Managing spatial uncertainty in remotely sensed data

The level of spatial uncertainty in geographical information systems (GIS) products is conditional on a number of operations on the spatial database prior to final output. Although stochastic and deterministic models have been used to predict the levels of uncertainty in digital map products through GIS operations such as overlay and buffering, the extent to which each operation contributes to overall product uncertainty is neither well known nor understood. An alternative approach is to examine the level of spatial uncertainty in the output product and then to improve the accuracy of this output by applying a rule-based model. In this paper, a conceptual rule-based model is proposed to effectively manage spatial uncertainty by examining the interrelationships between and within polygons resulting from the overlay of multiple independent human interpreter realisations of a satellite image. In a previous paper by Allan (1999), a method was proposed to extract information about these polygons from the overlay of these realisations for inclusion in the database and an approach was suggested in handling the resultant uncertainty polygons. Further work has seen the development of structural knowledge (rules) and procedural knowledge (operators) for managing these uncertainty polygons based on their likelihood of belonging to a particular thematic class. The adoption of such an approach is based on the semantic, geometric and spatial relations within and among uncertainty polygons generated from these realisations.