Content uploaded by Isaac Ullah
Author content
All content in this area was uploaded by Isaac Ullah on May 31, 2016
Content may be subject to copyright.
331
Advances in Archaeological Practice 3(4), 2015, pp. 331–350
Copyright 2015© The Society for American Archaeology
DOI: 10.7183/2326-3768.3.4.331
331
ABSTRACT
The data from older archaeological surveys are incredibly important resources, often containing our only information about sites that
have been destroyed or that are now inaccessible. These surveys occurred before the advent of GPS technology, however, so their
spatial accuracy is often uncertain. Many types of locational errors accumulate in such “legacy” datasets, so using them in modern
GIS-based spatial analyses is frequently problematic. Many of the sources of error can be identified and quantified, however, and
systematic and random errors (derived mainly from Cartesian, rounding, and human error) can largely be mitigated by scanning the
original field maps, georectifying the maps to trusted imagery, and then digitizing sites directly. The remaining “mislocation” errors
derive from difficulty identifying locations in the field. The original survey notes may contain clues about mislocation error, but it is
impossible to mitigate these errors without re-recording site locations with more accurate survey instruments. Instead, I advocate the
use of GIS-based models to estimate the influence of specific surveying practices on site location accuracy. These models can provide
a standardized, quantifiable measure of mislocation error in a legacy dataset, which can help guide its use in modern GIS analyses
that require accurate site locations.
Los datos de prospecciones arqueológicas antiguas son recursos de una enorme importancia, pues contienen a menudo la única
información disponible sobre yacimientos que han sido destruidos, o que resultan inaccesibles en la actualidad. Sin embargo,
estas prospecciones tuvieron lugar antes de la llegada de la tecnología GPS, por lo que su precisión espacial es a menudo incierta.
Los tipos de errores de localización que se acumulan en este tipo de datos “heredados” son numerosos, por lo que su uso en
análisis espaciales modernos basados en SIG es, con frecuencia, problemático. No obstante, muchas de las causas de estos errores
pueden ser identificadas y cuantificadas y, tanto los errores sistemáticos como aleatorios (derivados principalmente del Cartesiano,
el redondeo y los errores humanos), pueden ser mitigados en gran medida por medio del escaneo de los mapas de campo
originales, su georrectificado a partir de imágenes de confianza, y la posterior digitalización de los sitios. Los restantes “errores
de ubicación” tienen su origen en la dificultad a la hora de situar los puntos en el campo. Los diarios y anotaciones originales de
las prospecciones pueden contener pistas sobre dichos errores de ubicación, pero resulta imposible mitigar su efecto sin volver a
registrar las ubicaciones de los yacimientos con instrumentos más precisos. En su lugar, yo abogo por el uso de modelos basados
en SIG para estimar la influencia de determinados métodos específicos de exploración en la precisión de la localización de los
sitios. Estos modelos pueden proporcionar una medida estandarizada, cuantificable de los errores de ubicación en un conjunto de
datos heredado, lo que puede ayudar a guiar su uso en los análisis modernos de SIG, que requieren ubicaciones exactas de los
yacimientos.
Integrating Older Survey
Data into Modern Research
Paradigms
Identifying and Correcting Spatial Error
in “Legacy” Datasets
Isaac I. T. Ullah
332
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Integrating Older Survey Data into Modern Research Paradigms (cont.)
The ability to locate a site in the eld easily,
accurately, and precisely is taken for granted in
modern archaeological survey. Global Positioning
System (GPS) site location data allow for high-
resolution spatial analyses with Geographic
Information System (GIS) software, enabling
research on settlement patterns, predictive
modeling, and human interaction with (and
through) the environment (e.g., Barton et al. 2010;
Llobera 2003; Ullah 2011; White and Surface-
Evans 2012). However, these analyses require that
site location data be accurate and reasonably
precise in order to produce meaningful results. It is
important to note that accuracy and precision are
not the same thing. The accuracy of a site location
measurement is the degree to which recorded
coordinates actually match the location of the
surveyor when they were recorded (Goodchild
1993; Goodchild and Gopal 1989). The precision
of a site location measurement is the range of
error induced from the method of recording site
coordinates (Goodchild 1993). It is important to
note that a survey dataset can contain individually
imprecise site locations, but still be quite accurate
on the whole, and vice versa.
There is a large body of knowledge about the effects of error in
certain types of base GIS data used in archaeological research,
such as rasterized elevation models (Goodchild and Gopal
1989; Stanislawski et al. 1996) or categorical types of vectorized
data (Goodchild et al. 1992). Also well-studied are the effects
of algorithmic error in specic GIS procedures used by archae-
ologists, such as viewshed analysis (Maloy and Dean 2001) or
vector buffer-generation (Shi et al. 2003). There is not, however,
a comprehensive treatment of the propagation of error from
the way the spatial record of archaeological survey data is col-
lected. However, Ullah and Bergin (2012) have shown that small
differences in site location can lead to widely different outcomes
in least-cost path analyses in a GIS. The same is likely true of vis-
ibility analyses, site catchment analysis, and a host of other GIS-
based analyses commonly used by archaeologists (see Wheatley
and Gillings 2002).
The availability of highly accurate GPS coordinates for archaeo-
logical sites is a recent phenomenon. Although the United
States (U.S.) government opened up the military’s “Navstar-
GPS” system to civilian use in 1993, it intentionally scrambled
that GPS signal until May 2, 2000 (Bray 2014; Theiss et al. 2005).
1
Prior to this date, site coordinates were recorded in the eld
from topographic maps and aerial photographs. Their accuracy
was subject to the abilities of eld surveyors to visually cor-
relate landmarks between the real world and their representa-
tions on these media.
2
Doing so, however, is a non-automated
mental operation that involves the use of basic surveying tools
and a knowledge of geometry, as well as intuition and experi-
ence (Golledge 1999). It combines mathematical measurement
with the experience of being in a landscape, and site loca-
tions recorded in this manner are products of this convolu-
tion. Despite the (not insignicant) level of technical expertise
possessed by early archaeological surveyors, site coordinates
recorded in the pre-GPS era are subject to a host of errors
beyond their control. Complicating matters further, older,
inherited datasets are typically manipulated, reorganized, and
transformed by every archaeologist who uses them. As the
datasets pass through generations of archaeologists, knowledge
of previous manipulation can be lost—especially after they are
(digitally) archived. Thus, in the form that they are typically inher-
ited, legacy spatial datasets can be unreliable for use in research
requiring accurate spatial information.
It may be tempting to simply write off legacy datasets, but in
many cases older surveys may contain the only existing record
for sites that have since been destroyed or obscured by sub-
sequent human or natural disturbance (Holtorf 2001; Witcher
2008). Or, due to political or environmental conditions, it may
now be impossible or impractical to revisit the survey region to
re-record site locations with modern techniques (e.g., Galaty
and Watkinson 2004; Stone and Zimansky 1992). Even if a revisit
is possible, budgetary constraints might restrict the amount of
re-recording that can be accomplished. Thus, we may have no
other choice but to use legacy datasets in many parts of the
world.
The purpose of this paper is to discuss how error was intro-
duced into spatial records from pre-GPS archaeological survey,
to discuss methods of correcting some of these errors, and to
introduce a method for modeling the amount of error that may
remain in the corrected data. The procedures laid out herein are
designed to bring legacy survey data into modern GIS-based
spatial research to the fullest extent possible, while remain-
ing explicitly mindful of their limitations. I focus on pedestrian
surveys and less obtrusive sites, mainly from the prehistoric peri-
ods. I do not discuss issues of survey goals, sampling methodol-
ogy, biases in identication of certain types of sites, or survey
methodology unrelated to the gathering of site coordinates, as
these issues have been dealt with elsewhere (e.g., Athanasso-
poulos and Wandsnider 2004; Banning 1996; Rosen 1992; Wilkin-
son 2004; Witcher 2008). Issues of data-mining or integrating
the non-spatial portions of legacy data are likewise dealt with
by other authors (e.g., Atici et al. 2013; Witcher 2008). Finally,
although I do not focus on the curation of legacy datasets in
online or other digital repositories, there are several projects
that are actively doing so, such as the Comparative Archaeology
Database (Drennan et al. 2014), Open Context (Kansa 2010), the
Digital Archaeological Record (Kintigh 2006), and the Digital
Index of North American Archaeology (Wells et al. 2014), among
others.
Legacy Survey Data from Central Jordan
Throughout the paper, I will reference several “legacy” survey
datasets from central Jordan, including the Archaeological
333
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Survey of the Kerak Plateau (ASKP) (Miller 1991), the Southern
Ghors and Northeast Arabah Survey (SGNAS) (MacDonald and
Amr 1992; MacDonald et al. 1988), the Tala-Busayra Archaeo-
logical Survey (TBAS) (MacDonald et al. 2000; MacDonald et
al. 2004; MacDonald et al. 2001),
3
the Archaeological Survey of
Central Moab (ASCM) (Miller 1979), and the Southeastern Expe-
dition to the Dead Sea Plain (SEEDSP) (Rast et al. 1980; Rast and
Schaub 1981; Schaub and Rast 1984) (Figure 1). I pay particular
attention to survey data from the Wadi al-Hasa in central Jordan
(Figure 2), which I use as a more detailed case study to exem-
plify the proposed methodology. Wadi al-Hasa is an excellent
example because it was intensively surveyed in the pre-GPS era
by two separate teams, and the data has been used by several
subsequent researchers. The rst intensive pedestrian surveys
in Wadi al-Hasa were carried out by a team of archaeologists in
the Wadi el Hasa Archaeological Survey (WHS) led by Burton
MacDonald in the 1978, 1981, and 1982 eld seasons (MacDon-
ald 1982, 1988; MacDonald et al. 1980), which covered much of
the southern bank of the Wadi and associated uplands. Further
survey was carried out by Geoffrey A. Clark and colleagues on
the northern bank of the Wadi during the Wadi Hasa North Bank
Survey (WHNBS) in 1992 and 1993 (Clark et al. 1992; Clark et al.
1994), as part of the larger Wadi Hasa Paleolithic Project. Many
of the publications related to these projects were gathered
together into a two-volume set by Nancy Coinman (Coinman
1998; Coinman 2000), and the data have been used by several
subsequent researchers (e.g., Arıkan 2009, 2012; Hill 2002, 2004,
2006; Olszewski and Coinman 1998; Schuldenrein and Clark
1994, 2003). In fact, this paper was in part inspired by difculties
I encountered during my own eldwork in Wadi al-Hasa, where
I attempted to relocate several potential Neolithic sites located
by the earlier surveys during the Wadi al-Hasa Ancient Pastoral-
ism Project (WHAPP) (Ullah et al. 2008).
I planned the 2008 pilot season of the WHAPP using a digital
database of the WHS and WHNBS survey data that I had inher-
ited through connections at Arizona State University (ASU). The
database had been originally compiled by J. Brett Hill for his
own dissertation work (Hill 2002). Hill had painstakingly digitized
all of information from the original site forms of the two surveys,
including site type, likely time periods of occupation, landscape
features, notes on artifact collection, and general notes about
the site and its location. Importantly, the database included site
coordinates in the Universal Transverse Mercator (UTM) Coordi-
nate Reference System (CRS), which, I believed, would make it
easy to relocate sites with a hand-held GPS unit.
The goal of the WHAPP survey was to better understand the
pastoralist-landscape interactions east of the Jordan Rift during
the Neolithic and Chalcolithic periods using high-resolution GIS
analysis and Agent-Based modeling. The research plan of the
WHAPP required integrating the existing survey data from Wadi
al-Hasa with new survey to be conducted in the desert east of
the Hasa drainage. In the summer of 2008, I undertook a short,
two-week pilot season focused on relocation and reinvestigation
of 16 of the previously discovered potential Neolithic pastoral-
ist sites. I exported the UTM coordinates for these 16 sites from
the database into a consumer-grade GPS unit (a Garmin eTrex
Summit
®
), which I brought into the eld. Expecting some degree
of inaccuracy in the coordinates, I also brought the survey notes
and high-resolution satellite imagery into the eld. Despite
these preparations, I was able to “relocate” only eight sites in
two weeks of intensive eldwork. Later analysis would show that,
in reality, only one of those eight sites was actually a site from
my list. Of the remainder, ve were other sites identied by the
WHS or WHNBS, and two were previously unknown. The issue
was four-fold: (1) the coordinates from the database were clearly
inaccurate; (2) I did not know the amount of error to expect prior
to entering in the eld; (3) the topography of some parts of the
region is fairly uniform, making landmark reference difcult; and
(4) the density of sites in some parts of the region made it very
difcult to know that a specic site had been (re)located.
The difculties experienced during the pilot season made it
clear that a primary preliminary goal of the WHAPP must be
to assess the accuracy of the inherited spatial dataset for the
region, apply any possible corrections to this data, and ascertain
the degree to which even the corrected data could be used for
high-resolution research. The spatial record of the Wadi al-Hasa
surveys are therefore good case studies with which to exemplify
the best practices of addressing spatial error in any legacy data-
set. The original spatial records of these surveys are available
from the data portal for the Middle Eastern Geodatabase for
Antiquities (MEGA), Jordan (Getty Conservation Institute 2015).
THE PROCESS OF PRE-GPS
COORDINATE GATHERING
Determining a discovered site’s geographic coordinates in the
pre-GPS era was a non-automated, manual, and mental process
that relied on the accuracy of the surveyor’s measurements, the
quality of the available maps and imagery, and the surveyor’s
ability to read and interpret the maps and imagery (Golledge
1999). Thus the process was both quantitative and qualitative at
the same time, which exposed it to multiple sources of error.
Quantitative measurements were collected with manual
methods of varying accuracy and precision. Typically, a coordi-
nate x was achieved by triangulating the bearings to two or
more distant landmarks through a handheld transit or sighting
compass. The bearings were subject to sighting error and mag-
netic anomalies, which introduce typical errors of about .5 to 3
compass degrees (Lovallo et al. 1994). More accurate tripod-
mounted transit theodolites or an alidade and plane table were
sometimes used for triangulation during pedestrian survey, but
these were more often used only during limited “high inten-
sity” surveys in the vicinity of long-term excavation projects.
For example, the SEEDSP used a theodolite to map sites in the
vicinity of the major sites of Bab edh-Drah, Numeira, and es-Sa
(Schaub, personal communication 2014). Single bearing and
distance measurements from one landmark or previously identi-
ed site were used when not enough triangulation points were
available. Distance measurements introduced still more error,
as they were typically measured by counting paces, with survey
chains or long tape measures, or via rod-height triangulation
(Compton 1985).
Surveyors were aware of the limited accuracy of typical eld
measurement techniques, and coordinate xing was aided
through qualitative correlations between the surveyor’s senses
(e.g., their view of the local topography and landcover) and the
representation of features on maps or remotely sensed images
(Meilinger et al. 2007; Ottosson 1988). In other words, contour
334
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
FIGURE 1. Map of central Jordan, showing the approximate boundaries of several archaeological survey projects conducted
in the pre-GPS era. SEEDSP = Southeast Expedition to the Dead Sea Plain, ASKP/ASCM = Archaeological Survey of the Kerak
Plateau/Archaeological Survey of Central Moab, WHS = Wadi Hasa Survey, WHNBS = Wadi Hasa North Bank Survey, SGNAS =
Southern Gohrs and Northern Arabah Survey, TBAS = Taleh-Busayra Survey. See text for citations about these projects. Inset
shows location of larger map.
335
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
FIGURE 2. Map of the Wadi al-Hasa region, showing sites discovered during the Wadi Hasa Survey (WHS), Wadi Hasa North
Bank Survey (WHNBS), and the Wadi Hasa Ancient Pastoralism Project (WHAPP).
336
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
lines, map symbols, place names, or shades of gray had to be
intuited into real landscape features using mental models, and
these models then had to be translated to a surveyor’s visual
perception of the environment via pattern matching (Blades
and Spencer 1987, 1990; Ottosson 1988). These models—such
as the “rule of V’s” (Miller and Westerback 1989), estimations of
slope from contour densities, recognition of the distinctiveness
of local landscape features such as peaks (localized topographic
high points), spurs (sloping ridges extending orthogonally from
the main topographic trend), and saddles (the topographic low
between two adjoining peaks), and interpretation of shadows
and textures on imagery—were individually cultivated through
pedagogic and actual-use experience and so often varied
considerably in their application between researchers (Chang
et al. 1985; Gilhooly et al. 1988; Hegarty et al. 2006; Ishikawa
and Kastens 2005; Ishikawa and Montello 2006; Kozlowski and
Bryant 1977; Liben and Downs 1993; Pick Jr. and Thompson
1991;Thorndyke and Stasz 1993). Both quantitative and qualita-
tive locational information were inuenced by the experience
of walking in a landscape (Thorndyke and Hayes-Roth 1982).
Factors such as the amount of landscape visible from a particular
location, the experienced changes in scale of various land-
scape features during a traverse, changes in lighting conditions
throughout the day, vegetation, weather conditions, or even
fatigue from walking uphill can all affect or inuence a survey
measurement or the mental processes of pattern matching
(Ishikawa and Montello 2006; Meilinger et al. 2007). Although
their application would differ from surveyor to surveyor, such
qualitative survey assessments would still be conducted using
similar cognitive processes, which, importantly, are modelable
(Hegarty et al. 2006; Ishikawa and Montello 2006; Thorndyke
and Stasz 1980). This is important because it offers a way to
use quantiable data (maps, map symbols, landscape features)
to understand error associated with different cognitive map-
ping processes. If such models are encoded in a set of GIS
procedures, they can produce standardized outputs that can be
compared against each other to assess the relative magnitude
of error that might derive from a particular cognitive mapping
process for different sites in a landscape.
A nal aspect of non-digital site-location recording was the
recording of coordinates in a particular CRS (e.g., Latitude/
Longitude or UTM) in a site record or site-location log. Once a
location was plotted on a map, coordinates were read from CRS
markings on the margins of the map, typically using handheld
rules to line up site points with CRS grid markings (Compton
1985). This was often complicated by (1) torn, clipped, or other-
wise modied eld maps; (2) maps with text and numbers writ-
ten in alphabets foreign to the survey team; (3) use of unfamiliar
or highly localized CRS and units of measurement; and (4) pro-
jection and/or alignment issues between different map series.
Surveyors knew this was an imprecise, inconsistent technique,
and so they often rounded coordinates in an effort to provide
a minimum of consistency between measurements. While this
technically increased the average accuracy of the site coordi-
nate database, it did so at the expense of its precision, and the
potential accuracy of the individual measurements. Because of
this, unless sites were very large or obtrusive, surveyors often
recorded only the estimated center point of the site, together
with an approximation of the site size. Finally, after the eldwork
was completed (and often many years later), paper records had
to be converted to digital. This is most typically achieved by
manual data entry, which can lead to considerable error due to
typos, transpositions, or misreadings of coordinates.
This discussion makes clear that there are many potential
sources of error in the spatial record of legacy survey data. Some
of this error is relatively easy to identify and can be corrected
through basic data-cleaning protocols. The error that remains
can be modeled so that we can understand how it may propa-
gate into spatial analyses of the legacy data. To achieve these
things, however, we need to understand how particular errors
enter into the data.
CORRECTABLE ERRORS
Sources of Correctable Errors
There are two types of errors that are generally correctable if
the original paper survey records and metadata are still avail-
able. The rst of these are systematic mapping errors, which
derive from phenomena that affect coordinates in a regular way.
Common sources of systematic mapping errors include shifted
coordinate reference systems in base maps, errors introduced
by rounding the coordinates, and errors in subsequent reprojec-
tion of the coordinates. These errors can be common if base
maps in a region are in a highly localized coordinate system
(e.g., the Palestine Grid of the Southern Levant), if base maps
are themselves shifted, if the wrong geographic datum was used
in reprojection, or if the scale of base maps is too coarse.
The second type of correctable errors are random recording
errors, which generally derive from human errors introduced
during the reading of site coordinates from survey maps into site
records or during manual data entry of coordinates into elec-
tronic databases. Random recording errors can derive from rules
held nonparallel to the coordinate grid, misreads of coordinates
on map margins, errors in communication between readers
and recorders, transposition (e.g., of “easting” with “north-
ing”), dyslexia, and typographical errors in the data entered into
electronic formats. Random recording errors will generally be
found in any pre-GPS survey dataset, but could be particularly
pervasive if (1) numbers are written in numeral systems unfamil-
iar to surveyors (e.g., in the Near East, maps often use Eastern
Arabic numerals); (2) if eld teams are multi-lingual; (3) if reading
and recording of coordinates from maps were conducted by
multiple eldworkers (i.e., many different people were involved
in the collection of the coordinates); (4) if the maps contained
markings for more than one CRS; or (5) if the base map scale is
too large for the project area., thereby making it more difcult to
read the coordinates correctly.
Mitigating Correctable Errors
Mitigation of the potentially correctable errors is dependent
on the availability of the original survey records and ancillary
information about the survey team and methodology. This
is important because much of this information is never pub-
lished. Methods sections are often very brief and generalized.
For example, authors will describe the sampling frame of the
survey, but not the mapping techniques. This is compounded
by the fact that many publications, at best, only perfunctorily
address sources or potential magnitude of error in the data. For
example, sometimes references to the source of base maps or
337
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
aerial photographs are not published, as was the case with the
ASKP and SEEDSP projects, which did not publish map refer-
ences. The WHS, WHNBS, SGNAS, TBAS, and ASCM, however,
did publish this important information. In another example, it
was common practice for archaeological surveyors to round site
coordinates to the nearest 10 or 100 m, but this fact was very
rarely reported in survey publications. For example, among our
case studies, ASKP reported that they rounded their coordi-
nates, but WHNBS, WHS, SGNAS, TBAS, ASCM, and SEEDSP
did not report whether rounding occurred.
Much of this information can be learned or inferred from what
Witcher (2008) calls the “metadata” of the survey—the unpub-
lished notes, records, and memories of surveyors. However,
legacy datasets have typically been manipulated by perhaps
several generations of researchers, and may be inherited
without this metadata. For example, subsequent reprojection of
coordinates that were initially recorded as rounded values would
mask the fact that they were rounded. Quite typically, legacy
datasets are inherited in a form that seems accurate (e.g., as
an electronic database with precise-looking coordinates), and
locational errors become apparent only after site coordinates
are compared to more accurate spatial data (e.g., well-rectied
high-resolution imagery, GPS-gathered site location data, etc.).
As an example of this, prior examination of the preliminary
survey reports of the WHS and WHNBS (Clark et al. 1992, Clark
et al. 1994; MacDonald 1988; MacDonald et al. 1980) had not
made clear to me that the surveyors had originally recorded all
coordinates in the Palestine Grid (PalGrid) CRS, which is a local-
ized CRS commonly used on mid-twentieth-century maps of
the region. The coordinates had clearly been reprojected to the
UTM CRS that I had used while planning the WHAPP. I discov-
ered this only after I obtained the copies of the original survey
forms and maps, which were curated at ASU in the personal
archive of Geoff Clark. Once I obtained these records, the scope
of the potential source of errors became clear. The maps were
composed of several smaller quads that had been physically cut
and taped together and so were missing a signicant amount
of marginalia such as coordinate markings and map reference
information. Some of the maps were in English, and some were
in Arabic, including the numerical notation system (Eastern
Arabic numerals). Most of the quads had both Latitude/Longi-
tude and PalGrid CRS lines. However, without the marginalia, it
was difcult to discern them. Complicating things further, it was
also clear that the WHS quads were of a different map series
than the WHNBS maps. In the WHS survey report (MacDonald
1988; MacDonald et al. 1980), MacDonald cites the use of the
1:25,000 scale maps prepared by the Ministry of Economy and
U.S.A. Operations Mission to Jordan (1955) from 1953 stereo
aerial imagery for all but the extreme western edge of the survey
region, for which he used the 1:50,000 scale Series K737 maps
from the United States Army Map Service (1961). In the WHNBS
survey report (Clark et al. 1992; Clark et al. 1994), Clark cites the
1989-series 1:25,000 scale maps prepared by the Royal Geo-
graphical Society of Jordan (1989). Discussions with some of the
original survey team members and examination of the original
survey paper records also made it clear that different recording
strategies were employed in the two surveys. A key piece of
information that emerged from this investigation was that the
surveyors rounded coordinates, greatly reducing their precision
and, often, their accuracy.
In the case study example above, going back to the origi-
nal survey records made it clear that signicant amounts of
locational errors had been introduced between the time of the
original location xing on eld maps and the encoding of the
digital database I had inherited. Thus, the rst course of action
that should be taken to mitigate correctable error is to obtain as
many primary sources of information as possible, including all
available preliminary reports, original survey records and maps,
and, if possible, personal communication with members of the
original survey team. Because correctable errors enter into the
data during reading or transcription of site coordinates from
CRS markings on eld maps, it is particularly important to obtain
the original maps used during survey.
4
These errors can be fully
mitigated by digitizing site coordinates directly from a georec-
tied scan of the original eld map. I should stress, however,
that the scanned maps should not be rectied using the CRS
reference marks or grids of the maps themselves, given that the
CRS markings themselves may be a signicant source of error
because the CRS grid may be awed, warped, or misaligned.
Maps should therefore be rectied to high-resolution imagery
or topographic data via feature matching, using landmarks that
are not likely to have moved in the intervening time between
the creation of the map and the referencing dataset (e.g.,
avoid using highly active stream channels or ephemeral trails as
landmarks).
To illustrate the types and numbers of correctable spatial errors
that may typically be expected in uncorrected legacy survey
data, I have made available a spatially corrected set of the WHS/
WHNBS survey data in an online repository (Ullah 2015a). This
new data set was made by extracting corrected coordinates for
the WHS and WHNBS sites following the methodology laid out
above. I scanned and georectied all of the original survey maps
to LandSat imagery (United States Geological Survey [USGS]
2003). With the help of two work-study students, all of the sites
marked on the maps were digitized. The students kept notes
on potential misreads and other sources of error in the digitiza-
tion process. Figures 3a and 4a show example portions of the
rectied WHS and WHNBS survey maps, respectively, with newly
digitized site points compared to the points from the inherited
database. To understand the nature of these correctable errors
in each dataset, I undertook a directional analysis of the offset
between the original and corrected points. Directional analysis
of the WHS data (Figure 3b) shows little consistency between
sites, indicating a dominance of random recording errors in
the original database. A slight preferential offset to the north
and northeast indicates that some systematic mapping errors
were also present. The WHNBS dataset, however, shows a very
consistent eastward offset (Figure 4b), indicating that system-
atic errors were most prevalent in the original database. I also
undertook an analysis of the offset distances, which also showed
differences between the two surveys (Figures 3c and 4c). The
offsets were consistently larger in the WHS data, indicating that
coordinates were likely rounded to fewer signicant digits in the
WHS (e.g., the nearest 100 m), whereas the WHNBS team either
attempted not to round the data or rounded only to the nearest
10 m.
338
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
FIGURE 3. Analysis of offset between the original site locations in the WHS database and the site locations newly digitized
from the original eld map: (a) a portion of the scanned eld map, showing the offset (green lines) between database points
(red) and the newly digitized points (blue); (b) a rose diagram showing the frequency of different angles of offset between the
two sets of points; (c) a histogram showing the frequency of different distances of offset between the two sets of points.
339
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
FIGURE 4. Analysis of offset between the original site locations in the WHNBS database and the site locations newly digitized
from the original eld map: (a) a portion of the scanned eld map, showing the offset (green lines) between database points
(red) and the newly digitized points (blue); (b) a rose diagram showing the frequency of different angles of offset between the
two sets of points; (c) a histogram showing the frequency of different distances of offset between the two sets of points.
340
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
INHERENT ERRORS
Once the correctable errors are removed, the spatial record of
the legacy survey dataset will be as accurate as it possibly can
be. However, there may still remain signicant locational errors
that cannot be corrected because they are inherent to the data
itself. These errors will propagate through any spatial analyses,
and so must be understood and reported.
Sources of Inherent Errors
Inherent errors are the errors that remain after systematic map-
ping errors and random recording errors are corrected, and they
enter into the data during the process of xing a location to plot
on a map or imagery. This can occur as mismeasurement with
survey tools. Mismeasurement can derive from survey devices if
they are inherently low precision (e.g., hand-held transits) or low
accuracy (e.g., pacing) tools, or if they are improperly calibrated
(e.g., an incorrect declination setting on a compass). They can
also derive from misuse of the tools (e.g., improper orientation
of compass to map) or other human error (e.g., reading bearings
incorrectly). Inherent errors also stem from landform misidenti-
cation, which is an incorrect correlation between a real landform
and a pattern map or image. Misidentication is more likely to
occur when local topography is indistinct, highly repetitive, or
uniform and when few distant landmarks are visible. This is exac-
erbated by unclear or distorted maps or images, or when the
map scale is too large to show distinctive landscape features.
The likelihood of a misidentication is also increased if vegeta-
tive cover obscures the view of the surrounding landscape, or if
a surveyor is fatigued or inexperienced in reading topographic
maps or interpreting imagery. Importantly, these factors may
also increase the potential for human errors of mismeasurement
as well.
The inherent spatial error of a legacy survey dataset can be
directly measured only in comparison to a more accurate control
dataset (e.g., GPS waypoints). This is a problem, however, as
researchers are often using a legacy dataset precisely because
they cannot obtain newer, more accurate spatial records for a
study area. Even in the case where it is possible for a researcher
to gather a set of comparative GPS locations, that sample may
be too small or may be spatially or temporally biased. For exam-
ple, the comparative GPS sample gathered during the WHAPP
provides only six locations with which to assess the spatial error
that remains in the combined sample of over 1,500 WHS and
WHNBS sites after digitization. Analysis of the angular and
distance offset between the GPS coordinates and the digitized
map locations of the six WHAPP sites shows no consistency (Fig-
ure 5), but the sample is too small (and biased towards Neolithic
sites) to represent the range locational errors that may remain.
This is likely to be true for most legacy survey datasets, so it is
important to nd alternative methods of estimating or modeling
the amount of locational errors remaining in the dataset.
Modeling Inherent Error in Legacy
Datasets
A general qualitative estimation of the inherent error in the data-
set may be obtained through an assessment of those aspects of
the survey metadata relating to surveying conditions—especially
any information about survey conditions, surveyor mentality and
FIGURE 5. Analysis of offset between the GPS coordinates of six WHAPP sites and their coordinates as digitized from the
original eld map: (a) a rose diagram showing the frequency of different angles of offset between the two sets of points; (b) a
histogram showing the frequency of different distances of offset between the two sets of points.
341
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
experience, and surveyor familiarity with the landscape. This
type of general information about the survey can help provide
an overview of the general reliability of locations recorded
by the survey project, but does little to help differentiate the
amount of spatial error contained in any specic coordinate
gathered by the survey team. Conscientious surveyors recorded
possible mislocation errors in their unpublished eld notes.
Classication of descriptive phrases from these notes (e.g.,
“This site is a small lithic scatter on a large at terrace”) might
be used to generate a qualitative typology of mislocation error.
The typology could, for example, qualify such characteristics as
vagueness of landform identication or obtrusiveness of the site
according to a categorical scale. These categorical scores could
be combined with quantitative data from the survey notes, such
as site size, or number of referenced landmarks, to come up with
a general estimation of site location “reliability.” Often, however,
these types of survey data and metadata are incomplete, were
never recorded, or are no longer helpful. For example, many of
the notes included with the WHS were referenced to other sites
found during the survey (e.g., “This site is 100 m east of site
421”), and so are useful only if the real location of the referenced
site is known. Moreover, in very large datasets, such as with the
over 1500 combined WHS/WHNBS sites, the effort of manually
coding these types of indices may not be commensurate to the
information gained. At the very least, however, a general read of
the survey metadata and unpublished notes provides a gestalt
understanding of the types of error that might remain.
A more efcient, standardizable way to assess inherent errors is
to use external models derived from theory about human way-
nding, mental processes of orientation in landscapes, and map-
reading. This theory is discussed in the “Process of Pre-GPS
Coordinate Gathering” section, above. I advocate that models
of these processes be encoded as a series of GIS procedures to
assess mislocation error in a systematized and largely automated
manner. By using GIS tools to quantify theory-derived expecta-
tions of how the sources and amplitudes of mislocation error
change over space, expectations about the effect of this kind
of error are encoded as a series of digital maps, which can be
combined and inter-compared. This provides a spatially explicit
“reliability index” for all parts of a survey area, and these values,
uploaded to the site database, provide a standardized estimate
of the relative reliability of recorded site locations. Similar meth-
ods have been used to understand the error in other types of
GIS data (e.g., Hunter and Goodchild 1995).
To operationalize this method, we need to create abstract, sim-
plied, digital representations of the mental models surveyors
used to locate sites and the actual experience of surveying.
Here, I present GIS workows to model four different aspects
of the pre-GPS survey experience. Necessarily, these GIS
models cannot depict the entire experience of surveying, but
they nevertheless provide a standardized approximation of the
effect of particular types of mislocation errors. In the spirit of
the open data movement that champions the creative reuse of
extant archaeological data (Beale 2012; Beck and Neylon 2012;
Kansa 2012; Morgan and Eve 2012), I have elected to create
these models with the free and open-source GRASS GIS suite
(GRASS Development Team 2015). Using open software like
GRASS allows the methodology to be more completely tailored
(i.e., from the source code up), promotes intellectual equality
by removing nancial obstacles, allows full transparency, and
encourages critique and enhancement of the methods and tools
(Ducke 2012). I deploy the four models in the context of the
Wadi al-Hasa case study by using the freely available, void-lled,
90-m resolution topographic data for the Wadi al-Hasa region
from the Shuttle Radar Topography Mission (SRTM) (USGS
2012) to show example results. The horizontal resolution of the
data was interpolated in GRASS to 15 m via regularized spline-
tension interpolation (Mitasova and Hoerka 1993; Mitasova and
Mitas 1993) before subsequent analysis.
The rst workow models the way surveyors interpret the spac-
ing of contour lines on the original survey map. The density of
contour lines on topographic maps depicts the steepness of the
earth’s surface; the more dense the spacing of contour lines, the
steeper the actual topography should be. The model approxi-
mates a portion of the implicit mental ltering process at work
when a surveyor looks to match contour lines on a map to actual
topography of a landform (Gilhooly et al. 1988). Specically, the
model assumes that denser contour lines relate to more easily
identiable landforms, because more morphological infor-
mation is available to the surveyor. This allows them to more
condently identify a particular point on a particular landform.
In other words, the model predicts that a point on very steep
or rugged topography is easier to accurately identify on a map
than is a point on at topography. The model is operationalized
by extracting vector contour lines from elevation data (e.g., the
SRTM elevation raster for the Wadi al-Hasa region) at the same
interval as the original survey map (e.g., 10 m) using the GRASS
module r.contour. Points are sampled evenly along the contour
lines using the GRASS module r.to.points. The points are then
used to create a kernel density map with the v.kernel tool. This
creates a digital map that quanties the relative density of con-
tours on the original survey map (Figure 6a).
The second workow models the way surveyors orient them-
selves to landmarks when obtaining quantitative location xes
for places on the landscape. This is operationalized though
cumulative viewshed analysis (CVA), which produces a map
of inter-visibility between many viewpoints and all the other
pixels on a map (Lake et al. 1998; Llobera 2003; O’Sullivan and
Turner 2001; Tabik et al. 2013; Wheatley 1995). When applied
to archaeological sites as the originating viewpoints, CVA
is useful for modeling how ancient people may have under-
stood and used a landscape. This workow, however, uses
CVA in a different application. During survey, sites locations
are triangulated using a small number of distinct landmarks
(e.g., local topographic peaks). CVA can be used to model the
inter-visibility between these landmarks and all other points on
the landscape, which provides an estimation of the surveyor’s
ability to triangulate a position at all points on the landscape.
To operationalize this, points are digitized for all peaks marked
on the original survey map. These points, and the rasterized
topography (e.g., SRTM data), are used as input for a CVA
using a custom GRASS add-on module called r.viewshed.
cva. I created the r.viewshed.cva module with a structure that
facilitates this type of analysis, and it can be freely downloaded
from the GRASS add-on repository (Ullah 2015b). The mod-
ule calculates individual viewsheds for each input point (e.g.,
peak), and then cross-tabulates each viewshed so that the
resultant map encodes the number of peaks that can be seen
from each pixel (Figure 6b). Areas with few visible landmarks
would be harder to triangulate on a topographic map.
342
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
The next workow models the way surveyors experience the
landscape as they move through it. The experience of move-
ment in a landscape is important to the way an individual forms
a cognitive map of a region, and so plays an important role in
human waynding and orientation (Fitzjohn 2007; Hegarty et
al. 2006; Thorndyke and Hayes-Roth 1982). It can be modeled
through analysis of topographic prominence, which is a measure
of localized changes in topography at different horizontal scales
(Christopherson 2003; Llobera 2001). Prominence studies have
been used in archaeological studies that seek to understand
and model how the morphological character of a landscape
could have inuenced the way ancient peoples experienced and
moved through it. This idea is directly applicable to the experi-
ence of surveyors traversing transects. Highly prominent features
are likely to be noticed by a surveyor and, thus, are potentially
more likely to be identied on a map. The workow operational-
izes this by calculating the average elevation in different sizes
of moving window neighborhoods using the GRASS module
r.neighbors. This creates a series of elevation maps averaged at
different horizontal scales, which are each subtracted from origi-
nal elevation (e.g., the Wadi al-Hasa SRTM data) using the raster
map calculator, r.mapcalc. The “Average Prominence” is then
calculated in using the r.series module to calculate the average
of those maps (Figure 6c).
The nal workow models the way surveyors match landforms
to the shape of contour-line patterns on topographic maps. The
horizontal shape of contour-lines represents the curvature of
landscape features over space. Surveyors interpret these pat-
terns through a cognitive process called representational cor-
respondence (Liben and Downs 1993), which helps them match
particular contour-line shapes to specic topographic features.
Landforms that are composed of rapidly changing slopes (e.g.,
peaks, spurs, or saddles) create unique, compact, contour-line
shapes when represented on topographic maps, and so are
easier to identify than other landforms. The workow operation-
FIGURE 6. Example output for the Wadi al-Hasa area of the four GIS-based models of the cognitive process of xing locations
in the pre-GPS era: (a) output from a model based on contour-line density; (b) output from a model based on the number of
visible landmarks; (c) output from a model based on the average topographic prominence; (d) output from a model based on
average landform curvatures. For clarity, only a small portion of the entire Wadi al-Hasa region is shown.
343
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
alizes this idea by measuring topographic curvatures (change
in slope) with the GRASS module r.slope.aspect. The module
calculates the curvatures of the landscape in the direction of
steepest slope (“principle” or “prole” curvature) and in the
direction perpendicular to the direction of steepest slope along
the horizontal plane (“secondary” or “tangential” curvature).
Multiplying the two curvatures together creates a map in which
high values denote the parts of the landscape where land-
form shape changes rapidly in both the vertical and horizontal
dimensions (Figure 6d). Rapid change of both types of curvature
generally translates to distinct, easily identiable patterning of
contour lines on a topographic map.
While each of the individual models produces interesting results
on its own (Figures 6a-d), it is more useful to combine them
as a holistic “reliability index” for each pixel of the digitized
landscape. The most straightforward way to achieve this is to
standardize the output from each individual model to a common
scale (in the cases here, I chose a scale of 0–1) and then to cal-
culate the average of all the measures across the survey region.
The workow achieves this using a simple averaging formula
in the GRASS raster math module, r.mapcalc. Figure 7 shows
the map of reliability index as calculated for the Wadi al-Hasa
region from the four surveying models described above. This
particular map was created as a simple unweighted average, but
any of the input models could be weighted more heavily than
the others if there is reasonable justication to do so. The survey
metadata and the unpublished survey notes are the appropri-
ate sources of information about potential weightings of the
different factors (e.g., if there were many notes mentioning the
difculty of determining locations on at areas, then the contour
density model might be weighted more heavily than the others).
Finally, I note that although the four models described in this
section should be generally applicable to most legacy survey
datasets, they are not the only possible ones that could be
employed. For example, modeling the density of vegetation
might be important for how it obstructs views of landmarks, or a
combination of vegetation density and surface roughness could
model ground visibility and how that affects the identication
of site boundaries. Modeling variation in surface albedo might
simulate the recognizability of landscape patches on aerial pho-
tographs. A model derived from the proximity of recorded site
locations to roads or trails might be useful for surveys that were
non-systematic or vehicular. The effect of these particular addi-
tional factors appears to be minimal in the Wadi al-Hasa case
study (for example, Wadi al-Hasa is very sparsely vegetated), but
if these kinds of additional factors are deemed important for a
particular survey, then care should be taken to ensure that they
are modeled in a way that matches the conditions of the original
survey as closely as possible. As an example, if the density of
vegetation is thought to be important, then the digital represen-
FIGURE 7. A sample Reliability Index map for the Wadi al-Hasa case study, created as the standardized average of the maps
shown in Figure 6. Sites from the WHS and WHNBS surveys are shown, with sampling radii based on the reported site sizes.
For clarity, this gure shows only a portion of the entire Wadi al-Hasa region.
344
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
tation should be classied based on remotely sensed imagery
dating to the period of the original survey. Thus in a survey done
in the early 1980s, coarser Landsat MSS data would be preferred
over ner, but more recent Landsat TM or ETM+ data (Haack et
al. 1987).
Applying the Reliability Index to Legacy
Datasets
The location reliability map created with the workow presented
in the preceding section should inform subsequent analyses of
the spatial data. To do so, we must rst upload reliability values
for each site to a eld in the site database. If sites were digitized
as vector point locations, then a sampling radius (e.g., relative
to site size) should be employed in this process. A sampling buf-
fer can be created in GRASS with the v.buffer module, using a
vector point map of site locations, with a column of average site
dimensions to determine the size of the sampling radii for each
site (e.g., Figure 7). If the site locations were originally digitized
as polygons, they can be used directly. In either case, an average
index value should be calculated within the sampling radius or
polygon, using the GRASS module v.rast.stats. Doing so corrects
for issues of scale related to site size and raster resolution and
ensures that a reasonable estimate is uploaded for all sites.
Once the values have been uploaded, it is useful to compare
them to the spread of all possible values in the Reliability Index
raster map. As an example, the Reliability Index values for the
corrected WHS/WHNBS sites are made available in the new
online repository of these data (Ullah 2015b). Table 1 presents
univariate statistics about the range of reliability index values for
all pixels of the Wadi al-Hasa SRTM and compares them to the
range of reliability index values for all WHS and WHNBS sites.
These data show that the average reliability score in the project
area is only about 34 percent of the maximum possible score.
The average reliability score of the WHS sites matches this, but
the average score of the WHNBS sites are slightly lower at about
32 percent of the maximum possible score. These summary
statistics also provide a measure with which to assess the score
of any individual site, or subset of sites when conducting spatial
analyses with the dataset. When accurate, highly precise site
locations are needed, the analysis should be constrained to a
sample of sites with relatively high reliability index values.
DOING SPATIAL ANALYSIS WITH
LEGACY DATASETS
The techniques described in this paper provide what I believe
are “best practices” to obtain the highest possible quality
spatial records from older archaeological survey work, so that
they can be used in modern spatial research. I present here
an example of using legacy data in a modern spatial analysis
procedure—a cumulative viewshed analysis—that benets from
accurate, precise site coordinates.
Before beginning the analysis, it is important to understand
the beginning impact of inherent errors on the output of the
routine. I queried the WHS site-database to extract all Neolithic-
period lithic scatters to use as input for the CVA analysis. This
resulted in a sample of 30 sites, with an average reliability index
score of .282 (± .073) at the corrected site locations. This score is
similar both to that of the greater WHS database as a whole and
to that of the rest of the landscape. Thus, although likely subject
to some inherent error, the corrected spatial record of this sub-
sample of sites is not highly suspect. However, the median site
size of these Neolithic lithic scatters is fairly small (about 850 m
2
),
and so it is still possible that any remaining locational error will
propagate into the results of the CVA.
To understand better the effect of locational errors on the results
of the CVA, I conducted four separate CVA routines. The rst CVA
used the original, uncorrected coordinates from the inherited
WHS database. This point set contains all the correctable errors
that had entered into the data. The second used site coordinates
as digitized from the scanned, georectied survey maps. This
point set is spatially corrected, but likely still contains some inher-
ent error. The third routine used a perturbation of the corrected
coordinates that randomly shifted the coordinates by an amount
drawn from a uniform-random distribution with a maximum offset
of 50 m. This point set contains induced error, roughly equivalent
to what might be expected by rounding coordinates to the near-
est 100 m. The nal routine used a set of 30 new points randomly
chosen from within the WHS survey boundaries. This point set
acts as a comparative “control” sample.
TABLE 1. Univariate Statistics Describing the Range of Reliability Index Values for All Pixels of the Wadi al-Hasa DEM, All
Sites Discovered by the WHS, and All Sites Discovered by the WHNBS
All Pixels WHS Sites WHNBS Sites
Mean .283 .288 .269
Median .268 .277 .258
Variance .003 .003 .002
Standard deviation .055 .052 .044
Range .686 .321 .283
Minimum .150 .200 .201
Maximum .836 .521 .484
n 6089320 980 484
Note: Reliability index values for sites were sampled in a radius around the corrected site coordinate. These radii corresponded to the reported size of
each site.
345
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Using each of the four point sets as viewing locations, the four
CVA routines returned maps of the number of sites that can
see each pixel of the input DEM (Figure 8). I used the site-size
radii described in the previous section to upload these values
at each site location for each point set. Univariate statistics for
the CVA values for each point set are shown in Table 2. Firstly,
CVA undertaken with the random “control” point set produced
radically different values than the other three point sets. This
indicates that even the uncorrected data produced results that
are archaeologically signicant. However, CVA undertaken with
the uncorrected and perturbed point sets overpredicts the aver-
age amount of inter-visibility between sites. Correlation statistics
(Table 3) show that these differences are systemic, and not due
to one or two large differences. Thus, the correction procedure
improves the results of the CVA, but we also know that some
inherent spatial error may remain that affects the total accuracy
of the CVA routine.
CONCLUSIONS AND FUTURE
DIRECTIONS
If archaeology has become a largely digital endeavor (Zubrow
2006), where does that leave data gathered in the pre-digital
era? These older datasets grow ever more important as damage
to the archaeological record accumulates and as access to eld
sites becomes more and more restricted. Digital geospatial
technologies have radically changed the scale of spatial analyses
that are possible within the budgets of typical archaeologi-
cal research projects. For example, Lin and colleagues (2014)
leveraged on-the-ground survey with GPS and crowd-sourced
digital survey with high-resolution satellite imagery to nd and
accurately digitize over 2.3 million archaeological features in a
region of over 6,000 km
2
in Mongolia. Modern technologies,
such as smartphones and tablets, are putting customized, GPS-
enabled software literally in the palms of modern archaeologists
while they conduct eldwork (e.g., Austin 2014; Cascalheira et al.
2014; Goodale et al. 2013; and see Harris 2012 for an excellent
discussion of challenges and opportunities arising from use of
these technologies). Advanced GIS analyses, such as GIS-based
site catchment modeling (Pullar and Springer 2000; Ullah 2011),
construction of ancient “visualscapes” (Llobera 2003; Wheatley
1995), and construction/evaluation of various predictive models
(Graves McEwan 2012; Ladefoged et al. 2009), are now common
in landscape archaeology, but are predicated on an accurate
spatial record. This study has raised signicant concerns about
the spatial record of older survey data in the form that they are
typically received. In the case of the WHS and WHNBS survey
data, digitizing the original survey maps identies and mitigates
many of the spatial errors that had accumulated in the inherited
survey database, and this is likely true for most pre-GPS survey
data. The accuracy of the corrected data are, however, largely
dependent upon the scale and quality of the original base maps,
the style of survey and measurement, the relative distinctness of
FIGURE 8. A sample output of a Cumulative Viewshed Analysis undertaken with a small sample of the WHS sites. The location
and sampling radii of sites are indicated. The purple dots are the original uncorrected site locations, the yellow dots are the
corrected site locations, and the blue dots are the perturbed versions of the corrected site locations. The underlaying CVA
map was generated from the corrected site locations. For clarity, this gure shows only a portion of the entire Wadi al-Hasa
region.
346
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
landscape features on which sites are located, the number and
quality of easily identiable nearby landmarks in the survey area,
and other physical attributes of the landscape that enhance
or impede the process of archaeological survey. Returning to
the original “survey metadata” and the unpublished notes and
maps is an essential step in identifying remaining errors, but it
is often made difcult by missing or inadequate records. I have
presented an alternative approach to assessing the amount of
error that may remain. Using GIS-based models to estimate
the severity of these inherent spatial errors allows a standard-
ized, comparable, and quantiable measure of condence in
the applicability of spatially corrected legacy survey datasets in
analyses that require spatial accuracy. Hopefully, the procedures
discussed in this paper will encourage archaeologists to make
fuller use of “legacy” data in ongoing research programs and
to be more aware of the limitations and benets of these older
datasets for modern archaeological research methods.
Acknowledgments. I would like to acknowledge those archae-
ologists who conducted survey in the pre-GPS era. Their skill
and perseverance in the eld provides us with invaluable records
that cannot be replicated. In particular, I would like to thank
Geoffrey A. Clark, Edward B. Banning, and R. Thomas Schaub
for providing access to original eld records and for indulging
my questions about their pioneering survey work in central Jor-
dan. Sadly, R. Thomas Schaub passed away shortly before this
paper went to press. He will be missed. I would like to acknowl-
edge C. Michael Barton for general support and advice about
this project. Funding for the initial eldwork of the WHAPP was
provided through NSF grant BCS-410269 and was conducted
under permit from the Department of Antiquities of the Hash-
emite Kingdom of Jordan, during the period of August 16, 2008
to August 31, 2008. Research for this paper was facilitated by
a fellowship in the Center for Comparative Archaeology at the
University of Pittsburgh.
Data Availability Statement. Topographic data used in this
paper derive from SRTM scenes SRTM3N31E035V2, SRT-
M3N31E035V2 (USGS 2012). The survey maps were georectied
to LandSat ETM+ scene: LE71740392003006EDC03 (USGS 2003).
These data are made freely available by the USGS. The custom
GRASS GIS modules developed in the course of this research
are available as free add-ons for GRASS GIS (GRASS Develop-
ment Team 2015), and sourcecode is available at: https://svn.
osgeo.org/grass/grass-addons/. They are published there under
the GNU license (free for all types of use, modication, and
distribution, as long as credit is given and source code remains
open). The original WHS and WHSNBS survey data are avail-
able through the Middle Eastern Geodatabase for Antiquities
(MEGA), Jordan, data portal (Getty Conservation Institute 2015).
The corrected WHS and WHSNBS survey data are available at
http://dx.doi.org/10.6084/m9.gshare.1404216, under the cre-
ative commons license (free to use and modify for any purpose,
as long as credit is given).
TABLE 2. Table of Univariate Statistics for CVA Analyses of Neolithic Lithic Scatters from the WHS.
Original Corrected Perturbed Random
Mean 2.983 2.687 2.712 1.546
Median 2.321 2.000 2.000 1.107
Variance 4.242 3.391 3.647 .580
Standard deviation 2.060 1.842 1.910 .761
Range 6.529 6.140 5.211 2.487
Minimum .471 .860 .789 .513
Maximum 7.000 7.000 6.000 3.000
n 30 30 30 30
Note: Statistics are reported for CVA undertaken with the original, uncorrected dataset, the spatially corrected dataset, and a 50-m uniform-random
perturbation of the corrected dataset (simulates rounding errors). Statistics are also reported for a randomly generated “control” point set. CVA results
are reported as the number of other sites that can be seen from each site.
TABLE 3. Correlation Matrix Comparing the CVA Results for CVA Analyses of WHS Neolithic Lithic Scatters.
Original Corrected Perturbed Random
Original 1.000
Corrected .884 1.000
Perturbed .887 .942 1.000
Random -.214 -.293 -.370 1.000
Note: Compared are the results for the original, uncorrected dataset, the spatially corrected dataset, a 50-m uniform-random perturbation of the
corrected dataset (simulates rounding errors), and a randomly generated “control” point set.
347
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
REFERENCES CITED
Arıkan, Bülent
2009 Reorganization and Risk: Environmental Change and Tribal Land
Use in Marginal Landscapes of Southern Jordan. Unpublished Ph.D.
Dissertation, School of Human Evolution and Social Change, Arizona State
University, Tempe, Arizona.
2012 Don’t Abhor Your Neighbor for He Is a Pastoralist: The GIS-
based Modeling of the Past Human–environment Interactions and
Landscape Changes in the Wadi el-Hasa, West-central Jordan. Journal of
Archaeological Science 39:2908–2920.
Athanassopoulos, Efe F., and LuAnn Wandsnider
2004 Mediterranean Archaeological Landscapes: Current Issues.
University of Pennsylvania Museum of Archaeology and Anthropology,
Philadelphia.
Atici, Levent, Sarah Whitcher Kansa, Justin Lev-Tov, and Eric C. Kansa
2013 Other People’s Data: A Demonstration of the Imperative of
Publishing Primary Data. Journal of Archaeological Method and Theory
20:663–681.
Austin, Anne
2014 Mobilizing Archaeologists: Increasing the Quantity and Quality
of Data Collected in the Field with Mobile Technology. Advances
in Archaeological Practice: A Journal of the Society for American
Archaeology 2:13–23.
Banning, Edward B.
1996 Highlands and Lowlands: Problems and Survey Frameworks for
Rural Archaeology in the Near East. Bulletin of the American Schools of
Oriental Research 301:25–45.
Barton, C. Michael, Isaac I. Ullah, and Helena Mitasova
2010 Computational Modeling and Neolithic Socioecological Dynamics:
A Case Study from Southwest Asia. American Antiquity 75:364–386.
Beale, Nicole
2012 How Community Archaeology Can Make Use of Open Data to
Achieve Further Its Objectives. World Archaeology 44:612–633.
Beck, Anthony, and Cameron Neylon
2012 A Vision for Open Archaeology. World Archaeology 44:479–497.
Blades, Mark, and Christopher Spencer
1987 Young Children’s Strategies When Using Maps with Landmarks.
Journal of Environmental Psychology 7:201–217.
1990 The Development of Three- to Six-Year-Olds’ Map Using Ability:
The Relative Importance of Landmarks and Map Alignment. The Journal
of Genetic Psychology 151:181–194.
Bray, Hiawatha
2014 You Are Here: From the Compass to GPS, the History and Future of
How We Find Ourselves. Basic Books, New York.
Cascalheira, Joao, Celia Goncalves, and Nuno Bicho
2014 Smartphones and the Use of Customized Apps in Archaeological
Projects. The SAA Archaeological Record 14(5):20–25.
Chang, Kang-Tsung, James Antes, and Thomas Lenzen
1985 The Effect of Experience on Reading Topographic Relief
Information: Analyses of Performance and Eye Movements. The
Cartographic Journal 22:88–94.
Christopherson, Gary L.
2003 Using ARC/GRID to Calculate Topographic Prominence in an
Archaeological Landscape. Electronic document, http://www.casa.arizona.
edu/MPP/topo_promo/, accessed July 30, 2015.
Clark, Geoffrey A., Michael P. Neeley, Burton MacDonald, Joseph Schuldenrein,
and Khairieh Amr
1992 Wadi al-Hasa Paleolithic Project-1992: Preliminary Report. Annual of
the Department of Antiquities of Jordan 36:13–23.
Clark, Geoffrey A., Deborah I. Olszewski, Joseph Schuldenrien, Nazmieh Rida,
and James D. Eighmey
1994 Survey and Excavation in Wadi Al-Hasa: A Preliminary Report of
the 1993 Field Season. Annual of the Department of Antiquities of Jordan
38:41–55.
Coinman, Nancy R. (editor)
1998 The Archaeology of the Wadi al-Hasa, West-Central Jordan,
Volume 1: Surveys, Settlement Patterns and Paleoenvironments. Vol. 1.
Anthropological Research Papers. Arizona State University, Tempe.
Coinman, Nancy R.
2000 The Archaeology of the Wadi al-Hasa, West-Central Jordan, Vol. 2:
Excavations at Middle, Upper, and Epipaleolithic Sites in the Hasa. Arizona
State University Anthropological Research Papers 52. Arizona State
University, Tempe.
Compton, Robert R.
1985 Geology in the Field. Wiley, New York.
Drennan, Robert D., Bryan K. Hanks, Adriana Maguiña-Ugarte, and Alexander
J. Martín
2014 Comparative Archaeology Database. University of Pittsburgh,
Center for Comparative Archaeology. Electronic document, http://www.
cadb.pitt.edu/, accessed May 19, 2015.
Ducke, Benjamin
2012 Natives of a Connected World: Free and Open Source Software in
Archaeology. World Archaeology 44:571–579.
Fitzjohn, Matthew
2007 Viewing Places: GIS Applications for Examining the Perception of
Space in the Mountains of Sicily. World Archaeology 39(1):70–83.
Galaty, Michael L., and Charles Watkinson
2004 The Practice of Archaeology under Dictatorship. In Archaeology
under Dictatorship, edited by Michael L. Galaty and Charles Watkinson,
pp. 1–17. Springer, New York.
Getty Conservation Institute
2015 Middle Eastern Geodatabase for Antiquities (MEGA), Jordan
(2007–2014). Electronic document, http://www.megajordan.org, accessed
May 19, 2015.
Gilhooly, Kenneth J., Michael Wood, Paul R. Kinnear, and Caroline Green
1988 Skill in Map Reading and Memory for Maps. The Quarterly Journal
of Experimental Psychology Section A 40:87–107.
Golledge, Reginald G.
1999 Waynding Behavior: Cognitive Mapping and Other Spatial
Processes. JHU Press, Baltimore.
Goodale, Nathan, David G. Bailey, Theodore Fondak, and Alissa Nauman
2013 iTrowel: Mobile Devices as Transformative Technology in
Archaeological Field Research. The SAA Archaeological Record
13(3):18–22.
Goodchild, Michael F.
1993 Data Models and Data Quality: Problems and Prospects. In
Environmental Modeling with GIS, edited by Michael F. Goodchild,
Bradley O. Parks, and Louis T. Steyaert, pp. 94–103. Oxford University
Press, Oxford.
Goodchild, Michael F., and Sucharita Gopal
1989 The Accuracy of Spatial Databases. CRC Press, London.
Goodchild, Michael F., Sun Guoqing, and Yang Shiren
1992 Development and Test of an Error model for Categorical Data.
International Journal of Geographical Information Systems 6:87–103.
GRASS Development Team
2015 Geographic Resources Analysis Support System. Electronic
document, http://grass.osgeo.org, accessed May 19, 2015.
Graves McEwan, Dorothy
2012 Qualitative Landscape Theories and Archaeological Predictive
Modelling—A Journey through No Man’s Land? Journal of Archaeological
Method and Theory 19:526–547.
Haack, Barry, Nevin Bryant, and Steven Adams
1987 An Assessment of Landsat MSS and TM Data for Urban and Near-
urban Land-Cover Digital Classication. Remote Sensing of Environment
21:201–213.
Harris, Trevor M.
2012 Interfacing Archaeology and the World of Citizen Sensors:
Exploring the Impact of Neogeography and Volunteered Geographic
348
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Information on an Authenticated Archaeology. World Archaeology
44:580–591.
Hegarty, Mary, Daniel R. Montello, Anthony E. Richardson, Toru Ishikawa, and
Kristin Lovelace
2006 Spatial Abilities at Different Scales: Individual Differences in
Aptitude-Test Performance and Spatial-Layout Learning. Intelligence
34:151–176.
Hill, J. Brett
2002 Land Use and Land Abandonment: A Case Study from the
Wadi Al-Hasa, West-Central Jordan. Unpublished Ph.D. dissertation,
Department of Anthropology, Arizona State University, Tempe.
2004 Land Use and an Archaeological Perspective on Socio-Natural
Studies in the Wadi Al-Hasa, West-Central Jordan. American Antiquity
69:389–412.
2006 Human Ecology in the Wadi Al-Hasa: Land Use and Abandonment
through the Holocene. University of Arizona Press, Tucson.
Holtorf, Cornelius J.
2001 Is the Past a Non-renewable Resource? In Destruction and
Conservation of Cultural Property, edited by Robert Layton, Peter
G. Stone, and Julian Thomas, pp. 286–297. One World Archaeology.
Routledge, New York.
Hunter, Gary J., and Michael F. Goodchild
1995 Dealing with Error in a Spatial Database: A simple Case Study.
Photogrammetric Engineering and Remote Sensing 61:529–537.
Ishikawa, Toru, and Kim A. Kastens
2005 Why Some Students Have Trouble with Maps and Other Spatial
Representations. Journal of Geoscience Education 53:184.
Ishikawa, Toru, and Daniel R. Montello
2006 Spatial Knowledge Acquisition from Direct Experience in the
Environment: Individual Differences in the Development of Metric
Knowledge and the Integration of Separately Learned Places. Cognitive
Psychology 52:93–129.
Kansa, Eric C.
2010 Open Context in Context: Cyberinfrastructure and Distributed
Approaches to Publish and Preserve Archaeological Data. The SAA
Archaeological Record 10(5):12–16.
2012 Openness and Archaeology’s Information Ecosystem. World
Archaeology 44:498–520.
Kintigh, Keith
2006 The Promise and Challenge of Archaeological Data Integration.
American Antiquity 71:567–578.
Kozlowski, Lynn T., and Kendall J. Bryant
1977 Sense of Direction, Spatial Orientation, and Cognitive Maps.
Journal of Experimental Psychology: Human Perception and Performance
3:590–598.
Ladefoged, Thegn N., Patrick V. Kirch, Samuel M. Gon III, Oliver A. Chadwick,
Anthony S. Hartshorn, and Peter M. Vitousek
2009 Opportunities and Constraints for Intensive Agriculture in
the Hawaiian Archipelago Prior to European Contact. Journal of
Archaeological Science 36:2374–2383.
Lake, Mark W., Patricia E. Woodman, and Stephen J. Mithen
1998 Tailoring GIS Software for Archaeological Applications: An Example
Concerning Viewshed Analysis. Journal of Archaeological Science
25:27–38.
Liben, Lynn S., and Roger M. Downs
1993 Understanding Person-Space-Map Relations: Cartographic and
Developmental Perspectives. Developmental Psychology 29:739–752.
Lin, Albert Yu-Min, Andrew Huynh, Gert Lanckriet, and Luke Barrington
2014 Crowdsourcing the Unknown: The Satellite Search for Genghis
Khan. PLoS ONE 9(12): e114046.
Llobera, Marcos
2001 Building Past Landscape Perception With GIS: Understanding
Topographic Prominence. Journal of Archaeological Science
28:1005–1014.
2003 Extending GIS-Based Visual Analysis: The Concept of Visualscapes.
International Journal of Geographical Information Science 17(1):25–48.
Lovallo, Matthew J., Kurt C. Vercauteren, Naomi C. Hedge, Eric M. Anderson,
and Scott E. Hygnstrom
1994 An Evaluation of Electronic versus Hand-Held Compasses for
Telemetry Studies. Wildlife Society Bulletin 22:662–667.
MacDonald, Burton
1982 The Wâdi el-Hasa Survey 1979 and Previous Archaeological Work
in Southern Jordan. Bulletin of the American Schools of Oriental Research
245:35–52.
1988 The Wadi el Hasa Archaeological Survey 1979–1983, West-Central
Jordan. Wilfrid Laurier University Press.
MacDonald, Burton, and Khairieh Amr
1992 The Southern Ghors and Northeast’Arabah Archaeological Survey.
JR Collis, Department of Archaeology and Prehistory, University of
Shefeld.
MacDonald, Burton, Edward B. Banning, and Larry A. Pavlish
1980 The Wadi el Hasa Survey 1979: A Preliminary Report. Annual of the
Department of Antiquities of Jordan 24:169–83.
MacDonald, Burton, Andrew Bradshaw, Larry Herr, Michael Neeley, and Scott
Quaintance
2000 The Tala-Busayra Archaeological Survey: Phase 1, 1999. Annual of
the Department of Antiquities of Jordan 44:507–511.
MacDonald, Burton, Geoffrey A. Clark, and Michael Neeley
1988 Southern Ghors and Northeast Araba Archaeological Survey 1985
and 1986, Jordan: a Preliminary Report. Bulletin of the American Schools
of Oriental Research 272(1988):23–45.
MacDonald, Burton, Larry G. Herr, Michael P. Neeley, Traianos Gagos, Khaled
Moumani, and Marcy Rockman
2004 The Tala-Busayra Archaeological Survey 1999–2001, West-Central
Jordan. Archaeological Reports. American Schools of Oriental Research,
Boston, Massachusetts.
MacDonald, Burton, Larry G. Herr, Michael P. Neeley, Scott Quaintance, and
Andrew Bradshaw
2001 The Tala-Busayra Archaeological Survey: Phase 2 (2000). Annual of
the Department of Antiquities of Jordan 45(2001):395–411.
Maloy, Mark A., and Denis J. Dean
2001 An Accuracy Assessment of Various GIS-based Viewshed
Delineation Techniques. Photogrammetric Engineering and Remote
Sensing 67:1293–1298.
Meilinger, Tobias, Christoph Hölscher, Simon J. Büchner, and Martin Brösamle
2007 How Much Information Do You Need? Schematic Maps in
Waynding and Self Localisation. In Spatial Cognition V Reasoning,
Action, Interaction, edited by Thomas Barkowsky, Markus Knauff, Gérard
Ligozat, and Daniel R. Montello, pp. 381–400. Lecture Notes in Computer
Science 4387. Springer Berlin Heidelberg.
Miller, J. Maxwell
1979 Archaeological Survey of Central Moab: 1978. Bulletin of the
American Schools of Oriental Research:43–52.
1991 Archaeological Survey of the Kerak Plateau, Conducted during
1978–1982 under the Direction of J. Maxwell Miller and Jack M. Pinkerton.
ASOR Archaeological Reports 1. Scholars Press, Atlanta.
Miller, Victor C., and Mary E. Westerback
1989 Interpretation of Topographic Maps. Merrill Publishing Company,
Columbus, Ohio.
Ministry of Economy and U.S.A. Operations Mission to Jordan
1955 The Hashemite Kingdom of Jordan, 1:25,000. Ministry of Economy
and the U.S.A. Operations Mission to Jordan, Amman.
Mitasova, Helena, and J. Hoerka
1993 Interpolation by Regularized Spline with Tension: II. Application to
Terrain Modeling and Surface Geometry Analysis. Mathematical Geology
25:657–669.
Mitasova, Helena, and L. Mitas
1993 Interpolation by Regularized Spline with Tension: I. Theory and
Implementation. Mathematical Geology 25:641–655.
349
November 2015 | Advances in Archaeological Practice | A Journal of the Society for American Archaeology
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Morgan, Colleen, and Stuart Eve
2012 DIY and Digital Archaeology: What Are You Doing to Participate?
World Archaeology 44:521–537.
Olszewski, Deborah L., and Nancy R. Coinman
1998 Settlement Patterning during the Late Pleistocene in the Wadi
al-Hasa, West-Central Jordan. In The Archaeology of the Wadi Al-Hasa,
West-Central Jordan, 1:177–204. Arizona State Anthropological Research
Papers, Tempe.
O’Sullivan, David, and Alasdair Turner
2001 Visibility Graphs and Landscape Visibility Analysis. International
Journal of Geographical Information Science 15:221–237.
Ottosson, Torgny
1988 What Does It Take to Read a Map? Cartographica: The International
Journal for Geographic Information and Geovisualization 25(4):28–35.
Pick Jr., Herbert L., and William B. Thompson
1991 Topographic Map Reading. Electronic document, http://oai.
dtic.mil/oai/oai?verb=getRecord&metadataPrex=html&identier=
ADA238026, accessed July 30, 2015.
Pullar, David, and Darren Springer
2000 Towards Integrating GIS and Catchment Models. Environmental
Modelling and Software 15:451–459.
Rast, Walter E., and R. Thomas Schaub
1981 The Southeastern Dead Sea Plain Expedition: An Interim Report of
the 1977 Season. Vol. 46. The Annual of the American Schools of Oriental
Research. American Schools of Oriental Research, Boston.
Rast, Walter E., R. Thomas Schaub, David W. McCreery, Jack Donahue, and
Mark A. McConaughy
1980 Preliminary Report of the 1979 Expedition to the Dead Sea Plain,
Jordan. Bulletin of the American Schools of Oriental Research, accessed
July 30, 2015.
Rosen, Steven A.
1992 Nomads in Archaeology: A Response to Finkelstein and
Perevolotsky. Bulletin of the American Schools of Oriental Research
287:75–85.
Royal Geographic Society of Jordan
1989 Hashemite Kingdom of Jordan, 1:25,000. The Royal Geographic
Society of Jordan, Amman.
Schaub, R. Thomas, and Walter E. Rast
1984 Preliminary Report of the 1981 Expedition to the Dead Sea Plain,
Jordan. Bulletin of the American Schools of Oriental Research 254:35–60.
Schuldenrein, Joseph, and Clark, Geoffrey A.
1994 Landscape and Prehistoric Chronology of West-Central Jordan.
Geoarchaeology 9:31–55.
2003 Prehistoric Landscapes and Settlement Geography Along the Wadi
Hasa, West-Central Jordan, Part II: Towards a Model of Palaeoecological
Settlement for the Wadi Hasa. Environmental Archaeology 8:1–16.
Shi, Wenzhong, Chui Kwan Cheung, and Changqing Zhu
2003 Modelling Error Propagation in Vector-based Buffer Analysis.
International Journal of Geographical Information Science 17:251–271.
Stanislawski, Lawrence V., Bon A. Dewitt, and Ramesh L. Shrestha
1996 Estimating Positional Accuracy of Data Layers Within a GIS Through
Error Propagation. Photogrammetric Engineering and Remote Sensing
62:429–433.
Stone, Elizabeth C., and Paul Zimansky
1992 Special Report: Mesopotamia in the Aftermath of the Gulf War.
Archaeology 45(3):24.
Tabik, S., E.L. Zapata, and L.F. Romero
2013 Simultaneous Computation of Total Viewshed on Large High
Resolution Grids. International Journal of Geographical Information
Science 27:804–814.
Theiss, Adam, David C. Yen, and Cheng-Yuan Ku
2005 Global Positioning Systems: An Analysis of Applications, Current
Development and Future Implementations. Computer Standards &
Interfaces 27:89–100.
Thorndyke, Perry W., and Barbara Hayes-Roth
1982 Differences in Spatial Knowledge Acquired from Maps and
Navigation. Cognitive Psychology 14:560–589.
Thorndyke, Perry W., and Cathleen Stasz
1980 Individual Differences in Procedures for Knowledge Acquisition
from Maps. Cognitive Psychology 12:137–175.
Ullah, Isaac I. T.
2011 A GIS Method for Assessing the Zone of Human-Environmental
Impact around Archaeological Sites: A Test Case from the Late Neolithic
of Wadi Ziqlâb, Jordan. Journal of Archaeological Science 38:623–632.
2015a Wadi Hasa Ancient Pastoralism Project, Data Archive, Electronic
document, http://gshare.com/articles/Wadi_Hasa_Ancient_Pastoralism_
Project/1404216, accessed July 30, 2015
2015b r.viewshed.cva. Electronic document, https://svn.osgeo.org/grass/
grass-addons/grass7/raster/r.viewshed.cva/, accessed July 30, 2015.
Ullah, Isaac I. T., and Sean M. Bergin
2012 Modeling the Consequences of Village Site Location: Least Cost
Path Modeling in a Coupled GIS and Agent-Based Model of Village
Agropastoralism in Eastern Spain. In Least Cost Analysis of Social
Landscapes: Archaeological Case Studies, edited by Devin A. White and
Sarah L Surface-Evans, pp. 155–173. 1st ed. University of Utah Press, Salt
Lake City.
Ullah, Isaac I. T., Joseph Schuldenrein, and Mark Smith
2008 Preliminary Report of the 2008 Season of the Wadi Hasa Ancient
Pastoralism Project. Manuscript on le, Department of Antiquities of
Jordan, Amman.
United States Army Map Service
1966 Jordan, 1:50,000 Scale Topographic Map. Series K737. Army Map
Service, Washington, D.C.
United States Geological Survey (USGS)
2003 LANDSAT ETM+ Scene: LE71740392003006EDC03. L1G,
Orthorectied, Terrain Corrected. USGS Earth Resources Observation and
Science (EROS) Center, Sioux Falls.
2012 Shuttle Radar Topography Mission, Scenes: SRTM3N31E035V2,
SRTM3N31E035V2. Version 3, 3 Arc-Seconds. USGS, US Dept. of the
Interior, Washington, D.C.
Wells, Joshua J., Eric C. Kansa, Sarah W. Kansa, Stephen J. Yerka, David G.
Anderson, Thaddeus G. Bissett, Kelsey Noack Myers, and R. Carl DeMuth
2014 Web-Based Discovery and Integration of Archaeological Historic
Properties Inventory Data: The Digital Index of North American
Archaeology (DINAA). Literary and Linguistic Computing 29:349–360.
Wheatley, David
1995 Cumulative Viewshed Analysis: A GIS-Based Method for
Investigating Intervisibility, and Its Archaeological Application. In
Archaeology and Geographic Information Systems: A European
Perspective, edited by Gary Lock and Zoran Stancic, pp. 171–186.
Routledge, London.
Wheatley, David, and Mark Gillings
2002 Spatial Technology and Archaeology: The Archaeological
Applications of GIS. Taylor and Francis, London.
White, Devin A., and Sarah L. Surface-Evans (editors)
2012 Least Cost Analysis of Social Landscapes: Archaeological Case
Studies. 1st ed. University of Utah Press, Salt Lake City.
Wilkinson, T. J.
2004 The Disjunction between Mediterranean and Near Eastern Survey:
Is it Real? In Mediterranean Archaeological Landscapes: Current Issues,
Museum of Archaeology and Anthropology, University of Pennsylvania,
Philadelphia, edited by Efe Athanassopoulos and LuAnn Wandsnider,
pp. 55–68. University of Pennsylvania Museum of Archaeology and
Anthropology, Philadelphia.
Witcher, R. E.
2008 (Re)surveying Mediterranean Rural Landscapes: GIS and Legacy
Survey Data. Internet Archaeology 24.
350
Advances in Archaeological Practice | A Journal of the Society for American Archaeology | November 2015
Integrating Older Survey Data into Modern Research Paradigms (cont.)
Zubrow, Ezra B.W.
2006 Digital Archaeology. In Digital Archaeology: Bridging Method and
Theory, London, edited by Thomas L. Evans and Patrick Daly, pp. 10–31.
Routledge, New York.
NOTES
1. The intentional scrambling of civilian GPS signals by an executive order
signed in 1996 by President Clinton was ofcially ended on May 2, 2000.
All civilian GPS signals had a maximum spatial accuracy of ± 100 m up
to the May 2000 date. The intentional randomization of the civilian GPS
signal was called “selective availability” and is not the same as the error
induced by atmospheric aberrations that are removed in “differential”
GPS systems that receive external radio-broadcast corrections to these
aberrations.
2. From 1993–2000, it was at least possible use a handheld GPS unit to help
orient oneself and to get a rough estimate of the site location prior to the
manual process of coordinate identication.
3. Of the surveys listed here, only the TBAS occurred in the GPS era, but
its rst season occurred before the end of “selective availability” for
consumer GPS units, and so was subject to the ± 100 m error of early
civilian GPS signals.
4. Note that, in some cases, it may be impossible to nd the original eld
maps. In these cases, it may be possible to use “publication” maps that
accompany reports, articles, or manuscripts related to the eldwork,
especially if these are large-format “pull-out” maps (e.g., like those that
accompany the ASKAP and WHS monographs). In any case, because they
are “cleaned” prior to publication, such maps will be less detailed than
original eld maps and will mask clues to other sources of error.
AUTHOR INFORMATION
Isaac I. T. Ullah n Arizona State University School of Human Evolution and
Social Change, PO Box 2402, Tempe, AZ 85287-2402 (iullah@asu.edu)
