ABSTRACT
The guiding principle in the development of error handling in geographic information (GI) science has been that of fitness for use (Chrisman, 1983). Fitness for use aims to supply enough information upon the error characteristics of a data set to allow a data user to come to a reasoned decision about the data’s applicability to a given problem in a given situation. The concept of fitness for use is congruent with the NCDCDS concept of truth in labelling, which rejects the formulation of prescriptive and arbitrary thresholds of quality in favour of the provision of detailed information, in the form of the five quality elements. However, the widespread acceptance of the NCDCDS quality elements does not necessarily bear any relation to their suitability for this task: to describe the quality of any data set so as to allow a user to assess fitness for use. Little or no empirical work exists to determine how well the NCDCDS ‘famous five’ perform in different scenarios and so their applicability to all situations is, at the very least, open to question. It is argued here that the success of the NCDCDS proposal as a standard is not indicative of its usefulness as a basis for the implementation of what has been termed an error-sensitive GIS (Unwin, 1995). The aim of this chapter is to explore the implementation of an error-sensitive GIS using a new approach to data quality. The approach used is inclusive of existing data quality standards, whilst allowing database designers, rather than system designers or standards organisations, to decide how best to describe the quality of their data. 53
5.1.1 Exhaustiveness
Anecdotal evidence, at least, suggests the five NCDCDS elements of spatial data quality are not exhaustive. The International Cartographic Association (ICA) Commission on Spatial Data Quality accepts the NCDCDS quality elements, yet feels the need to augment these with semantic and temporal accuracy. In ‘Elements of spatial data quality’, Morrison (1995) points out that it was ‘more important to place solid definitions of the seven components in the literature than to attempt to be totally complete at this time’. Whilst often accepting the NCDCDS core of quality elements, many researchers have suggested the use of a number of other data quality elements, such as source and usage (Aalders, 1996), detail (Goodchild and Proctor, 1997) and textual fidelity (CEN/TC287, 1996). Even a cursory examination, then, reveals considerably greater than five quality elements. Further, it is worth noting that there is no firm agreement on the actual definitions of many of the elements mentioned (Drummond, 1996). It is unreasonable to expect any data quality standard to be exhaustive; there will always be eventualities and situations for which the standard was not intended. Therefore, the design of errorsensitive GIS should allow enough flexibility to accommodate not simply existing data quality standards, but the redefinition of these standards and the use of novel, user defined data quality elements.
