Identifying Land Cover Change Using a Semantic Statistical Approach

Abstract

Land resource inventories are often subject to changes in methodology due to technological innovation and developments in policy (Comber et al., 2002, 2003a). The net result is that different data nominally purporting to record the same processes may do so in different ways, identifying different objects, using different labels, and different definitions. Subjectivity derives from the use of different pre-processing techniques, classification algorithms, and class definitions. There is an increasing recognition of this phenomenon within remote sensing, acknowledging the subjective nature of much information classified from remotely sensed data. This is exemplified by the increased acceptance of concepts such as fuzzy classifications, soft accuracy measures, as described by Shalan et al. (Chapter 2, this volume) and fuzzy land cover objects Lucieer et al. (Chapter 6, this volume).