STOCHASTIC CELLULAR MODEL FOR LOWLAND URBAN DEVELOPMENT

An urban growth model using stochastic cellular simulation was developed with motivation to understand the consequence of zone management policies in lowland cities. The model could integrate the growth, decline, spread, intensification, and protected areas of the urban growth into a single generalization of both the Eden and the p-models. Calibration strategy was demonstrated using historical aerial photographs of Saga city, Japan. A special well known characteristic of lowland cities is its sensitivity from the fluctuating water levels. Flood and storm water are commonly regarded as the most frequent and widespread natural hazard for such places. In connection with urban development, the improvement regulation of zone management is one of the most comprehensive and long-term solutions for hazard mitigation. The overall aim is to reduce the risks involved in the present occupation of flood-prone land and to deter further invasion of such area (Smith and Ward, 1998). To make such policy of zone management effective, an urban development model is needed. Since real field experiments in urban development is impossible, numerical experiment using computer simulation can be utilized to comprehend the effect of zone management policies and to predict the long term effect of several urban development scenarios. In the last decades, urban development modeling has attracted many researchers in urban planning fields because it may be used as laboratories for exploring ideas about how cities work and change over time (Torrens and Ou0027Sullivan, 2001). (Clarke et al, 1997) proposed an urban development model using simple growth that an occupied cell has at least three neighbors will become a new developed cell if it can pass constraints of repeating spread and slope. This growth has analogy to the spread of fire in the forest. New growth location is always selected at a random location that can pass some constraint. Spontaneous growth is growth wherein a new seed can be put at a random location that has at least one neighbor and quite flat. However, the Clarke model relies heavily on ad hoc solution through combinations of many unrelated and independent sub models such as road, slope, seed cells, and protected areas. (Batty, 1991) uses the Diffusion Limited Aggregation (DLA) model to analog urban growth. The unconstrained real city growth however, is more similar to a circle rather than the tentacles-shaped DLA. (White and Engelen, 1993) uses fractal land-use structure and calibrates the model by simply matching the fractal dimension of simulated city with the map of the real city. (Benguigui, 1995) proposed to model the city as binary value of developed or undeveloped cell. The growth rule is similar to a simple cellular growth of Eden model with additional parameter number of visit p, before the cell is developed. A higher p value tends to make dispersion or unconnected development clusters. The dispersion phenomenon is related to the spread of the city growth toward scattered clusters rather than aggregated clusters. This model, however, do not incorporate the intensity of the development. The model could not distinguish which part of the city has more development than the other. The literature of cellular growth model can be traced back to 1961 where Murray