2

First principles

2.1 Introduction
The power of GIS, as with other computer programs, can be deceptive: visually
impressive but ultimately meaningless results can appear unassailable because of
the sophisticated technologies used to produce them (Eiteljorg 2000). The famil-
iar adage ‘garbage in, garbage out’ is particularly applicable to GIS, and one of
our primary aims throughout this book is to provide guidance on how to use this
technology in ways to strengthen and extend our understanding of the human past,
rather than to obfuscate it. In this chapter we start by providing an overview of the
‘first principles’ of GIS: the software and hardware requirements, geodetic and car-
tographic principles, and GIS data models. These provide the conceptual building
blocks that are essential for understanding what GIS is, how it works, and what
its strengths and limitations are. Although some of these ‘first principles’ may be
familiar to readers who are experienced in cartography and computer graphics, we
nevertheless provide a thorough review of each as they yield the foundation on
which we build in later chapters.

2.2 The basics

2.2.1 GIS functionality
What does a GIS do? Simply providing a definition of GIS and referring to its
abilities to capture and manipulate spatial data doesn’t provide much insight into
its functionality. More informative is to break some of the basic tasks of a GIS into
five groups: data acquisition, spatial data management, database management, data
visualisation and spatial analysis. Some of the routine tasks performed under these
headings are outlined in Fig. 2.1 and described in Box 2.1.
While each of these tasks are important in themselves, above all GIS should
be considered as both an integrated and as an integrating technology that pro-
vides a suite of tools that help people interact and understand spatial information.
It is important to stress that although the origins of GIS are strongly rooted in
digital cartography, GIS is not just about ‘maps’ nor is it necessarily only about
the digital manipulation of the sorts of information and methods that are usually
depicted on maps (cf. Longley et al. 1999). The use of GIS has a much broader
contribution to make in terms of understanding spatial and even space–time rela-
tionships between natural and anthropogenic phenomena (Couclelis 1999). Indeed,
it is increasingly common to make the distinction between the software tools used
to process geospatial data (GIS), and a geographic information science (‘GISc’)

11

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 12 First principles

Fig. 2.1 The five main groups of tasks performed by GIS (after Jones 1997, Fig. 1.2).

that is concerned both with the more fundamental conceptual issues of spatial and
space–time relationships as well as the impact geospatial technologies are having
within the humanities and social sciences (Marble 1990; Curry 1998; Forer and
Unwin 1999; Johnston 1999; Longley et al. 2005).

2.2.2 Geographic information

There is considerable overlap between the aims of the disciplines of archaeology
and geography as both share an interest in exploring and interpreting the spatial
structure and organisation of human societies at scale from the micro to macro (e.g.
Clarke 1977). We thus treat the term ‘geographical’ as an inclusive one that tran-
scends the discipline of geography (Couclelis 1999). ‘Geospatial information’ (GI)
can therefore be broadly defined as information about natural and anthropogenic
phenomena and their relationships with each other. Geospatial information can also
describe micro-scale phenomena, such as the patterning of erosion on the facades
of historic buildings or the distribution of cut marks on bone (e.g. Marean et al.
2001; Abe et al. 2002). Most archaeological data – whether artefacts, ecofacts,
features, buildings, sites or landscapes – have spatial and aspatial attributes that can
be explored using GIS (Fig. 2.2). These attributes include:
r A spatial location that tells us where the information is in either a local or global context.
A location can be defined by a qualitative term such as ‘in Texas’ or ‘next to the river’
or quantitatively using map coordinates. A large component of social-science research
uses qualitative positioning data such as counties, cities, census districts or postcodes to
form the unit of analysis. Qualitative locations are less commonly used in archaeology

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.2 The basics 13
Box 2.1 GIS tasks and descriptions

The acquisition of spatial data GIS is a software platform for the acquisition and integration
of spatial datasets. Spatial data include, but are certainly not limited to, topographic maps,
site locations and morphology, archaeological plans, artefact distributions, air photography,
geophysical data and satellite imagery, all of which can be integrated into a common analytic
environment.
Spatial data management GIS uses sophisticated database management systems for the storage
and retrieval of spatial data and their attributes. This might involve the transformation of
map coordinate systems to enable data collected from different sources to be integrated,
the building of vector topologies, the ‘cleaning’ of newly digitised spatial datasets, and the
creation of geospatial metadata.
Database management A major strength of GIS is that it provides an environment for linking
and exploring relationships between spatial and non-spatial datasets. For example, given a
database on the provenance of a sample of projectile points, and another database that contains
information on the morphology of the same points, they can be linked in such a way that it
becomes possible to look for spatial patterns in points’ morphological variability. Database
management, involving conceptual and logical data modelling, is thus an important part
of GIS, as is database construction and maintenance to ensure that the spatial and aspatial
components of a dataset are properly linked.
Spatial data analysis GIS also provides the ability to undertake locational and spatial analysis
of archaeological data, as well as tools for examining visibility (viewsheds) and movement
(cost-surfaces) across landscapes. Much work in GIS involves the mathematical combination
of spatial datasets in order to produce new data that may provide insight into natural and
anthropomorphic phenomena. These range from ecological models that provide predictions
of soil suitability for agriculture or erosion potential, or predictive models of potential site
location. Tools for geostatistical modelling of spatial data to create, for example, continuous
surfaces from a set of discrete observations are also available. GIS can also be a route to
the computer simulation of human behaviour and decision making in different types of
environments.
Spatial data visualisation GIS has powerful visualisation capabilities used for viewing spatial
data in innovative ways (such as thematically or for ‘fly-throughs’ in three dimensions) that can
suggest potential patterns and routes for further analysis. GIS also provide cartographic tools
to help produce hard-copy paper maps. Many GIS packages also facilitate the publication of
interactive map data on the Internet.

although we may, at times, use locations such as parish, county or survey region. More
frequently we use quantitative location data in the form of map coordinates. These
include global geographic locational systems, with latitude and longitude being the most
common, or national, regional or locally defined Cartesian metric coordinate systems.
r A morphology that defines the shape and size of an object, such as ‘straight’ or
‘100 m2 ’. Qualitative or quantitative descriptors can be recorded as attribute data by,
for example, recording the size of an archaeological site or the shape of a distribution.
Alternatively, it is possible to record spatial morphology directly by mapping the size
and shape of a phenomenon, such as an archaeological site on a map. For certain
analytical or visual purposes, morphology might be drawn directly on a map, such as
the arrangement of a skeleton or the shape of a distribution of artefacts.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 14 First principles

Spatially Classification
related to

Aspatially Archaeological Spatial attributes

related to Phenomenon

Temporally Aspatial attributes

related to
Fig. 2.2 The possible spatial and aspatial characteristics of archaeological data.
A burial, for example, is spatially related to other adjacent burials, aspatially related to
other burials of that gender or age group, temporally related to other burials of that
period, might be classified by the specific position of the body, will have a spatial
location and morphology, and possess any number of aspatial attributes (e.g. lists of
grave goods and qualitative/quantitative characteristics of the skeleton). After Jones
1997, Fig. 2.8.

r Information about spatial association and interaction that describes spatial relation-
ships, such as ‘path a crosses path b’, ‘from settlement p one can see settlement q’ or
‘site k is 100 m east of fresh water’. As we discussed in Chapter 1, some types of spatial
associations are referred to as topological, such as when we talk about path or road
connections. Topological relationships are also described as orientation-independent
because only the connective relationships between objects are important and not their
orientation or spatial location. Orientation-dependent or directional relationships are
those that use relational directions, such as above, below, in front, behind, or the
cardinal directions east, west, south, north (Jones 1997, p. 25).
r Temporal relationships that describe the date and/or associated features in relative
terms, like ‘contemporary with’, ‘later than’, ‘earlier than’, etc. Temporal relationships
can be important for ensuring that particular types of analysis, such as settlement
patterns, are undertaken only on contemporaneous sites.
r One or more aspatial attributes that describe the nature of the object. This might
consist of a biography of a site or object, information about the colour and raw material
of an object, the time of day that a field was fieldwalked, the shape of the cross-section
of a feature, or the estimated age of a burial.

The ability to associate aspatial data with spatial objects means that it is pos-
sible to explore the spatial characteristics of non-spatial data. For example, given
a database of handaxes that records their spatial provenance and aspatial morpho-
logical attributes (e.g. weight, size, shape, raw material, reduction stage, etc.) it is
possible to explore the relationship between their location and their other charac-
teristics. This is an extremely important ability of GIS that has found application
in many areas of archaeological research.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.2 The basics 15
2.2.3 Components of a GIS
A GIS is a computer-dependent technology. In addition to the computer itself, there
are a number of other important components to a GIS. The most important ones
are:
Software In order to qualify as a GIS the software must have: (i) a spatial database that
stores and manages spatial objects; (ii) some mechanism of linking attribute data to
these spatial objects, either as an internal function of the GIS package, or by provid-
ing functions that enable access to external database systems; (iii) a ‘geoprocessing
engine’, which permits the manipulation and analysis of the geospatial information
stored in the spatial and attribute databases. None of the many GIS packages cur-
rently available perform all tasks equally well. The choice of software consequently
needs to be made with respect to several factors, including the tasks it is needed for,
what operating system it has to run under (e.g. UNIX or UNIX-like systems such as
MacOS X, Linux, Irix, or Solaris; or one of the versions of Microsoft Windows) and
the size of the budget for software, hardware and training costs. A large number of
packages are available – too many for us to attempt to review – each with their own
strengths and weaknesses in terms of ease of use and the range of analytical tools they
offer. An afternoon’s research on the web will provide a reasonable grounding in the
range of software options. If cost is a primary concern, it is worth knowing that one
of the more powerful GIS packages, GRASS GIS,1 is available free of charge under
an open-source licence. Excellent comprehensive commercial GIS packages include
Idrisi,2 the ArcGIS suite of programs,3 and MapInfo,4 and all may offer discounts for
educational users.
Hardware In addition to the computer that runs the software, which could range from
a small palmtop computer to a large institutional mainframe, there are several other
hardware components that are essential to making a GIS work. These can be divided
into two groups. The first consists of input devices, which might be limited to the
keyboard and mouse supplied with the computer, but could extend to digitising tablets,
flatbed and roll scanners, digital surveying equipment such as global positioning
system (GPS) devices and Total Stations, or geophysical sensors. Chapter 5 discusses
the various methods for acquiring digital data in some detail. The second group
consists of the output devices needed for viewing and sharing information. A computer
monitor is the basic piece of display hardware but, with the obvious exception of the
WWW, it is not a very convenient device for distributing information to other people.
Some type of printer, from standard letter devices to larger colour plotters, is needed
for producing the maps, graphs and tables that GIS routinely produces. We review
map production and spatial data communication in Chapter 12.
People GIS operators are the most crucial part of the system as they are responsible for
the design and analysis of spatial datasets. A GIS is never a fully objective process –
data and questions can rarely be simply ‘fed’ to a GIS and useful results returned – so
it is essential that the specialists responsible for digitising, processing and analysing
data are closely integrated with both project design and data collection. This is less
of an issue when one researcher is conducting both the project design, data collection
and analysis, but in large research projects or commercial archaeological units it is

1 http://grass.itc.it. 2 www.clarklabs.org.
3 www.esri.com. 4 www.mapinfo.com.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 16 First principles
important to ensure that GIS analysts are included at the earliest stages of the project
design to prevent any disjuncture between the envisioned project aims and outcomes.
A GIS will rarely contribute in any meaningful way if it is tacked-on as an extra and
handed to a ‘GIS-person’ who has no real understanding of the original goals of the
project in question.

2.3 Cartographic principles

2.3.1 Maps, digital cartography and GIS
A major element of GIS is the visualisation, management and analysis of spatial
data presented in the form of digital maps. It is consequently important to emphasise
that all maps, whether paper or digital, simplify the world and present an abstract
model of spatial phenomena. Maps can be divided into two basic types:
Topographic maps provide general information about the physical surface of the earth,
including natural and human-made features like roads, rivers, settlements and eleva-
tion. These exist at a variety of different formats and scales, each suited to particular
purposes. Navigational air-charts, for instance, are compiled at a scale which is useful
for pilots (1 : 500 000) and emphasise topography, settlements, restricted air-space and
airports. In the UK, the Ordnance Survey produce a variety of different topographic
maps (in both paper and digital formats) showing elevation, natural and cultural land-
scape features (including archaeological and historical sites and monuments), roads,
towns and villages that are suitable for a range of different applications. The US Geo-
logical Survey produce equivalent maps for the USA and most countries have similar
organisations (e.g. Canada’s Centre for Topographic Information, and Geoscience
Australia).
Thematic maps provide specific information about a single feature of the landscape or
environment, or display information about a single subject. When the data values vary
continuously through space it is common to display them on isarithmic maps, which
use lines to connect points of constant numeric value, such as elevation (contours),
temperature (isotherms), precipitation (isohyets) or even frequencies of hailstorms
(isochalazes). Other themes are more likely to be displayed on choropleth maps,
which use shading or symbols to display average values of information in different
areas, such as vegetation, geology or numbers of artefacts collected in a survey unit.

To emphasise the differences between traditional paper maps and the dynamical
interface that GIS offers it is worth noting some of the constraints of the former (cf.
Longley et al. 1999, p. 6). Paper maps differ from GIS because they are:
Static The dynamic space-time interactions between objects cannot easily be depicted
(e.g. changes in population and settlement patterns, or environmental change). A GIS
offers the advantage of enabling exploration of the dynamics of temporal patterning.
The University of Sydney Archaeological Computing Lab’s TimeMap project5 is an
excellent example of this form of dynamic mapping.
Two-dimensional Multidimensionality cannot be easily depicted on paper. Multivariate
spatial data and the three-dimensional representation of topography benefit from
multidimensional forms of display available in GIS (e.g. Portugali and Sonis 1991;
Couclelis 1999).

5 www.timemap.net.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.3 Cartographic principles 17
Flat Representing a curved three-dimensional surface, such as the Earth, in two-
dimensions often introduces significant distortion in spatial measurements (see below)
and GIS provides facilities for improving this.
Precise The traditional methods of cartographic representation do not allow for the
depiction of imprecise, ‘fuzzy’, boundaries (that occur between, for example, vegeta-
tional zones, cultural boundaries, etc.). While this remains a problem for some forms
of spatial representation in GIS, there are more possibilities for working with less
clearly defined boundaries than paper maps traditionally offer.
Difficult to update Once committed to paper, a map is fixed and can only be updated
by producing a new map, whereas a digital map may be updated continuously – even
in real-time.
Difficult to relate to non-spatial data The attributes of the objects on traditional maps
have to be coded and further information can only be found by reference to a gazetteer.
A GIS has several advantages over non-digital systems with regards to attribute data:
in particular, a GIS offers more comprehensive data retrieval, ease of update and an
ability to explore data patterns more quickly compared with its paper counterpart.

A further major advantage of a GIS over traditional mapping is that a GIS permits
the organisation of different components of the same map into different thematic
map layers (and thus often referred to as thematic mapping), which is the basic way
that spatial data are organised within a GIS environment. In practice this means
that in one GIS digital display many different elements may be combined, each
of which can be individually turned on or off, queried, modified, reclassified and
edited. Many analytical functions, such as spatial queries, can operate across one
or more layers depending on the need of the GIS analyst. Map layers, or subsets
of individual layers, can also be combined to produce new maps at will, providing
potential insight into relationships between elements on different themes.

2.3.2 Map projection systems and geodetic datums

A basic property of a map is that it is has a spatial context – more properly, geo-
referenced – by implicitly or explicitly referring to positions on the Earth’s surface.
Obviously with many maps a precise and absolute spatial context is not important;
a quick sketch of the route to a friend’s house serves a purpose even though it may
be inaccurate and relative. However, when precision and absolute spatial context
are important, then an explicit system of measurement is required. As the Earth is a
complex shape this is not a trivial process and the science of geodesy is concerned
with the measurement of the morphology of the Earth’s surface. The shape of the
Earth is best approximated by a flattened sphere, referred to as either an ellipsoid or
geoid, and positions on it can be defined using polar or geographical coordinates
(Fig. 2.3). Geographical coordinate systems define degrees, minutes and seconds
north or south of the equator as latitude and degrees, minutes and seconds east or
west from Greenwich Prime Meridian as longitude.
This is an elegant and simple solution for locating positions on the planet. It is
less suitable for representing the surface of the Earth on a two-dimensional plane,
for example, on a paper map or computer screen. The name given to a system used

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 18 First principles

f q x

Fig. 2.3 Polar coordinates. The circle of the sphere in the x, y-plane is the equator, and
in the x, z-plane it is the meridian. If p is an arbitrary point on the surface of the Earth,
then the angle defined by θ is therefore longitude, and the angle defined by φ is latitude
(after Worboys 1995, p. 143).

to display areas of the Earth’s round surface on a flat map is map projection, which
involves a mathematical transformation of the units of longitude and latitude (i.e.
graticules) to a flat plane. Essentially, a flat map of a large area of the Earth’s surface
cannot be produced without some form of projection. When mapping areas at the
continental or international scale the transformation from three to two dimensions
causes profound distortion and spatial error in particular types of measurement. At
national and regional scales or larger, the distortion arising from projection to a
flat surface causes fewer problems, and national and state mapping agencies have
established projections for minimising error within their own boundaries. At very
small scales, what we might term subregional or local, the surface of the Earth can
be regarded as flat and grid systems can be established and used without reference to
geodetic correction. Note here the use of the terms ‘large scale’ and ‘small scale’, as
this can be a source of confusion. Large scale generally refers to scales of 1 : 50 000
or greater (e.g. 1 : 25 000, 1 : 5000, etc.), and small scale to maps with scales smaller
than 1 : 50 000 (e.g. 1 : 100 000, 1 : 1 000 000, etc; Thurston et al. 2003, p. 37).
Many forms of map projection have been developed for both global and national
mapping purposes and most GIS programs will support many or all of the common
ones (GRASS, for example, supports some 123 different projections). Projection
systems may be grouped into a projection family of which there are three main
ones, conical, azimuthal and cylindrical, defined according to how the sphere is
projected onto a flat surface (for a mathematical discussion see Iliffe 2000). Each
projection family has either a line of tangency or two lines of secancy that define
where the imagined projection surface comes into contact with the Earth, and where
there is correspondingly the least distortion (Fig. 2.4). All projections will distort

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.3 Cartographic principles 19

Fig. 2.4 A conical projection with two lines of secancy (left) and one line of tangency
(right). The point(s) of contact are also referred to as standard parallels.

one or more of the parameters of distance, direction, scale, conformality (shape)

and area, although each projection family attempts to minimise distortion in one or
two parameters at the expense of increasing it in others.
In addition to the three projection families, there are four projection groups
defined on the basis of how this distortion is managed. Conformal or orthomorphic
projections preserve the 90◦ intersection of lines of latitude and longitude to ensure
correct angle measurements between points, but in so doing distort area measure-
ments. Equal-area projections preserve area calculations, so that the multiplication
of the two edges of rectilinear features represented on a map and globe will be iden-
tical (but the properties of shape, angle and scale are then distorted). Projections
that maintain distances between one or more pairs of points are described as equi-
distant projections. Any given equidistant projection will only apply to measure-
ments taken in a certain direction: sinusoidal equidistant projections, for example,
enforce the measurements parallel to the equator, but distort measurements parallel
to the meridian. True-direction projections maintain the correct angle from any line
measured from the centre of the projection to any other point on the map.
A projection is defined by the combination of a family and then a projection
type. For example, a conical projection can be conceptualised as fitting a cone over
one of the polar regions as depicted in Fig. 2.4, which is then cut along a meridian
as in Fig. 2.5.
The result is a map in which the lines of longitude are straight and convergent,
and lines of latitude are concentric arcs. The line of tangency on conic projections
is referred to as the standard parallel and distortion increases the further one moves
away from this line. The amount of distortion can, however, be controlled by altering
the spacing of the lines of latitude; if evenly spaced then the projection will be
equidistant along the north–south axis (equidistant conic projection); if compressed
at the northern and southern ends, then the projection becomes equal-area (Albers
equal-area conic projection).
Azimuthal (or planar) projections represent the Earth’s surface on a flat plane
using a single point of contact rather than a line of tangency (Fig. 2.6). Azimuthal

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 20 First principles

Fig. 2.5 Albers equal-area conical projection with one line of tangency (left) and a
meridian (dashed line). The resulting map is to the right, showing the lines of latitude
as concentric arcs.

Fig. 2.6 Azimuthal projection with a point of contact at the North Pole. The resulting
map has radiating lines of longitude, and concentric lines of latitude. Angle and
distance measurements taken along the lines of longitude remain accurate.

(or planar) projections are usually used to map the poles although in theory they can
occur anywhere on the Earth’s surface. If polar, then the projection is conformal
with concentric lines of latitude and radiating lines of longitude. Area distortion
occurs as one moves away from the poles, but directions and linear distances from
the centre point to any other point on the map are accurate.
Cylindrical projections are conformal and so 90◦ angles are maintained between
the lines of latitude and longitude (Fig. 2.7). Measurements along the line of tan-
gency are equidistant but at further distances from this line area measurements
become increasingly distorted.
The most common cylindrical projection is the Mercator Projection, which
uses the equator as its line of tangency and scales the y-dimension (latitude) to
reduce the distortion at polar extremes. This projection gives a very misleading
view of the world as movement away from the equator causes areas towards the top
and bottom of the map to become disproportionately large in area (Snyder and
Voxland 1989, p. 10). The Transverse Mercator Projection (TM projection),
invented by Johann Lambert (1728–1777), rotates the cone 90◦ so that a merid-
ian becomes the line of tangency. This distorts measurements in the east–west
axis but maintains north–south measurements better than the standard Mercator
Projection. The TM Projection is one of the standard ways of mapping the globe.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.3 Cartographic principles 21

Fig. 2.7 Cylindrical projection with a line of tangency corresponding to the equator
and a meridian (dashed line). The resulting map is to the right, showing the lines of
latitude as parallel lines.

Finally, the Universal Transverse Mercator Projection (UTM) is a twentieth-

century modification of the TM Projection that divides the world into 60 vertical
zones, each of which are 6◦ of longitude wide. There is a central meridian in each of
these 60 zones that minimises measurement distortion in the east–west to approx-
imately 1 m in every 2500 m (Robinson et al. 1995; DeMers 1997, pp. 63–64).
Each zone is divided into rows of 8◦ latitude (12◦ in the northernmost section)
which equates to a 100 000-m wide grid square. The central meridian is given a
false easting value of 500 000 m to eliminate the need for negative numbers when
specifying east–west coordinates. For the same reason the equator is given a north-
ing value of 0 m for measurements in the northern hemisphere, and 10 000 000 m
for measurements in the southern hemisphere. Universal Transverse Mercator co-
ordinates are given by first specifying the zone and then the easting (with 6 digits
for 1 m precision) and northing (with 7 digits for 1 m precision). The UTM projec-
tion is very popular in GIS and related geospatial technologies like remote sensing
because of its global application, minimal distortion and metric coordinate system.
Most GPSs are able to record locations in UTM coordinates, making it an ideal
system for spatial data collection when a local grid system is not available.
In addition to the projection system used to make the map, it is also important
to be aware of which mathematical approximation of the shape of the Earth was
used for the construction of a map. The Earth is not an exact ellipsoid, since the
surface is not smooth and the poles are not equidistant from the equator. Polar
coordinates of latitude and longitude are therefore calculated using a mathematical
approximation of the Earth’s shape and its centre. Several different approximations
have been calculated, often for a specific region of the planet. Clarke’s 1866 cal-
culations formed the basis for the 1927 datum of North America (North American
Datum 27, or NAD27). NAD27 is being replaced by satellite-derived measure-
ments of an ellipsoid called NAD83 but many organisations still use measurements
and locations using the earlier geodetic datum. The Geodetic Reference System

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 22 First principles
(GRS80), World Geodetic System 84 (WGS84) and European Terrestrial Refer-
ence System (ETRS89) are more recent recalculations of the ellipsoid used in
Europe. National mapping agencies generally use whichever ellipsoid calculations
most closely fit their needs. For example, most national mapping in Great Britain
uses the OSGB36 Datum based on the 1830 Airy ellipsoid, although the Ordnance
Survey have adopted the ETRS89 ellipsoid for more recent mapping derived from
GPS receivers.

Coordinate transformation and reprojection

Maps that share a projection system (e.g. UTM) but are based on different ellipsoids
(e.g. WGS84 versus NAD27) are not compatible, nor are maps that use different
projection systems (e.g. Transverse Mercator versus State Plane) but share the same
ellipsoid (e.g. NAD27). For example, the physical distance between two points that
have identical geographical coordinates, but one based on NAD27 and the other
based on NAD83, can be as much as 100 m apart in the USA. For data from multiple
map sources to be combined, the maps must share a common projection and geodetic
reference system. If this is not the case the projections and/or reference system must
be altered through a process called secondary transformation or reprojection. To
compute the transformation, fairly specific information is required about the existing
and desired projection systems. Most GIS packages provide tools to transform maps
from one geodetic datum to another, and dedicated software tools are also available
to help convert between NAD27 and NAD83 (see, for example, the directory of
software on the US National Geodetic Survey’s website6 and their online conversion
tool7 ). Details about the ellipsoid and projection system used in the construction of
a map are typically printed in the corner or are contained in an associated metadata
record for digital data (see Chapter 13 for further discussion of metadata elements).

2.3.3 National and regional grid systems

Many GIS programs use geographic coordinates of latitude and longitude as the
basis for regional maps (most often as decimal degrees where minutes and seconds
are converted to decimal units so that, for example, 30 minutes 30 seconds is equal
to 0.508 of a degree). While decimal degree systems can work well in GIS packages
that are able to manage the corrections for spatial measurement, a Cartesian system
based on metric units, such as UTM or national (military) grid system, is often a
better choice because of the advantages it offers for calculating distances and areas.
When a two-dimensional Cartesian grid system is used for mapping, east–west mea-
surements are located on the horizontal x-axis and called eastings, and north–south
measurements are located on the vertical y-axis and are called northings (Fig. 2.8).
At larger scales (e.g. greater than 1 : 1 000 000) most national or regional mapping
systems provide metric planar coordinates alongside, or in place of, latitude and
longitude. Metric planar coordinates are used in the global UTM projection, the

6 www.ngs.noaa.gov. 7 www.ngs.noaa.gov/cgi-bin/nadcon.prl.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.3 Cartographic principles 23

50

y-axis (northings)
p (67, 31)
67

31

0,0 x-axis (eastings) 100

Fig. 2.8 A two-dimensional Cartesian coordinate system. Point p is located by

reference to its distance from a 0, 0 datum in the x- and y-planes (respectively referred
to as the ‘eastings’ and ‘northings’).

50
y-axis (northings)

p (67, 31)
c
b

q (34,18) a

0,0 x-axis (eastings) 100

the linear distance between points p and q (c), Pythagoras’

Fig. 2.9 To calculate √
theorem is used: c = a 2 + b2 . As a and b are
√ known (a = x p − xq = 33,
b = y p − yq = 13), the calculation is c = 332 + 132 = 35.5.

US State Plane system, British National Grid and in most other national grids.
National grid systems, such as the US State Plane system, are often better choices
for regional mapping projects because the ellipsoid is often selected to maximise
spatial accuracy for the specific area covered by that particular system. In parts of
the world where national or military grids are unavailable, then UTM is an excellent
choice. We must emphasise again our warning from the previous section regarding
the inevitable and significant spatial errors that will result from combining data
derived from maps with different projections and/or ellipsoids.
Metric planar systems have the important and crucial advantage of allowing the
easy calculation of distance and area. For example: linear distance measurements
can be calculated using Pythagoras’ theorem (Fig. 2.9); polygon areas can be

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 24 First principles
calculated using a system that first breaks the shape into smaller trapezia and then
sums their individual areas to derive the total area; and the geometric centre of a
polygon (its centroid) can be found by taking the mean of the coordinates of all
vertices that define the polygon (for alternatives, see Jones 1997, p. 66; Burrough
and McDonnell 1998, p. 63).

2.4 Data models and data structures: the digital representation

of spatial phenomena
How does a GIS represent spatial data? The roots of GIS originate with the devel-
opment of automated mapping in the middle of the last century. In the late 1950s
some of the basic computer algorithms for handling geographic information were
developed, including the principles for digital cartography, at about the same time
that technology had developed to incorporate computer graphics (e.g. Tobler 1959).
The Canada Geographic Information System, developed in 1963 to manage natural
resources, was a natural outcome of these developments and qualifies as the first
GIS. It was followed soon after by the development of other systems that were capa-
ble of automated mapping (Foresman 1998; Tomlinson 1998). Automated mapping
offered considerable time savings over traditional paper methods by providing faster
and more accurate facilities for the management and updating of spatial data. These
early systems relied on point, line and polygon ‘geographic primitives’, which still
form the building blocks of modern vector-based GIS.
A GIS works by manipulating the digital representations of real world entities.
However, a GIS only has a finite set of resources with which to replicate the infinitely
complex world and, as a consequence, the digital representations used by GIS are
necessarily schematic and generalised. The representation of elements of reality in
this way is referred to as a data model. In GIS, data models tend to be very simple
representations of reality, although as we shall see in later chapters, simple models
may become the building blocks for more complex models that are designed to
quantify relationships between different entities.
As we saw in Chapter 1, GIS represent spatial data using one or both of the
entity and continuous field data models. These are usually implemented as vector
and raster data structures, respectively. Raster and vector data structures store,
represent and manipulate spatial information in very different ways. Certain types of
entities are more typically represented in one format or the other, although much of
the data that archaeologists routinely encounter can ultimately be represented using
either structure. Until recently, these two data structures were virtually mutually
exclusive: GIS programs tended to rely on either one or the other, forcing users to
make a decision as to which they would use. Today most GIS permit the mixing of
both raster and vector data as separate thematic map layers, giving users the freedom
to decide on the most appropriate structure without necessarily using a different
program. The following sections outline the differences between the vector and
raster data structures, and provide some examples of ways that different sorts of
data are handled by each format.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.4 Data models and data structures 25

Points Lines Polygons

Fig. 2.10 The three vector ‘geographic primitives’ of points, lines and polygons.

2.4.1 The vector data structure

A vector is a mathematical term that refers to one or more coordinates used to
define an object in Cartesian space. In the vector structure, real-world entities are
represented using one of three geometrical primitives: points, lines or polygons.
Each primitive is defined using one or more x, y-coordinate pairs called vertices,
and are thus described as discrete objects because of their precisely defined locations
and boundaries (Fig. 2.10). Vertices that are located at the ends of discrete lines, or
at their intersections, are called nodes.
For example, points are zero-dimensional objects (for they have no length or
breadth) defined by a single coordinate pair, and lines and polylines (often also
referred to as arcs or edges) are one-dimensional vectors (having the property
of length, but not breadth) defined by two or more coordinate pairs. Polygons,
or areas, are two-dimensional objects defined by three or more coordinate pairs.
Three-dimensional objects are referred to as volumes, but despite the fact that
CAD systems routinely use three-dimensional vector objects, the three-dimensional
vector structure has not yet been widely implemented in GIS.
The discrete nature of every vector object means that, in addition to possessing
its own unique spatial location and morphology, it is a trivial process to provide
each vector object with an identification (id) number. On the basis of this unique
identifier, each and every object can thus be linked to a set of additional non-spatial
attributes that describe additional properties of that object. These properties most
often consist of real-world quantitative and/or qualitative variables that give the
vector object meaning within the GIS (Fig. 2.11).

Vector topology
An extremely important concept that underlies the vector structure is the geomet-
rical relationships between vector objects, referred to as topology. The analysis
of topological relationships is explored more fully in Chapter 11, so here it is
sufficient to note a few basic concepts. Firstly, topological relationships define

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 26 First principles
Vector objects Attribute data

polygon id soil type

1 1 clay
2
2 clay with flint
3 chalk
3

Fig. 2.11 Vector objects linked to attribute data. In this example, each polygon has a
unique id number that links it directly to an attribute table that defines the soil type
represented by that polygon.

the connections and relationships between vector objects rather than their spatial
location. For example, when two roads cross each other, two different topological
relationships can potentially exist between those entities. If the lines simply cross
without sharing a node, the lines are not topologically connected. This is equivalent
to a road crossing another via an underpass and it is not possible to get from one
road to another at the point of intersection. If the roads do share a node, they are
topologically linked. In this case, it would be equivalent to the two roads meeting
at an intersection. Topological relationships are therefore defined by the presence
of shared nodes between vector objects. In practice, many GIS require nodes at
both crossing and meeting points, in which case additional methods must be used
to provide adequate topological information (see Chapter 11).
Topological relationships also define how polygons relate to each other. For
example, two adjacent polygons, perhaps representing separate parcels of land or
survey zones, are topologically related if they share one or more nodes or arcs
in common (Fig. 2.12). Without common nodes this relationship does not exist,
and the polygons then must either overlap and/or have a gap between them. It
is entirely possible that they intentionally overlap or have a gap to reflect a real-
world spatial relationship; but more usually adjacent polygons have an assumed,
if not actual, topological relationship. The calculation of spatial relationships and
properties of vector objects is not a trivial process, and is dependent both on the
data structure and accuracy of the dataset. During the data collection phase and
particularly during the process of digitising vector objects, care should to be taken
to ensure that topological relationships are properly maintained and defined. Many
vector GIS programs have ‘clean-up’ routines that can be used to create topologies
between objects automatically (Chapter 5).
Some geodatabases, such as ArcGIS, provide a set of topological rules to ensure
that vector objects are always related in appropriate ways. For example, polygons
that define survey areas might have a ‘Must Not Overlap’ rule, so that any instances

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.4 Data models and data structures 27

b
a c d

l
1
k 2
m
e
n
3 f
j
g

i h

Fig. 2.12 Three topologically related polygons. Polygons 1, 2 and 3 share arcs (edges)
defined by nodes cm, mj and mg.

where this occurs are identified and the appropriate action taken (e.g. the over-
lapping area is subtracted from one polygon, or a new polygon defined by the
overlapping area is created).
Topological accuracy also makes for more efficient storage of vector data as
vector objects can then share data. Some GIS systems take advantage of this when
storing the geometric definitions by only recording an arc (and its vertices) once,
and then defining its relationship to polygons. In Fig. 2.12, for example, arcs cm,
m j and mg need only be stored once instead of twice for each of the polygon
boundaries they define. On large, complex, polygonal maps such as those routinely
encountered with soil or geological series, this can result in a significant saving
of storage space and computational time, an issue examined in further detail in
Chapter 4.

2.4.2 The raster data structure

Unlike vector graphics, which use coordinate geometry to define the spatial param-
eters of objects, raster graphics use a grid matrix of equally sized cells or pixels
to represent spatial data (Fig. 2.13). Raster maps are therefore defined only by the
number of rows and columns in the grid and the size of each pixel in terms of
actual area covered. Each cell also has a value associated with it that represents the
attribute status of the object at that location. In a digital elevation model (DEM),
for example, each cell has a quantitative value that signifies the mean elevation
across the area defined by that pixel, whereas the pixels in a vegetation map may
be coded to reflect modal vegetation type. The reliance on pixels and the use of a
single attribute per pixel may appear to be very limiting in comparison to the vector
structure, but within the simplicity of the raster structure lies its strength. Raster

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 28 First principles

Point Line Polygon

Fig. 2.13 Point, line and polygon primitives as represented on a raster grid.

datasets are easily combined and mathematically manipulated as computers can

process and display raster data considerably more quickly than vector data because
of the efficiency with which they can store and handle grid data. The simplicity
of the structure does not reduce its functionality for, as we shall see in the next
chapter, the raster data structure can be used to model some extremely complex
spatial phenomena.
A critical variable in the raster structure is the size of the cells, since they define
the resolution of a map by providing a minimum unit of representation. Whereas
a raster map depicting density of archaeological sites across a large region may
consist of a grid of pixels that each represent an area of a square kilometre or more,
a raster map showing the density of artefacts across a site may use pixels that define
an area of a square metre or less. Although computer processing speed and storage
space is continuously increasing, there are nevertheless some practical restrictions
on how much information a typical desktop computer can efficiently process. This,
of course, varies with the specification of the computer, but loss of performance may
be noticed when the total number of pixels is several million or more (e.g. a raster
map representing an area of 50 × 50 km with individual pixel sizes of 10 m2 will
require a grid of 5000 columns by 5000 rows and therefore storage of information
for 25 000 000 pixels). Decisions relating to resolution need to be made very early
on in the model-building process. In the next chapter we show that such decisions
can have important consequences for interpretation of the results of analysis and
for making sense of spatial patterns.

2.4.3 Choosing a data structure

There are many instances where vector systems provide the only sensible means
of answering a specific set of questions, or handling the sorts of multiscalar and
precisely defined data that one might be interested in exploring. On the other hand,
raster data is suitable for powerful spatial modelling, is able to represent continuous
datasets more smoothly, and provides image analysis and classification routines
suitable for aerial photography and satellite imagery. Many modern GIS systems

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.4 Data models and data structures 29
are to a large extent ‘hybrid’ and offer capabilities for manipulating both raster and
vector datasets. As a result it is now common to find both data structures being used
in a single program environment, thus reducing the choice between the raster versus
vector structures to that of appropriateness for the particular needs and questions at
hand. From the philosophical perspective adopted in Chapter 1, the vector structure
is generally most appropriate when the subject matter has been conceived using the
entity model of space. Conversely, the raster structure is usually a better choice if
the subject matter has been conceived as a continuous field.

Advantages and disadvantages of the vector structure

Advantages of the vector structure A major advantage of the vector structure is its
spatial precision. Real-world entities can be drawn and positioned with an accuracy
restricted only by practical limitations such as the precision of the recording equip-
ment. Artefacts, features, sites and other archaeological entities can be integrated
in a single environment, each mapped with as much spatial detail as is required for
analysis. While centimetre-scale precision would not be required for the analysis
of the spatial distribution of sites across a large study region, this level of precision
would, however, be essential for the study of the distribution of chert flakes on a
knapping floor. Finding the balance between spatial precision and the minimum
scale of analysis is crucial: most importantly to prevent spatial errors from influ-
encing pattern recognition. In practice, it is important to recognise that increased
resolution also means increased file sizes, with a corresponding burden on stor-
age and processing time. Another significant advantage of the vector structure is
that vector objects are maintained as distinct entities and can be easily linked to
attribute data records in an internal or external database. For this reason, vector-
based GIS programs have traditionally led the way in terms of database integration,
as complex attribute queries can be performed with relative ease. A vector map may
therefore act as a window into a database, in which each object is described in great
detail.

Disadvantages of the vector structure Vector objects are computationally demand-

ing. Every vertex and node of a vector map must be stored in computer memory and
drawing vector objects requires a considerable amount of processor time. For this
reason, vector data are often much slower to generate on a computer screen than
raster data. The manipulation of vector data is correspondingly intensive; spatial
queries involving, for example, the calculation of areas of overlap within a large
set of polygons needs considerable computer-processing unit (CPU) time. Vector
data also impose properties onto real-world objects that do not necessarily cor-
respond with reality. The most important imposition is ‘boundedness’. Although
many real-world objects do indeed have precise and discrete boundaries, certain
types of data are more ‘fuzzy’ and do not lend themselves to the hard, precise, edges
of vector objects. As vector data cannot readily deal with fuzziness or imprecision,
this can result in artificial precision in some scenarios. An issue related to fuzzy

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 30 First principles
boundaries is the implied non-varying state of an attribute across a vector object.
For example, a polygon used to represent a discrete survey area may possess an
attribute value that represents the density of artefacts in that enumeration unit.
This implies a continuous distribution of artefacts in that area, which in reality is
rarely the case (e.g. Fig. 12.2). An additional attribute could be used to express
the variability within a polygon, but there is no simple way spatially to map con-
tinuous change with vector objects. Elevation is therefore inherently difficult to
represent using discrete vector objects such as points or lines: contour lines, for
example, give an indication of topographic variation at set intervals, but it can be
difficult to predict elevation values between the lines. There are special vector struc-
tures called triangulated irregular networks (TINs) that overcome these difficulties
(Chapter 6), but the case remains that some types of data are less well suited to the
vector structure.

Advantages and disadvantages of the raster structure

Advantages of the raster structure The speed at which raster data can be processed
offers advantages for some applications involving very large datasets and there
are several other key areas in which the raster structure can offer advantages over
vector formats. Firstly, raster data are very good for mapping continuously varying
phenomena, such as elevation, as the continuous cell-based structure is akin to a
continuously varying surface. The raster structure is also very good at represent-
ing real-world entities that have fuzzy boundaries. For example, a distribution of
artefacts collected in a ploughed field could be represented more realistically by
using raster cells that show the changing density of material rather than a single
polygon that arbitrarily defines the site’s area with a single density value. When
this type of information is crucial, then the raster data structure offers a clear
advantage. Secondly, raster datasets can be mathematically manipulated and com-
bined more easily than polygon maps, making it an exceedingly powerful tool for
spatial modelling. A simple model of agricultural potential, for example, may be
constructed by combining data from several different sources, such as raster maps
of elevation, slope, aspect, soil drainage and soil type in a process called map
algebra. Thirdly, aerial photographs, satellite images and geophysical surveys pro-
duce data in raster formats, and the image processing that is often needed to enhance,
classify and make sense of these sorts of data can only be performed in a raster
environment.

Disadvantages of the raster structure There are three major disadvantages of the
raster structure: its fixed resolution, its difficulty in representing discrete entities and
its limited ability to handle multiple attribute data. The first problem arises when
data collected at different scales need to be integrated. Combining multiscalar
datasets could be seen as introducing additional problems regardless of the data
model, and might ideally be avoided, but in practice there are many instances when

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 2.5 Conclusion 31

Fig. 2.14 Representing complex curves with raster data can be problematic. The box
on the far left shows five vector polylines. The centre box shows the same lines using a
10 × 10 raster grid (i.e. 100 cells). On the far right the resolution has been increasing
to a 20 × 20 grid (i.e. 400 cells). This improves the representation, but the raster map
still suffers from being blocky and from lost detail.

data collected at different scales must be combined. Field survey data, for example,
often mix scales of representation from the larger survey unit (such as a field) to
site-based artefact collections where more detail is collected. The representation
of multiscalar data is difficult in raster systems and the combination of raster data
collected at different scales often results in having to default to the smaller scale and
loosing detail. Secondly, problems can arise with representing complex boundaries
using raster data because of the inherent limitations of grid data for representing
tightly curved objects. Unless the cells are very small in relation to the object being
represented and the storage size correspondingly increased, curved lines always
will be blocky in appearance (Fig. 2.14).
For this reason complex shapes, such as contour lines, are better modelled using
vector objects. Finally, raster data have always been difficult to connect to attribute
tables. Although some GIS programs, notably Idrisi and GRASS, provide a facility
for linking raster data to a database, in practice this is often more cumbersome than
the embedded attribute tables that vector-based GIS programs provide. The raster
data structure thus has limitations for the management and querying of multiscalar
spatial datasets.

2.5 Conclusion
Geographical information systems (GIS) are a powerful technology that offer a
host of analytical possibilities for investigating the spatial organisation of culture
and human–environment relationships. These ‘first principles’ of GIS only define
the starting point for exploring the complexity of the human use of space with GIS.
In fact, many of these first principles are being constantly challenged by research
that is pushing beyond the constraints of two-dimensional mapping to use GIS to
model space–time relationships more adequately than the basic vector and raster

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press

 32 First principles
building blocks presented in this chapter. Nevertheless, GIS is – for the time being
at least – still reliant on cartographic principles and a reductionist tendency that
restricts its range of possibilities for representing and interpreting the real world.
Within these limitations, however, there is still a very broad range of ways that GIS
can be used to develop an understanding of human culture in a spatial framework,
and the next chapter provides some real-world examples of how GIS can and does
work in archaeology.

https://doi.org/10.1017/CBO9780511807459.002 Published online by Cambridge University Press