Uncertainty and Error

Resolution as a Cause of Uncertainty

Many causes of uncertainty and error in spatial information have been mentioned in the preceding discussion, but one important cause of profound uncertainty that is often incorrectly treated as error is that of scale or resolution of a data set. Scale as such refers to the ratio between a distance on a map and the same distance on the ground. This concept has little or no meaning in an age of spatial databases, although the terminology persists, even for products that have never been presented as paper maps. When a mapping scale is associated with a geographical information product (rather than a hard-copy map) it conveys an idea of the resolution of the information: the smallest discernible object in the database, for example, or the possible width of a line object on the ground when it was drawn as a line in the cartographic product from which the information was digitized—or even in a cartographic product that might be generated from the information. It is more honest, and directly informative, to quote the areal or linear dimensions of the discernible objects in the information. This is sometimes referred to as the minimum mapping unit, and it should be noted that the minimum mapping unit of a categorical data set held in raster format is not necessarily the same size as the raster grid. The Land Cover Map of Great Britain (1990), for example, has a 25 m grid size, but a 2 hectare minimum mapping unit.

Categories of information mapped at different resolutions are different. Thus, in soil maps, the information collected for data storage at one scale is different from that collected at another scale; in England and Wales, the so-called soil series is mapped at 1:50,000 and larger scales for restricted areas, while the associations of soil series (the next level of aggregation) are mapped at 1:250,000. This means that any analysis over the whole country can answer questions suitable only for soil associations, [Page 494]not series, and any attempt to analyze at the series level is flawed.

...