distributed Geospatial
Data Management,” FOSS4G, Open Source Geospatial Foundation (OSGeo), 2016.
https://av.tib.eu/media/20348 (accessed Jul. 22, 2020).
[53] J. O. Sexton et al., “Landsat Tree Cover Continuous Fields.” University of Maryland,
College Park, 2015.
18
[54] Y. Vasyunin and V. Lobanov, “Paititi Research: Data Processing Scripts for Earth Engine.”
figshare, 2021. doi: 10.6084/m9.figshare.14421125.v1.
[55] M. Buchhorn et al., “Copernicus Global Land Service: Land Cover 100m: collection 3:
epoch 2019: Globe.” Zenodo, 2020. doi: 10.5281/zenodo.3939050.
[56] “NASA Shuttle Radar Topography Mission Global 1 arc second.” NASA LP DAAC, 2013.
doi: 10.5067/measures/srtm/srtmgl1.003.
[57] D. Visentin et al., “Prehistoric landscapes of the Dolomites: Survey data from the highland
territory of Cadore (Belluno Dolomites, Northern Italy),” Quaternary International, vol.
402, pp. 5–14, 2016, doi: 10.1016/j.quaint.2015.10.080.
[58] V. A. Durán et al., “‘To and fro’ the southern Andean highlands (Argentina and Chile):
Archaeometric insights on geographic vectors of mobility,” Journal of Archaeological
Science: Reports, vol. 18, no. February, pp. 668–678, 2018, doi:
10.1016/j.jasrep.2017.05.047.
[59] C. Stanish et al., “Tiwanaku trade patterns in southern Peru,” Journal of Anthropological
Archaeology, vol. 29, no. 4, pp. 524–532, 2010, doi: 10.1016/j.jaa.2010.09.002.
[60] D. A. Contreras, “How far to conchucos? A GIS approach to assessing the implications of
exotic materials at chavín de huántar,” World Archaeology, vol. 43, no. 3, pp. 380–397,
2011, doi: 10.1080/00438243.2011.605841.
[61] K. Kompatscher and M. Hrozny Kompatscher, “Dove piantare il campo: modelli
insediativi e di mobilità nel Mesolitico in ambiente alpino,” Preistoria alpina, vol. 42, no.
42, pp. 137–162, 2007.
[62] S. Caracausi, G. L. F. Berruti, S. Daffara, D. Bertè, and F. Rubat Borel, “Use of a GIS
predictive model for the identification of high altitude prehistoric human
frequentations. Results of the Sessera valley project (Piedmont, Italy),” Quaternary
International, vol. 490, no. November 2017, pp. 10–20, 2018, doi:
10.1016/j.quaint.2018.05.038.
19
APPENDIX A. LIST OF THE KNOWN ARCHAEOLOGICAL SITES
Coordinates: longitude, latitude (WGS84)
Type
Name
-78.18911009041936211 -9.54916617589478633
Inca
unknown
-65.88649965857210589 -22.51079371023869768
Inca
unknown
-72.11288030339787269 -13.30098187108892382
Inca
unknown
-71.96997944139567949 -13.50830249957862073
Inca
unknown
-73.95403729365048662 -13.65389959829188626
Inca
unknown
-77.07198444695121964 -12.06723787728948238
pre-Inca
unknown
-77.07215274547407091 -12.06227902089199588
pre-Inca
unknown
-74.01429306297312394 -13.60065919632519638
Inca
Acllahuasi
-65.84857780000000105 -22.02524539999999931
Inca
Calahoyo
-78.42842652190149977 0.00018832541572125
Inca
Catequilla
-74.85800133182270599 -12.73478481678160179
Inca
Chunkana
-71.56055627202657377 -16.44512874384819412
pre-Inca
Complejo arqueologico Kakallinca
-71.56353417186322474 -16.44021836937946190
pre-Inca
Complejo arqueologico Kasapatac
-71.56254170001484738 -16.44353695077025890
pre-Inca
Complejo arqueologico Tres Cruces
-77.07269238428330027 -11.98450875741795763
pre-Inca
Huaca Aznapuquio
-77.07107192555619690 -11.88957033583235123
Inca
Huaca Tambo Inga
-76.81635102476359123 -9.87541968860685415
Inca
Huánuco Pampa
-72.06351646239625097 -13.36353541577751258
Inca
Machu Qollqa
-71.60638889999999890 -14.96861110000000039
Inca
María Fortaleza
-71.56500409999999590 -14.92476779999999970
pre-Inca
Mauk'allaqta
-72.19708230495820089 -13.33048485382195203
Inca
Moray
-65.88572323756328331 -22.51232257670934445
Inca
Moreta
-75.17824304409330693 -14.55299338856019098
Inca
Nazca Lines
-72.92796036703208529 -13.11109954629611174
Inca
Ñusta Hispana
-78.99691147679186543 -2.90740317403922788
Inca
Parque Arqueologico Pumapungo
-72.42502207191016339 -13.23102744054985891
Inca
Patallacta
-72.53152510035695855 -13.20632265580255371
Inca
Phuyupatamarka
-72.26088324196463475 -13.25638700781221324
Inca
Pinkuylluna
-71.84319517809181832 -13.40821846561960662
Inca
Pisaq
-70.68378789999999867 -34.06834589999999707
Inca
Pucará del Cerro Grande de La Compañía
-71.96216028138753984 -13.48335108060471832
Inca
Puka Pukara
-75.82896960048169888 -13.70519710552497905
Inca
Pukatampu
-76.93599554467552082 -12.04962003679565008
Inca
Puruchuco
-71.97175117653151233 -13.51034624781295612
Inca
Q'enqo Chico
-71.97049307681248820 -13.50901159324859790
Inca
Q'enqo Grande
-72.26210343998327801 -13.61450273616862638
Inca
Qollmay
-74.86079831300708065 -12.73796948248669381
Inca
Qorimina
-73.96390900000000101 -13.83710969999999918
Inca
Raqa Raqaypata
-71.36929837921640285 -14.17496004399198917
Inca
Raqch'i
-74.06619411790684637 -13.96872877319371398
Inca
Ruinas de Inti Watana
-78.43087192407350017 0.01335577461163674
Inca
Rumicucho
-72.50139498618257505 -13.22845600726906667
Inca
Runkuraqay
-71.99244283544612699 -13.52935116256297654
Inca
S.A. Pokenkancha
-71.99469277501910369 -13.53154425677604600
Inca
S.A. Puquin / Hermanos Ayar
-71.98042427751055072 -13.50732942924198987
Inca
Saksaywaman
-71.25511125581158467 -29.88552875250002216
Inca
Sitio Paleoindio de Asentamiento y Necrópolis El
Olivar
20
-71.96736911450786067 -13.47888675402604974
Inca
Tampu Mach'ay
-75.68501510000000110 -11.47505529999999929
Inca
Tarmatambo
-74.85795984135754111 -12.73757134621275355
Inca
Uchkus Inkañan
-72.93186783394334327 -13.09827929063190410
Inca
Vitcos
-72.49564965592215060 -13.02076404242771446
Inca
Wamanmarka
-73.20710649185696184 -12.90207456860550117
Inca
Willkapampa
-72.42322342843154104 -13.23537539901991522
Inca
Willkarakay
-72.53640625789245178 -13.19293752860655111
Inca
Wiñay Wayna
-71.98914906044983297 -13.53419116329233951
Inca
Z.A. Qhataq'asapatallaqta
-76.77242694332446149 -12.50955480261609587
Inca
Zona Arqueológica Bandurria
-66.08736319999999864 -19.81218150000000122
Inca
Zona Arqueologica Chaquilla
-76.90104949435638559 -12.25696776933438059
pre-Inca
Zona Arqueológica Monumental Pachacámac
-74.06795083729583951 -13.91477942317083993
Inca
chancas
Zona Arquologica de Cachipata
21
APPENDIX B. SLOPE STEEPNESS/SETTLEMENT SUITABILITY MAP
It predicts suitable areas of the search for archaeological sites, which are depicted in green color.
22
APPENDIX C. CLOSER LOOK TO THE AOI 3
Natural-color image of the AOI 3 from PlanetScope
satellite, acquired Sep 28, 2019, by courtesy of
Planet Labs, Inc.
Settlement Suitability Map. A potential large site is
located in the “green” zone with a high suitability
score, mostly defined by a low slope grade.
Overview of the AOI 3 and its surroundings in 3D: NASA SRTM3 SRTMGL1 digital elevation model
with superimposed PlanetScope image, acquired Sep 28, 2019, by courtesy of Planet Labs, Inc.
Red markers correspond to potential archaeological features identified in the imagery.
