Content uploaded by Christopher Mark Nicholson
Author content
All content in this area was uploaded by Christopher Mark Nicholson on Sep 26, 2016
Content may be subject to copyright.
Egeland et al. / GIS and Paleoanthropological Site Selection
89
INTRODUCTION
e timing and nature of the initial hominid dispersals
from Africa during the Plio-Pleistocene (here 2.0-1.5
million years ago [MYR]) is an issue of great interest
for paleoanthropology. However, the biological, tech-
nological, and ecological context of these dispersals
remains cloudy due largely to a paucity of Eurasian
paleoanthropological sites dating to this time period.
Indeed, there are only a handful of well-accepted
Plio-Pleistocene sites from Eurasia: Dmanisi in the
Republic of Georgia at 1.77-1.81 MYR (de Lumley
et al. 2002), the Nihewan and Yuanmou basins of
China at 1.66-1.70 MYR (Zhu et al. 2008), and the
Indonesian island of Java at least 1.66 MYR (Sangiran)
but perhaps as early as 1.81 MYR (Mojokerto) (Larick
et al. 2001; Swisher et al. 1994). Although the Levant,
given its geographic location, is the most logical extra-
African source of dispersing hominid populations, the
earliest well-accepted occupations there ('Ubeidiya
in Israel) date to somewhat later in time at 1.4 MYR
(Belmaker et al. 2002).
Plio-Pleistocene sites are extremely rare, and sites
preserved in high-integrity depositional contexts
are even more so. In fact, the rich early Pleistocene
component at Dmanisi was itself unearthed more-or-
less accidentally during the excavation of a medieval
fortress (Djaparidze et al. 1989). As fortunate as this
discovery was, survey efforts informed by ecologi-
cally relevant variables such as vegetation, geography,
topography, and geology may not only increase the
chances of finding paleoanthropological sites, but
will also help place hominid occupations into a
broader environmental context. Here we describe
an approach to identify target areas for paleoanthro-
pological survey. is method uses GIS to integrate
data from archaeology and ecology to identify high
potential areas for intensive ground survey. As an
example, we present pre- and post-survey data from
a new paleoanthropological research project in
northern Armenia.
FiElD notE
Using GIS and Ecological Variables to Identify
High Potential Areas for Paleoanthropological Survey:
An Example from Northern Armenia
C P. E
C M. N
B G
Journal of Ecological Anthropology
90
Vol. 14 No. 1 2010
PREDICTIVE MODELING USING GIS
DATA
Predictive models assume that the locations of sites
are at least partially influenced by modern or pre-
historic environmental factors such as vegetation,
distance to water, or topographic setting (e.g., Mehrer
and Wescott 2006). For example, remote sensing data
have been successfully used to identify high potential
geological strata for paleoanthropological survey in
East Africa (Asfaw et al. 1990; Harmand et al. 2009).
e greater affordability of digital data and the abil-
ity of GIS to integrate and manipulate numerous
datasets now permit relatively sophisticated remote
predictive modeling. As described below, the isolation
of possible hominid dispersal routes and—within
these dispersal corridors—areas that are likely to
contain evidence of early hominid activity, allows
for more focused pedestrian survey.
NORTHERN ARMENIA AS A HIGH
POTENTIAL SURVEY REGION
Current evidence indicates that by the early Pleis-
tocene, hominids had traveled between 1,000 and
5,400 miles from their African homeland (Carbonell
et al. 2008). However, this seemingly widespread
occurrence does not necessarily mean that hominid
populations were distributed evenly across Eurasian
landscapes, especially during the initial stages of
dispersal. It is therefore possible that hominids used
particular corridors that contained favorable ecologi-
cal conditions for their expansion. erefore, the first
step is the identification, in a very broad sense, of
potential survey regions.
A theoretical dispersal path was constructed between
the Levant and the earliest well-accepted evidence for
hominid occupation outside of Africa—Dmanisi.
Any origin point in the area provides the same results;
FIGURE 1: Regional map showing origin (‘Ubeidiya, Israel) and destination
(Dmanisi, Georgia) points for the Cost Path Analysis.
Egeland et al. / GIS and Paleoanthropological Site Selection
91
in the analysis presented here, the site of 'Ubei-
diya in Israel was used. A simple cost path analysis
(CPA) model was employed, which determines the
path from a source to a destination using a series of
algorithms that take into account impediments to
travel (e.g., Hare 2004). Assuming that populations
will select a path that minimizes the cost (energy) of
travel, the goal of the application was to identify a
least cost path (LCP). is function was performed
in ArcMap 9.3 using the Spatial Analyst with two
input raster layers: the cost raster and the back link
raster. e cost raster was represented by modern ter-
rain (derived using digital elevation models [DEM]),
while the back link raster retraced the least-costly
route from the destination to the source over the
cost distance surface. Using these two raster layers,
an algorithm calculated a single path of raster cells
that is the “cheapest” cumulative route relative to cost
(i.e., slope). Once the slope and back link rasters were
created, ArcMap performed the cost path analysis to
create a raster layer of the least cost path, which was
then converted to a vector file for display.
Based on modern terrain, the cheapest route between
'Ubeidiya and Dmanisi runs northeast across Syria,
into eastern Turkey and skirts along the northwestern
border of Armenia (Figure 1). Once in the Lesser
Caucasus of northern Armenia, the least cost path
passes north across the Tashir Plateau before termi-
nating at Dmanisi. Because regional topography has
changed somewhat over the past two million years
(see below), this cost path analysis was not meant to
predict the precise location of paleoanthropological
sites; rather, as mentioned above, it served to isolate
potential survey regions. at the cost path analysis
matched well with the distribution of known Lower
Paleolithic occurrences in northern Armenia supports
the presumption that the region was an important
corridor for the movement of early hominid popula-
tions (Figure 2).
FIGURE 2: Map of northern Armenia (inset) and northeastern Armenia with the location
of geographic features, previously identied Paleolithic sites, and the Least Cost Path.
http://scholarcommons.usf.edu/jea/vol14/iss1/8 | DOI: http://dx.doi.org/10.5038/2162-4593.14.1.8
Journal of Ecological Anthropology
92
Vol. 14 No. 1 2010
Although systematic data for the Plio-Pleistocene of
the Lesser Caucasus is only beginning to emerge (e.g.,
Roiron et al. 2007), paleoenvironmental consider-
ations further buttress this assertion. Perhaps most
importantly, the site of Dmanisi clearly indicates that
the Lesser Caucasus could accommodate hominid
habitats during the Plio-Pleistocene. It has even
been suggested that the region served as a refugium
during colder time periods (Gabunia et al. 2000). In
addition, many of the intermontane depressions of
the Lesser Caucasus were filled by large freshwater
lakes during the late Pliocene. Pleistocene volcanism
eventually fragmented these lakes into smaller lacus-
trine basins (Lededev et al. 2008a, 2008b; Sayadyan
2006a, 2006b). e potential presence of lake-mar-
gin and alluvial environments of Plio-Pleistocene age
in the region is especially significant given that Dma-
nisi itself is thought to have been in close proximity
to a lake (Gabunia et al. 2000), and early hominid
occupation of well-watered habitats such as riparian
woodlands and lake-margins is well-documented in
East Africa at both Olduvai Gorge and sites in the
Turkana Basin (Hay 1976; Rogers et al. 1994).
e next step was to identify specific areas in the
Lesser Caucasus for focused pedestrian survey. As
Figure 2 shows, there are several paleoanthropologi-
cal sites documented on the Tashir Plateau that lie
along the dispersal path calculated by the cost path
analysis. However, many of these and other known
sites in the region document hominid occupation
only back to the early middle Pleistocene—which
post-dates the earliest dispersals from Africa—and
tend to lack materials that provide reliable dates
(e.g., volcanic material and/or well-preserved fauna)
(Doronichev 2008). e closest area within the high
potential dispersal region (as determined by the cost
path analysis) that preserves alluvial, lacustrine, and,
most importantly, datable volcanic deposits spanning
much of the Plio-Pleistocene, is the Debed River Val-
ley of northeastern Armenia. e Debed was there-
fore considered to be an attractive area for identifying
new paleoanthropological sites. Particularly striking
was the lack of paleoanthropological sites in and
along the valley (Figure 2), which is related directly
to a lack of prior paleoanthropological research in
the area. GIS was therefore used to conduct a site
suitability analysis for the Debed River Valley.
Land Cover Type No. of Occurrences LST Score1
14 - Rain-fed croplands 11 48
20 - Mosaic croplands/vegetation 23 100
30 - Mosaic vegetation/croplands 22 96
50 - Closed broadleaved deciduous forest 15 65
110 - Mosaic forest/shrubland/grassland 1 4
1LST = Linear Scale Transformation
TABLE 1: Land cover categories used in the site suitability analysis. All LST scores were scaled
to the maximum value (23) to derive suitability scores. See text for full explanation.
Egeland et al. / GIS and Paleoanthropological Site Selection
93
Aspect (Degrees) No. of Occurrences LST Score1
23-67 5 38
68-112 8 62
113-157 11 85
158-202 11 85
203-247 4 31
248-292 13 100
293-337 12 92
338-360 6 46
TABLE 2: Aspect categories used in the site suitability analysis. All LST scores were scaled
to the maximum value (13) to derive suitability scores. See text for full explanation.
TABLE 3: Slope categories used in the site suitability analysis. All LST scores were scaled
to the maximum value (29) to derive suitability scores. See text for full explanation.
Slope (Degrees) No. of Occurrences LST Score1
0.0-0.5 29 100
0.6-1.0 17 59
1.1-1.5 11 38
1.6-2.0 12 41
2.1-2.5 1 3
2.6-3.0 2 7
TABLE 4: Elevation categories used in the site suitability analysis. All LST scores were scaled
to the maximum value (31) to derive suitability scores. See text for full explanation.
Elevation (Meters) No. of Occurrences LST Score1
0-1000 19 61
1000-2000 31 100
2000-3000 16 52
3000+ 6 19
1LST = Linear Scale Transformation
http://scholarcommons.usf.edu/jea/vol14/iss1/8 | DOI: http://dx.doi.org/10.5038/2162-4593.14.1.8
Journal of Ecological Anthropology
94
Vol. 14 No. 1 2010
SITE SUITABILITY ANALYSIS
Site suitability analysis enters variables into a com-
puter model that geographically displays areas that
are most (and least) likely to preserve sites based on
numerical suitability scores (the higher the score,
the more conducive an area is for site identification).
e location of previously identified Paleolithic sites
in northern Armenia (n = 72; see Figure 2) was used
to identify predictive variables for site location. e
variables most closely associated with site location
were slope, aspect, elevation, land cover, and
proximity to rivers. For the GIS analysis, polygon
data for each variable were converted from shapefiles
to raster files. As an example, consider land cover:
five categories coincided with previously identi-
fied Paleolithic sites (Table 1). Using a linear scale
transformation (LST; Malczewski 1999), numerical
values for each land cover category were assigned
based on the number of sites that occurred in a
particular category. For land cover, known Paleo-
lithic sites were most often associated with mosaic
croplands/vegetation (a total of 23 times). Because
this represented the highest frequency of asso-
ciations, croplands/vegetation received a suitability
score of 1 and all subsequent scores were scaled to
this value. e linear scale transformation values
for each variable were summed using the raster
calculator, averaged to remove potential outliers,
and multiplied by 100. is resulted in a composite
suitability score that ranged from 0 (lowest suit-
ability) to 100 (highest suitability). In general, the
highest suitability scores were associated with areas
located near rivers with low slope and relatively open
vegetation (i.e., cropland). Tables 2-4 summarize
the LST scores for aspect, slope, and elevation. A
2 km buffer was constructed along major rivers to
assign distance-to-water scores.
FIGURE 3: Raster map of site suitability scores for the Debed River Valley
and the location of identied paleoanthropological sites.
Egeland et al. / GIS and Paleoanthropological Site Selection
95
e calculated raster values were reclassified into three suit-
ability categories: Unsuitable, Suitable, and Very Suitable.
Suitable was defined as the mean suitability score of the
previously identified sites (= 65) with a range equal to the
standard deviation of the previously identified sites (SD
= 15.4). is provided a range of 50-81 for the Suitable
category. Scores below 50 were then defined as Unsuitable
and scores above 81 as Very Suitable. ese values were
then used to produce a raster map to visualize the potential
location of paleoanthropological sites in the Debed River
Valley (Figure 3), which in turn served to focus survey
efforts. It quickly became clear that, based on the site suit-
ability analysis, the northernmost stretch of the Debed near
the Georgian border had the highest potential to preserve
paleoanthropological sites.
POST-SURVEY RESULTS
During the summer of 2009, preliminary survey was
conducted along the Debed River Valley between its
confluence with the Dzoraget River in the south to the
Georgian border in the north, a distance of approximately
60 km. Limited field time precluded a complete and sys-
tematic survey of the entire 60 km stretch, so, guided by
the suitability analysis, the survey team was transported to
high potential localities by vehicle after which pedestrian
survey was carried out. A total of 25 new sites spanning
the Lower Paleolithic through the Upper Paleolithic were
identified (Table 5). As can be seen in Table 5, a majority
of the sites were discovered—as predicted by the suitability
analysis—along the lower Debed near the border with
Georgia (Figure 3). Two of these sites (Haghtanak 3 and
Ayrum 2) preserved Oldowan-type chopper forms that may
be associated with a Plio-Pleistocene hominid occupation
(Egeland et al. 2010).
e concentration on Suitable and Very Suitable areas in
the Debed River Valley was an effective survey strategy,
and the remote GIS analysis certainly maximized field
time. However, there are some limitations to the study as
currently conceived. First, the goal of this initial round
of research was simply to identify the presence of paleo-
anthropological material. Survey of the valley in general
and at each site in particular was by no means systematic
TABLE 5: List of Paleolithic sites
identied in the Debed River Valley
during the summer of 2009 and
associated site suitability scores.
Suitability scores below 50 are
considered Unsuitable scores
between 50-81 are considered
Suitable, and scores above 81
are considered Very Suitable.
Site Site Suitability
Score
Lchkadzor 49
Akori 1 59
Haghtanak 3 59
Arevatsag 2 60
Vahagni 1 60
Bagratashen 5 61
Akori 2 64
Arevatsag 1 66
Debedavan 3 67
Haghtanak 2 68
Ptghavan 3 69
Haghtanak 1 71
Bagratashen 4 71
Ptghavan 4 77
Ayrum 1 78
Debedavan 1 78
Debedavan 2 78
Haghtanak 4 81
Ayrum 3 82
Ayrum 2 84
Bagratashen 1 88
Bagratashen 3 88
Bagratashen 2 89
Ptghavan 1 91
Ptghavan 2 92
http://scholarcommons.usf.edu/jea/vol14/iss1/8 | DOI: http://dx.doi.org/10.5038/2162-4593.14.1.8
Journal of Ecological Anthropology
96
Vol. 14 No. 1 2010
and it therefore cannot be determined at this point
what percentage of each suitability category was
surveyed. It can be said, however, that several areas
with high suitability scores have yet to be surveyed.
Second, the data on modern landscape variables
were relatively coarse-grained and, importantly,
lacked a temporal dimension. Environmental recon-
structions are available for the middle Pliocene in
formats easily incorporated into a GIS (Salzmann et
al. 2008). Unfortunately, these data are simply too
coarse to be of much use for an analysis at the scale
presented here. More detailed data on a number
of paleogeographic and paleoecological variables
and how they would affect the predictive model-
ing are needed. e spatial extent of Pleistocene
lakes throughout the Lesser Caucasus would be
particularly useful in this context. Finally, it must
be realized that remote GIS predictive modeling,
while providing a useful guide for site identifica-
tion, is no substitute for (and can be modified by)
on-the-ground experience. Consider the site of
Lchkadzor, which is the one locality that scored in
the Unsuitable category (though only by a single
point). e site is a diffuse lithic scatter located on
the relatively steep slopes of a small foothill over-
looking the Debed. e sedimentary outcrops that
prompted further investigation at Lchkadzor were
only identified when the survey team was on the
ground investigating a high potential area nearby.
Future work will aim to address these issues more
fully. Nevertheless, the results of this study indicate
that paleoanthropological survey can benefit from
predictive modeling using the integration of envi-
ronmental variables and GIS.
Charles P. Egeland, Department of Anthropology,
University of North Carolina at Greensboro,
cpegelan@uncg.edu
Christopher M. Nicholson, Water Resources Data
System, Department of Civil and Architectural Engi-
neering, University of Wyoming, cnichol5@uwyo.edu
Boris Gasparian, Institute of Archaeology and
Ethnography, National Academy of Sciences of the
Republic of Armenia, borisg@virtualarmenia.am
ACKNOWLEDGMENTS
is is Paper No. 1 in the Lori Depression Paleoan-
thropological Project publication series. Members of
the project not appearing as authors include Dmitri
Arakelyan, Ryan M. Byerly, and Robert Ghukasyan.
Sincere thanks to Dr. Pavel Avetisyan and the In-
stitute of Archaeology and Ethnography (National
Academy of Sciences of the Republic of Armenia) for
supporting this research. anks to Dina Zardaryan
(Institute of Archaeology and Ethnography) for
help in dating materials and to Samvel Nahapetyan
(Department of Cartography and Geomorphology,
Yerevan State University) for providing important
information on the geomorphology of the Debed
River Valley. Funding for the Lori Depression Paleo-
anthropological Project has been provided through a
grant from the National Science Foundation (BCS-
0936385) to Charles P. Egeland.
REFERENCES CITED
A, B., C. E, D. H, T.D. W,
G. WG.
1990 Space-based imagery in paleoanthropo-
logical research: An Ethiopian example.
National Geographic Research 6:418-434.
B, M., E. T, S. C,
O. B-Y.
2002 New evidence for hominin presence in
the Lower Pleistocene of the southern
Levant. Journal of Human Evolution
43:43-56.
Egeland et al. / GIS and Paleoanthropological Site Selection
97
C, E., M. M, X.P. R,
J.M. B C, F. B,
J. R, R. S, J. V.
2008 Eurasian gates: The earliest human
dispersals. Journal of Anthropological
Research 64:195-228.
L, H., D. L, G. F,
T. G, C. P, C. F,
J. G, T. S, P. V.
2002 40Ar/39Ar dating of the Dmanisi (Georgia)
hominid-bearing volcanic ash level (
Layer VI): 1.81Ma. Comptes Rendus
Palevol 1:181-189.
D, V., G. B, T. B,
L. G, A. J, N. K,
E. K, D. L,
G. M, N. M, M. N,
E. P, H. S,
D. S, D. T,
M. T, A. V.
1989 Der altpaleolitische Fundplatz Dmanisi
in Georgien. Jahrbuch des Römisch-Ger-
manisches Zentralmuseum 36:67–116.
D, V.B.
2008 The Lower Paleolithic in eastern Europe
and the Caucasus: A reappraisal of the
data and new approaches. PaleoAnthro-
pology 2008:107-157.
E, C.P., B. G, D. A,
R.M. B, R. G,
C.M. N.
2010 New data on the Paleolithic settlement of
the Lori Depression, northern Armenia.
Abstracts of the PaleoAnthropology Society
2010 Meetings: A9.
G, L., A. V, D. L.
2000 The environmental contexts of early
human occupation of Georgia (Trans-
caucasia). Journal of Human Evolution
38:785-802.
Hare, T.S.
2004 Using measures of cost distance in the
estimation of polity boundaries in the
Postclassic Yuatepec Valley, Mexico.
Journal of Archaeological Science
31:799-814.
H, S., D. DG, L. S, J. L,
S. M, I. D, M. O.
2009 Nouveaux sites paléolithiques anciens en
République de Djibouti: bilan prélimi-
naire de prospections récentes dans le
Bassin du Gobaad, Afar central.
Comptes Rendus Palevol 8:481-492.
H, R.L.
1976 Geology of the Olduvai Gorge. Berkeley,
CA: University of California Press.
L, R., R.L. C, Y. Z, S,
S, Y. R, F. A, M. R,
M. H.
2001 Early Pleistocene 40Ar/39Ar ages for Ba-
pang Formation hominins, Central Jawa,
Indonesia. Proceedings of the National
Academy of Sciences 98:4866-4871.
L, V.A., S.N. B, O.Z. D,
G.T. V.
2008a Geochronology of Pliocene volcanism in
the Dzhavakheti Highland (the Lesser
Caucasus). Part 1: Western part of the
Dzhavakheti Highland. Stratigrafiya,
Geologicheskaya Korrelyatsiya 16:204-
224.
http://scholarcommons.usf.edu/jea/vol14/iss1/8 | DOI: http://dx.doi.org/10.5038/2162-4593.14.1.8
Journal of Ecological Anthropology
98
Vol. 14 No. 1 2010
L, V.A., S.N. B, O.Z. D,
G.T. V.
2008b Geochronology of Pliocene volcanism in
the Dzhavakheti Highland (the Lesser
Caucasus). Part 2: Eastern part of the
Dzhavakheti Highland. Regional geologi-
cal correlation. Stratigrafiya, Geologiches-
kaya Korrelyatsiya 16:553-574.
M, J.
1999 GIS and Multicriteria Decision Analysis.
New York, NY: John Wiley and Sons.
M, M.W., K.L. W, E.
2006 GIS and Archaeological Site Location
Modeling. Boca Raton, FL: CRC Press.
R, M.J., J.W.K. H, C.S. F.
1994 Changing patterns of land use by
Plio-Pleistocene hominids in the
Lake Turkana Basin. Journal of Human
Evolution 27:139-158.
R, P., I. G, S. J,
S. N, V. O, J-J. C.
2007 “Paleoenvironmental reconstruction of
Pleistocene fluvio-lacustrine landscapes in
Armenia by multiproxy studies: Geomor-
phology, K/Ar chronology, paleomagne-
tism, leaf flora and pollen analysis,” Paper
presented at the 4th International Limno-
geology Congress. Barcelona, Spain.
S, U., A.M. H, D.J. L,
P.J. V, D.J. H.
2008 A new global biome reconstruction and
data-model comparison for the Middle
Pliocene. Global Ecology and Biogeography
17:432-447.
S, Y.V.
2006a Regional stratigraphic scheme and paleo-
geographic events of the Late Miocene,
Pliocene and Quaternary in Armenia.
Doklady Earth Sciences 407:198-201.
S, Y.V.
2006b Upper Miocene, Pliocene, and Quaterna-
ry stratigraphic reference sections of large
intermontane depressions in Armenia.
Doklady Earth Sciences 407:217-219.
S III, C.C., G.H. C, T. J,
A.G. G, A. S.
1994 Age of the earliest known hominids in
Java, Indonesia. Nature 263:1118-1121.
Z, R.X., R. P, Y.X. P, H.T. Y,
L.Q. L, X. Z, X. G, L.W. C, F. G,
C.L. D.
2008 Early evidence of the genus Homo in
East Asia. Journal of Human Evolution
55:1075-1085.
