Natural resource inventories are more and more based on quantitative spatial methods. The discrete classes of vegetation, soil, land cover etc. can be quantified by using membership maps rather than only crisp delineations. GIS methods to visually explore fuzziness and uncertainty of the prediction maps are clearly needed. The paper presents two relatively new techniques for visualisation of multiple membership maps and uncertainty of spatial prediction methods: Pixel Mixture (PM) and Colour Mixture (CM). The PM assigns classes to sub-pixels relatively to the membership value, while CM calculates average colour taking the memberships as weights. The PM technique appears to be more simple, "user-friendly" and easier to interpret. The CM technique requires a special fuzzy legend that also reflects taxonomic relations and can be therefore considered a double-continuous approach to mapping. The derived saturation of the map calculated using the CM technique gives us insight into the spatial and taxonomic confusion of our map. Both PM and CM can be both applied to visualize the confusion index or relative error together with the maps of estimates. In this case we used brightness correction and white pixels to include the relative uncertainty and confusion. These techniques will be further on developed to produce operational maps to be used in decision-making.
@InProceedings{ hengl:2002:PACM,
author = {Tomislav Hengl and Dennis J.J. Walvoort and Allan Brown},
title = {Pixel (PM) and Colour Mixture (CM): GIS Techniques
for Visualisation of Fuzziness and Uncertainty of
Natural Resource Inventories},
booktitle = {Proceedings of the 5th International Symposium on
Spatial Accuracy Assesment in Natural Resources and
Environmental Sciences (Accuracy 2002)},
pages = {300--309},
year = {2002},
}