There is a new growing interest in the development and in the use of integrated land use and transport planning models in France. In this paper, we describe the steps of a current project which aims to integrate UrbanSim, a flexible land use model, and METROPOLIS, a dynamic traffic model, and to apply this integrated model to Paris region. We shortly present the two models and the common architecture then we describe the fastidious but crucial step of collecting input data and calibration data for the study area. Paris region is one of the most important metropolises in the world: 12,000 km2, 11 millions inhabitants and 5 millions jobs. Most interactions between the land use dynamics and the transportation dynamics are taken into account in the short, middle and long term. All of this consists in a pioneering and innovative work, for a region where urban planning and fiscal policies are very important. UrbanSim is a land use model developed at the University of Washington (USA). It is based mainly on three logit models (households and jobs localization choices and development type choice models) and a hedonic regression model (land price model). The data structure is based on a large grid which partitions the whole Paris region with 50 000 square cells by 500 meters. This high level of spatial resolution is really original in France but requires a huge amount of data and spatial analysis that we have performed thanks to the GIS tool. METROPOLIS is a dynamic transportation model developed at the University of Cergy-Pontoise (FRANCE). It provides the user surplus as the measure of accessibility. This measure takes into account the time-dependent congestion situation of the transportation system. On the other hand METROPOLIS can differentiate the users by their value of time and desired arrival time and some other behavioral parameters. The roads network contains more than 16,000 links, the transit network contains about 4,000 links. An architecture bas been designed to integrate these two models within a coherent framework. A prototype of interface has been developed which allows input and output data to be exchanged in an automatic feedback process. We use different sources to build the input database: general census, numerical land use database (cover of 400,000 parcels classified into 83 different types), regional travel survey, the notary database of real-estate transactions, local land use plans, commercial and offices surfaces data, income tax files, … Since none of these sources is perfect, we had to develop innovative methods to realize data fusion and mixed databases. For example, we localize the 11,000 households of the travel survey in the grid, or we associate the attribute of income from the tax files to the attributes of household in the general census. The second database concerns the calibration data. For each of the four models of UrbanSim, we have developed a significant sample of individual observations from four sources: the general census, the travel survey, the land use evolution database and the notary database of real-estate transactions. These files will be used to estimate the models thanks to an econometric software. We choose as period of calibration 1990 – 1999. We plan to achieve our project in the end of 2005.
