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.
111 C h A P T e r 3 Purposes Travel demand forecasting model systems are only able to test the effects of policies and assumptions which have been explic- itly included when the system is designed and implemented; they are not intrinsically sensitive to the increasingly broad range of transportation policies and improvements of interest to decision makers. While most regional models are sensitive to large-scale assumptions about land use and demographics, few are sensitive to more detailed assumptions about pricing poli- cies or to traffic or travel demand management strategies. Even when models have the capability to address these types of poli- cies, they are typically not sufficiently sensitive to the dynamic interplay between travel behavior and network conditions by time of day, and are not able to reasonably represent the effects of road pricing, travel demand management, and other policies. A key goal of the SHRP 2 C10A project is to make operational a dynamic, integrated travel demand model with a fine-grained, time-dependent network, and to demonstrate the modelâs performance through sensitivity tests and policy analyses. Sensitivity testing of model systems involves the evaluation of the effects of changes in model inputs on model outputs. Although sensitivity testing of model systems can be performed in many ways, two approaches to sensitivity testing of travel demand forecast models are often employed. In the first method, the sensitivity of individual model components is eval- uated by adjusting model inputs and documenting the effect on outputs. Elasticities are calculated and evaluated relative to established standards. In the second method, the focus of the sensitivity testing is on the overall model system. The C10A project took the latter approach and focused on reporting the sensitivities of the model system. Sensitivity Tests A key motivating force behind the SHRP 2 C10A project is the need to address transportation policies that are being consid- ered in metropolitan planning organizations (MPOs) around the United States. These policies are not adequately addressed by the current state-of-the-practice travel- forecasting models, so the integrated modeling tool developed for this project seeks to improve how these policies are addressed. To assess the increased sensitivity of the integrated model system, a set of tests was designed, implemented, and evaluated. These tests were designed to illustrate the unique capabilities of the model system and included the following: ⢠Pricing. Pricing strategies are the costs imposed on travelers using certain roads, traversing certain screenlines, or travel- ing to certain areas (tolling, cordon pricing, or area pricing). These costs may either be fixed or vary by time of day, or they may respond to congestion. Two types of pricing tests were evaluated as part of this effort. In the first, a number of scenarios were defined in which freeway tolls varied by time of day. In the second, a number of scenarios were defined in which auto operating costs were modified from a baseline condition. ⢠Travel demand management. TDM approaches incorporate a wide range of strategies aimed at changing travel behav- ior to reduce congestion and improve mobility. Examples include increasing the number of people who work at home and their frequency of doing so; adjusting work schedules to facilitate travel in off-peak, less congested conditions; or increasing the number of people who carpool to work. This sensitivity testing focused on the impacts of a flexible work schedule in which workers worked fewer days but longer hours on those days. The overall time spent in work activities was held fixed. ⢠Operations. Operational strategies, also known as trans- portation system management (TSM) also address a wide range of projects and changes, including bottleneck improve- ments, corridor improvements, and parking strategies. For this project, the sensitivity testing focused on a scenario in which signals were coordinated along three primary regional corridors with the goal of reducing bottlenecks and improving the overall traffic flow. Model Sensitivity Testing
112 The sensitivity tests documented in this report were per- formed using the Burlington implementation of the model system. Use of this smaller region allowed for more rapid test- ing and debugging of a greater number of scenarios. To ensure a sufficient congestion-related delay on the Burlington regional network (which does not have significant congestion), the socioeconomic inputs to the model system were scaled up by 50%. This increase in the population, employment, and all related inputs exceeds the forecast growth for 2030 in the Burlington region. Two points about the charts and tables that illustrate the results of the sensitivity tests are worth noting. First, DaySim and TRANSIMS were used to generate summa- ries of travel demand and network performance measures, respectively. Second, many of the charts employ time of day along the x-axis to highlight one of the distinguishing features of the integrated model system: the exchange of information between DaySim and TRANSIMS by detailed time of day. Pricing Freeway Tolling The first set of sensitivity test scenarios evaluated using the model system involved assessing the effects of freeway tolling by time of day. For these sensitivity tests, a set of three scenarios were evaluated and compared with the baseline alternative. These scenarios were based on pricing alternatives tested in a pricing experiment conducted by the Puget Sound Regional Council in Seattle, Washington, to observe travel behavior and better understand regional pricing analyses. In the baseline alternative, no costs were assessed at any time. In the Pricing_3 scenario, a fixed $0.25/mi charge was assessed for anyone using the freeways during the peak periods. In the Pricing_4 scenario, the fixed peak charge was maintained, and a fixed $0.10/mi charge was added in midday. Finally, in the Pricing_ 5 scenario, the fixed peak-period charge was increased to $1.00/ freeway mi and the fixed midday charge was increased to $0.50/ freeway mi. When testing the sensitivity of the model system, testing extreme cases or scenarios such as Pricing_5 is often use- ful, even if they are unlikely ever to be implemented in reality. Given the structure and linkages of the DaySim and TRANSIMS models, one would expect that increases in tolls on facilities at certain times of day will result in overall increases in user costs (unless these tolls are optimized, which was not per- formed as part of these tests). These changes may be reflected in decreases in the overall level of activity generation through the upward feedback of aggregate logsum measures, although this effect would likely be small. The effect on the distribution of travel demand by time of day would likely be more pro- nounced, with travelers choosing to reduce travel distances dur- ing the tolled time periods. The effect on mode choice would likely be small because of the relatively few transit service options offered in Burlington. And overall freeway volumes would likely decrease noticeably during the peak periods. Figure 3.1 through Figure 3.3 show that the expected changes were all observed in the model system outputs. Fig- ure 3.1 shows the difference in total trips by time of day relative to the baseline alternative. Pricing_3 shows declines in travel during the a.m. and p.m. peaks, when freeway tolling is in place, but little change during the midday, which is untolled. Pricing_4 and Pricing_5 both show declines in travel during all tolled time periods, with higher tolls resulting in greater reduc- tions in travel. Interestingly, all three pricing scenarios show pronounced increases in travel demand during the evening, suggesting that travelers reschedule activities to occur when no tolls are charged and fewer scheduling constraints are present. -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 03 :0 0 04 :0 0 05 :0 0 06 :0 0 07 :0 0 08 :0 0 09 :0 0 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 00 :0 0 01 :0 0 02 :0 0 PRICING_3 PRICING_4 PRICING_5 Figure 3.1. Difference in trips from base scenario, by hour of day and freeway tolling scenario.
113 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 03 :0 0 04 :0 0 05 :0 0 06 :0 0 07 :0 0 08 :0 0 09 :0 0 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 00 :0 0 01 :0 0 02 :0 0 BASE-WORK PRICING_5-WORK BASE-SOCREC PRICING_5-SOCREC Figure 3.2. Percentage of daily work and social/recreational trips by time of day, by freeway tolling scenario. 0 500 1000 1500 2000 2500 0: 00 1: 00 2: 00 3: 00 4: 00 5: 00 6: 00 7: 00 8: 00 9: 00 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 30-minute time period BASE PRICING_3 PRICING_4 PRICING_5 Figure 3.3. Hours of delay, by time of day and freeway tolling scenario.
114 Figure 3.2 illustrates that the trip making by time-of-day affects different purposes differently, with the work purpose distribu- tions (blue) relatively unaffected, but the social/recreational distributions (red) shifting noticeably out of the peaks and into the evening. Finally, Figure 3.3 illustrates that the network- based total delay is higher than the base in all scenarios, as the tolling induces travelers to shift onto more capacity- constrained surface facilities. An analysis by facility type indi- cated that most of this additional delay accumulates on minor arterials. This result is likely caused by the coarseness of the sensitivity test: the specific temporal and spatial extent of con- gestion on the freeway system did not inform the design of the tolling scheme. Thus, some peak location congestion was likely alleviated as a result of the tolling; however, in many locations across the broad tolling time periods, increased costs were not offset by travel time reductions because of the low levels of baseline congestion. Auto Operating Costs The second set of sensitivity test scenarios evaluated using the model system involved assessing the effects of changing auto operating costs. Auto operating costs are represented in the model system as an average cost per mile experienced by trav- elers. These costs are reflected in both the DaySim demand model, through the inclusion of these monetary costs in the utility specifications of a number of component models, and in the TRANSIMS model through the inclusion of these costs in the generalized costs used for path building. For these sensitivity tests, a set of three scenarios was evalu- ated and compared with the baseline alternative. The baseline alternative assumes a cost of $0.12/mi. A lower cost scenario of $0.06/mi (AOC_X05) was tested as well as two higher costs scenarios of $0.24/mi (AOC_X2) and $0.60/mi (AOC_X5). Given the structure and linkages of the DaySim and TRANSIMS models, in general, increases in auto operating costs would most likely be manifest in overall lower levels of auto ownership, which is a long-term choice that exists toward the top of the DaySim model system and influences subsequent decision making. Higher auto costs would likely have only marginal effects on activity generation and time of day but might have more pronounced effects on overall trip distances (shorter) and on mode choices (more transit). Table 3.1 and Table 3.2 confirm the expected effects on auto ownership and overall activity generation. Table 3.1 shows that when auto operating costs decline the share of households choosing to maintain 0 vehicles also declines; as the costs increase, the share of 0-vehicle households also increases. Table 3.2 illustrates changes in regional tour making by pur- pose and demonstrates that lower costs slightly increase tour making for discretionary purposes while higher costs slightly decrease tour making by purpose. However, the reductions in tour making by purpose are not consistent across pur- poses, with mandatory work and school tours declining while discretionary purposes such as personal business and social/ recreational purposes are seemingly unaffected. By this mea- sure, personal business and social-recreational trips are less discretionary than meals, escorting passengers, and shopping. Consistent with expectations, Figure 3.4 indicates little sys- tematic difference in changes in trips by time of day across the three auto operating cost scenarios; Figure 3.5 illustrates a slight Table 3.1. Household Auto Ownership Shares, by Auto Operating Cost Scenario Autos per Household BASE (%) AOC_X05 (%) AOC_X2 (%) AOC_X5 (%) 0 5.5 5.3 5.7 6.5 1 36.7 36.8 36.7 36.3 2 39.5 39.6 39.5 39.3 3 13.1 13.1 13.0 12.9 4+ 5.2 5.2 5.2 5.0 Table 3.2. Tours, by Purpose and by Auto Operating Cost Scenario Purpose BASE AOC_X05 AOC_X2 AOC_X5 AOC_X05 BASE AOC_X2 BASE AOC_X5 BASE Work 116,928 117,475 115,898 114,321 1.00 0.99 0.99 School 44,011 44,246 43,449 41,906 1.01 0.98 0.96 Escort 41,011 42,022 41,610 41,028 1.00 0.99 0.99 PersBus 45,877 45,756 45,549 45,635 1.00 1.00 1.00 Shop 38,841 39,432 38,210 37,525 1.02 0.97 0.98 Meal 15,908 16,001 16,021 15,794 1.01 1.00 0.99 SocRec 47,181 47,436 46,999 47,076 1.01 0.99 1.00 Total 350,728 352,368 347,736 343,285 1.00 0.99 0.99 Note: Columns 6, 7, and 8 show the ratio between the tours by scenario and the tours in the base.
115 increase in shorter trips and decrease in longer trips associated with the highest assumed auto operating costs. Figure 3.6 shows per capita changes in vehicle miles traveled (VMT) associated with the highest assumed auto operating costs relative to the base, indicating that the most pronounced decreases in VMT are associated with areas located at the periphery of the region. That result reflects the overall higher levels of baseline VMT in those areas and may also reflect boundary effects, in which resi- dents at the edges of the modeled region may be forced to travel further to implement their daily activity patterns. In addition, Figure 3.6 demonstrates the parcel-level spatial resolution used in DaySim. The pattern of per capita VMT increases and decreases may illustrate the effect of the Monte Carlo simu- lation method used in the simulation. A unique feature, as well as limitation, of the spatially and temporally disaggregate model system is that simulation sampling methods used in con- junction with static user equilibrium network assignment are not as easily employed. Travel Demand Management The third set of sensitivity test scenarios evaluated using the model system involved assessing the effects of a travel demand management strategy. TDM approaches are intended to change travel behavior to reduce congestion and improve mobility. They include increasing the frequency and numbers of people who work at home and adjusting work schedules to travel in off-peak, less congested conditions. Because DaySim predicts -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 03 :0 0 04 :0 0 05 :0 0 06 :0 0 07 :0 0 08 :0 0 09 :0 0 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 00 :0 0 01 :0 0 02 :0 0 AOCX05 AOCX2 AOCX5 Figure 3.4. Difference in trips from base scenario, by hour of day and auto operating cost scenario. 0.0% 20.0% 40.0% 60.0% 80.0% 0 5 10 15 20 25 30 35 40 45 BASE-ALLPURP AOC_X5 Figure 3.5. Trip length frequency distribution (in miles), by auto operating cost scenario.
116 Figure 3.6. Per capita changes in VMT between baseline and x5 auto operating cost scenario.
117 the daily activity pattern of each individual in the region, it can be used to reflect the effect of workers working fewer days but longer hours. However, this sensitivity is purely scenario-based. DaySim cannot identify which policies will be most effective at affecting flexible work schedules, though it can estimate the impact on individual travelersâ activity patterns and schedules and on the overall transportation system performance assum- ing that an effective policy is in place. To represent this effective policy, model parameters influencing the work tour and trip generation as well as work durations were modified to repre- sent a shift to working fewer days but more hours, holding the total aggregate time in work activities constant. For this sensitivity test, a single scenario was evaluated in which workers shifted from a 5-day ~7.5-hour workweek to a 4-day ~9 hour workweek. Given the structure and linkages of the DaySim and TRANSIMS models, in general, overall lev- els of activity generation would likely be lower, although the declines in work-related travel might be offset by increases in travel for discretionary purposes. Clearly, shifts in the distri- bution of travel by time of day due to the lengthened workday should be expected. Changes in the destination and mode choices would likely be marginal, though the time-of-day changes should be manifest in volumes by time of day on the roadway network. Note that for this sensitivity test, the adjusted work activity duration distribution represents an analystâs qualitative judg- ment about a potential distribution, which should ideally be informed by more empirical analysis of observed changes in work-tour durations. Table 3.3 demonstrates the impact on the tour patterns of full-time workers, illustrating that as work tours decline, full-time workers tend to make more personal business, social/recreational, and shop tours. Changes in travel by time of day are evident in summaries of the DaySim travel demand model outputs and are also manifest in summaries of network performance by time of day. For example, Figure 3.7 shows a reduction in hours of delay on major arterials asso- ciated with implementation of an effective alternative work Table 3.3. Full-Time-Worker-Tours, by Purpose and TDM Scenario Purpose Original Adjusted Adj/Orig Work 94,408 78,472 0.83 School 115 140 1.22 Escort 8,070 9,023 1.12 PersBus 13,519 16,848 1.25 Shop 10,531 12,938 1.23 Meal 3,817 3,842 1.01 Soc/Rec 13,076 14,360 1.10 Workbased 27,949 23,211 0.83 Total 171,485 158,834 0.93 0 200 400 600 800 1000 0: 00 1: 00 2: 00 3: 00 4: 00 5: 00 6: 00 7: 00 8: 00 9: 00 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 30-minute time period BASE TDM Figure 3.7. Hours of delay on major arterials, by TDM scenario and time of day.
118 schedule policy; this reduction occurs across all types of facilities throughout the region. Operations Signal Progression Operational strategies, also known as transportation system management (TSM), can address a wide range of projects and changes, including bottleneck improvements, corridor improvements, and parking strategies. For this project, the sensitivity testing focused on a signal progression scenario in which signals were coordinated along three primary regional corridors with the goal of reducing bottlenecks and improv- ing the overall traffic flow. The DaySim-TRANSIMS model system provides sensitivity to these improvements. Tradi- tional travel demand forecast models cannot typically rep- resent such improvements because of the modelsâ linkage with traditional static network assignment methods that lack detailed network operation attributes, as well as their coarse temporal resolution. Given the limited geographic extent of these improvements, little change should be expected in any of the aggregate regional statistics measuring activity gen- eration, time of day, or destination and mode choice. How- ever, one would expect to see local level changes would be more likely, reflecting improved speeds along the targeted corridors. The initial model results showed some reductions in delay by facility type, particularly during the peak periods, as shown in Figure 3.8 for major arterials. However, closer inspection of the speed profiles along the three targeted corridors showed more mixed results, with the signal progression producing bet- ter speeds in some corridor directions and worse speeds in other corridor directions. In implementing these operational tests, a significant number of iterations were required to estab- lish a set of baseline signal timings for the entire region that produced reasonable performance results. Before establish- ing new baseline timings, all the corridor progression changes resulted in significant simulation problems. This occurrence likely reflects that such assumptions are necessarily more detailed in geographic and temporal scopeâand ultimately more consequential for the performance of the network simu- lation. As others have noted, the sensitivity of dynamic traffic assignment (DTA) and traffic microsimulation models to detailed inputs suggests that users will encounter distinct chal- lenges when attempting to incorporate these assumptions in a forecasting mode, especially at a regional scale. Of all the sce- narios evaluated as part of this sensitivity testing, the signal progression scenario required the greatest amount of time and resulted in the least interpretable results. 0 200 400 600 800 1000 0: 00 1: 00 2: 00 3: 00 4: 00 5: 00 6: 00 7: 00 8: 00 9: 00 10 :0 0 11 :0 0 12 :0 0 13 :0 0 14 :0 0 15 :0 0 16 :0 0 17 :0 0 18 :0 0 19 :0 0 20 :0 0 21 :0 0 22 :0 0 23 :0 0 30-minute time period BASE PROGRESSED Figure 3.8. Hours of delay on major arterials, by operations scenario and time-of-day.
