Lost Worlds: A predictive model to locate submerged archaeological sites in SE Alaska, USA

Abstract

The primary objectives of this research are to develop and test an archaeological settlement pattern model designed to identify areas of high archaeological potential on the continental shelf of Southeast (SE) Alaska. Sea level history and glacial geology suggest that the archaeological record prior to 10,600 cal BP 2 (9,400 RCYBP 3) may be submerged on the continental shelf. To locate and test for sites older than 10,600, it is essential to extend archaeological survey to the continental shelf in areas that were either unglaciated refugia or areas that were deglaciated during the interval between 18,000 -10,600 cal BP (15,000 – 9,400 RCYBP). This research facilitates the exploration and interpretation of the origins and character of early maritime adaptations along the Northwest Coast (NWC) of North America. Archaeological sites document continuous occupation of SE Alaska following sea level rise above modern levels around 10,600 cal BP. 2 cal BP = calibrated using Calib 6.0 (http://calib.qub.ac.uk/ calib/calib.html). IntCal 09 (Reimer et al. 2009) 3 RCYBP = radiocarbon years before present The few documented earlier occupations are small interior sites that include locations where people occasionally engaged in hunting bears at hibernacula. These sites, which date between 12,500 and 10,600 cal BP, demonstrate that the region was occupied at times of lower sea level (Dixon 1999; Fedje et al. 2008; Fedje and Mathews 2005). They support the possibility of submerged coastal sites located on the continental shelf where maritime subsistence resources were abundant. High potential areas are defined based on synthesis and interpretation of archaeologically and ethnographically documented settlement patterns applied to reconstructions of the submerged landscape. Both inductive and deductive modelling methods were utilized (Verhagen and Whitley 2012); the scale of measurement is interval or ratio and both the analytic and the systemic contexts were assessed to develop the model (Kohler 1988: 35-37, Schiffer 1972). The model was produced at 500 RCY intervals from 10,790 to 18,100 cal BP (9,500 to 15,000 RCYBP).

Full text

678
The primary objectives of this research
are to develop and test an archaeological
settlement pattern model designed to identify
areas of high archaeological potential on the
continental shelf of Southeast (SE) Alaska.
Sea level history and glacial geology suggest
that the archaeological record prior to 10,600
cal BP2 (9,400 RCYBP3) may be submerged
on the continental shelf. To locate and test for
sites older than 10,600, it is essential to extend
archaeological survey to the continental shelf
in areas that were either unglaciated refugia or
areas that were deglaciated during the interval
between 18,000 -10,600 cal BP (15,000 –
9,400 RCYBP). This research facilitates the
exploration and interpretation of the origins
and character of early maritime adaptations
along the Northwest Coast (NWC) of North
America.
Archaeological sites document continuous
occupation of SE Alaska following sea level rise
above modern levels around 10,600 cal BP.
Corresponding author: krbm@unm.edu
2 cal BP = calibrated using Calib 6.0 (http://calib.qub.ac.uk/
calib/calib.html). IntCal 09 (Reimer et al. 2009)
3 RCYBP = radiocarbon years before present
The few documented earlier occupations are
small interior sites that include locations where
people occasionally engaged in hunting bears
at hibernacula. These sites, which date between
12,500 and 10,600 cal BP, demonstrate that
the region was occupied at times of lower sea
level (Dixon 1999; Fedje et al. 2008; Fedje and
Mathews 2005). They support the possibility
of submerged coastal sites located on the
continental shelf where maritime subsistence
resources were abundant.
High potential areas are dened based on
synthesis and interpretation of archaeologically
and ethnographically documented settlement
patterns applied to reconstructions of the
submerged landscape. Both inductive and
deductive modelling methods were utilized
(Verhagen and Whitley 2012); the scale of
measurement is interval or ratio and both
the analytic and the systemic contexts were
assessed to develop the model (Kohler 1988:
35-37, Schier 1972). The model was produced
at 500 RCY intervals from 10,790 to 18,100 cal
BP (9,500 to 15,000 RCYBP).
Lost Worlds: A Predictive Model to Locate
Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone and E. James Dixon
University of New Mexico, USA
Andrew D. Wickert
University of Colorado, USA
Abstract:
The archaeological record of the northern Northwest Coast (NWC) extends to approximately 12,200 cal
BP; however, much of the habitable area dating to before 10,600 cal BP is now submerged. Recent research
indicates that large areas of Southeast Alaska and western British Columbia were glaciated from 21,000
to 17,000 cal BP, albeit with refugia (unglaciated areas able to support life) existing along the coast. By
16,000 cal BP, much of the region was deglaciated and ecologically viable for human habitation. This
project develops and tests a model to identify high potential areas for the occurrence and preservation of
archaeological sites on the continental shelf of Southeast Alaska. The paleolandscapes are developed from
bathymetric data and paleoecological data.
Keywords:
Marine Survey, Underwater Archaeology, SE Alaska, Sea Level, Paleolandscape

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
679
1. Study Area
The study region is part of the NWC
culture area of North America that extends
from the south end of Haida Gwaii (Queen
Charlotte Islands), BC to the Gulf of Alaska
in the north (Ames and Maschner 1999, 18),
crossing the international boundary between
Canada’s British Columbia (BC) and the US
(Alaska). This research is centred (study area)
on the Alexander Archipelago in SE Alaska,
but also draws on previous work in Haida
Gwaii. The model covers the entire study area
including areas that are currently submerged
and subaerial.
The Alexander Archipelago is a chain of
over 10,000 islands and islets in SE Alaska.
The islands rise to a maximum elevation of
approximately 1100 m (3600 ft). The largest
island is Prince of Wales Island (POWI) (Fig.
1). The Alexander Archipelago is the traditional
home to Native American groups of Tlingit,
Haida, and Tsimshian. The region is a coastal
temperate rainforest and has a rich supply of
maritime resources including shellsh, sh,
marine birds, and marine mammals, along
with a variety of terrestrial plants and animals
(O’Clair et al. 1992). The shorelines contain
numerous embayments and coves that harbour
estuarine environments and biotic resources
(Moss 1998, 91). The submerged nearshore
landscape was previously glaciated and is
characterized by fjords and submerged valleys
that have been drowned by post Last Glacial
Maximum (LGM) sea level rise.
There are 11 large rivers in southeast
Alaska. These rivers are all on the mainland;
there are no major river systems on the islands
in the Alexander Archipelago (Johnson et
al. 2008, 3-5). The regional absence of large
sediment laden rivers indicates that no
mechanism is present to route signicant
amounts of terrigenous sediment to the ocean.
This is important because heavy sediment
loads can deeply bury archaeological sites on
the continental shelf, making them dicult to
detect and sample.
1.1 Glaciation
Based on coral records from Barbados,
the LGM occurred approximately 21,000 cal BP
(Peltier and Fairbanks 2006, 3326). However,
the record for the northern NWC appears to
be diachronous. Evidence from Haida Gwaii
suggests the maximum glacial extent was
reached at approximately 19,000 cal BP (16,
000 RCYBP) (Blaise et al. 1990, 292). In SE
Alaska, the maximum glaciation occurred
between 29,000 and 18,000 cal BP (25,000
– 15,000 RCYBP) (Clague et al. 2004, 86).
This large temporal range largely results from
limited research in the region (Kaufman and
Manley 2004, 22; Mann and Hamilton 1995,
459).
In the early twentieth century, glacial
geologists believed that the Cordilleran ice sheet
owed across SE Alaska and its continental
Figure 1. Northern NWC including Alexander
Archipelago, Study Area, and pre-9000 cal BP
archaeological sites.

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
680
shelf as a contiguous ice mass (Antevs 1929,
651; O’Clair et al. 1992, 21). More recent work,
however, shows that the Cordilleran ice sheet
merged with local valley glaciers to create
discrete ice lobes that owed westward to the
sea. Between these ice lobes existed refugia,
ice-free areas that continued to support life
(Carrara et al. 2007, 232; Heaton and Grady
2003; Kaufman and Manley 2004, 22; Mann
1986). The presence of faunal remains dating
to the LGM in both the Alexander Archipelago
and Haida Gwaii document the presence
of Late Wisconsin age refugia (Heaton and
Grady 2003; Fedje and Mathewes 2005).
Paleoenvironmental research suggests that
refugia may have played a signicant role in
biotic colonization following deglaciation,
in conjunction with subsequent northward
migration of plant species following deglaciation
about 16,340 cal BP (14,000 RCYBP; Ager
and Rosenbaum 2007; Ager et al. 2010). This
evidence indicates that these environments
could have supported humans migrating along
the NWC during the Late Wisconsin (Heaton
and Grady 2003; Shafer et al. 2010).
1.2 Sea Level
During the LGM, eustatic (i.e. globally
averaged) sea level was approximately 120 m
below modern levels (Peltier and Fairbanks
2006). Local relative sea level history for SE
Alaska is complicated by the glacial history.
The load and westward ow of the Cordilleran
Glacier on and adjacent to the region produced
dierential subsidence (glacial isostatic
depression) and uplift (in a glacial forebulge).
The complexity of the regional lithospheric
structure precludes simple isostatic calculations;
instead, “hinge” points between uplifting and
subsiding areas are drawn empirically from
local relative sea level data. This requires sea
level curves to be regionalized based on these
hinge areas. Baichtal and Carlson (2010) have
compared the sea level record for the outer
islands of the Alexander Archipelago to that
of Josenhans (et al. 1997) and Hetherington
and Reid (2003) for Haida Gwaii, and nd that
they are similar (Fig. 2). As Haida Gwaii and
the Alexander Archipelago were both formed
by the geologically recent accretion of new
material to the North American continental
margin, it is reasonable to assume that they
have similar lithospheric properties that allow
them to isostatically respond in a comparable
way (Anderson 1991).
The sea level history for Haida Gwaii
was determined using subaerial and marine
cores, which record the transition from fresh to
Figure 3. Change in sea level from 18,100 cal BP (15,000
RCYBP) to 10,160 cal BP (9,000 RCYBP) in 1000 RCY
increments.
Figure 2. Sea level curves.

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
681
marine sediments (Fig. 2: light grey line with
diamonds; Josenhans et al. 1997). At the LGM,
Haida Gwaii was raised due to a combination
of isostatic uplift and eustatic sea level fall. In
the early postglacial, Haida Gwaii lowered as
the forebulge subsided. In the early Holocene,
local relative sea level rose above the modern
sea level by approximately 15 m. Finally, during
the late Holocene, sea levels stabilized (Fedje
et al. 2005, 24). Hetherington (et al. 2004)
calculated the forebulge on Hecata Strait and
Queen Charlotte Sound to be an upwarp of over
100 m, and that it lasted until 9,700 BP. The
total sea level change for Haida Gwaii was 150
m from the LGM to 5000 BP (Fig. 2). There
were some small (few metres) dierences
between the northern and southern areas of
Haida Gwaii (Hetherington and Reid 2003:
Fig. 2 circles).
Baichtal and Carlson (2010) have been
working to develop a sea level curve for the
outer islands of the Alexander Archipelago
(Fig. 2: dark line with squares, Fig. 3). Carlson
(2007) focuses on the elevations
of shell middens to dene sea level
history. She identies three terrace
levels on the west side of POWI
where people could have lived. The
Upper Terraces are 16–21 m above
modern sea level and would have been
occupied during the highest sea level
(9,400 to 5000 RCYBP). The middle
terraces are between 8.5 and 13 m
above modern sea level and would
have been occupied from 5,000 to
2,000 RCYBP. Finally, the lower or
modern terraces are 6–7 m above
modern sea level and would have
been occupied from 2,000 RCYBP to
present. Baichtal and Carlson (2010)
have identied 430 shell middens,
of which 231 have been dated. This
provides reliable dates for the part
of the sea level curve that is above
modern sea level. However, there are
currently only two data points below
modern sea level. The rst is the -165
m point, which Baichtal refers to as a terrace.
Carrara (et al. 2007, 235) indicates there was
a minimum depression of the sea oor in Sitka
Sound of 160 m. Barron et al. (2009) analysed a
dated sediment core (EW0408-11JC) from the
Gulf of Esquibel (Fig. 1). The core documents
the change from freshwater to salt-water
at 10,600 RCYBP based on a radiocarbon
determination run on a marine shell. Applying
the circa -600 year marine reservoir correction
from Fedje et al. (1996), the date of inundation
can be estimated to approximately 10,000
RCYBP or 11,325 to 11,775 cal BP. The core
provides the second point below modern sea
level documenting local sea level rise (Baichtal
and Carlson 2010, Barron et al. 2009). When
sea level was signicantly lower, the Gulf of
Esquibel would have been a fresh water lake
(Baichtal and Carlson 2010).
1.3 Archaeology and Ethnography
The earliest archaeological sites in the
northern NWC demonstrate that the regional
Region Site Component Mean of Calibrated
Age Ranges
SE Alaska
Ground Hog Bay 2 Lower 11,528
Hidden Falls110,157
49 PET 408 (On Your
Knees Cave)
Human
Remains1
11,212
Bone Tool 12,129
Chuck Lake Loc 1
(midden)
9,204
Haida Gwaii
K1 Cave112,650
Lyell Bay East 9,906
South 10,241
Echo Bay19,916
Richardson1 10,442
Arrow Creek 2110,584
Gaadu Din Cave 1112,683
Gaadu Din Cave 2112,480
Werner Bay 12,481
Kilgii Gwaii110,511
BC Mainland Namu (ElSx1) 11,049
1 Average of several mean calibrated age ranges
Table 1. Average calibrated 14C Dates for sites older than 9000 cal
BP (Lee 2007, 32, 46).

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
682
archaeological record extends to the time
of LGM sea level rise. Table 1 depicts mean
calibrated ages for the oldest component at
sites that are older than 9000 cal BP (Fig. 1).
The mean ages of these sites were calculated
following the same method as Lee (2007), but
are updated to use the IntCal 09 calibration
curves (Reimer et al. 2009).
The archaeological implications of sea
level are important. Early researchers primarily
relied on bathymetric maps to evaluate and
explore the sea oor (Bailey and Flemming
2008, 2153). However, the advent of GIS
modelling enables bathymetry to be more
precisely correlated with other data sets, or
layers. In total, there are 7,447 archaeological
sites in the project’s current database. Alaska
State Historic and Preservation Ocer (SHPO)
and BC Archaeology Branch provided the
archaeological site data (AK SHPO 2009, BC
Arch Branch 2009). Of these sites, 1077 are
within the projects study area. These sites are
above mean low water; there are currently
no underwater archaeological sites recorded
within the study area, except for presumed
shipwrecks.
For each of the model parametres, site-
specic values were collect for the entire
region. These values are slope, aspect, distance
from stream, lakes, tributaries, coast and other
archaeological sites, and coastal sinuosity. Each
parametre utilized the modern environment.
The sites within the study area also were
compared to the appropriate temporally
reconstruction parametres for each of the
model inputs. Originally processed in ArcGIS
using the “near” tool in proximity analysis,
the GRASS “v.what.rast” was more ecient
at calculating values from the raster layers
developed for the model. The values calculated
derived the weight and values used in the model
via an inductive modelling approach (Kohler
1988, 37; Verhagen and Whitley 2012).
Ethnographic data were utilized to better
approximate values and weights for the model
utilizing a deductive modelling approach
(Kohler 1988, 37; Verhagen and Whitley 2012).
The purpose of using the ethnographic data in
this research is twofold. First, this information
provides insights for establishing a framework
for environmental conditions that are important
for human subsistence (Maschner 1992). In
other words, similarities in site locations tend
to exist due to basic human needs and resource
distributions; demonstrated by the statistical
analysis of known archaeological sites with the
region. Secondly, ethnographic data provide
insights into the seasonal round as it pertains
to human land use based on the distribution
of resources. NWC people used a system of
ownership, or stewardship, to ensure returns
and adequate storage of resources. They also
responded to the uncertainties of resource
abundance by incorporating an amount of
exibility and widening their resources;
for example, shifting from mainly salmon
to ratsh, shellsh and deer in years when
salmon harvests were not as abundant (Moss
2011,78). This information can be applied to
paleolandscapes to help determine areas most
likely to have been used by people in ancient
times.
2. Methods
The archaeological predictive model
(Fig. 4) was developed in three stages. Stage 1
is the assembly of the ArcGIS database. Stage
2 can be divided into two types of products:
intermediate and nal. Intermediate products
are the raster data sets derived from Stage 1
data. These include slope; aspect; type of coast;
coastal sinuosity; and, distance from coast, fresh
water, tributary junctions, and archaeological
sites. The second stage is derived using metrics
that were computed in GRASS GIS (GRASS
Development Team 2012) and in ArcGIS 10. The
nal products of stage 2 are weighted overlays
that depict high potential areas for the probable
occurrence of archaeological sites based on the
geomorphic characteristics documented for
known sites. The third stage of the modelling

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
683
process is testing. This was conducted in
two ways. The rst is through statistical tests
including cross-validation, Kvamme’s Gain,
and spatial autocorrelation. The second test
was through marine geophysical surveying
and bottom sampling. The surveys included
multibeam sonar, side-scan sonar, sub-bottom
proles, ROV (remotely operated vehicle),
and bottom sampling (Van Veen sampler).
The dierent geophysical survey instruments
provide a view of the sea oor used in a similar
way to an aerial photograph (multibeam and
side scan sonars), and a prole of the sediments
below the sea oor (sub-bottom) similar to
ground penetrating radar. The ROV provides
an interactive view of the seaoor with video
camera and the ability to collect small objects
using a robotic arm. The model incorporates
an iterative process where model values are
rened to reect testing results and can be
regenerated.
2.1 Bathymetry and DEM
Bathymetry is the topography of the sea
oor. Unlike land features, there are few sources
from which to download compiled elevation
data of the surface of sea oor. Initially, a one
arc-second grid le (EPOPO1) was downloaded
from the NOAA website. This le had too
many data errors to be useful for this analysis.
Regional data from early soundings, multibeam
Figure 4. Flow chart of three stages for developing the
model.
Figure 5. Distribution of data points included in the
seamless DEM (both bathymetry and land topography).
Figure 6. Comparison of spline and IDW DEM methods.
A) Spline DEM. B) IDW DEM. C) IDW minus spline
DEMs with land. Comparison was conducted between
the IDW and a Spline version of the DEM to determine if
there would be a signicant difference in results between
the two different methods.

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
684
sonar surveys, and other surveys in the region
were downloaded as xyz les from the hydrology
section of the NOAA website (NOAA 2011).
These les were processed using a Python script
that converted the data to point shape les.
Other bathymetric data were purchased from
Scientic Fishers Inc. (SciFish Inc.). These
included coastal points and bathymetry points
throughout the study area. The NOAA data
were merged in ArcGIS 10.0 with the coastal
and bathymetric data from SciFish Inc. to form
a comprehensive bathymetric dataset used for
modelling.
The adjacent terrestrial topography was
downloaded from the USGS website (NED 2)
(Gesch et al. 2002) These DEMs were merged
to form a single land surface for the study
area with a resolution of 25 m. The DEM
was converted to points at 25 m intervals. To
create the nal seamless DEM including both
bathymetry and land, point les were merged
together to form a large (over 40 million point)
data le (Fig. 5). This le was converted to a
5 m resolution raster using Inverse Weighted
Distance (IWD) in ArcGIS’s 3D analyst toolbox
(Fig. 6b).
The Spline DEM (ArcGIS 10 3D analysis
tool) was generated at 5-metre resolution. Figure
6A is the Spline DEM using the same colour
ramp as gure 6B. Figure 6C was generated
by subtracting the Spline DEM from the IDW
DEM. Land was added as a semi-transparent
green for reference. The two DEMs are very
similar other than a few smaller anomalies. The
larger anomalies are outside the focus of this
study. This comparative analysis demonstrates
the two DEMs are analogous, and the IDW
DEM is adequate for this research.
The IDW DEM and the multibeam data
collected in May 2012 will be compared.
Preliminary analysis indicates that the two
data sets are very similar. Variables for the
archaeological site predictive model were
generated from the IDW DEM.
2.2 Water
An important factor for human settlement
is the distances from streams, lakes, and
tributary junctions. These fresh water-related
variables constitute 35% of the nal high
potential model (Table 4). A suite of topologic
and hydrologic analyses was used to predict
where streams and lakes might have existed
prior to inundation by the ocean.
Drainage paths were calculated from
the IDW DEM using an improved, highly-
ecient least-cost-path search (Metz et al.
2011) in GRASS GIS (GRASS Development
Team 2012). Flow accumulation was calculated
using a constant value for precipitation
minus evapotranspiration of 4×10-5 mm/s,
characteristic of the region based on the TraCE-
21K paleoclimate model (Liu 2009; He 2011). A
0.1 m3/s discharge threshold, based on records
from gaging stations in SE Alaska, were used to
dene the headwaters of streams. For each time
period, all of the stream paths were clipped
to the paleoshorelines. The shorelines were
dened as the edge between the land and open
water. The streams are routed through regions
that were locally below sea level and could
represent brackish wetlands or lakes. Tributary
junctions were determined by nding the ends
of stream segments that joined.
Archaeological site potential around
streams was ranked from 1 (lowest) to 5
(highest) in consecutive buer rings at 100,
500, 1000, 2000, and 3000 m. The 500 m
buer was ranked higher than the 100 m buer
based on statistical analysis of the distance
from archaeological sites in the region. The
median distance from sites to water sources
is 418 m. Higher outlier values could be
petroglyphs, pictographs, or specic resource
patches or other types of sites such as caves or
lithic sources that do not necessarily require
water resources.
Possible lake locations were generated
in a two-step process. First, depressions were

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
685
identied in GRASS GIS with the basin lling
algorithm from the TerraFlow project (Arge,
Toma, and Vitter 2000; Arge et al 2001; Arge et
al 2003). This algorithm typically is used to ll
pits caused by data errors in the DEM, but also
lls enclosed depressions such as lake basins.
The goal of the second step was to extract only
those depressions that were lake basins. The
depressions were classied as lakes based on
their size and lack of compactness, the latter
being given by:
where N is non-compactness, P is
perimetre, and A is area. N equals one for a
circle, and increases as the shape becomes more
oblong and complex (GRASS Development
Team). Lakes tend to be circular, and small
basins are more likely to be DEM artefacts, so
larger and more compact regions are preferred.
Lakes were dened based on a good visually-
estimated match between modern lakes and
those computed by the DEM, which was found
for depressions that had an area >0.5 km2 and
a non-compactness of <5, or an area of >4 km2
and a non-compactness of <2. Archaeological
site potential with distance from lakes was
ranked using the same buer sizes and ranks
as streams.
2.3 Coasts
Coastlines are perhaps the most important
resource procurement areas for the NWC. In
historic and proto-historic times, ethnographic
accounts indicate that the largest native
settlements were located in sheltered bays or
harbours chosen for a good canoe-landing beach
(de Laguna 1960, 30; Mears 1790, 109). Winter
villages containing permanent houses were
situated in locations where “the people prized
a view of the more open water across which the
canoes of their friends or their enemies might
be seen approaching” (de Laguna 1960, 30).
The summer shing camps could be located
further up bays at salmon streams (de Laguna
1960; Moss 1992, 7). Thus, it is not just the
proximity to the coast, but the shape of coast
and character of the oshore environment that
is important.
The coastline was reconstructed at
500 RCY intervals based on the regional sea
level curve using a contour line at 0 m from
the DEM. This contour line was assumed to
approximate the paleocoastline. The coastal
sinuosity algorithm was then applied to the
paleocoastlines. The mean distance of recorded
archaeological sites from the coast is 676 m;
however, 50% of the sites are within 176 m from
the coast and 75% are within 550 m.
Mackie and Sumpter (2005) utilized
shoreline intricacy to analyse settlement
patterns on Haida Gwaii. They dened four
categories of shoreline intricacy: linear,
sinuous, elaborate, and intricate in increasing
complexity. Their premise was that greater
shoreline length near a site would allow for a
more readily accessible intertidal and subtidal
zone, and thus greater potential for food
production and higher biodiversity (Mackie
and Sumpter 2005, 350-351). They conclude
that early sites (9400-9500 BP) existed on
more “elaborate coasts” and late sites (2000-
200 BP) were more evenly distributed between
the shoreline intricacy categories.
Figure 7. Coastal sinuosity is measured by taking the line
length of the coast (lc) and dividing it by the diameter of
the circle (d). The diameter used was 3 km.

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
686
This analysis measured coastal sinuosity
using adjacent 3 km diametre circles along the
reconstructed coast. The line length of the coast
(lc) was divided by the diametre of the circle
(d), which normalizes the values to the chosen
circle of size (Fig. 7). Values were produced for
the study area coastline that range from linear
(1) to sinuous (8) with a mean of 3.55 (Fig.
8). This is a new algorithm developed for this
study.
A proximity analysis (ArcGIS 10: Analysis
Tools, Proximity, near) was run to determine
the distance between the location of the
recorded archaeological sites and coastal
sinuosity values. The mean coastal sinuosity
near archaeological sites is 2.72. A t-test
was run between all of the coastal sinuosity
values and coastal sinuosity values nearest to
archaeological sites. The p-value for the t-test is
less than 0.01, which demonstrates that the two
sinuosity values are dierent. The average or
mean coastal sinuosity in the region is 3.55, but
for archaeological sites is 2.72. This suggests
that people may have selected less complex
coastlines for settlement and subsistence
purposes. The coastal sinuosity values for the
region were classied into high, medium, and
low (Table 2) based on the proximity analysis
from archaeological sites to coastal sinuosity.
Distance from the paleoshoreline was combined
with the coastal sinuosity variable to form a
single ranked variable within the model. The
high, medium and low ranks were then buered
in 100, 500, 1000, and 2000 m groups. (See
Table 4 for the ranked values.)
2.4 Other Variables
From the DEM, slope and aspect were
generated using ArcGIS’s 3D analyst. The best
resolution for computing slope and aspect was
found to be 10 m. Table 3 has the ranking of
the slope and aspect used in the high potential
model.
The nal variable used in the model was
distance between archaeological sites. Distance
from known archaeological sites was buered
and ranked in 100, 500, 1000, and 5000 m.
The highest weight was given to the 500 m
bin. Using the concept of home range, distance
from known archaeological sites is used as a
variable to increase the weight for areas known
to contain sites. Twenty-ve percent of the sites
in the study area are within 70 m of another site
and 50% of the sites are within 416 m.
Value Rank
< 1.5 Low
1.5 - 2.5 High
2.51 - 4.5 Medium
> 4.5 Low
Slope
Degrees Value
0- 2 ° 5
2-5 ° 4
5-10 ° 3
10-20 ° 2
20 °+ 1
Aspect 130°-275° 2 W, SW, S, SE
276°-129° 1 E, NE, N, NW
Figure 8. Coastal sinuosity of Shakan Bay area at 11,000
RCYBP.
Table 2. Coastal complexity
ranks in relation to the
location of recorded sites.
Table 3. Example of ranking of variables (based on
Maschner 1992).

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
687
Variable
Type
Variable Degrees Value Weights Metadata
Landscape
Slope 0 - 2 5 30% Calculated from DEM at 10 m resolution;
values based on Maschner (1992)
2.001 - 5 4
5.001 - 10 3
10.001 - 20 2
20.001 + 1
Aspect 130 - 275 2 5% Calculated from DEM at 10 m resolution;
values based on Maschner (1992)
275.001
-129.9999
1
meters
Water
Distance from streams 100 4 20% Calculated in GRASS; values based on
statistical analysis of archaeological sites
within region (at appropriate coastline)
500 5
1000 3
2000 2
3000 1
Distance from lakes 100 4 10% Calculated in GRASS; values based on
statistical analysis of archaeological sites
within region (at appropriate coastline)
500 5
1000 3
2000 2
3000 1
Distance from coast H 100 9 9% Calculated in GRASS; values and ranking
(high, medium, or low) based on statistical
analysis of archaeological sites within region
(at appropriate coastline). (Totals 18%)
500 8
1000 5
2000 4
M 100 8 6%
500 7
1000 4
2000 3
L 100 7 3%
500 6
1000 2
2000 1
Distance from tributary 100 5 5% Calculated in GRASS; values based on
statistical analysis of archaeological sites
within region (at appropriate coastline)
500 4
1000 3
2000 2
3000 1
Archaeological
Distance from archaeological
sites
100 3 12% Calculated in ArcGIS; values based on
statistical analysis of distance from nearest
known archaeological site (inuences the
model towards current land surfaces)
500 4
1000 2
5000 1
Total 100%
Table 4. Weights used to create nal high potential model.

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
688
2.5 High Potential Models
Each of the variables was converted
into raster format, incorporating the buers
reclassied by the associated value (Table 4).
In ArcGIS 10.0 spatial analyst, the weighted
overlay tool was used to create high potential
models for each time slice. The weights used
were a combination of values derived from the
statistical analysis of the known site locations in
the region and the study area, the ethnographic
literature, and from other models (Hamilton
and Larcombe 1994; Maschner 1992). The nal
values of the model ranged from 0 to 4. Based
on the distribution of results, moderately high
potential was determined to be value 3 and high
potential was determined to be value 4 (Fig. 9
a-c).
The nal results were added using ArcGIS
spatial analyst math tool. This produced a
combined model that covered the entire study
area from 10,790 to 18,100 cal BP (9,500 to
15,000 RCYBP). The results ranged from 0 to
44 (Fig. 9 b-d). Moderately high potential was
determined to be above 23, and high potential
was determined to be above 30. These values
are slightly below the 20% threshold used in
the individual models to account for the lower
values further out on the continental shelf. This
nal model accurately incorporates the coasts,
lakes, rivers, and archaeological sites in high
potential areas. Both the individual temporal
intervals (in 500 RCY intervals) and the
combined model are being tested and utilized.
3. Results and Discussion
To test the model statistically, Kvamme’s
Gain statistic was utilized (Kvamme 1988,
329; Mink, Stokes, and Pollack 2006, 235).
The known 1077 sites and a 1000 randomly
generated collection of points were used to
calculate gain statistics (Table 5) on the modern
environmental variables. All of the known
archaeological site locations produced positive
gain values. The random locations produced
negative gain values for the moderately high
potential and the combined high potential
(values 3 and 4). The random locations did
not produce any values for high potential
sites (value 4). The combined model was not
utilized in the gain test because it was created
for locations that are currently underwater; this
means that there was little, if any, gain using
sites that are on the modern land surface. The
Model
Values
Data Set Gain
Statistic
Predictive
Utility (gain)
3 Known sites 0.5300 Positive
Random
locations
-3.5361 Negative
4 Known sites 0.9895 Positive
Random
locations
- None
3+4 Known sites 0.5352 Positive
Random
locations
-3.5373 Negative
Figure 9. Model results. A) Map of Shakan Bay at 11,000
RCYBP as an example of a time slice results. B) Map of
Shakan Bay with the sum of all the model results showing
nal high potential model. C) Histogram of model results
for 11,000 RCYBP. D) Histogram of model results for all
time slices combined. The histograms show the full values
of the weighted overlay.
Table 5. Gain statistics for known archaeological sites
and random points.

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
689
goal of the iterative process is continually to
rene the model, increasing predictability and
eliminating additional areas of lower potential.
Underwater predictive modelling
presents several unique challenges. The
greatest challenge was to develop a DEM from
various sources and eld-testing the DEM.
Field survey is one of the primary means of
testing archaeological settlement models,
but is dicult and expensive in underwater
applications. This model investigates depths
down to 160 m (525 ft), recreational diving
is safe only to 43 m (140 feet) and even then
divers only stay at that depth for seven minutes.
Technical and commercial divers have dierent
limits and requirements, but training and
equipment can be very expensive. In addition to
the logistical constraints for recreational divers,
the state of Alaska requires special training for
archaeological SCUBA diving. SCUBA is not a
viable option for this project, and survey must
be conducted using geophysical instruments,
ROVs, and bottom sampling.
Geophysical surveying is not considered
equivalent to pedestrian surveys in terrestrial
archaeology, and it is not equivalent in submerged
archaeology either. Multibeam sonar and sub-
bottom proler provide information equivalent
to similar instruments employed in terrestrial
geophysics. The results of the surveys must
be ground truthed and/or veried using other
methods. It is essential to include the means to
verify and evaluate anomalies when conducting
submerged archaeology at depths below the
limits of conventional diving, such as using
ROV video camera investigation and sampling.
Like the geophysical surveys, the ROV requires
specialized technicians and is limited. The
particulate matter in the water column distorts
views, and the eld of view is often very small,
especially in deep, dark locations. These make
mosaicking imagery from the ROV dicult.
The ROV’s mechanical arm recovers only small
objects from the sea oor, and the Van Veen
samplers used collect no more than the top 25
centimetres of sediment. The costs associated
with underwater archaeology are high,
especially for archaeological budgets. Support
vessels capable of deploying an ROV and Van
Veen sampling are often too large for the slow
speed required for multibeam, sidescan, and
subottom sonar surveys and often multiple
vessels are required. Research conducted in
remote areas further increase logistic costs.
Because marine archaeological surveying
is expensive, models are needed to optimize
research eorts. Models to locate submerged
archaeological sites have been produced
in other parts of the world (Dixon 1979;
Ejstrud 2003; Evans and Keith 2011; Faught
2004; Fedje and Christensen 1999; Ganey,
Thomson and Fitch 2007; Momber 2000;
Ruppe 1988). As modelling progresses, it
has become increasingly apparent that they
must be regionally specic and tailored to the
geologic and paleoecological parametres of the
region for which they are being developed. This
requires generating precise regional DEMs,
sophisticated understanding of local land/sea
level relationships, in-depth knowledge of the
regional archaeology, and accurate paleographic
and paleoecological reconstructions.
The model has already proven useful for
identifying specic locales for eld survey.
Although no conclusive archaeological sites
have been located, several anomalies have been
identied that require additional eld-testing
to determine conclusively whether or not they
are archaeological sites. The iterative process
is designed to increase accuracy and eciency
over time, with the ultimate goal to identify
submerged sites, and extend the archaeological
record seaward, into early Holocene and late
Wisconsin times.
Acknowledgements
This research was supported by the
National Science Foundation, Oce of Polar
Programs award numbers 0703980 and
1108367. The authors would also like to

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
690
thank Sealaska Heritage Institute and the two
anonymous reviewers.
References
Ager, T. A., and J. G. Rosenbaum. 2007. “Late Glacial-
Holocene Pollen-Based Vegetation History from Pass
Lake, Prince of Wales Island, Southeastern Alaska.”
Professional Paper 1760-G: Studies by the US Geological
Survey in Alaska.
Ager, T. A., P. E. Carrara, J. L. Smith, V. Anne, and J.
Johnson. 2010.“Postglacial vegetation history of Mitkof
Island, Alexander Archipelago, Southeastern Alaska.”
Quaternary Research 73 (2): 259-268.
AK SHPO. 2009. “Alaskan Sate Historic Preservation
Ocer - archaeological site database.” Accessed January
2009.
Ames, K. M., and H. D. G. Maschner. 1999. Peoples of the
Northwest Coast: Their Archaeology and Prehistory.
New York: Thames and Hudson.
Anderson, R. G. 1991. “Contributions to the geology
and geophsyics of northwestern British Columbia and
southeastern Alaska.” Canadian Journal of Earth
Sciences 28: 835-839.
Antevs, E. 1929. “Maps of the Pleistocene Glaciations.”
Bulletin of the Geological Society of America 40: 631-720.
Arge, L., J. S. Chase, P. N. Halpin, L. Toma, J. S. Vitter, D.
Urban, and R. Wickremesinghe. 2001. “Flow computation
on massive grids.” In Proceedings of the ACM Symposium
on Advances in Geographic Information Systems, edited
by W. G. Aref. Atlanta.
Arge, L., J. S. Chase, P. N. Halpin, L. Toma, J. S. Vitter, D.
Urban, and R. Wickremesinghe. 2003. “Flow computation
on massive grid terrains.” GeoInformatica, International
Journal on Advances of Computer Science for Geographic
Information Systems 7 (4): 283-313.
Arge, L., L. Toma, and J. S. Vitter. 2000. “I/O-ecient
algorithms for problems on grid-based terrains.”
In Proceedings of the 2nd Workshop on Algorithm
Engineering and Experimentation. ALENEX .
Baichtal, J. F., and R. J. Carlson. 2010. “Development
of a Model to Predict the Location of Early Holocene
Habitation Sites along the Western Coast of Prince of
Wales Island and the Outer Islands, Southeast Alaska.”
Paper Presented at the Alaskan Anthropology Association
37th Annual Meeting, Anchorage.
Bailey, G. N., and N. C. Flemming. 2008. “Archaeology
of the continental shelf: Marine resources, submerged
landscapes and underwater archaeology.” Quaternary
Science Reviews 27: 2153-2165.
Barron, J. A., D. Burky, W. E. Dean, J. A. Addison,
and B. Finney. 2009. “Paleoceanography of the Gulf
of Alaska during the past 15,000 years: Results from
diatoms, silicoagellates, and geochemistry.” Marine
Micropaleontology 72: 176-195.
BC Arch Branch. 2009. British Columbia Archaeology
Branch, Victoria, BC. Victora.
Blaise, B., J. J. Clague, and R. W. Mathewes. 1999. “Time
of Maximum Late Wisconsin Glaciation, West Coast of
Canada.” Quaternary Research 34: 282-295.
Carlson, R. 2007. “Current Models for the Human
Colonization of the Americas: the Evidence from Southeast
Alaska.” MA Diss. Cambridge University.
Carrara, P. E., T. A. Ager, and J. F. Baichtal. 2007. “Possible
refugee in the Alexander Archipelago of southeastern
Alaska during the late Wisconsin glaciation.” Canadian
Journal of Earth Sciences 44: 229-244.
Clague, J. H., R. W. Mathewes, and T. A. Ager. 2004.
“Environments of Northwestern North America before
the Last Glacial Maximum.” In Entering America:
Northeast Asia and Beringia before the Last Glacial
Maximum, edited by D. B. Madsen, 63-94. Salt Lake City:
The University of Utah Press.
de Laguna, F. 1960. The Story of a Tlingit Community:
A Problem in the Relationship between Archaeological,
Ethnological, and Historical Methods. Washington:
Bureau of American Ethnology.

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
691
Dixon, E. J. 1979. “A predictive model for the distribution
of archaeological sites on the Bering Continental Shelf.”
PhD Diss. Brown University.
Dixon, E. J. 1999. Bones, Boats, and Bison: Archaeology
and the First Colonization of Western North America.
Albuquerque, University of New Mexico Press.
Ejstrud, B. 2003. “Modelling the coast. Shoreline
displacement in Vendsyssel since the Early Atlantic
transgression.” Working paper. Accessed April 2012.
http://web.sdu.dk/ejstrud/forskning/GIS/coastmodel.
html#data.
Esposito, G. S., and J. L. Smith. 2007. “Cultural Resource
Survey in the Navy Timber Sale Project Area, Etolin
Island, Alaska.” Wrangell: MS on le, Tongass National
Forest.
Evans, A., and M. E. Keith. 2011. “Potential Contriutions
of a Maritime Cultural Landscape Approach to Submerged
prehistoric Resources, Northwestern Gulf of Mexico.” In
The Archaeology of Maritime Landscapes, edited by Ben
Ford, 163 -178. New York: Springer.
Faught, M. K. 2004. “The Underwater Archaeology of
Paleolandscapes, Apalachee Bay, Florida.” American
Antiquity 69 (2): 275-289.
Fedje, D. W., and H. Josenhans. 2000. “Drowned forests
and archaeology on the continental shelf of British
Columbia, Canada.” Geology 28 (2): 99-102.
Fedje, D. W., and R. W. Mathewes (ed). 2005. Haida
Gwaii: Human History and Environment from the
Time of Loon to the Time of the Iron People. Vancouver:
University of British Columbia Press.
Fedje, D. W., and T. Christensen. 1999. “Modeling
paleoshorelines and locating early Holocene coastal sites
in Haida Gwaii.” American Antiquity 64 (4): 635-652.
Fedje, D. W., H. Josenhans, J. J. Clauge, J. Vaughn Barrie,
D. J. Archer, and J. R. Southon. 2005. “Hecate Strait
Paleoshorelines.” In Haida Gwaii: Human History and
Environment from the Time of Loon to the Time of the
Iron People, edited by D. W. Fedje and R. W. Mathewes,
21-37. Vancouver: University of British Columbia Press.
Fedje, D. W., Q. Mackie, D. McLaren, and T. Christensen.
2008. “A projectile point sequence for Haida Gwaii.”
In Projectile Point Sequences in Northwestern North
America, edited by R. L. Carlson and M. P. R. Magne, 19-
40. Burnaby, Simon Fraser University Archaeology Press
Publication.
Ganey, V., K. Thomas, and S. Fitch (ed). 2007. Mapping
Doggerland: The Mesolithic Landscapes of the Southern
North Sea. Oxford: Archaeopress.
Gesch, D., M. Oimoen, S. Greenlee, C. Nelson, M. Steuck,
and D. Tyler. 2002. “The National Elevation Dataset.”
Photogrammetric Engineering and Remote Sensing 68
(1): 5-11.
GRASS Development Team. 2012. “Geographic Resources
Analysis Support System (GRASS) Software, Version
6.4.1.” Accessed June 2012. http://grass.osgeo.org:Open
Source Geospatial Foundation.
Hamilton, S., and L. Larcombe. 1994. “Cultural Hertiage
Resource Predictive Modelling Project: vol 1: Introduction
to the Research.” Thunder Bay, Report Prepared for
the Ontario Ministry of Natural Resources, Lakehead
University.
He, F. 2011. “Simulating Transient Climate Evolution of
the Last Deglaciation with CCSM3.” PhD Diss. University
of Wisconsin.
Heaton, T. H., and F. Grady. 2003. “The Late Wisconsin
Vertebrate History of Prince of Wales Island, Southeast
Alaska.” In Ice Age Cave Faunas of North America, edited
by B. W. Schubert, J. I Mead, and R. W. Graham, 17-53.
Bloomington: Indiana University Press.
Hetherington, R., and R. G. B. Reid. 2003. “Malacological
insights into the marine ecology and changing climate of
the late Pleistocene – early Holocene Queen Charlotte
Islands archipelago, western Canada, and implications for
early peoples.” Canadian Journal of Zoology 81: 626-661.
Hetherington, R., J. Vaughn Barrier, R. G. B. Reid, R.
MacLeod, and D. J. Smith. 2004. “Paleogeography,
glacially induced crustal displacement, and Late
Quaternary coastlines on the continental shelf of British
Columbia, Canada.” Quaternary Science Reviews 23:

CAA2012 Proceedings of the 40th Conference in Computer Applications and Quantitative Methods in Archaeology,
Southampton, United Kingdom, 26-30 March 2012
692
295-318.
Johnson, J. A., B. A. Andres, and J. A. Brissonette. 2008.
“Birds of the Major Mainland Rivers of Southeast Alaska.”
General Technical Report PNW-GTR-739: United States
Deparment of Agriculture.
Josenhans, H., D. Fedje, R. Pienitz, and J. Southon. 1997.
“Early Humans and Rapidly Changing Holocene Sea
Levels in the Queen Charlotte Islands - Hecate Strait,
British Columbia, Canada.” Science 277: 71-74.
Kaufman, D. S., and W. F. Manley. 2004. “Pleistocene
Maximum and Late Wisconsinan Glacier Extents across
Alaska, U.S.A..” In Quaternary Glaciations Extents and
Chronology Part 2: North America, edited by J. Ehlers
and P. L. Gibbards , 9-27. Amsterdam: Elsevier.
Kohler, T. A. 1988. “Predictive Locational Modeling:
History and Current Practice.” In Quantifying the Present
and Predicting the Past: Theory, Method, and Application
of Archaeological Predictive Modeling, edited by W. J.
Judge and L. Sebastian, 19-96. Denver: US Department of
the Interior, Bureau of Land Management.
Kvamme, K. L. 1988. “Development and Testing of
Quantitative Models” In Quantifying the Present and
Predicting the Past: Theory, Method, and Application
of Archaeological Predictive Modeling, edited by W. J.
Judge and L. Sebastian, 325-428. Denver: US Department
of the Interior, Bureau of Land Management.
Lee, C. M. 2007. “Origin and Function of Early Holocene
Microblade Technology in Southeast Alaska, USA.” PhD
Diss. University of Colorado.
Liu, Z., B. L. Otto-Bliesner, F. He, E. C. Brady, R. Tomas,
P.U. Clark, A.E. Carlson, J. Lynch-Stieglitz, W. Curry, E.
Brook, D. Erickson, R. Jacob, J. Kutzbach, and J. Cheng,
2009. “Transient simulation of last deglaciation with a
new mechanism for Bolling-Allerod warming.” Science
325: 310-314.
Mackie, A. P., and I. D. Sumpter. 2005. “Shoreline
Settlement Patterns in Gwaii Hanas during the Early and
Late Holocene.” in Haida Gwaii: Human History and
Environment from the Time of Loon to the Time of the
Iron People, edited by D. W. Fedje and R. W. Mathewes,
337-371. Vancouver: University of British Columbia Press.
Mann, D. H. 1986. “Wisconsin and Holocene glaciation of
southeast Alaska.” In Glaciation in Alaska—The Geologic
Record, edited by T. D. Hamilton, K. M. Reed, and R. M.
Thorson, , 237-265. Anchorage: Alaska Geological Society.
Mann, D. H., and T. D. Hamilton. 1995. “Late Pleistocene
and Holocene Paleoenvironments of the North Pacic
Coast.” Quaternary Science Reivews 14: 449 - 471.
Maschner, H. D. G. 1992. “The origins of hunter and
gatherer sedentism and political complexity: A case study
from northern Northwest Coast.” PhD Diss. University of
California.
Mears, J. 1790. Voyages made in the years 1788, and 1789
from China to the north west coast of America. London:
Logographic press.
Metz, M., H. Mitasova, and R. S. Harmon. 2011. “Ecient
extraction of drainage networks from massive, radar-
based elevation models with least cost path search.”
Hydrology and Earth System Sciences 15(2): 667-678.
Momber, G. 2000. “Drowned and deserted: A submerged
prehistoric landscape in the Solent, England.”
International Journal of Nautical Archaeology 34 (1):
86-99.
Moss, M. L., “Relationships between maritime cultures
of Southern Alaska: Rethinking culture area boundaries.”
Arctic Anthropology 29 (2) (1992), 5-17.
Moss, M. L. 1998. “Northern Northwest Coast Regional
Overview.” Arctic Anthropology 35 (1): 88-111.
Moss, M. L. 2011. Northwest Coast: Archaeology as Deep
History. Washington: Society of American Archaeology.
NOAA National Geophysical Data Center. 2011. NOS
Hydrographic Survey Data. Accessed April 2009. http://
www.ngdc.noaa.gov/mgg/bathymetry/hydro.html, 2011.
O’Claire, R. M., R. H. Armtsrong, and R. Carstensen.
1992. The Nature of Southeast Alaska. Anchorage: Alaska
Northwest Books.

Lost Worlds: A Predictive Model to Locate Submerged Archaeological Sites in SE Alaska, USA
Kelly R. Monteleone, E. James Dixon and Andrew D. Wickert
693
Peltier, W. R., and R. G. Fairbanks. 2006. “Global glacial
ice volume and Last Glacial Maximum duration from an
extended Barbados sea level record.” Quaternary Science
Reviews 25 (23-24): 3322-3337.
Reimer, P. J., M. G. L. Baillie, E. Bard, A. Bayliss, J. W.
Beck, C. Bertrand, P. G. Blackwell, C. E. Buck, G. Burr, K.
B. Cutler, P. E. Damon, R. L. Edwards, R. G. Fairbanks, M.
Friedrich, T. P. Guilderson, K. A. Hughen, B. Kromer, F. G.
McCormac, S. Manning, C. Bronk Ramsey, R. W. Reimer,
S. Remmele, J. R. Southon, M. Stuiver, S. Talamo, F. W.
Taylor, J. van der Plicht, and C. E. Weyhenmeyer. 2009.
“calib 6.0.” Radiocarbon 51.
Ruppe, R. J. 1988. “The location and assessment of
underwater archaeological sites.” In Wet site archaeology,
edited by B.A. Prudy, 55-68. Telford: Telford Press.
Schier, M. B. 1972. “Archaeological Context and Systemic
Context.” American Antiquity 37 (2): 156-165.
Shafer, A. B. A., C. I. Cullingham, S. D. Cote, and D. W.
Coltman. 2010. “Of glaciers and refugia: A decade of study
sheds new light on the phylogeography of northwestern
North America.” Molecular Ecology 19: 4589-4621.
Smith, J. L. 2011. “An update of Intertidal shing structures
in Southeast Alaska.” Alaska Journal of Anthropology 9
(1): 1-26.
Smith, J. L., and G. S. Esposito. 2002. Cultural Resource
Investigations in The Scott Peak Project Area, Kupreanof
Island, Alaska. Petersburg Ranger District: Tongass
National Forest.
USGS Earthexplorer Server. 2009. Accessed April 2009.
http://edcsns17.cr.usgs.gov/EarthExplorer/.
Verhagen, P., and T. G. Whitley. 2012. “Integrating
Archaeological Theory and Predictive Modeling: A Live
Report from the Sciene.” Journal of Archaeological
Method and Theory 17 (4): 1-52.