Stochastic Analysis of Urban Development

Abstract

Two major drawbacks prevent the wide application of Markov processes to land use and other types of planning analysis: the lack of temporal homogeneity of the transition mechanism, and the lack of independence of spatial data. In order to avoid these difficulties Gilbert (1972) proposed the use of the nonhomogeneous, two-state, continuous time Markov process and the semiMarkov process. In this paper we expand upon Gilbert's discussion and illustrate the calibration of the nonhomogeneous (birth and death) model while commenting on the empirical intractability of the semiMarkov process.