Suitability assessments of candidate locations for a particular land-use are typically uncertain, as future changes in land-use patterns may have an impact on performances while these changes are hard to predict. Most planning support tools for location decisions do not take such uncertainty into account. To address this problem of uncertainty, we propose a ‘Bayesian decision network’ approach. In this approach, the possible courses of actions, causal knowledge and preferences of a decision maker are represented in a network of causal relationships between a set of variables. Uncertainty about future land-use developments, which may influence outcomes of location decisions, can be represented as conditional probabilities in these networks. However, estimating these probabilities for a given study area is not a trivial problem, as the space of all possible future scenarios is approximately infinitely large. In this paper, we propose and test a sampling method for this estimation purpose. As an illustrative case, we specify a decision network model of a retail location planning problem, investigate parameters of the sampling method and explore the extent to which different ways of coping with uncertainty affects outcomes of decisions.
|Number of pages||16|
|Journal||Computers, Environment and Urban Systems|
|Publication status||Published - 2007|