Remote sensing of land cover classes as type 2 fuzzy sets

In recognition of the idea that land cover classes are vague phenomena, a number of researchers have advocated modelling of Land Cover as fuzzy sets, as opposed to Boolean sets. Once vagueness is admitted, it is not sufficient to stop at modelling land cover simply as a fuzzy set, because any threshold map (α-cut) of the fuzzy set is itself a Boolean set. Rather it may be appropriate to model it as a type 2 fuzzy set recognising the existence of higher order vagueness. Using the example of estimating land cover extents in an area of the forest–savanna ecotone in Bolivia, this paper presents the results and possible advantages of a type 2 fuzzy analysis of land cover. The consideration of type 2 fuzzy sets reveals the depth of the uncertainty, and allows the analyst to be specific about minimum, maximum, and average extents of land cover types and even to report fuzzy area itself as a fuzzy number and to justify descriptive qualifications of the results. The nuanced analysis, which is possible with this approach, seems an advance for the scientific community concerned with the conceptualisation of land cover classes, but it may be a nuisance for those involved in policy matters.