A place in history: a guide to using GIS in historical research

CHAPTER 2: THE WORLD AS VIEWED THROUGH A GIS

2.1 Introduction

This chapter gives a detailed description of the data models that GIS software packages use to store their data. Little of what is said here is directly historical; however, it is important to present a brief overview of the way in which GIS models the world. The chapter also establishes the terminology used throughout this book, as this varies between different authors and different software packages. There are many good introductions to GIS that are appropriate for historians. In 1991 Maguire, Goodchild and Rhind edited a large collection of papers on GIS in two volumes that rapidly became the standard reference work in GIS (Maguire et al 1991). In 1999 a completely re-written second edition of this book was published edited by Longley, Goodchild, Maguire and Rhind (Longley et al 1999). Most of the first edition of the 'big book', as it became known, was then made available online. In 2001 the four authors responsible for the second edition produced a monograph on GIS that provides a wide ranging introduction to GIS at a more affordable price than the edited works (Longley et al 2001). Other good introductions include Chrisman (1997), Heywood et al (1998) and Martin (1996a).

As described in Chapter 1, GIS software uses a conventional database, termed attribute data (sometimes termed 'non-spatial data'). Each record in the attribute data is given a spatial reference using coordinates: these coordinates are termed the spatial data. There are two main types of GIS data model: in vector data models the spatial data consist of points, lines or polygons; while in raster data models the study area is usually sub-divided into square pixels, although other regular tessellations such as hexagons or triangles can also be used. In a vector model, therefore, space is sub-divided into discrete features, while the raster model attempts to represent space as a continuous surface. There are also other ways of sub-dividing the world into surfaces such as triangular irregular networks (TINs), often used to represent relief through digital terrain models (DTMs, also known as digital elevation models or DEMs).

Figure 2.1: Abstracting the real world into layers for the study of book history
Source: MacDonald and Black 2000, 510.

Rather than store all their information about a study area in a single structure, the GIS separates the world into different layers (or coverages or themes) with each layer representing a different type of information. Most GIS texts give examples such as data on relief being one layer, data on the road network another, data on settlements another, data on rivers another, and so on. MacDonald and Black (2000) give an historical example. They are interested in exploring the development of print culture in the 19th century. They argue that the spread of books was partly due to factors directly related to books themselves, including sources of production such as printers and sources of supply such as libraries and bookshops. In addition the spread of books was also influenced by broader factors such as the transport network and demographic factors such as migration. Figure 2.1 shows how they can abstract the real world into layers in order to create a GIS that will allow them to explore book culture.

Understanding that GIS models the world through spatial and attribute data, that spatial data is made up of a small number of crude graphic primitives, and that data on different themes are brought together using layers is critical to an understanding of GIS. This basic model fundamentally defines what can and cannot be done within a GIS and the strengths and weaknesses of all GIS operations. There is an alternative approach to modelling the world in GIS using an object-oriented approach. Here features are not subdivided into layers but are instead grouped into classes and hierarchies. Although there are theoretical advantages to doing this, few commercially available GIS software packages have fully implemented this approach. For this reason object-oriented models will not be discussed in this guide.