Some Reflections on the Transition to Activity-based Models in the U.S.
M Bradley, Mark Bradley Research and Consulting, US; J Bowman, Transportation Systems and Decision Sciences, US; P Vovsha, PB America, US
We provide examples of applications of activity-based models in the US, and suggestions as to how the transition to activity-based models can be made easier and more attractive for those who will be responsible for using them.
It has now been nearly ten years since the first in a family of activity-based disaggregate microsimulation models of travel demand was applied in the US. Since that time, models in this same family have been applied in the cities of Portland, New York, San Francisco, Columbus, and Sacramento, and models are now under development for Denver, Oakland, Atlanta, Dallas,Lake Tahoe and other city and state government agencies. Although each of these model systems has its own distinguishing features, their similarities outweigh their differences, especially when contrasted to the traditional 4-step modeling approach.
The claims behind the new activity-based models have been that they are superior to the 4-step approach in a number of ways, including:
? They are sensitive to a wider range of policies and demographic shifts.
? They provide more realistic and accurate forecasting sensitivities/elasticities.
? They are able to accommodate a much finer level of disaggregation ? temporally, spatially, demographically, and in terms of typology of activities.
? They are able to represent time-of-day shifting and activity scheduling effects.
? They are able to represent detailed land use patterns and the effects on non-motorised travel.
? They provide results that can be used in a wider variety of contexts, including environmental justice analysis, traffic microsimulation models, and land use microsimulation models. (In particular a natural integration of traffic, demand, and land-use models based on the dynamic microsimulation concept looks to be a promising and realistic avenue for future development.)
? They are less of a ?black box? and more intuitive to users and policy makers.
? They can take advantage of recent advances in GIS and computing capabilities.
Now that there is evidence and experience from the application of a number of such models in the US, it is useful to step back for a moment and consider whether or not the models appear to be living up to these claims. In cases where they are, we provide examples of what has been done with the models that could not have been done with more conventional models. (The first successful examples of activity-based model applications include various environmental impact studies, road-pricing projects and policies, large-scale rail / LRT/ BRT transit projects, and others.) In cases where they are not, we discuss to what extent this is due to the models themselves and to what extent it is due to the way that they have been implemented and used (or not used) in practice. We also discuss what can be done by those who create such models and by those who use them to take maximum advantage of their capabilities.
The authors have been involved in the creation of all of the activity-based model systems currently used for applied regional forecasting in the US, and thus we are certainly not the most objective reviewers, but we can provide a unique perspective on both the technical and practical issues involved in moving these new modeling techniques into practice. In particular, we focus on some of the challenges involved in applying microsimulation-based models within the pre-existing framework of demand/supply equilibration (i.e. traffic assignment) and model calibration and validation methods. We close the paper by providing ideas as to how the transition to activity-based models can be made easier and more attractive for those who will be responsible for using them, and with a discussion of how the U.S. experiences relate to European forecasting contexts.
Association for European Transport