Simulating Land use Development through a Stochastic Allocation Procedure in Johannesburg, South Africa

Abstract

This paper describes a method of forecasting demographic and economic change at a spatially disaggregate level that is compatible with the requirements of a conventional transport model. The method was developed and tested in the city of Johannesburg, South Africa, as part of a scenario-planning exercise to assess potential land use and transport interventions. The procedure does not model land use development decisions explicitly but uses a geographic information system–based multi-criteria analysis approach incorporating the factors influencing the development of residential and nonresidential land uses, the constraints of land availability, and the guiding effects of government policy. Allocation occurs in discrete time steps, allowing the dynamic evolution of outcomes to be modeled in a nonequilibrium framework. It operates in connected mode with the transport model, taking accessibility changes as input into subsequent land use allocations. Monte Carlo simulation is used to approximate randomness in the location decision outcome, thus providing some sense of the variability of the outcomes that may occur at a zonal level. The method is illustrated for the city of Johannesburg for a 30-year planning horizon. It is shown to be particularly useful in developing societies where data to establish more elaborate behavioral models are absent or too costly to collect.