Dynamic, Integrated Model System: Sacramento-Area Application, Volume 1: Summary Report (2014)

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.

16 Development of the Integrated Model This chapter describes the individual components that make up the C10B integrated model and how they were used in the new integrated model. The first section provides details on the model components: SACSIM (including DaySim), DynusT, FAST-TrIPs, and MOVES. The next section describes the revisions made to these models as part of the SHRP 2 C10B project. The third section provides information about the integration of the components. Original Models This section describes the original versions of the models that are the components of the C10B integrated model. SACSIM This section presents a brief summary of the Sacramento regional travel simulation model (SACSIM). SACOG uses this travel demand model in the preparation of transportation plan analyses. Complete documentation of SACSIM can be found in SACOG et al. (2008). SACSIM, the original version of which was completed in 2007, is one of the first activity-based models developed in the United States. While SACOG has more recently updated SACSIM, the original 2007 version—which was in use throughout the first part of Project C10B—is used in the C10B integrated model. The flow of SACSIM is displayed in Figure 2.1. In activity-based models, person travel is modeled from a set of activities that require travel. The activity-based compo- nent of SACSIM, the person-day activity and travel simulator (DaySim), is implemented at the parcel level. The model flow for DaySim is shown in Figure 2.2. Besides DaySim, SACSIM includes other submodels, includ- ing commercial vehicle, external travel, and airport passenger ground access submodels. The major components of SACSIM are summarized in the following paragraphs. A population synthesizer (referred to as “representative pop- ulation generator,” or PopGen, in Figure 2.1) creates a popula- tion database which is the basis for the activities and travel simulated in DaySim. The database comprises person records, drawn from Census Bureau Public Use Microdata Sample (PUMS) households in the Sacramento region. For each scenario, the population data set is consistent with regional residential, employment, and school enrollment forecasts in quantity, location, and key demographic variables such as age and income. Population data sets are generated for each forecast land use alternative and are treated as input files for testing transportation scenarios. The population data set can be modi- fied directly (e.g., changing locations of specific households, changing income or age characteristics) to test the effects of dif- ferent land use forecasts or demographic trend assumptions. Within DaySim, long-term choices (work location, school location, and auto ownership) are simulated for each member of the population. DaySim creates a 1-day activity and travel sched- ule for each person in the population, grouping activities requir- ing travel into “tours” beginning and ending at the person’s home. For each tour and each segment (trip) of each tour, desti- nation, mode, and time-of-day choice at the half-hour level are simulated. The main output of DaySim is a list of all tours made by the synthetic population, including the trips on each tour. In the version of SACSIM currently used by SACOG, the trips from the DaySim outputs are aggregated into trip matrices and combined with predicted trips for what is referred to in the C10B integrated model as “exogenous travel.” Exogenous travel includes airport passenger ground access and egress travel, external travel, and commercial vehicle traffic. The exog- enous travel is generated as zone-to-zone origin–destination matrices for four broad time periods. The aggregation pro- cess creates time- and mode-specific trip matrices, in person trips for transit assignment and vehicle trips for highway assignment. The highway assignment model loads the trips from these matrices onto the highway network using a conventional C h a p t e r 2

17 static equilibrium highway assignment process. A conven- tional static transit assignment process is used to load the transit person trips onto the transit network. The assign- ment process is performed for four broad time periods, representing the a.m. peak, midday, p.m. peak, and night periods. (For the C10B integrated model, these processes of aggregating to trip tables and performing static high- way and transit assignments were replaced by DynusT and FAST-TrIPs.) SACSIM iterates until convergence is achieved. Conver- gence is defined as a model’s internal consistency of major data items (e.g., trip tables, traffic volumes, and level-of- service matrices) used throughout the model process. DynusT The dynamic traffic simulation and assignment model DynusT (Dynamic Urban Systems in Transportation) is designed and implemented to perform simulation-based dynamic traffic assignment (DTA) and associated analysis. It is capable of performing DTA on regional-level networks over a long simulation period, making it particularly well-suited for Source: SACOG et al. (2008). Figure 2.1. SACSIM model system.

18 regional-level modeling such as regional transportation planning, corridor studies, integration with activity-based models, and mass evacuation modeling. This section briefly describes DynusT as implemented for the SHRP 2 C10B project; Volume 2 (Chiu et al. 2014) describes in detail the traffic simulation component of DynusT that captures capacity constraints, congestion, and queue propagation and allows the generation of time-dependent level-of- service (LOS) measures that are closer to traffic theory. DynusT determines the shortest-path route for each driver, a concept that is described as “user equilibrium.” DynusT consists of two main modules: traffic simulation and traffic assignment. Vehicles are created and loaded into the network based on their respective origins and follow a specific route to their intended destinations. The large- scale simulation of networkwide traffic is accomplished through the mesoscopic simulation approach that omits intervehicle car-following details while maintaining realistic macroscopic traffic properties (e.g., speed, density, and flow). More specifically, the traffic simulation is based on the Anisotropic Mesoscopic Simulation (AMS) technique (see Chiu et al. 2010) that calculates a vehicle’s speed from the traffic conditions ahead of the vehicle. Specifically, at each simulation interval, a vehicle’s speed is determined by a speed-density curve, the density being the number of vehi- cles per mile per lane within a limited forward distance. Figure 2.2. DaySim hierarchy and flow.

19 The traffic assignment module of DynusT consists of two algorithmic components: a time-dependent shortest-path (TDSP) algorithm and a time-dependent traffic assignment, or routing. The TDSP algorithm determines the time-dependent shortest path for each origin, destination, and departure time; the traffic assignment component selects a route for each driver following some heuristic rules that lead to approximate user equilibrium, a condition in which each driver has selected the least-cost path available. After shortest paths have been calculated and a route choice has been made, all the vehicles are simulated. DynusT uses the time gap between a vehicle’s simulated travel time and the vehicle’s available shortest-path time to assess the level of convergence. If the average time gap for all the vehi- cles in the simulation is small enough, DynusT terminates and outputs networkwide LOS measures; otherwise it contin- ues iterating between its two models until convergence is achieved. Although DynusT continues to evolve, the version included in this project was completed in 2012. This version included some enhancements made as part of this project to the exist- ing DynusT version at the time. A key enhancement was the simulation of vehicles in the presence of transit vehicles with or without bus pullouts. As illustrated in Figure 2.3, when a bus pullout is present and a transit vehicle resides in the pull- out, the passerby vehicles’ speed-influencing regions (SIR) remain unchanged. On the other hand, without the pullout the stopped transit vehicle typically blocks one traffic lane, creating a temporal blockage to the following traffic steam. The departure from each stop involves different rules for fre- quency or schedule-based transit. The main difference is that for schedule-based transit operation, a transit vehicle needs to be held until the scheduled departure time if the transit vehicle is still ahead of schedule after boarding and alighting. Such vehicle holding is unnecessary in frequency-based operation. Another enhancement to DynusT was made to consider that in the C10B integrated model, demand is generated from a tour-based travel model (DaySim). Before this information can be used for traffic simulation purposes, it must be manipulated to meet DynusT’s specific network and demand inputs. DynusT demand inputs take two forms: (1) the typi- cal origin–destination (O-D) table for specified time periods and (2) vehicle and path files. In general, the exogenous travel components (truck, external, and airport vehicle trips) yield O-D demand files given diurnal factors, while tour/trip records yield vehicle and path demand files. Generating DynusT’s vehicle demand file is a more involved process because it requires detailed trip information as opposed to O-D demand files that simply require O-D and diurnal factors. Examples of this mandatory information include household identification (ID), traveler/person ID, tour/ trip ID, origin–destination parcels/points, origin–destination zones, mode choice, travel time, value of time, and arrival/ departure time. The purpose of this information is to repre- sent a trip as realistically as possible within DynusT’s node- link–based network and context. Examples of this “realistic” representation not only entail correct zone vehicle generation or destinations but, most important, also ensure that a spe- cific person’s trip reaches its destination before his or her next trip (tour) is generated. This instance is usually prevalent in networks with congestion or disruption or trips that are sequenced immediately after one another. FAST-TrIPs This section provides a brief summary of FAST-TrIPs as implemented in the SHRP 2 C10B integrated model. The companion document (Chiu et al. 2014) provides more details of the implementation. FAST-TrIPs interfaces with DynusT and also connects with the DaySim activity-based model. FAST-TrIPs is a regionwide dynamic transit assignment model that determines an individual-specific transit route for each transit traveler in the system, taking into account pub- lished transit schedules and transit vehicle run times that are congestion responsive and are provided by the traffic simu- lation component of DynusT. FAST-TrIPs deals with both tran- sit-only and park-and-ride trips and is able to maintain multiple constraints associated with activity time-windows and the choice of modes in multimodal travel tours. DynusT and FAST-TrIPs interoperate with each other to provide a model system in which the highway and transit assignments influence each other and are based on the same set of LOS variables. FAST-TrIPs is divided into two main submodules: transit assignment and simulation. The transit assignment submod- ule plays the role of passenger assignment for given O-D pairs. For assigning transit passengers for the O-D pairs, a trip-based shortest path model is utilized by searching for a feasible path on each O-D pair. The assigned passengers, including their Figure 2.3. SIR areas with (a) and without (b) bus pullouts.

20 paths, are given to and simulated through the transit simula- tion submodule in FAST-TrIPs. During the simulation, experienced arrival and departure times of transit vehicles are used to simulate boarding and alighting of passengers, considering transfers and other com- ponents (such as walking and waiting). Each passenger’s tra- jectory (i.e., experienced path) is recorded, and dwell time for each transit route is calculated as a function of the boardings and alightings at each stop. Results of the simulation are used in the next iteration of auto-transit vehicle simulation and are also fed back to the activity-based model in the next global iteration for updating the demand. FAST-TrIPs has an intermodal functionality embedded in its two submodules. It is capable of assigning and simu- lating the intermodal passengers in a mixed environment, modeling these movements for auto and transit passengers. The intermodal model consists of a park-and-ride assign- ment model for individual tours, a transit assignment and transit simulation model for the transit portion of the tour, and an interface with DynusT for the auto assignment and simulation. MOVES The Motor Vehicle Emission Simulator (MOVES) is the cur- rent regulatory mobile source emissions model developed by the EPA. MOVES serves as a single comprehensive system for estimating emissions from on-road mobile sources and is officially approved for developing state implementation plans (SIPs) and regional or project-level transportation confor- mity analyses. MOVES is designed to estimate emissions at scales ranging from individual roadways and intersections to large regions. MOVES is a database-driven model. The inputs, outputs, default vehicle activities, base modal emission rates, and all intermediate calculation data of MOVES are stored and man- aged in MySQL databases (see Figure 2.4 for an example). MOVES model functions query and manipulate MySQL data Figure 2.4. Sample emissions data table in MOVES MySQL database.

21 pursuant to scenario parameters specified in a graphical user interface (see Figure 2.5). This design provides users with flexibility in constructing and storing their own database under the unified framework in MySQL. The MOVES model incorporates input data that include vehicle fleet composi- tion, traffic activities, and meteorology parameters at the macro-, meso-, or microscale, and conducts modal-based emissions calculations using a set of model functions. The outputs of emissions inventories or emissions factors are functions of modal-based vehicle emission rates and detailed vehicle activities specified for the desired geographic scale. The MOVES model represents a fundamental shift in the methodology used to estimate on-road vehicle emissions from that of its predecessors (e.g., the MOBILE6 model, which used average speed as the only traffic-related variable to estimate vehicle emissions). MOVES is a modal emissions model in which emissions are calculated based on vehicle- specific power (VSP) derived from second-by-second vehicle performance characteristics for various driving modes (e.g., cruise and acceleration). The modal nature of the MOVES methodology allows the model to, in principle, more accu- rately estimate emissions at analysis scales ranging from those associated with individual transportation projects to large regional emission inventories. Since MOVES was first released in 2005, EPA has been working to refine the model; example improvements over time include updated modeling data, calculation functions, and feature improvements. After the development of two intermediate versions of MOVES (MOVES2004 and MOVES- HVI Demo), EPA released Draft MOVES2009, MOVES2010, MOVES2010a, and MOVES2010b versions, which provide enhanced modeling functions, updated data sources, and bug Figure 2.5. MOVES graphical user interface, geographic bounds page.

22 fixes (see http://www.epa.gov/otaq/models/moves/index.htm for EPA’s MOVES documentation). The University of Illinois, Chicago, and Sonoma Technology, Inc., applied the most recently available MOVES versions during the course of the C10B project (i.e., MOVES2010 and MOVES2010a, released in December 2009 and August 2010, respectively). The MOVES- DynusT integration and data processing approaches are valid for all recent MOVES versions, including MOVES2010b, released by EPA in April 2012. revisions to Original Models for Shrp 2 C10B To meet the objectives of the SHRP 2 C10B project, some revisions were made to the original models that make up the integrated model. This section describes these revisions. Incorporation of Variable Value of Time One of the shortcomings of aggregately applied models is the need to use average values across broad market segments for key parameters. One example is the use of a single value of travel time for each segment for decisions about mode and route choice. The value of time determines the extent to which travelers will pay to save travel time. This is an impor- tant factor in estimating how many and which travelers might use toll roads or managed lanes. In conventional models, and even in existing disaggre- gately applied activity-based models, trade-offs between cost and time are based on relative cost and time parameters. In mode choice models, the parameters may vary by tour or trip purpose and by income level, but the assumed value of time is the same for each traveler within a purpose/income-level segment. In the aggregate route choice (highway assignment) models used in nearly all regions, the value of time is a param- eter that may vary by vehicle class; but these classes are usu- ally defined only by vehicle type (auto, truck) and vehicle occupancy level. Some newer models have begun to incorpo- rate additional segmentation by tour/trip purpose and income level, effectively matching the type of segmentation used in mode choice. The main drawback to this segmentation approach is that individual values of time can vary substantially within these market segments. This variation may lead to inaccurate esti- mates of who would use priced roadways. Because the values of time for segments are averages, they do not include the extremes of very high or low values of time. Furthermore, seg- mentation used to define values of time may coincide with the segmentation needed for analysis of model results. For exam- ple, if a planner wishes to estimate the impacts of a toll road project on low-income travelers, assuming that everyone in that segment behaves the same can produce unreliable results. Because an activity-based model simulates each indi- vidual, an individualized value of time for each individual can be simulated from a probability distribution. This has been done with other activity-based models, such as the SF-CHAMP model (Sall et al. 2010) maintained by the San Francisco County Transportation Authority (SFCTA). However, since nearly all activity-based models use aggre- gate static assignment procedures, segmentation and aver- aging are still required, and the effects of the individual values of time cannot be used in the highway assignment process. Because it incorporates both activity-based demand mod- eling and traffic simulation, the C10B integrated model pro- vides an opportunity to use individual values of time throughout the modeling process. The methods incorporated into the integrated model are described in this section. The original DaySim models are documented in the SACSIM documentation (SACOG et al. 2008). Model Specifications The approach used was to revise the tour mode choice model in DaySim, preserving as much of the existing model as pos- sible. The original DaySim model has models for five tour purposes: work, school, escort, other, and work-based tours. Each model is a nested logit model except the escort purpose, which was estimated as a multinomial logit (MNL) model. The revised models preserve all of the alternative-specific variables included in the original DaySim mode choice mod- els. The only changes to the models are the specification of network LOS variables (e.g., cost and travel time) and the addition of variable value of time (VOT). A key attribute of the original DaySim mode choice models is how out-of-vehicle time (OVT) is specified. Walk and bike access/egress times for transit have the same impact on modal utility of transit alternatives as walk and bike times for non- motorized modes have on modal utility of walk and bike modes. However, walk and bike speeds can vary widely across individuals and depending on terrain and accessibil- ity. Moreover, individuals may perceive nonmotorized modes in different ways from motorized modes of travel. Thus, in the revised model, walk and bike times were treated separately from other motorized mode travel times (as is the case in many other mode choice models). The new specification removed the walk and bike mode travel times from the OVT variables and created two new variables: a walk distance variable and a bike distance variable. These new variables were nonzero only for the walk and bike modes. Since the network skim variables attached to the survey data did not associate any OVT with automobile modes, the new specification has OVT variables for transit modes only.

23 Variable VOT is achieved in the new mode choice models by specifying a distribution for the in-vehicle time (IVT) coefficient, in this case a lognormal distribution. With a fixed cost coefficient, the VOT distribution can be described easily. Instead of estimating a fixed coefficient for OVT in the new models, the ratio of OVT to IVT (typically in the range of 2.0 to 3.0) was estimated. This means the coeffi- cient for OVT also follows a lognormal distribution but is determined by the IVT distribution and the ratio of OVT to IVT. Because of the lack of travel cost variations in the survey data set used to estimate the original tour mode choice mod- els (there are no toll roads in the Sacramento region), VOT distributions were transferred from the San Francisco region. SFCTA and its consultants used stated preference data to esti- mate distributions of VOT for mode and time-of-day choice (Sall et al. 2010). VOT distributions were estimated for four income-level segments. The SFCTA model was chosen as the basis for the C10B work for two reasons. First, it was conducted recently and in a nearby region similar in many ways to the Sacramento region. Second, the implications of the estimated VOT distributions seem reasonable. Figure 2.6 shows the estimated VOT distri- butions for each of four income categories specific to manda- tory travel purposes. For nonmandatory travel purposes, mandatory VOTs are multiplied by a factor of two-thirds. It is important to note that the only parameters imported to DaySim mode choice models from the SFCTA models are those related to the distributions shown in Figure 2.6. Ratios of OVT to IVT were not taken from the SFCTA model, nor was the scale of the SFCTA model. All parameters related to non-LOS variables were estimated using the Sacramento esti- mation data set. Model Estimation Results The model estimation results are shown in a set of tables, as follows: • Table 2.1. Work Tour Mode Choice Estimation Results, • Table 2.2. School Tour Mode Choice Estimation Results, • Table 2.3. Escort Tour Mode Choice Estimation Results, • Table 2.4. Home-Based Other Tour Mode Choice Estima- tion Results, and • Table 2.5. Work-Based Subtour Mode Choice Estimation Results. The first column of each table (Applicable Modes) indicates to which modes each coefficient relates. The following abbre- viations are used in the tables: • DT—drive to transit; • WT—walk to transit; • DA—drive alone; • S2—shared ride, 2 occupants; • S3—shared ride, 3+ occupants; • BI—bicycle; • WK—walk; and • SB—school bus (school tours only). Table 2.6 summarizes model fit statistics for the five mod- els. Details of the estimation process, including several issues encountered, can be found in Lemp et al. (2011). Source: Sall et al. (2010). Figure 2.6. SFCTA work tour value-of-time distributions.

24 Table 2.1. Work Tour Mode Choice Estimation Results Applicable Modes Variable Coefficient t-stat Level-of-Service DA, S2, S3, DT, WT Cost-Income < $30,000 -0.1498 Constr. DA, S2, S3, DT, WT Cost-Income: $30,000–60,000 -0.1022 Constr. DA, S2, S3, DT, WT Cost-Income: $60,000–100,000 -0.0862 Constr. DA, S2, S3, DT, WT Cost-Income > $100,000 -0.0700 Constr. DA, S2, S3, DT, WT Cost-Missing Income -0.0462 -1.15 DA, S2, S3, DT, WT Mean IVT (min) -0.0150 Constr. DA, S2, S3, DT, WT Coefficient of Variation of VOT 1.065 Constr. DT, WT Ratio Wait to IVT 2.50 Constr. DT, WT Ratio Walk/Bike to IVT 3.00 Constr. BI Distance (mi) -0.302 -7.72 WK Distance (mi) -0.956 -7.82 Mode-Specific DT Constant -3.120 -4.96 DT HH fewer cars than workers -1.191 -1.36 DT Drive time/total IVT -0.831 -0.60 WT LRT walk access 2.292 2.49 WT Constant -2.926 -6.39 WT, DT Mixed-use density at destination 0.0109 6.54 S3 Constant -2.175 -5.50 S2, S3 HH #children < age 5 0.361 2.67 S2, S3 HH #children age 5–15 0.283 4.40 S2, S3 HH #nonworking adults 18+ -0.122 -1.25 S2, S3 Log of auto distance (mi) -0.201 -3.99 S3 One-person HH -1.088 -3.02 S3 Two-person HH -1.255 -4.07 S2 Constant -1.700 -4.63 S2, S3 No cars in HH -2.791 -4.17 S2, S3 HH fewer cars than drivers 0.527 3.53 S2 One-person HH -0.725 -2.71 S2, S3 Escort stop purpose/#tours in day 3.479 9.10 S2, S3 Other stop purposes/#tours in day 0.339 2.29 DA Constant 0.760 2.23 DA HH fewer cars than workers -1.093 -5.80 DA HH income < $25,000 -0.709 -3.87 DA Escort stop purpose/#tours in day -2.124 -4.95 DA Other stop purposes/#tours in day 0.127 0.92 (continued on next page)

25 BI Constant -2.914 -5.24 BI Male 1.068 3.12 BI Age > 50 -0.769 -2.22 BI Davis zones 2.818 7.79 BI Mixed-use density at origin 0.0105 2.66 WK Male -0.717 -2.12 WK Mixed-use density at origin 0.00661 1.85 All Mode nesting parameter 0.773 5.14 Table 2.1. Work Tour Mode Choice Estimation Results (continued) Applicable Modes Variable Coefficient t-stat (continued on next page) Table 2.2. School Tour Mode Choice Estimation Results Applicable Modes Variable Coefficient t-stat Level-of-Service DA, S2, S3, WT Cost-Income < $30,000 -0.1947 Constr. DA, S2, S3, WT Cost-Income: $30,000–60,000 -0.1328 Constr. DA, S2, S3, WT Cost-Income: $60,000–100,000 -0.1121 Constr. DA, S2, S3, WT Cost-Income > $100,000 -0.0910 Constr. DA, S2, S3, WT Cost-Missing Income -0.0585 Constr. DA, S2, S3, WT Mean IVT (min) -0.0130 Constr. DA, S2, S3, WT COV of VOT 1.065 Constr. DA, S2, S3, WT Ratio OVT to IVT 2.20 Constr. BI Distance (mi) -0.445 -5.47 WK Distance (mi) -0.711 -10.42 Mode-Specific SB Constant -1.295 -4.05 SB Child < age 5 -0.666 -0.82 SB Adult age 18+ -3.735 -3.61 WT Constant -2.653 -5.05 WT No cars in HH 1.314 2.38 WT HH fewer cars than drivers 0.662 1.80 WT Child < age 5 -4.000 Constr. WT Adult age 18+ 1.721 4.00 WT Child age 16–17 1.229 2.65 WT Mixed-use density at origin 0.0120 2.57 WT Mixed-use density at destination 0.00590 1.31

26 S3 Constant -0.0168 -0.05 S3 One- or two-person HH -1.096 -4.36 S2 One-person HH -1.224 -1.15 S2 Constant -0.568 -1.81 S2, S3 No cars in HH -2.116 -3.54 S2, S3 HH income < $25,000 -0.605 -3.20 S2, S3 HH income: $25,000–50,000 -0.402 -2.83 S2, S3 Child < age 5 1.447 2.53 S2, S3 Escort stop purpose/#tours in day 1.450 5.00 S2, S3 Other stop purposes/#tours in day 0.258 2.41 DA Constant 1.725 4.40 DA HH fewer cars than drivers -1.245 -5.07 DA HH income < $25,000 -1.408 -4.26 DA HH income > $75,000 0.490 1.81 DA Child age 16–17 -1.878 -7.47 DA Escort stop purpose/#tours in day -2.352 -2.56 DA Other stop purposes/#tours in day 0.297 1.38 BI Constant -2.213 -5.29 BI Male 0.693 2.41 BI Davis zones 3.152 10.07 BI Adult age 18+ 0.837 2.55 WK Intersection density at origin 0.00782 2.00 All Mode nesting parameter 0.850 Constr. Table 2.2. School Tour Mode Choice Estimation Results (continued) Applicable Modes Variable Coefficient t-stat Table 2.3. Escort Tour Mode Choice Estimation Results Applicable Modes Variable Coefficient t-stat Level-of-Service S2, S3 Cost-Income < $30,000 -0.2995 Constr. S2, S3 Cost-Income: $30,000–60,000 -0.2043 Constr. S2, S3 Cost-Income: $60,000–100,000 -0.1724 Constr. S2, S3 Cost-Income > $100,000 -0.1400 Constr. S2, S3 Cost-Missing Income -0.0900 Constr. S2, S3 Mean IVT (min) -0.0200 Constr. S2, S3 COV of VOT 1.065 Constr. WK Distance (mi) -3.071 -5.41 (continued on next page)

27 Mode-Specific S3 Constant -0.830 -1.01 S3 HH #children < age 5 0.908 6.28 S3 HH #children age 5–15 0.465 7.60 S3 HH #children age 16–17 -0.371 -2.85 S2 Constant 0.0284 0.03 S2, S3 No cars in HH -6.096 -4.69 WK Age > 50 -0.664 -0.89 WK Intersection density at destination 0.0178 2.23 WK HH #children < age 5 1.041 2.83 WK HH #children age 5–15 0.447 2.18 WK HH #children age 16–17 -1.621 -2.64 All Mode nesting parameter 1.00 Constr. Table 2.3. Escort Tour Mode Choice Estimation Results (continued) Applicable Modes Variable Coefficient t-stat Table 2.4. Home-Based Other Tour Mode Choice Estimation Results Applicable Modes Variable Coefficient t-stat Level-of-Service DA, S2, S3, WT Cost-Income < $30,000 -0.2696 Constr. DA, S2, S3, WT Cost-Income: $30,000–60,000 -0.1839 Constr. DA, S2, S3, WT Cost-Income: $60,000–100,000 -0.1552 Constr. DA, S2, S3, WT Cost-Income > $100,000 -0.1260 Constr. DA, S2, S3, WT Cost-Missing Income -0.0810 Constr. DA, S2, S3, WT Mean IVT (min) -0.0180 Constr. DA, S2, S3, WT COV of VOT 1.065 Constr. DA, S2, S3, WT Ratio OVT to IVT 2.70 Constr. BI Distance (mi) -0.192 -6.36 WK Distance (mi) -1.200 -17.75 Mode-Specific WT Constant -4.569 -6.60 WT No cars in HH 3.009 4.12 WT Intersection density at origin 0.00744 1.44 WT Mixed-use density at destination 0.00593 1.32 WT Shopping tour -1.3488 -1.45 WT Meal tour 1.600 2.08 (continued on next page)

28 S3 Constant -0.916 -3.13 S2, S3 HH #children < age 5 0.483 4.55 S2, S3 HH #children age 5–15 0.0785 1.62 S2, S3 HH #nonworking adults 18+ 0.168 3.80 S2, S3 Log of auto distance (mi) 0.204 6.05 S3 One-person HH -2.769 -12.10 S3 Two-person HH -1.500 -16.45 S2 Constant -0.892 -3.08 S2, S3 No cars in HH -0.816 -2.03 S2, S3 HH fewer cars than workers -0.305 -1.25 S2 One-person HH -1.301 -9.74 S2, S3 Escort stop purpose/#tours in day 1.249 3.16 S2, S3 Other stop purposes/#tours in day 0.343 2.32 S2, S3 Shopping tour 0.191 2.22 S2, S3 Meal tour 1.710 11.37 S2, S3 Social/recreational tour 0.427 4.44 DA Constant 0.778 2.74 DA HH fewer cars than drivers -0.618 -6.80 DA Escort stop purpose/#tours in day -0.796 -1.91 DA Other stop purposes/#tours in day 0.185 1.24 BI Constant -3.976 -8.17 BI Male 0.770 2.56 BI Age > 50 -0.416 -1.38 BI Davis zones 2.296 6.67 BI Intersection density at origin 0.00453 1.08 BI Mixed-use density at origin 0.00977 2.23 BI Social/recreational tour 0.606 1.81 WK Age > 50 -0.322 -1.67 WK Davis zones 0.993 3.36 WK Intersection density at origin 0.0055 2.64 WK Meal tour 1.112 3.15 WK Social/recreational tour 0.969 4.70 All Mode nesting parameter 0.850 Constr. Table 2.4. Home-Based Other Tour Mode Choice Estimation Results (continued) Applicable Modes Variable Coefficient t-stat

29 Table 2.5. Work-Based Subtour Mode Choice Estimation Results Applicable Modes Variable Coefficient t-stat Level-of-Service DA, S2, S3, WT Cost-Income < $30,000 -0.2995 Constr. DA, S2, S3, WT Cost-Income: $30,000–60,000 -0.2043 Constr. DA, S2, S3, WT Cost-Income: $60,000–100,000 -0.1724 Constr. DA, S2, S3, WT Cost-Income > $100,000 -0.1400 Constr. DA, S2, S3, WT Cost-Missing Income -0.0900 Constr. DA, S2, S3, WT Mean IVT (min) -0.0200 Constr. DA, S2, S3, WT COV of VOT 1.065 Constr. DA, S2, S3, WT Ratio OVT to IVT 2.80 Constr. BI Distance (mi) -0.202 -0.64 WK Distance (mi) -1.314 -8.08 Mode-Specific WT Constant -4.094 -5.08 S3 Constant -2.612 -2.64 S2 Constant -3.710 -3.74 S2, S3 Drive alone to work 2.115 2.37 S2, S3 Shared ride to work 2.265 2.59 DA Constant -4.092 -2.93 DA HH income < $25,000 -0.607 -1.31 DA HH income: $25,000–50,000 -0.288 -1.22 DA Drive alone to work 4.243 3.32 DA Shared ride to work 3.163 2.49 BI Constant -11.380 -2.96 BI Male 2.200 0.70 BI Davis zones 8.506 3.23 BI Bike to work 7.500 Constr. WK Mixed-use density at origin 0.00670 2.80 WK Walk to work 5.500 Constr. All Mode nesting parameter 0.750 Constr. Table 2.6. Model Fit Statistics Measure Work School Escort Other Work-Based Observations 3,063 1,484 877 4,526 573 Log likelihood 1,961.7 1,825.4 -603.9 4,306.4 -572.2 Log likelihood @ zero 4,993.1 2,560.8 -897.8 7,293.1 -950.1 Log likelihood constants only 2,870.1 2,246.4 -744.9 5,244.8 -655.7 Pseudo Rho squared @ zero 0.607 0.287 0.327 0.410 0.398 Pseudo Rho squared constants only 0.317 0.187 0.189 0.179 0.127

30 Incorporation of Reliability A method was developed for including reliability into the C10B analysis framework. This section describes the method and how it was implemented in DynusT. The reliability procedure is based at the link level, not the O-D level. The primary purpose was to get reliability esti- mates as an output from the model, as additional perfor- mance measures. However, it is noted that as an input to traveler behavior models, it is the trip reliability that should ideally be used. The method developed for incorpo- rating reliability was a compromise based on a number of constraints: • The scenario method—as explored in SHRP 2 Projects L04, L08, and several previous studies—was ruled out because it would involve multiple runs of the model for each improvement type tested, and run time of the model is high. (The scenario approach is based on defining mul- tiple runs for studying a single improvement type, each made with varying input levels for the factors affecting reliability, such as incidents and demand.) Furthermore, developing scenarios for incidents and work zones on a regional basis is problematic: Where and when to do they start? Focusing on an individual facility would have helped with this problem; but the model only deals with the reli- ability of trips on that facility, not regionally. This is a big issue moving forward in incorporating reliability into regional models. • Project L04 developed a vehicle trajectory processor for simulation models which would have been useful—it could have been used to develop trip-based reliability; but the project schedules did not coincide. The reliability pro- cedure needed to be easily accommodated by SACSIM without any adjustments of recalibration. Therefore, the project team opted for an approach that is based on using indirect measures for assessing reliability. This method is based on the idea that travelers perceive each minute of travel under different conditions with a certain weight [see, for example, Small et al. (1999) and Levinson et al. (2004)]. The concept was originally developed to account for travelers valuing a unit of time under congestion more highly than uncongested time. The project team adapted this approach by assuming that the weight associated with perceived travel time was the reliability component of travel on a link, adjusted for the reliability ratio so that it equilibrates with average travel time. This results in a travel time value that is inflated over what it otherwise would be, a “travel time equivalent.” In the traditional weighting approach, the travel time weights are scaled to increase with increasing link volume-to-capacity (v/c) level. Because unreliability increases as base congestion grows, the travel time equivalents also increase with v/c level. The activity-based model portion of the SACSIM model treats the travel time equivalent in the same man- ner as it would an average travel time without the need for internal adjustments, mechanically speaking, that is. Functionally, how this inflated travel time would affect a model that has been calibrated to average travel time only is unknown. In the future, it will be desirable to account for reliability directly in the traveler behavior modeling process. Quantifying Reliability As an input, reliability affects travelers’ decisions about trip making and the choice of destination, mode, and route. It can be thought of as an extra impedance to travel over and above the average travel time generally used in demand models. Note that the original model’s definition of average travel time is based solely on recurring (demand and capacity) con- ditions. Considering reliability means that nonrecurring sources of congestion factor into the process. The concept of “extra impedance due to unreliable travel” is probably the best way to incorporate reliability into the modeling structure as an input. SHRP 2 Project L04 (Stogios et al. forthcoming) used this approach, in which the impedance on a link can be captured as a generalized cost function that includes both the average travel time and its standard deviation (which is used as the indictor of reli- ability). Because Project L04 was not complete at the time of the relevant work in Project C10B, this project used travel time equivalents. To apply this method, a method must exist for predicting the standard deviation of travel time. SHRP 2 Project L03 (Cambridge Systematics, Inc., et al. 2013) developed such methods from empirical data, using the Travel Time Index (TTI) as the dependent variable. The TTI is defined as the ratio of the actual travel time to the travel time under free- flow conditions, or equivalently: (2.1)TTI FreeFlowSpeed ActualSpeed= Equation 2.1 is a generalized equation for TTI. The follow- ing discussion defines several versions of the TTI for use in reliability estimation. In addition to the TTI calculation, free- flow speed is required for estimating delay. In DynusT net- works, each link is specified with a free-flow speed, so such a value can be readily used for TTI calculation. Because of limitations of the procedures being adapted here, the smallest time period for which travel time per- formance measures can be calculated is 1 h. The same com-

31 putation applies for a different time period, such as 30 min, but with a different aggregation/average period. The equations for versions of the TTI follow, as Equations 2.2 through 2.10. Performance measures for urban freeways th Percentile TTI MeanTTI( )= + p95 1 3.6700 ln (2.2) th Percentile TTI MeanTTI( )= + p90 1 2.7809 ln (2.3) th Percentile TTI MeanTTI( )= + p80 1 2.1406 ln (2.4) = (2.5)0.8601MedianTTI MeanTTI ( )= −p0.71 1 (2.6)0.56StdDevTTI MeanTTI Performance measures for signalized arterials ( )= + p95 1 2.6930 ln (2.7)th Percentile TTI MeanTTI ( )= + p80 1 1.8095 ln (2.8)th Percentile TTI MeanTTI = (2.9)0.9149MedianTTI MeanTTI ( )= −p0.3692 1 (2.10)0.3947StdDevTTI MeanTTI MeanTTI is the grand (overall) mean. Since it was devel- oped from continuous detector data, it includes all of the pos- sible influences on congestion (e.g., incidents and inclement weather). Currently, DynusT only provides an estimate of recurring congestion related to volume and capacity (bottle- necks). Therefore, a MeanTTI based on current DynusT out- put cannot be used. The following method should be used to estimate the true MeanTTI. The method uses the DynusT output to estimate recurring delay and a sketch planning method to estimate incident delay, then combines them. The steps are these: 1. Compute the recurring delay for each link in hours per mile from the simulation model (Equation 2.11): RecurringDelay AverageTravelRate FreeFlowSpeed( )= − 1 (2.11) where AverageTravelRate is the inverse of the DynusT speed. 2. Compute the delay due to incidents (IncidentDelay) in hours per mile using the lookup table for a 1-h period from the ITS Deployment Analysis System (IDAS) User Manual (Cambridge Systematics, Inc., and ITT Industries 2001). This requires the ratio and the number of lanes. The lookup table is shown in Table 2.7. This is the base incident delay. If incident management programs have been added as a strategy or if a strategy lowers the incident rate (fre- quency of occurrence), then the “after” delay is calculated as follows (Equation 2.12): D D R Ra u f d( ) ( )= − −p p1 1 (2.12)2 where Da = adjusted delay (hours of delay per mile); Du = unadjusted (base) delay (hours of delay per mile, from the incident rate tables); Rf = reduction in incident frequency expressed as a fraction (with Rf = 0 meaning no reduction, and Rf = 0.30 meaning a 30% reduction in incident fre- quency); and Rd = reduction in incident duration expressed as a frac- tion (with Rd = 0 meaning no reduction, and Rd = 0.30 meaning a 30% reduction in incident dura- tion). Changes in incident frequency are most commonly affected by strategies that decrease crash rates. However, crashes are only about 20% of total incidents. So, a 30% reduction in crash Table 2.7. Incident Delay Rates: IDAS Delay Rates for 1-Hour Peak (Vehicle-Hours of Incident Delay per Vehicle-Mile) Volume-to-Capacity Ratio Number of Lanes 2 3 4+ 0.05 3.44E-08 1.44E-09 4.39E-12 0.10 5.24E-07 4.63E-08 5.82E-10 0.15 2.58E-06 3.53E-07 1.01E-08 0.20 7.99E-06 1.49E-06 7.71E-08 0.25 1.92E-05 4.57E-06 3.72E-07 0.30 3.93E-05 1.14E-05 1.34E-06 0.35 7.20E-05 2.46E-05 3.99E-06 0.40 0.000122 4.81E-05 1.02E-05 0.45 0.000193 8.68E-05 2.34E-05 0.50 0.000293 0.000147 4.93E-05 0.55 0.000426 0.000237 9.65E-05 0.60 0.0006 0.000367 0.000178 0.65 0.000825 0.000548 0.000313 0.70 0.001117 0.000798 0.000528 0.75 0.001511 0.001142 0.00086 0.80 0.002093 0.001637 0.00136 0.85 0.003092 0.002438 0.002115 0.90 0.005095 0.004008 0.003348 0.95 0.009547 0.007712 0.005922 ≥1.0 0.01986 0.01744 0.01368

32 rates alone would reduce overall incident rates by 0.30 × 0.20 = 0.06. 3. Compute the overall MeanTTI, which includes the effects of recurring and incident delay: Remember that Equation 2.1 (TTI = FreeFlowSpeed/ ActualSpeed) is a general equation for TTI. TTI can also be computed as: ActualTravelTime FreeFlowTravelTime ActualTravelRate FreeFlowTravelRate or To be able to use Equations 2.2 through 2.10, an estimate of the overall mean TTI from a distribution of TTIs (which are just converted travel times) is needed. The overall mean TTI includes all sources of congestion because the equations were based on a year of data at each location. For simplicity, it is assumed that the mean TTI has two components: a recurring mean (from DynusT) and an incident mean (from IDAS). To use the IDAS numbers, which are in terms of delay, everything must be converted into delay and then con- verted back to TTI. Rewriting the original Equation 2.12 yields Equations 2.13A and 2.13B: = (2.13A)MeanTTI MeanTravelRate FreeFlowTravelRate 1 1 (2.13B)MeanTTI v v t t d v d v v v f f f f = = = = From MeanTTI = t t f , it follows that MeanTTI = t t t tf f f = + θ , where q is the total delay (in hours), defined as the additional travel time on top of the free-flow travel time, which is the sum of recurring delay qr and incident induced delay qi; that is, q = qr + qi. Consequently, = = + θ = + θ = + θ + θ = + −     + θ     = + −     + θ     1 1 1 1 1 1 1 1 MeanTTI t t t t t t d v v d v v v v f f f f r i f f i f f f i The final equation becomes this: = + θ   1 1 MeanTTI v v i f This essentially means that MeanTTI is the ratio of the sum of the recurring congestion-induced trip rate and the inci- dent-induced trip rate, to the free-flow trip rate. The term qi is the delay due to incidents (IncidentDelay) and is proposed using the IDAS table in Table 2.7. The table estimates the vehicle-hour of incident delay per vehicle-mile based on v/c ratio for a two-, three-, and four-plus–lane facility. To facilitate the implementation of this table in DynusT, the table is transformed into three polynomial equations that best fit the tabulate data. Each of the fitted curves has an R2 value of at least 0.99, meaning that using this approach is consistent with the original data, but the polynomial equations speed up the value lookup using v/c values. The three equations are shown and graphed in Figures 2.7 through 2.9. The following is an example for a three-lane roadway: • Free-flow speed = 60 mph; • DynusT speed = 45 mph; and • IDAS delay = 0.000798 h/mi (from Table 2.7 or Figure 2.7). Figure 2.7. Incident delay for two-lane roadways.

33 Figure 2.8. Incident delay for three-lane roadways. Figure 2.9. Incident delay for four-plus–lane roadways. = + = 1 45 0.00798 1 60 1.381MeanTTI Note that since the SHRP 2 L03 equations predict the TTI, the travel time can be computed as follows (Equation 2.14): = p (2.14)TravelTime TTI FreeFlowSpeed At the time the reliability calculations were incorporated into the C10B integrated model, coefficients for the reliabil- ity utility function had not yet been developed by Project L04. An alternate method is to compute travel time equiva- lents for reliability. For this purpose, empirical results devel- oped by Small et al. (2005) were used. The authors defined unreliability as the difference between the 80th percentile travel time and the 50th percentile travel time and found the value of unreliability to be approximately equal to the value of time. Based on this result, Equation 2.15 was used to calculate travel time equivalents for a trip: TTE MTT a 80%TT 50%TT( )= + −p (2.15) where TTE = the travel time equivalent on the link; MTT = the mean travel time (min); a = the Reliability Ratio (assumed value is 0.8); 80%TT = the 80th percentile travel time (min); and 50%TT = the 50th percentile travel time (min). MTT, 80%TT, and 50%TT are computed with the equations presented earlier. The “a” parameter reflects the value of unre- liability relative to mean travel time. Based on currently avail- able information, a value of 0.8 was used for this parameter. TTE is used as a replacement for the average travel time in the feedback loop to the activity model. It is basically an inflated value of travel time over the average that accounts for how travelers value reliability.

34 This completes the “input” (demand) side of reliability inclusion. To produce estimates of the economic impact of reliability, total equivalent delay is computed based on the TTE, as shown in Equation 2.16. (2.16) TotalEquivalentDelay TTE FreeFlowTravelTime VMTp( )= − Delay may be decomposed into passenger and commer- cial portions using different travel time equivalents and VMT values. Delay is valued with the usual unit costs for the value of (average) travel time applied to the travel time equivalent. The adjustment for reliability has already been made. DynusT Implementation Details The necessary inputs for the reliability calculation are specified in a newly created input file called “reliability_input.dat.” The contents are these: The first block after the headlines is the coefficients of the polynomial equations resulting from the incident-induced delay specified in Table 2.7. Blocks two and three are the coefficients of the equations specified for free- ways and arterials (Equations 2.2 through 2.10). The final block is the “a” reliability ratio. The file that includes the network skim data includes two columns in addition to the original field representing the mean travel time. These two columns are the toll-related cost, and the computed value a  (80%TT - 50%TT). These entries are in units of minutes. In this version the skim interval input has been changed to be part of the file epoch.dat. The first number in the first line is the skim output interval (30 min). Model Integration Integration of SACSIM and DynusT/FAST-TrIPs The outputs from SACSIM that are inputs to DynusT/FAST- TrIPs in the C10B integrated model are • The tour and trip rosters from DaySim; and • The trip tables representing exogenous travel. The tour and trip rosters already include most of the infor- mation required as input by DynusT, including the origin and destination of each trip and relevant traveler information, such as the simulated value of time (see the previous section on incorporating the variable VOT). The time of day is also pro- vided but only at the half-hour level for trips. In the C10B model, a random start time for each trip is simulated within the appropriate half-hour period. The conversion of the rosters to the input format required by DynusT is performed within the integrated model software. The exogenous travel trip tables must be converted to trip rosters for input to DynusT. This is done using existing pro- cedures for processing trip tables in DynusT. There are trip tables from SACSIM for each of four aggregate time periods (a.m. peak, midday, p.m. peak, and night). Departure time profiles from traffic count data were used to define start times for each trip in the roster. The C10B integrated model is run in an iterative manner until convergence is achieved, as discussed in Chapter 3. The term “big loop” is used to refer to an iteration that includes a complete run of SACSIM and a complete run of DynusT and FAST-TrIPs (which includes internal iterations of its own). Before each big loop after the first, the travel time informa- tion from DynusT and FAST-TrIPs from the previous big loop is fed back as input to SACSIM. The feedback process is somewhat complex because the traffic and transit simulation in DynusT/FAST-TrIPs repre- sents nearly continuous time while the inputs to DaySim are in half-hour intervals, and the inputs to the exogenous travel components of SACSIM are for the four broad time periods. Furthermore, each trip in DynusT/FAST-TrIPs has its own trajectory through the network with its travel time based on the conditions confronted continuously through its jour- ney. There is no single travel time from one point in the net- work to another in DynusT. A specialized process to compute the travel times to be fed back from one big loop to the next was developed for the C10B integrated model. Stated simply, the feedback process employed in the C10B model combines information from all relevant trajectories within a time period (half hour or broad period) to estimate an average time to use as input to SACSIM. The integrated model software executes this process. Integration of DynusT and MOVES A significant feature available in MOVES is the ability to sup- port quantitative project-level emissions assessments using detailed vehicle activity data. The MOVES project-scale analysis function is the most spatially resolved modeling level in MOVES; it calculates emissions from a single roadway link, a group of specific roadway links, and/or an off-network com- mon area (e.g., a transit terminal or park-and-ride lot). [See U.S. Environmental Protection Agency (2012) for additional information.] DynusT is capable of performing up to 24-h simulations of dynamic traffic assignment on roadway networks with sizes ranging from corridor to regional level. DynusT uses iterative interactions between traffic simulation and traffic assign- ment modules to provide detailed and finely resolved travel activity data, such as vehicle trajectories (i.e., when and where a vehicle is located), volume, speed, and density.

35 To take advantage of detailed activity data with improved temporal and spatial resolution, the project team developed a fine-grained integrated method that links MOVES to DynusT at the individual roadway link level. The MOVES-DynusT inte- gration is realized through data conversion functions that use DynusT activity outputs to generate MOVES project-scale inputs. This integration method ensures transition of data flow from DynusT to MOVES without manual intervention or addi- tional data preparation. Detailed DynusT data can be processed in two ways for MOVES project-scale modeling use: (1) drive schedules, in the form of second-by-second speed trajectories, and (2) operating mode distributions, in the form of vehicle run- ning time associated with operating modes defined by speed and vehicle-specific power (VSP) bins. Although MOVES accepts both types of data for calculating link-level emissions, using second-by-second speed trajectories may involve work- ing with a very large data file (especially when modeling a sizable roadway network, as opposed to a small group of roadway links) and significantly increases MOVES modeling time. Therefore, the MOVES-DynusT integration method mainly focuses on using DynusT data to generate operating mode distributions inputs for MOVES. Figure 2.10 presents a flowchart that illustrates the link- age between the MOVES and DynusT models and shows how data files are organized in the integration process. Dur- ing a simulation run that generates detailed vehicle activity data, DynusT uses an intermediate data file (move_input .dat) that contains required data items for calculating hourly operating mode distributions for each roadway link. The key data used by DynusT in this process include roadway parameters (e.g., link ID, hour, and road grade), vehicle type, vehicle count (used for developing traffic volumes), and speed for each simulation interval (for calculating VSP and vehicle operating modes). The operating mode distri- bution data and other processed travel activity data are organized into multiple data tables and used as MOVES input files. In addition to travel activity data converted from DynusT outputs, the MOVES model runs also require non- travel activity data inputs. These inputs—such as vehicle age distribution, fuel supply and formulation, inspection and maintenance program status, and meteorological data (e.g., temperature and relative humidity)—are prepared outside of the MOVES-DynusT integration framework and are con- sistent with DynusT scenarios (i.e., during the same time range and for the same geographic area). Three major data files associated with DynusT modeling runs are used to generate MOVES travel-activity input data: • Network.dat. This DynusT input file, including parameters that describe the roadway network configuration, is used to populate MOVES link attribute inputs, such as link ID, road type, and link length. • Speed.txt. This DynusT output file, including speeds on roadway links for each simulation interval and averaged over a number of intervals, is used to calculate vehicle- specific power and operating mode bins in MOVES. • OutAccVol.data. This DynusT output file, including cumu- lative number of vehicles that go through the midpoint of the link at each minute, is used to calculate vehicle running times and proportions for various operating mode bins. The core data processing to link MOVES and DynusT involves calculation of VSP and operating mode fractions. For each vehicle during a modeled hour, VSP was calculated using the following equation: ( ) ( ) ( ) ( )= × + × + × + + × θ ×sin (2.17) 2 3VSP A M v B M v C M v a g v where VSP = vehicle specific power in kilowatt/tonne; A = load coefficient in (kilowatt-second)/(meter-tonne); B = load coefficient in (kilowatt-second2)/ (meter2-tonne); Figure 2.10. Data flow and organization of MOVES and DynusT integration.

36 C = load coefficients in (kilowatt-second3)/ (meter3-tonne); M = mass of the vehicle in kilograms; g = acceleration due to gravity (9.8 m/s2); v = vehicle speed in meters per second; a = vehicle acceleration/deceleration in meters per second2; and sinq = road grade (fractional). Based on DynusT VSP calculations and speed data, the corresponding operating mode bins can be identified for MOVES use. DynusT volume data and simulation interval information were also used to calculate vehicle time distri- butions associated with various operating mode bins for each roadway link during an analysis hour. Once these oper- ating mode distribution data are ready for MOVES use, there are two steps to set up a MOVES project-scale modeling run. • Step 1. Create MOVES Runspec file. The MOVES Runspec file, typically generated through the MOVES graphical user interface, specifies a MOVES scenario run and contains the following model run information: 44 Description: Brief summary of the purpose of the mod- eled scenario; 44 Scale: Definition of the level of analysis (project-scale in this integration framework); 44 Time spans and aggregation level: Years, months, days, and hours, as well as aggregation by a specified time unit; 44 Geographic bound: Location to be modeled—for exam- ple, the county where the roadway links belong; 44 Vehicle types: Vehicle types as specified by engine type, fuel type, and other vehicle technologies (e.g., gasoline passenger car and gasoline passenger truck); 44 Road types: On-road roadway link or off-network link in urban/rural environment; 44 Pollutants and processes: Each pollutant that would be generated by one or more emission processes (e.g., running exhaust oxides of nitrogen); and 44 Additional user databases: Other user-specified infor- mation. • Step 2. Prepare and load MOVES input data through the MOVES project data manager (PDM) user interface. As shown in Figure 2.11, each tab in the PDM interface win- dow defines the data item required, including travel activity data tables (e.g., “Links,” “LinkSourceTypes,” and “OpmodeDistribution,” generated by processing DynusT data) and nontravel activity data tables (e.g., “AgeDistri- bution,” “Meteorology,” and “FuelSupply,” populated using the MOVES default database or other appropriate data sources). The project-scale MOVES modeling allows for emissions calculation for a given hour during a specific month and analysis year in a single MOVES run. To generate emissions estimates for multiple hours (e.g., daily emissions), batch mode features in MOVES must be employed. MOVES generates two types of emissions outputs, which are stored in a MySQL database: (1) emission inventories with quantity of emissions and/or energy consumption within a region (e.g., for the modeled roadway network) and time span and (2) emission rates with quantity of emissions per unit of activity (e.g., grams per mile). The C10B MOVES-DynusT inte- gration effort focused on using MOVES CO2 emission invento- ries for demonstration purposes. In summary, the project team developed a MOVES-DynusT integration framework with specific approaches to process DynusT travel activity data for MOVES project-scale model- ing runs. MOVES requires traffic-related input data at a reso- lution much higher than can typically be provided by traditional travel demand models. The integration method- ology developed under the C10B project allows for using detailed travel activity data, generated from DynusT, with improved temporal and spatial resolution, to develop modal- based emissions estimated with MOVES. The MOVES-DynusT integration framework and data processing approaches can potentially be used for modeling vehicle emissions at both the regional scale (e.g., a roadway network for a metropolitan area or county) and the project scale (e.g., a highway corridor or local transportation project).

37 Figure 2.11. MOVES project data manager interface and sample travel activity data inputs.