UrbanSim models are built using local data for each metropolitan area, and the parameters for each model are estimated using advanced statistical methods to ensure that the model actually reflects local conditions. In no case are models applied to a region using data or coefficients from another region. Finally, we provide tools to automatically search for the best-fitting model specifications, and to calibrate the models to observed data at a more aggregate level than the UrbanSim model predictions. For the census block models for the United States, we have already done all of these processes and the models are ready for use. For metropolitan areas anywhere in the world, we have options to implement UrbanSim models at a parcel level or a zone level. UrbanCanvas Modeler provides a web-based platform that provides additional tools to make it straightforward for new users to simply upload their data, and have our platform build and calibrate their UrbanSim models, preparing them for use in simulating policy scenarios.
Once the models are built and calibrated, they are ready to be used to evaluate alternative transportation plans and land use plans. Transportation plans are encoded in travel networks and need to be modeled by the user's own transport models as of now. Skims of the travel model network with congested travel times zone to zone, or generalized cost or composite utilities (logsums from the mode choice model, typically) can be used to inform UrbanSim about changes in the transportation system in different simulation years. Land use scenario inputs can come from comprehensive plans of local jurisdictions, or more generalized representations of development capacity. In UrbanCanvas Modeler, we provide easy to use visual interfaces to upload and edit inputs to UrbanSim scenarios. Once scenarios have been created, we provide the cloud infrastructure to simultaneously simulate as many scenarios as a user wishes.
As simulations are being run, UrbanCanvas Modeler provides feedback on progress. UrbanSim simulations run annually, over a simulation period specified by the user. Commonly the simulation period for long-range planning is 30 years. Smaller regions run typically in less than 30 seconds per year, while larger regions of 3-5 million population may take 2-4 minutes per year to simulate.
Once simulations are completed, users have access to a large number of indicators that can be created to better understand the results, at different levels of geography, and to evaluate the scenario from a policy perspective. In UrbanCanvas Modeler, users can visualize the indicators as 2D or 3D maps, charts, and tables. All in a web browser.