Comparison of Hybrid 4-Step and Activity-Based Models: Example for Jerusalem, Israel
Gaurav Vyas, Parsons Brinckerhoff, Peter Vovsha, Parsons Brinckerhoff, Danny Givon, Jerusalem Transportation Master Team
This paper compares the two families of travel demand models: Hybrid 4-step models and Activity-Based travel demand models developed for Jerusalem, Israel.
Aggregate 4-step trip-based models represented a standard and most used in practice paradigm of travel demand modeling for over half-century. These models are relatively easy to develop, they are helpful in understanding the general travel patterns (and specifically, commuting patterns) in the region, and they proved to be instrumental for decision makers in testing certain types of transportation infrastructure projects, and transportation policies. However, the scope of possible transportation policies that could be handled by 4-step models proved to be limited, and primarily, due to the explosion of the number of trip matrices that have to be created at the trip distribution and mode choice steps if additional segmentation details or explanatory variables are to be added. For example, in order to test the effect of transit passes holding on mode choice, there need to be two sets of trip matrices created – one for transit pass holders and another for non-holders for each existing segment (combination of trips purpose, car ownership group, income group, time-of-day period, etc). In addition to the practical limitations in policy analysis, aggregate 4-step models as has been intensively discussed in literature, are characterized by a limited behavioral realism in terms of the sequence of travel choices.
Introduction of the microsimulation approach to travel demand in 1990s proved to be revolutionary in many respects. With the advent of microsimulation approach, another family of travel demand models in practice emerged, referred to as Activity-Based travel demand Models (ABMs). The microsimulation approach in ABMs made a plethora of additional travel demand policies possible for analysis. The underlying central paradigm of travel modeling in ABMs is that people are fundamentally driven by participation in activities, and trips and tours emerge subject to spatial and temporal constraints associated with the activity participation. ABMs by design are behaviorally richer than 4-step models in terms of the explanatory variables and logical interconnections between multiple travel choices of the same individual. As the result, they also can accommodate a wider spectrum of tendencies in travel behavior and transportation policies that are intended to affect this behavior.
Another interesting family of travel demand models in practice is emerging that is designed to take an advantage of some of the individual behavioral aspects of ABMs that are easier to implement in a micro simulation fashion while leaving the rest of the model system at the simpler aggregate level pertinent to the 4-step models. This new family of travel models is commonly referred to as hybrid models. Hybrid models are characterized by development cost and overall level of complexity that is close to the conventional 4-step model. The model chain in hybrid model systems typically starts with a set of disaggregate models and eventually ends with aggregate mode choice model. The hybrid models are considered as a practical compromise between 4-step models and ABMs due the shorter development cycle and quite moderate resources for development compared to fully-fledged ABMs. A hybrid model is also a useful intermediate step on the way from a classic 4-step model to a complete ABM in terms of the learning curve for the model users and preparation of the necessary inputs.
Jerusalem Transportation Masterplan Team decided to develop a hybrid travel model as the first step of the ABM development. By the completion of the entire ABM, the agency became uniquely equipped with recently developed and well-calibrated travel models of both types – hybrid and advanced ABM. This paper compares the performance of these two types of travel models for the 1.1M Jerusalem region. These two models for Jerusalem were developed in parallel using the same basic socio-economic data, land use, and transportation networks. This common denominator helped in understanding the challenges and shortcomings of the hybrid model and the ABM as two travel modeling technologies available today in practice. In this effort these two models were put side by side and assessed with respect to their structure, development cost, spectrum of transportation projects and policies that can be modeled with each of the models, as well as the observed model sensitivity for each policy. The paper summarizes practical advantages of ABM over the hybrid model in certain important policy contexts.
Association for European Transport