Innovations in Travel Demand Modeling, Volume 2: Papers (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

CS recently completed an FHWA research project on time-of-day models that resulted in a methodology for time-of-day choice models, trip-based models, and activ- ity-based models. These were validated in case studies in Denver, Colorado, and San Francisco, California. The trip-based time-of-day modeling method was applied to a pricing scenario in the Denver region. Tolls were assumed on a (currently toll-free) 20-mi section of a circumferential freeway. Tolls were highest in the two peak periods (0.2 to 3.5 h long), with lower tolls in shoulder periods (1 to 3.5 h) and lowest tolls in off-peak periods. The time-of-day choice method estimated trips by time of day for half-hour periods. The application of the model for this scenario showed a modest amount of peak spreading resulting from implementation of the period-based tolls. The tour-based time-of-day modeling method was applied to a pricing scenario for downtown San Francisco and estimated trips by time of day for half-hour periods. A hypothetical $4.00 toll was applied for all auto trips enter- ing downtown San Francisco during the a.m. peak period (6:00 to 9:00 a.m.). The largest effect appears to be on mode choice. About 20% of the reduction in downtown trips is due to people choosing not to travel downtown at all, about 70% is due to changes is mode, and about 10% appears to be due to time-of-day shifts. These results seem reasonable, as many downtown travelers may not have the flexibility to change their travel times. For the Washington State Department of Transporta- tion, CS updated the time-of-day choice models by divid- ing the five main periods (a.m. peak, midday, p.m. peak, evening, and night) into 30-min subperiods in order to model peak-spreading behavior (5). In addition to auto travel time variations between periods, the model was structured to be sensitive to auto travel cost differences between periods (i.e., to emulate time-of-day-specific con- gestion pricing). The new time-of-day choice models were estimated for eight trip purpose and direction combina- tions, using 32 alternatives. These five time periods are used for transit, truck, and external trips. Auto trips are further subdivided into 32 time periods, as shown in Table 1. The auto time-of-day model uses highway travel times from each of the five time periods to predict travel for 30- min time periods. When estimating the time-of-day mod- els, the chosen time period for each trip will be based on the midpoint between the reported trip departure and arrival times. Trip tables are developed for each time period, purpose, mode, and direction. These are applied to networks by mode and time period in the trip assignment model. Multinomial logit (MNL) choice models were esti- mated for six home-based trip purpose and direction com- binations—home to work (HW), work to home (WH), home to shop (HS), shop to home (SH), home to other (HO), and other to home (OH). Two features were added to the time-of-day models to make them more sensitive to con gestion pricing: • The three periods where congestion occurs (a.m. peak, midday, p.m. peak) were fur ther divided into 30- min subperiods, in order to model peak-spreading behavior. Because it would be impractical to perform a separate traffic assignment for each 30-min period, the distribution of trips across the subperiods was based on travel times for the same five periods that are included in the current model. As the congested travel time in the “peak of the peak” increases relative to the free-flow travel time, the peak tends to flat ten, and a higher per- centage of peak travelers will travel in the shoulders of the peak. Gen erally, this type of effect is not symmetric because there are different constraints for traveling ear- lier as opposed to traveling later. In the a.m. peak, for example, we expect more workers to shift toward the earlier shoulder of the peak rather than the later shoul- der because many workers have to arrive at work before a specific time. • Second, in addition to auto travel time variations between periods, the model is sensitive to auto travel cost differences between periods, for instance from time- of-day-specific conges tion pricing. Because there are no data on such cost differences in the household survey, it is necessary to infer the sensitivity to travel cost by using the sensitivity to travel time mul tiplied by the appropri- ate value of time for each income group and travel pur- pose. The authors use the same values of time as in the mode choice models. 144 INNOVATIONS IN TRAVEL DEMAND MODELING, VOLUME 2 TABLE 1 Time-of-Day Choice Models A.M. Peak Midday P.M. Peak Evening Night 5:00 a.m.–5:29 a.m. 10:00 a.m.–10:29 a.m. 3:00 p.m.–3:29 p.m. 8:00 p.m.–10:59 p.m. 11:00 p.m.–4:59 a.m. 5:30 a.m.–5:59 a.m. 10:30 a.m.–10:59 a.m. 3:30 p.m.–3:59 p.m. 6:00 a.m.–6:29 a.m. 11:00 a.m.–11:29 a.m. 4:00 p.m.–4:29 p.m. 6:30 a.m.–6:59 a.m. 11:30 a.m.–11:59 a.m. 4:30 p.m.–4:59 p.m. 7:00 a.m.–7:29 a.m. 12:00 a.m.–12:29 p.m. 5:00 p.m.–5:29 p.m. 7:30 a.m.–7:59 a.m. 12:30 p.m.–12:59 p.m. 5:30 p.m.–5:59 p.m. 8:00 a.m.–8:29 a.m. 1:00 p.m.–1:29 p.m. 6:00 p.m.–6:29 p.m. 8:30 a.m.–8:59 a.m. 1:30 p.m.–1:59 p.m. 6:30 p.m.–6:59 p.m. 9:00 a.m.–9:29 a.m. 2:00 p.m.–2:29 p.m. 7:00 p.m.–7:29 p.m. 9:30 a.m.–9:59 a.m. 2:30 p.m.–2:59 p.m. 7:30 p.m.–7:59 p.m.