Development of Prototype UrbanSim Models

Development of Prototype UrbanSim Models


Z Patterson, M Bierlaire, EPFL, CH


A few applications of UrbanSim (by its developers) exist in the literature. This paper evaluates, based on experience with two UrbanSim applications, the effort to develop an UrbanSim model as being comparable to that of a traditional 4-step model.


Traditional (4-Step Methodology) transportation modelling does not explicitly evaluate the effect of transportation system performance on induced development or transportation demand. However, this feedback is increasingly identified as being necessary to properly evaluate future transportation demand, system performance, atmospheric emissions, etc.

Integrated transportation land-use modelling (Integrated Modelling) explicitly considers the effect of transport system performance on land development and land-use. Many integrated models currently exist. UrbanSim is one modelling system that has received a fair bit of attention since its development in the late 1990s.

UrbanSim is open-source, operates at very fine geographical scale and relies heavily on discrete choice models. The fine level of detail requires a great deal of data so that data collection and preparation can take up to two years. A few successful UrbanSim applications have been reported in the literature, but these have been produced primarily by the model developers. Given the potentially demanding requirements of using UrbanSim, planning organizations considering developing integrated models should like to know just what is involved in developing an UrbanSim model, and how difficult it is to do.

This paper provides a users? side view of the use of UrbanSim. The research involved the development of UrbanSim models for two different cities. In a ?data-poor? case, for Brussels in Belgium, aggregate land-use and demographic data were disaggregated and used in UrbanSim. This case is compared with a ?data-rich? case (of Lausanne in Switzerland) where plentiful disaggregate data were available.

The analyses provide a number of important conclusions with respect to the use and operation of UrbanSim. First, it reveals that the amount of work required to develop an UrbanSim model, while not being automatic, is comparable to what could be expected for the development of a four-step model. Second, it reveals that the use of aggregate and synthetic data (on households, jobs and buildings) can provide respectable results in an UrbanSim model. At the same time, some data turn out to be very critical in the operation of UrbanSim. In particular, sufficient historical development data (on which the real-estate development models are calibrated) is very important in the successful operation of UrbanSim. Moreover, the availability of accurate zoning data proves indispensable in assuring the reasonable placement of jobs, households and real-estate development.


Association for European Transport