An Agent-Based Model of Residential Mobility in the Tel-Aviv Metropolitan Area

BENENSON, Itzhak (bennya@post.tau.ac.il), OMER, Itzhak and PORTUGALI, Juval, Tel-Aviv University, Department of Geography and the Human Environment, 69978, Ramat-Aviv, Tel-Aviv, Israel

Key Words: agent-based modeling, GIS-based model, residential mobility, residential segregation

An attempt to construct an agent-based model of a residential mobility in Tel-Aviv is described. It is based on a hybrid of a PC-based GIS and High-Performance (HP) parallel computer. The GIS serves as a tool to represent and model immobile infrastructures, and, as a user's interface while the HP is a computation engine for simulating residential mobility. At the current stage of the research we have accounted for urban physical structure and residential distribution only (ignoring, for instance, the distribution of jobs).

An urban system in the model is decomposed into two compartments, i.e., physical properties and population, according to different time scales in which their dynamics take place. For example, in Tel-Aviv, during last 30 years, new dwellings appear at a yearly rate of 1 percent and below, while the rates of internal residential mobility, out-migration and in-migration, each remain at an annual level of 5 percent. To represent infrastructure, land parcels, buildings, and the links of the street networks are considered as basic units. They are organized as a set of layers within the high-resolution (vector) GIS of the Tel-Aviv metropolitan area. GIS serves as the ground upon which to place the model of residential mobility. Descriptions of urban population mobility are based on the description of the residential behavior of an individual decision-maker (householder).

Data on the city's inhabitants are based on the Israeli Central Bureau of Statistic (CBS) databases, which are partially available for supervised research. Model agents have the ability to estimate the state of the city and to behave in accordance with their own properties, the local information regarding the state of the neighborhood, and of the neighbors and the global information based on the state of the city as a whole. Based on this information, agents immigrate into the city, occupy and change residential locations there, and leave the city when conditions become unsatisfactory. When residing in the city, agents instantaneously re-estimate the available information on the neighborhood, and on the whole city, and react to changes by adjusting their properties and residential behavior. As a result, the dynamics of the entire urban system are described as the collective consequence of complex interactions of numerous immobile infrastructure units and mobile human agents. Several algorithms aimed to determine the neighboring buildings, based on visibility, adjacency, and the other geographic (geometric) relations among houses, streets, and open spaces and are implemented within infrastructure GIS. We assume that the influence of a city's global structure on an agent's properties increases with the level of residential segregation according to this property. Residential segregation is recognized and described in the model by means of local measures of spatial autocorrelation.

Simulation experiments are held for neighborhoods in the southern part of Tel-Aviv-Jaffo- with its heterogeneous population distribution. They demonstrate that if we introduce sufficiently strong tendencies of individual agents to depart from alienating neighbors, and to resettle in more fitting neighborhoods, then the likelihood clustering of agents of similar properties are establishing. In this way we can easily obtain spatial clusters of individuals of similar levels of income, origin, etc. The extension and form of the clusters is essentially influenced by the local topology of a city space. A distance-independent search entails non-diffusion dynamics of urban residential distribution. A relatively homogeneous group of agents can be established in the model far from the current residence of the majority of the group members.