ABSTRACT
The way in which the uncertainty in input data layers is propagated through a model depends on the degree of nonlinearity in the model’s algorithms. Consequently, it can be shown (Burrough and McDonnell, 1998) that some GIS operations in environmental modeling are more prone to exaggerate uncertainty than others, with exponentiation functions being particularly vulnerable. Also of influence are the magnitude of the input values and the statistical distribution of the datasets. It is generally assumed, often through lack of information, that the uncertainty in a data layer is normally distributed (Gaussian).
