Representing Uncertainty in Geographical Data

Geographical data is inherently uncertain. There are a variety of sources that contribute to this uncertainty. Operations performed on data during an analysis will most likely only increase this uncertainty. To understand the uncertainty in the results of an analysis, requires information about the uncertainty in the input data, and some idea about how various analysis operations propagate and alter uncertainty. This article describes various ways to represent this uncertainty, and discusses their applicability. Implicit representation involves constructing a model of the statistical properties of the uncertainty without taking the spatial distribution of the uncertainty into account. In a qualitative representation, only some relationships between the spatial objects in the data are present; since many instantiations of such a representation are possible, it can be said to represent uncertainty. Fuzzy, rough and bifuzzy classifications are three approaches that explicitly represent the spatial distribution of the uncertainty in the data.