Multiple Bayesian models for the sustainable city: The case of urban sprawl

Abstract

Several possible models of urban sprawl are developed as Bayesian networks and evaluated in the light of available evidence, also considering the possibility that further, yet unknown models could offer better explanations. A simple heuristic is proposed in order to attribute a likelihood value for the unknown models. The case study of Grenoble (France) is then used to review beliefs in the different model options. The multiple models framework proves particularly interesting for geographers and planners having little available evidence and heavily relying on prior beliefs. This last condition is very frequent in research on sustainable cities. Further options of multiple models evaluations are finally proposed.