Using information based rules for sliver polygon removal in GISs

Abstract

Of all the roles that GISs perform, their ability to successfully integrate cartographic information from a wide range of sources and scales to create a unified data base is perhaps one of the most important. Ensuring that information held is of a substantive quality however, is very much up to the user.

When data is obtained from several sources, a variety of topographical and categorical differences may be observable between the representations. Processes responsible for such disparities include survey errors, data capture techniques, machine processing to ensure that the data ‘fits’ the current data model being utilised and the temporal infidelity associated with cartographic material.

When each representation is viewed in isolation, such ‘discrepancies’, are often insignificant, however, when the data sets are integrated using polygon overlay techniques, such mismatches become flagged as errors, known as sliver polygons. Generally these correspond to a particular geometrical shape, and so the accepted policy among the majority of the large software vendors is to use this geometrical value as a basis for ‘removal’. Problems arise however when the geometry of the ‘sliver’ polygons correspond to polygons of differences between data sets that are worthy of retention, notably change, such as land use variation.

If the information or attributes of the sliver polygon are taken into account, rather than simply their geometry, the degree of uncertainty associated with their removal will be reduced and the quality of the overlaid material maximised. This contribution looks at the problems of data integration and sliver polygon removal and possible alternatives based on user defined rules.