Approaching human judgement in the automated comparison of categorical maps

Map comparison methods are used to express the similarity between maps. They are amongst others used in validation and calibration procedures of spatial models. It is found that traditional map comparison methods do not distinguish small differences from large differences, which leads to comparison results that are not in accordance with human judgment. In this paper a relatively new method for the comparison of categorical raster maps, the Fuzzy Kappa statistic, is compared to two other methods, which are much applied: the Kappa statistic and the Percentage-of-Agreement. It is found that the Fuzzy Kappa statistic approaches human judgement, as expressed in an Internet survey, better.