Can I map uncertainty?

The error matrices produce an overall statistic for error for the whole map and also for each class. However, these summary statistics apply to the whole map and show no geographic variation over the map. The fuzzy maps do indicate geographic trends in uncertainty (in the above example, there is uncertainty where there is a hatch over the most probable class). Are there other ways of showing the varying degree of uncertainty over a map?
 
Image processing techniques provide one way of doing this. The section on supervised classification How Do I Make A Map? explains that when habitat class signatures are applied to the raster layers the pixels are assigned on the basis of the highest probability. This is picked from the probabilities that have been calculated for all habitat classes for each pixel. With the standard classification routine these individual probabilities are not visible. However, they can be viewed as individual layers, one for each habitat class. They can also be used to determine the level of certainty with which a pixel is classified (the more evenly the probabilities are spread between the classes, the lower the certainty). Maps can then be prepared showing certainty from 1 (one class has a probability of 1, all the others have a zero probability) to 0 (all classes have an equal probability).
 
GMHM5-07_Probability_maps
 
 
Links to other sections:

All material variously copyrighted by MESH project partners 2004-2010

| List Access Keys | Mapping European Seabed Habitats (MESH) - Home | Overview | News | Work areas in the MESH Project | Partners | Contact Us | Search | Site Map | Partner Extranet |