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

CHAPTER 5: TIME IN HISTORICAL GIS

5.4 Methods of handling time in historical GIS

Although the temporal functionality included in most GIS software packages is usually very limited, there are many ways that time can be handled with a GIS. The best method to choose will depend on the nature of the individual researcher's data.

One simple way of handling time is to treat it as an attribute. Healey and Stamp do this in their study of regional economic growth in Pennsylvania (Healey and Stamp 2000). For both firms (represented as points) and railroads (represented as lines) the dates of their founding and closure are attached to the spatial features as part of the attribute data. In this way the development of the transport network and industrial development can be examined over time and the links between the two can be studied.

Figure 5.1: Time as an attribute
This example shows the history of three firms over time. Usually the data are in the form of annual output data, as happens consistently for firm 1. Where a firm opens or closes this can be represented using the attribute data: for example, firm 2 closes on 16 May, 1871. Changes to the names of firms can also be handled in this way: for example firm 3 has its name changed from Frasers to Bloggs in 1872.

The simplest way to implement this is with a single row of attribute data attached to each spatial feature. Multiple rows can also be attached to each spatial feature with each row having a start and end date. This allows us to handle complex situations, for example, where the aim is to monitor a firm's economic statistics, such as output, profit, and employment, but where the name and ownership of the firm also changes over time. A simplified example of this is shown in Figure 5.1. Handling time in this way allows spatial features to be created and abolished and their attributes to change over time. The limitation of this approach is that the location of features cannot change.

Where the temporal nature of the data is more explicitly spatial, different layers can be used to represent the situation at different dates. This is termed the key dates approach and is particularly suitable where spatial data are taken from source maps from different dates. A good example of this approach is taken by Kennedy et al. (1999) in their atlas of the Great Irish famine. The atlas uses census data to show demographic changes resulting from the famine. At its core are layers representing the different administrative geographies used to publish the censuses of 1841, 1851, 1861 and 1871. These layers are linked to a wide variety of census data from these dates. This allows sequences of maps to be produced showing, for example, how the spatial distribution of housing conditions and use of the Irish language change over time.

While this approach is simple and effective, it is only suitable for a limited number of dates or where change occurs at clearly defined times between periods of relative stability. More complex situations are more problematic. If, for example, a researcher wanted to create a database of changing administrative boundaries for an entire country, the key dates solution would be to digitise the boundaries for every date at which maps are available. There are two problems with this: first of all, where boundaries do not change the same line has to be digitised multiple times. This results in redundant effort and will inevitably lead to problems with sliver polygons (see Chapter 4). Secondly, it is unlikely to be possible to digitise the boundaries for every possible date, and while linking attribute data to spatial data for a nearby date may provide a good approximation of the actual boundaries, there will be some error introduced as a result. This can range from an incorrect representation of the administrative unit concerned, to either polygons with no attribute data or attribute data with no polygons.

These limitations have led researchers in various countries to attempt to build systems that are continuous records of boundary change. This allows the researcher to extract the boundaries for the appropriate date and link them to any suitable attribute data. Two distinct approaches can be identified to doing this: the date-stamping approach used by the Great Britain Historical GIS (Gregory and Southall 1998; 2000), and the space-time composite approach which was proposed as a theoretical structure by Langran (1992). This approach has been used by the Swedish National Topographic Database (Kristiansson 2000), the Belgian Quantitative Databank (Vanhaute 1994) and is proposed by Ott and Swiaczny for the Palatinate area of Germany (Ott and Swiaczny 2001).

Figure 5.2: Storing changing administrative boundaries: the Great Britain Historical GIS
A master layer consists of label points (representing administrative units) and lines (representing their boundaries). The attributes for both sets of features include date stamps. Features in existence when a type of unit was formed are date stamped 0/0/0000, while those in existence when they were abolished are date-stamped 0/0/5000. There is no topology on a master layer. This example shows how a boundary change between Anarea and Elsewhere on 1 September 1894 can be handled. Relevant features are selected from the master layer for a user-specified date and topology is then constructed. Source: Gregory and Southall 2000, 327.

The date-stamping approach handles time as an attribute in a manner similar to that described above; however, it is complicated by the need for polygon topology. Gregory and Southall (2000) cope with this by storing all their spatial data in what they term master coverages (i.e. master layers). These are layers that have both label points, representing administrative units, and lines, representing their boundaries. There is, however, no topology to link the two at this stage. In this structure, boundary changes can be handled in the manner shown in Figure 5.2, showing a transfer of territory from one unit to another. More complex changes, such as where an entire unit is created or abolished, can be handled using the label point attributes. Using this structure a user specifies a date and custom written software extracts the appropriate label points and lines and creates topology to form a polygon layer for that date.

Figure 5.3: Storing changing administrative boundaries: a space-time composite
This diagram shows the same change as Figure 5.2. Using a space-time composite the Least Common Geometry is a polygon layer. The attribute data are split into time periods and each polygon is assigned to an administrative unit for each time period. The user specifies a time period and a dissolve operation (see Chapter 4) is used to aggregate the polygons to form the administrative units in existence during the specified period.

The space-time composite approach creates administrative units through aggregating smaller polygons, by storing the unit that each polygon lies in at each date as attribute data. These smaller polygons are referred to as the Least Common Geometry (LCG). This can consist of low-level administrative units that are known to be stable over time, as in the Swedish system that uses parishes to create districts, municipalities, and counties. Where no such units are available, it can consist of polygons created as a result of boundary changes, as is proposed by the Palatinate system. In either case the basic structure is the same, as is shown in Figure 5.3. A dissolve operation is used to re-aggregate the polygons in the LCG to form the units in existence at the required time.