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

CHAPTER 3: ACQUIRING SPATIAL DATA

3.5 Error and accuracy

As was stated earlier, a digital representation of a paper map is at best of equal quality to the original map, but it will almost inevitably accrue some additional error or inaccuracy. It is important to distinguish between the different types of error and inaccuracy. Unwin uses a six-way classification as follows: error is the difference between reality and the digital representation of it; blunders are simply mistakes; accuracy is the closeness of results, computation or estimates to values accepted as true; precision is the number of decimal places given in a measurement which is usually far more than it can support; quality is the fitness for purpose of the data; and uncertainty which is a measure of the doubt or distrust that the data should be used with (Unwin 1995).

Scanning, digitising and geo-referencing are particular sources of locational error. To digitise a map it must first be placed completely flat on the digitiser or scanner. Even this is not always as easy as it sounds: maps may have been stored folded, paper warps over time, and so on. The accuracy of the scanning or digitising equipment itself is the next possible source of error, although if specialised (normally expensive) hardware is used this is usually only minimal. If head-up digitising is used there will be cumulative error as there is the error introduced by the scanning, and then error introduced by the digitising.

The next source of error is the user's involvement with the data capture process. With digitising, the person capturing the data has to place the puck or cursor over the point to be captured. Even a highly motivated and alert person will make minor positional errors. Where digitising is done for many hours by low-paid staff this increases the potential for both inaccuracy and blunders. Even with the best will in the world it is not always possible to capture the exact location accurately. Point symbols are not always a precisely defined point: for example railway stations on an OS 1:50,000 map are represented by a circle that is nearly 2mm in diameter. Line features are even more difficult. Digitising a line relies on the operator capturing each point at which the line changes direction. For gentle curves, such as are found on roads, rivers or contour lines, this is inevitably a subjective choice and no two operators digitising a line of this type will ever digitise exactly the same points to describe it.

A final source of error is the geo-referencing process. Coordinates measured from a map will have a certain amount of error in them. The locations of the reference points will also have some error. This means that the placement of every location on the layer will be slightly distorted. Most software packages will provide a measure of the error expressed as Root Mean Square (RMS) error. This is often expressed in both digitiser units and also real world units. It is recommended that the RMS error should not exceed 0.003 digitiser inches (ESRI 1994a). Even this standard is not always possible with historic maps, and it is important that minimum standards of accuracy are established as part of the data capture process and that these are documented (see Chapter 9). The key point to this section is that there are many issues associated with error and uncertainty in spatial data that may not be familiar to the historian. The problems these cause may seem daunting, however by facing up to the issues and being aware of their implications these should not cause serious difficulties to the research process.