Investigating accessibility indicators for feedback from a travel to a land use model

Activity locations such as work locations or leisure facilities are not uniformly distributed geographically. Also, the travel access to different locations is not uniform. It is plausible to assume that locations with easier access to other activity locations are more attractive than locations with less access. In consequence, urban simulation models such as UrbanSim use accessibility measures, such as ``number of jobs with 30 minutes by car', for several of their submodels. A problem, however, is that accessibility variables are not easy to compute within UrbanSim, for two reasons: 1) UrbanSim does not contain a travel model, and in consequence is not able to compute by itself the congestion effects resulting from land use decisions 2) The travel times are fed back from the travel model in the form of zone-to-zone travel time matrices. As is well known, such matrices grow quadratically in the number of zones. This limits the number of attributes that can be passed, for example different values for different times-of-day and/or for different activity purposes. These issues could be solved within UrbanSim, but only with considerable implementation effort. For that reason, it is important to consider how accessibility measures could be fed back from a travel model to UrbanSim. The present study will look at the question in how far location-based accessibility measures that are computed in the travel model and then fed back to UrbanSim could be used for this purposes. Those accessibility measures are no longer measures belonging to pairs of locations, but just belong to one location; a typical representative is a logsum term. In consequence, the number of entries now grows linearly in the number of locations, allowing much more freedom both in the number of considered locations and in the number of attributes that could be attached to every location that is considered in this way. This paper will address issues such as different spatial resolutions of such accessibility measures, comparisons between different accessibility measures, and computing times.