Effect of Smart Growth Policies on Travel Demand (2013)

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.

97 A P P e n D I x A The Built environment’s Impacts on Peak Auto Demand Performance Metrics There are a variety of performance metrics for evaluating the effect of the built environment’s impacts on peak auto demand. This section includes examples of metrics from state transpor- tation departments and metropolitan planning organizations (MPOs). Recent overviews of performance metrics, from the Pew Center on the States and the Rockefeller Foundation and from the Transportation Research Board’s Sustainable Trans- portation Indicators Subcommittee, are also discussed. The Florida Department of Transportation (Florida DOT) uses an in-house tool to inform highway expansion planning, and there are several performance metrics by which their tool evaluates projects. As described in Strategic Investment Tool (Florida DOT 2008), the Florida DOT uses five different stra- tegic investment tool (SIT) measures to evaluate projects, which are safety and security, system preservation, mobility, economic competitiveness, and quality of life. In the Florida DOT’s SIT, safety and security is measured by four categories. They are (1) crash ratio, (2) fatal crashes, (3) bridge appraisal rating, and (4) connection to military bases. System preservation is rated according to four mea- sures. These measures are (1) volume-to-capacity ratio, (2) truck volume, (3) vehicular volume, and (4) bridge condi- tion. Mobility is scored by nine measures: (1) connector loca- tion (evaluating a project based on its proximity to priority hubs and corridors), (2) volume-to-capacity ratio of a facility, (3) percent share of truck traffic relative to total traffic, (4) average annual daily traffic, (5) segment deficiencies that result in a system gap, (6) projected change in the volume-to- capacity ratio, (7) interchange operations (used only when evaluating interchanges), (8) bottlenecks and opportunities for grade separation, and (9) daily vehicle hours of delay. Eco- nomic competitiveness is measured by four indices. These indices are (1) demographic preparedness, (2) primary sector robustness, (3) tourism intensity, and (4) supporting facilities. Quality of life is assessed according to four measures, which are (1) land and social criteria (farmland impact, land use, and demographic impact); (2) geology criteria (sinkholes, histori- cal site, contamination); (3) habitat criteria (conservation preservation, wildlife); and (4) water criteria (flood plains/ flood control, coastal/marine, special designations, water quality, and wetlands). The MetroPlan Orlando (2009) 2030 Long Range Transporta- tion Plan analyzed a smart growth land use scenario that “emphasizes compact development, infill and redevelopment, mixing land uses, improved jobs to housing balance within compact urban travel sheds and configurations that support multi-modal transportation.” The effectiveness of this alterna- tive land use strategy was evaluated based on vehicle miles trav- eled (VMT), vehicle hours traveled (VHT), suburban expansion, and the utilization of commuter rail infrastructure. For the Delaware Valley Regional Planning Commission’s (DVRPC 2009) 2025 long-range transportation plan, Connec- tions: The Regional Plan for a Sustainable Future, alternative scenarios were compared for a variety of transportation per- formance metrics, including VMT, vehicle trips, crashes, peak period roadway speed, transit trips, person hours of delay, delay per capita, pedestrian trips, and bicycle trips. Metro, the Portland area MPO, articulates several transportation-related performance targets in its 2035 Regional Transportation Plan (Metro 2010). For congestion, the goal is to reduce 2035 vehicle hours of delay (VHD) by 10% relative to 2005. For travel, the goal is to reduce 2035 VMT by 10% compared to 2005. Metro is not expected to meet either of these targets. While small reductions in VMT are projected, they do not reach 10%. VHD are projected to increase dramati- cally, far above the target of a 10% reduction. The Washington State DOT produces an annual report ana- lyzing highway performance according to various metrics. For example, The 2010 Congestion Report describes several metrics for evaluating the performance of the transportation system. Performance Metrics and Tools

98 System-wide congestion indicators include VMT, VMT per capita, congested lane miles of highway, percent of highway system congested, VHD, and VHD per capita. Corridor-specific congestion indicators include the number of routes where the duration of the congested period improved, the number of routes where the average peak travel time improved, and the number of routes where 95% reliable travel time improved. A 2011 report published by the Pew Center on the States and the Rockefeller Foundation provides a high level overview of performance metrics that guide transportation decision making at the state level. The report, Measuring Transporta- tion Investments: The Road to Results, focuses on six goals that are both important and widely used across the country (Pew Center 2011). These six goals are safety, jobs and com- merce, mobility, access, environmental stewardship, and infra- structure preservation. The performance measures associated with these goals make up an inventory of the most commonly used metrics for assessing transportation systems in the 50 states and Washington, D.C.: 1. Safety: fatalities, injuries, crashes, infrastructure-related (hazard index, high crash areas), response to weather emergencies; 2. Jobs and commerce: jobs created, freight tonnage or ton- miles or by value, freight travel times/speeds, infrastruc- ture support for freight movement, business access to freight services; 3. Mobility: congestion/density, delay, travel times/speed, travel time reliability, accident response, transit on-time performance; 4. Access: access for elderly, disabled and low-income popu- lations, access to multi-modal facilities and services, access to jobs and labor, access to nonwork activities; 5. Environmental stewardship: emissions, fuel consumption/ alternative fuels, air quality, water quality, recycling; and 6. Infrastructure preservation: road condition, bridge condi- tion, remaining life of roads and bridges, rail system con- dition, transit vehicle condition. Performance metrics not only can help to chart a communi- ty’s progress but can also serve to entrench the status quo. One example is a recent table of metrics recommended by the Transportation Research Board’s Sustainable Transportation Indicators Subcommittee. It includes in its “most important (should usually be used)” category the following economic indicator: “Personal mobility (annual person-kilometers and trips) and vehicle travel (annual vehicle kilometers), by mode (nonmotorized, automobile and public transport)” (Litman 2010). While it is helpful to monitor the effects of the built environment on trip making, uncritically citing decreased auto trips and VMT as an indicator of economic loss to be guarded against may work against the goals of smart growth. Application Tools State DOT Strategies State DOT methods for addressing smart growth often take the form of a strategy. For example, the Florida DOT’s SIT is a methodology “for determining project priority and is applica- ble only to evaluating and setting priorities for highway capacity expansion projects” (Florida DOT 2008). There are three main SIT components: (1) a system viewer, which provides back- ground data, short- and long-term plan schedules, and a docu- ment library of former studies; (2) an analyzer, which evaluates performance measures; and (3) a reporter, which displays results in various formats graphical and interactive interfaces. The most relevant planning tool on the New York State Department of Transportation (NYSDOT) Smart Growth Program website is a qualitative checklist for the application of smart growth principles to proposed development projects. The eight sections of the smart growth checklist tool include (1) locating the proposed project near existing infrastructure; (2) providing a range of housing options; (3) protecting open space, farmland, and critical environmental areas; (4) provid- ing a mix of land uses; (5) providing multiple transportation and access choices; (6) designing for walkability and personal interaction; (7) respecting community character; and (8) plan- ning for economic and environmental sustainability. Although the Smart Planning Program is promoted by NYSDOT, its intention is to enable community members to determine “whether a proposed project is likely to contribute to the overall well-being” of their community. The Pennsylvania Department of Transportation, in its 2010 publication Improving the Land Use-Transportation Connection through Local Implementation Tools, states that “Effective com- prehensive plan implementation—most specifically within integrated transportation/land use elements—can enhance the function of the overall transportation system by promoting multi-modal travel and minimizing the demand for single occupancy trips that congest our system at peak travel times.” The following are listed as applicable tools for achieving these goals: access management, site design and roadway standards, traffic operations, zoning for mixed use and density, parking system management, transit revitalization investment districts, joint municipal zoning ordinances, urban growth areas and rural preservation, and zoning overlays. Comprehensive Land Use-Transportation Planning Tools There are a variety of commercially available comprehensive tools for land use-transportation planning. These tools include CommunityViz, Envision Tomorrow, I-PLACE3S, INDEX, Urban Footprint, Rapid Fire, MetroQuest, and TREDIS. Addi- tional land use-transportation tools, such as MXD-P, MXD-V,

99 direct ridership models (DRM), best management practices (BMP), and the Southern California Association of Govern- ments (SCAG) TDM Tool, are sensitive to the effect of trans- portation policies and development scenarios on travel demand. A matrix of the tools and their capabilities (verified by tool providers) is presented in Table A.1. Capabilities are noted by type as well as by scale, depending on their applica- bility to regions, subregions and corridors, or neighborhoods and communities. The following discussion is supplemented with additional coverage of tool characteristics and capabili- ties in topic-specific chapters (mobility by mode and purpose, induced traffic/growth, and smart growth and congestion topic areas) in the main report. These tools typically provide adequate representation of land use data and transportation facilities, as well as the rela- tionship between the built environment and travel demand. Less frequently included in these tools is the ability to reflect demand management, the influence of demand and supply on congestion, or feedback loops for determining induced growth or induced travel. These tools provide a wide range of metrics that is often specific to their area of focus. For example, Urban Footprint produces metrics related to local infrastructure costs, while the DRM estimates transit trips. Additional metrics may be available through customized programming of tools. Each of these tools has been used by at least a handful of MPOs and/or at a state level to perform interactive smart growth scenario evaluations of a broad array of social, eco- nomic, and environmental indicators. Many of the tools per- form analysis of transportation and other effects, while several (MetroQuest, TREDIS, CommunityViz) serve pri- marily as visualization platforms for standard transportation modeling. These tools may also be distinguished from one another by the scale at which they operate, the specific data they require, and the performance indicators they produce. In terms of scale, the different tools operate at one or more of the following levels: • Development project or transit station area TOD in a neighborhood or community (micro); • Subregional or corridor (meso); and • Regional or county (macro). Table A.1 identifies the analysis scale and data requirements of each of these application tools. Table A.2 includes the per- formance metrics that each of application tools will produce. For most prospective users, selection of the most appropriate tool would be a matter of selecting the tool that best addresses the scales of analysis and list of indicators desired and the available data, based on information in Table A.2, as well as logistical questions such as cost, resources required, and cus- tomer support. The data availability subject is addressed in general terms usually under consideration in smart growth scenario planning and evaluation: the land use aggregation level and unit of analysis, and the extent that the model rep- resents the regional transportation network. These tables also include a set of simpler evaluation tools that can be used to selectively produce quick-response indi- cators of the effects of land use and transportation strategies at various scales on specialized subsets of performance met- rics. Those tools are MXD-P (project/plan), MXD-V (vision/ region), DRM, BMP, and SCAG TDM Tool. These transportation–land use interactive effect tools are primarily spreadsheets, some with interactive dashboards, which have been used in local and regional smart growth analysis in various parts of the United States. In some cases these tools pivot from baseline analyses produced by more sophisticated analysis models. Their data requirements are much more limited than those of the multi-issue land use transportation planning tools previously described. Travel Demand Models In a recent set of guidelines, the California Transportation Commission (2010) provides the following summary of travel demand models: Travel demand models are statistical and algorithmic attempts to predict human travel behavior. They endeavor to forecast potential outcomes of various transportation scenarios. Travel demand models provide essential information about the region’s transportation system operations, conditions and performance and they are used to predict future transporta- tion needs. Typical factors that are included in travel demand models are a region’s demographic profile, general plan desig- nations, highway and transit networks, distribution of trips and existing travel patterns including morning and evening peak-hour travel demand, trip generation, and split among automobile (Single Occupancy Vehicle and High Occupancy Vehicle), transit, bicycle, and pedestrian modes of travel. (California Transportation Commission 2010, p. 35) Conventional four-step models remain the most common modeling approach to forecast peak auto demand. A conven- tional four-step model is based on the individual trip and defined by four steps: trip generation, trip distribution, mode choice, and trip assignment. Socioeconomic (household and population) data and/or land use data are translated into a.m. and p.m. peak period trips on highway networks and daily boarding on transit networks. Without significant enhance- ments or off-model adjustments, most four-step models can- not adequately produce hourly volumes and hourly speeds (TRB 2007). A review of the conventional travel forecasting process used in California and throughout the United States identi- fied a variety of limitations in the model systems regarding

100 101 Table A.1. Capabilities of Planning Tools for Evaluating Interactions between Land Use and Transportation Macro Regional or County Meso Subregional or Corridor Micro Neighborhood or Community C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M Land Use Representation Place types b b b b b b b b b b b b b b b b b b b b b b b Parcel-based b b b b b b b b b b b b b b b b b b b Grid-Cell-based b b b b b b b b b b b b b b b b b b Census block b b b b b b b b b b b b b b b b b b Traffic analysis zone b b b b b b b b b b b b Major Transport Net Representation Internal major multimodal net b b b b b b b b Shares data with network model b b b b b b b b b b b b b b b b b b b b Only local connectivity and transit stations b b b b b b b b b b b Relationships Addressed Built Environment ➔ Demand b b b b b b b b b b b b b b b b b b b b b b b b b b Demand Management ➔ Demand b b b b b b b b b b b b b b b b b Demand  Supply ➔ Congestion b b b b b b Feedback/Induced Growth b b b b Feedback/Induced Travel b b Freight b b Note: Comprehensive, multi-issue land use transportation planning tools: CV = CommunityViz, ET = Envision Tomorrow, iP = iPLACE3S, IN = INDEX, UF = Urban Footprint, RF = Rapid Fire, MQ = MetroQuest, and TR = TREDIS. Transportation/land use interactive effect tools: MXP = MXD-P (project/plan), MXV = MXD-V (vision/region), DRM = direct ridership models, BMP = best management practices, and TDM = SCAG TDM Tool. smart growth analysis. DKS Associates et al. (2007), in their assessment of models’ smart growth capabilities, describes the current limitations: 1. Few local jurisdictions in California use models that have sensitivity to smart-growth strategies. Most jurisdictions use models that (a) lack the capability to estimate transit use or carpooling; (b) do not include representation of walking or bicycling trips; and/or (c) do not allow for variation in vehicle trip rates based on land use density, mix, or design. 2. Local jurisdictions using Metropolitan Planning Organiza- tion (MPO) or Congestion Management Agency (CMA) travel demand models that have “moderate- to high- sensitivity” can capture some of the smart-growth sensitiv- ity, but to what degree is not clear. 3. Geographic information system (GIS) systems for local jurisdiction land use and transportation system characteris- tics are making it possible to bring more information into the Urban Transportation Modeling System (UTMS) modeling process, and that has the potential to increase smart growth sensitivity. This includes parcel-level land uses and GIS layers for street systems, bicycle routes, sidewalks, topography, envi- ronmentally sensitive areas, etc. GIS systems are also facili- tating the application of supplemental methods such as I-PLACE3S and INDEX. Because of the current lack of smart growth sensitivity in many models, research has been conducted to develop sup- plemental tools to provide the missing sensitivity. Over the past 15 years, a series of studies have used cross-sectional analyses of variations in travel patterns for zones in major metropolitan areas. These research efforts have documented how four key factors, referred to as the 4 Ds, influence the rate of vehicle use per capita (DKS Associates et al. 2007): • Density—population and employment per square mile; • Diversity—the ratio of jobs to population; • Design—pedestrian environment variables, including street grid density, sidewalk completeness, and route directness; and • Destinations—accessibility to other activity concentra- tions expressed as the mean travel time to all other destina- tions in the region. Research that resulted in the 4 Ds characteristics also pro- duced estimations of “elasticities” regarding vehicle travel per capita with respect to changes in each of the 4 D variables. These elasticities have been used in a variety of application tools to assess the potential vehicle travel reduction benefits of smart growth land use strategies (DKS Associates et al. 2007). The DKS Associates study defines three ranges of modeling improvement regarding sensitivity to smart growth strategies, ranging from low sensitivity to high sensitivity (DKS Associates et al. 2007). Among the high-sensitivity models are those commonly referred to as tour- or activity-based models. Activity-based models are more sensitive to transportation policies, such as pricing, parking, or demand management, than trip-based models. This sensitivity arises from linking travel together over the course of the day in such a way that a policy that influences a round trip (such as the cost of parking at the destination) will be sensitive to all aspects of that round trip. The California Transportation Commission (CTC) con- cludes its guidelines as follows: Additional research and development attention is being directed to tour/activity-based modeling, an approach which is believed to be a significant advance over the traditional trip- based modeling approach. Tour/activity-based models better recognize the complex interactions between activity and travel behavior. These models require more information on travel activity, particularly travel time, focusing on the trip chains and the sequences of activities in the chain, and need more detailed

102 103 data on person and household travel characteristics. These mod- els also require significant time investments in data assembly and model development and resources, which are major challenges typically best addressed by the largest MPOs. Because of these formidable challenges, only a handful of major MPOs across the country are in the relatively early stages of tour/activity-based model development and/or implementation. The mainstream and the state-of-the-practice in travel demand modeling still remains the traditional 4-step trip-based models. However, there are significant add-ons and enhancements to this approach that can improve land use/transportation assessment capabilities. (California Transportation Commission 2010) Examples from the CTC of significant add-ons and enhance- ments for assessing land use/transportation interaction include postprocessing model outputs where models are insensitive to certain policies or factors (such as the Ds) and include feedback loops that account for the effects of congestion on mode choice, induced demand, and induced growth (California Transporta- tion Commission 2010). The recent TRB meta-analysis of advanced travel forecasting practices points out that SACOG selected an activity-based model, in part, due to its anticipated advantages in document- ing how the built environment affects travel decisions. The structure of four-step models can sometimes hinder the mean- ingful comparison of alternative land use scenarios at associ- ated with finer-grained changes. SACOG’s activity-based model was able to demonstrate, for one particular large devel- opment, how a denser development option produced less VMT than an alternative spread option. This approach could presumably extend to peak-hour congestion comparisons as well (TRB 2010). Travel Demand Models and Postprocessing Given the dearth of empirical evidence on smart growth and peak travel, large-scale, regional forecasting models might be the best framework available for tracing the travel demand impacts and congestion (reducing or inducing) effects of smart growth. Still, most large-scale models fail to capture the trip-reducing benefits of smart growth (Cervero 2006). Four- step models were never meant to estimate the travel impacts of neighborhood-scale projects or development near transit stops. Their resolution tends to be too gross to pick up fine- grained design and land use mix features of neighborhood- scale initiatives like new urbanism and TOD. For these and other reasons, it is often necessary to postprocess initial esti- mates to reflect more recent empirical evidence. Differences between the do-nothing versus do-something (i.e., smart growth) scenarios are the best gauge of traffic congestion impacts. Postprocessing normally involves pivoting off four-step model outputs, using elasticity to account for effects (such as those of land use variables) not specifically accounted for in models. Postprocessing has been used to fine-tune generic model estimates to reflect local conditions (Fehr & Peers 2005), assess alternative regional growth scenarios involving Table A.2. Performance Metrics of Planning Tools for Evaluating Interactions between Land Use and Transportation Macro Regional or County Meso Subregional or Corridor Micro Neighborhood or Community C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M C V E T iP IN U F R F M Q T R M X P M X V D R M B M P T D M Daily Vehicle Trips and VMT b b b b b b b b b b b b b b b b b b b b b Daily Transit Trips or Share b b b b b b b b b b b b b b b b b Vehicles by Purpose, Peak b b b b b b b b b b b b b VHT, VHD, Emissions, Energy b b b b b b b b b b b b b b b b b b b b b b b Traveler Cost b b b b b b b b b b Development Cost b b b b b b b b b b b b b b b b b Transportation System/Service Cost b b b b b Location Efficiency b b b b b b b b b b b Economy, Property Values, Jobs b b b b b b b b b b b b Environment and Equity b b b b b b b b b b b b b b b b Livability, Community Character b b b b b b b b b b b b b b b Building Energy Use, Emissions b b b b b Building Water Use, Emissions b b b b b Public Health Impacts, Costs b b b b b Local Infrastructure Costs (Capital, O&M) b b b b b Local/Jurisdictional Revenues b b b b b Land Consumption b b b b b Fiscal Impact b b Resource Usage, Waste Generation b b b Housing Affordability b b b Note: Comprehensive, multi-issue land use transportation planning tools: CV = CommunityViz (CV), ET = Envision Tomorrow, iP = iPLACE3S, IN = INDEX, UF = Urban Footprint, RF = Rapid Fire, MQ = MetroQuest, and TR = TREDIS. Transportation/land use interactive effect tools: MXP = MXD-P (project/plan), MXV = MXD-V (vision/region), DRM = direct ridership models, BMP = best management practices, and TDM = SCAG TDM Tool.

104 jobs-housing balance (Kuzmyak 2006), and predict daily traf- fic for land use and transportation options along proposed multi-modal corridors (Fehr & Peers 2004). In the case of the planned Legacy Parkway west of Salt Lake City, elasticities from national research on “Traveler Responses to Transporta- tion System Changes” were used to pivot off four-step fore- casts to refine estimates (Kuzmyak et al. 2003). One of the more notable examples of postprocessing was to study the travel impacts of redeveloping the Atlantic Station site in central Atlanta (Walters, Ewing, and Schroeer 2000). The Atlanta region’s nonconformity with federal clean air standards held up progress on the project by freezing federal financial assistance for supporting improvements, including a pedestrian bridge to a nearby subway station. The developer argued that a mixed-use infill project near rail transit would yield air-quality benefits by housing population that would otherwise live less centrally, and be more car-dependent. Con- sultants hired to estimate the travel impacts of the Atlantic Steel proposal quickly realized that the four-step model was not up to the task. Thus, four-step model outputs were post- processed. Studies from the San Francisco Bay Area (Cervero and Kockelman 1997) and metropolitan Portland (K. Lawton, personal interview, Sept. 20, 1998) found that the 3 Ds— density, land use diversity, and pedestrian friendly designs— reduced vehicle trip rates and VMT were used to adjust trip generation and mode-choice estimates. Through these modi- fications, the proposed Atlantic Steel location was estimated to reduce future travel by as much as 52% compared to a green- field location. Postprocessing results were pivotal in EPA’s decision to give the Atlantic Steel project a green light. Some of the major shortcomings of postprocessing approaches include: • Most adjustments are made only for the residential produc- tion end of trips, and do not take into account the effects of what is happening at the destination end, which obviously must affect the choice of destination (where that is an option) as well as choice of mode to access the destination (more alternatives to balanced 4 Ds locations, higher costs of driving/parking, less need for a car while at the site). • Some postprocessors estimate only change in VMT, which makes it virtually impossible to ascertain what is happen- ing on the surrounding road network. • Even those postprocessors that estimate changes in trips by mode (in addition to VMT) lack the capacity to account for what destinations in the trip table are being affected. • Most models do not differentiate between work and non- work trips, which appear to be affected by different socio- demographic and land use characteristics and at different magnitudes. • None of the postprocessor approaches differentiate travel by time of day. As a result of the above, the adjustments made through the postprocessor models miss a large part of the behavioral con- struct through which smart growth impacts travel choice. In general, it is anticipated that the predicted benefits are much less than would happen in reality. Mobility by Mode and Purpose Performance Metrics Although they do not typically differentiate by trip purpose, a growing number of transportation agencies have formulated performance metrics for multiple modes of travel. The Florida DOT developed the Quality/Level of Service Handbook in 2009 based on Highway Capacity Manual 2000 (2000), Transit Capacity and Quality of Service Manual, Bicycle level of ser- vice (LOS) Model, and Pedestrian LOS Model. The Bicycle LOS Model evaluates roadway segments and requires a variety of data including average daily traffic, percent heavy vehicles, number of lanes of traffic, posted speed limit, total width of pavement, on-street parking presence and occupancy, outside lane width, pavement condition, and presence designated bike lane. The Pedestrian LOS Model evaluates the width of the outside lane, the width of the shoulder, presence of on-street parking, presence and type of buffer between the walk and a roadway, buffer width, presence of a sidewalk, sidewalk width, traffic volumes, peak-hour factor, number of travel lanes, and average speed. Although each of the methodologies makes use of the LOS A–F scales, the meaning of A–F is not consistent across the modes. Smart Mobility 2010, produced by the California Depart- ment of Transportation (Caltrans) includes several smart mobility goals, including reliable mobility and location effi- ciency. Metrics for reliable mobility include travel times and costs by mode between representative origins and destinations, the day-to-day range of travel time variability between repre- sentative origins and destinations, and mode-specific assess- ments of the quality of service (multi-modal LOS). Metrics for location efficiency include supporting sustainable growth through compliance with regional performance standards; percentage of trips within a corridor or region occurring by high occupancy transit vehicle; households located 30 minutes by transit from employment, 20 minutes by car from employ- ment, and walking distance from schools; and the weighted travel time and cost between trip producers and attractors. The Denver Regional Transportation District’s Quality of Life Study (2008) provides another example of mobility met- rics by mode. Under the objective of improving travel choices and accessibility, several mode-specific measures are listed. Transit measures include access and egress mode, population within walking distance of transit, employment within walk- ing distance of transit, miles of rapid transit facilities, revenue

105 hours of advanced driver assistance service, and transit reve- nue hours. Auto metrics include park-and-ride capacity and utilization. Bicycle metrics include bike-on-bus usage, station bicycle access. Pedestrian metrics include station pedestrian access. Application Tools There is a small field of emerging tools for measuring perfor- mance by mode and trip purpose, including the 2010 High- way Capacity Manual, I-PLACE3S, and Urban Footprint. Recent federal research into multi-modal LOS analysis for urban streets (NCHRP Project 3-70) has resulted in publica- tion of a proposed set of methodologies to analyze LOS for auto, transit, bicycle, and pedestrian modes in Highway Capacity Manual 2010 (2010). The study conducted video laboratories and field surveys involving the general public from four urban areas and then developed a LOS model for each of the four modes (auto, transit, bicycle, and pedestrian). The models were calibrated and validated to observed data and were found to match the public’s perception better than the 2000 Highway Capacity Manual. The method provides an integrated LOS modeling system where changes to a single variable can be quickly evaluated for their effect on each modal LOS. I-PLACE3S is a model that uses real-time GIS to analyze and display the results of different land use scenarios. An option is available in I-PLACE3S to apply the 4 Ds (density, diversity, design, and destinations) to estimate travel behav- ior based on land use change. Specifically, I-PLACE3S can measure how different land use scenarios for a given travel network can affect travel behavior indicators such as VMT, vehicle trips per household, and mode choice, based on the 4 D factors. I-PLACE3S reports percent-change indicators that include transit and bike/walk shares. Urban Footprint uses GIS to create and evaluate physical land use-transportation investment scenarios. The model defines future scenarios through a common set of place types, a range of development types and patterns that varies from higher density mixed use, to single-use zones. Physical and demographic characteristics associated with the place types are used to evaluate each scenario’s impacts. The model produces travel behavior output metrics that include vehicle miles trav- eled, nonauto mode share, and related travel metrics. The MXD tool, mentioned in the tools summary (Table A.1) uses hierarchical modeling to estimate walking and transit use (for external trips) from mixed-use development (Ewing et al. 2011). Walking share of external trips is related to three types of D variables: diversity, destinations accessibility, and demographics. Transit use share of external trips is related to measures of design, destinations accessibility, distance to transit, and demographics. Travel Demand Models The modeling discussion in Chapter 3 alluded to the limita- tions of current models to accurately reflect built-environment characteristics. Similar limitations are evident in addressing the relationship between the built environment and the ten- dency to drive versus walk versus bike versus use transit. In response, a fifth D, distance to rail transit, has been used to accurately estimate transit use based on the built environment and other locally specific determinants of rail patronage (DKS Associates et al. 2007). Many four-step models do not model walking or bicycle travel, which makes it difficult to evaluate smart growth policies including transit-oriented development (TRB 2007). Within the past 10 years, however, more MPOs have incorporated bicycling and walking into the modeling scheme, by introducing a high degree of spatial resolution (i.e., smaller traffic analysis zones that reflect meaningful walking distances) (TRB 2007). Tour/activity-based models offer potential advantages in forecasting mobility by mode and purpose. For example, “Trip-chaining allows mode choice to consider the context of the trips. For example, transit must be available in both the departure and return period for it to be available, so there is an advantage to having a tour-based model that considers the level-of-service in both directions” (TRB 2010, p. 39). Induced Traffic and Induced Growth Performance Metrics The standard metrics used to gauge the degree of induced demand impacts are (a) percent growth in traffic attributed to induced demand over a defined time line and (b) elastici- ties of changes in travel demand as a function of changes in capacity, speed, or built-environment attributes, measured over the short, intermediate, or longer terms. Percent Growth in Traffic Attributed to Induced Demand Studies of impacts at the project level, which could be a specific road improvement or a specific smart growth strategy, typi- cally compare observed traffic counts either along a facility or within a defined impact zone to what would have been expected had the change not occurred. Expected volumes under the null might be based on trend extrapolation, travel demand fore- casts, or comparisons to a control corridor, facility, or neigh- borhood. Thus, if 10,000 ADT is recorded in a surrounding neighborhood prior to a TOD, and 2 years after the TOD open- ing an ADT of 14,000 is recorded, yet only 12,000 ADT is fore- casted (based on trend projections and accounting for the trips generated by the TOD itself), then the share of additional

106 traffic attributable to the TOD is assumed to be 50% - [(14,000 - 12,000)/(14,000 - 10,000)] = 0.50, or 50%. One problem with some before-and-after project-level analyses is they fail to sort out diverted trips from latent trips in gauging induced demand. Additionally, if matched-pair comparisons are conducted (e.g., comparing ADT trends in a TOD versus an otherwise comparable non-TOD setting), it is virtually impossible to find nearly identical projects in terms of income profiles, transit provisions, levels of regional acces- sibility, and other determinants of travel. Elasticities as a Function of Changes in Capacity, Speed, or Built-Environment Attributes By establishing a statistical relationship between travel out- comes and “stimuli” or “intervention,” be it a road expansion or a smart growth strategy, an elasticity can be measured as a general form shown in Equation A.1: Elasticity – % change in Travel Demand attributable to induced traffic % change in Intervention, as measured in speed, density, etc. (A.1)=           The tricky part of this formula is the numerator; that is, sepa- rating changes in traffic that can be assigned to induced traffic or growth impacts. This is normally done within an economet- ric framework involving the use of time series data and multi- ple regression methods to associate changes in travel demand to changes in the intervention, controlling for other factors (e.g., gasoline prices, transit service levels, unemployment rates) that influence travel over time. Mathematically, the elas- ticity derived from a regression model might appear as the beta coefficient (b) for a log-log model or the beta coefficient multi- plied by the ratio of means—b ∗ (X–/Y–)—for a linear model (also known as a mid-point elasticity). The ability to attribute induced demand impacts over time hinges on the ability to introduce a lag structure in the pre- dictive model. If the influences of higher densities on VMT are thought to be negative in the near term, however, some of these impacts might be eroded over the long term and then a distributed lag model might be introduced with the following form in Equation A.2: , , , . . . , , (A.2)–1 –2 –Y f D D D D Ct t t t t k t( )= where Y = VMT, D = density, C = control variables, and t = time series data point. These models normally assume that lag effects taper according to an exponential function, with the strongest influences occurring immediately and impacts atten- uating during longer lag periods (Hansen and Huang 1997; Noland and Cowart 2000; Fulton et al. 2000; Cervero and Hansen 2002; Cervero 2002, 2003). If higher densities are assumed to initially depress VMT (e.g., over Year 0 to Year 2) and some of these benefits erode thereafter (e.g., from Year 3 to Year k), then the model should estimate negative coefficients on Dt, Dt-1, Dt-2, and positive but smaller coefficients on Dt-3 to Dt-k (assuming the net impact of densities over the long run is a diminution of VMT). To the degree a distributed lag model is estimated by using a log-log model structure, then the net induced demand impact of higher densities, adjust for a rebound effect, would be the sum of the marginal coefficients across all lagged express of the variable D. Application Tools No standard, widely accepted kitbag of tools has emerged for estimating induced demand impacts of highway or tran- sit improvements, much less for gauging the second-order, rebound impacts of smart growth strategies. In the absence of such tools, the simplest approach to adjust for possible ero- sion of the traffic-reducing impacts of smart growth is to borrow from the experiences of others. As reviewed in this section, however, the compendium of empirical experiences in this area is quite slim and for many specific initiatives, be they neighborhood-level TOD or regional-scale jobs-housing balance, nonexistent. The best empirical numbers on possible second-order impacts of changes in the built environment are for the diver- sity dimensions of the 3 Ds (Cervero and Kockelman 1997) or 5 Ds (Ewing and Cervero 2001, 2010)—that is, mixed land uses. The direct traffic-reducing impacts of mixed land uses are typically accounted for in the “internal capture” factor, which according to the Institute of Transportation Engineer’s (ITE) Trip Generation manual is generally a small number, on the order of 3% to 5% of total generated trips (Ewing, Dumbaugh, and Brown 2001). A recent analysis of six U.S. regions with mixed-use suburban activity centers found an internal capture rate of 18%, which in combination with non-automobile external trips by walking or transit meant “a total of 29 percent of the trip ends generated by mixed-use development put no strain on the external street network and should be deducted from ITE trip rates for stand-alone sub- urban developments” (Ewing et al. 2011). NCHRP Report 684: Enhancing Internal Trip Capture Esti- mation for Mixed Use Developments (NCHRP 2011) provides an improved methodology to estimate how many internal trips will be generated in mixed-use developments—trips for which both the origin and destination are within the develop- ment. The methodology estimates morning and afternoon peak period trips to and from six specific land use categories: office, retail, restaurant, residential, cinema, and hotel. The 684 methodology is intended to be used at the project level and would therefore not be well suited to the MPO and state level of analysis employed in SmartGAP.

107 By using simple factor methods (more formally, sometimes called “postprocessing”), one can make a plausible, empiri- cally informed adjustment of internal captures accounting for the induced demand impacts of suburban, mixed-use devel- opment. Ascribing to the 18% internal capture factor of Ewing et al. (2011) and the finding of Sperry et al. (2010) that in the suburbs of Dallas around 26% of internal trips are induced, one could adjust the internal capture figure to account for second-order induced travel effects downward to 13.3% - [(0.18) ∗ (1 - 0.26)] = 0.133. By way of example, assume a suburban mixed-use activity center with the following land use program is proposed: (a) 300 apartment units, (b) 50,000 square feet of general office space, (c) 100,000 square feet retail shopping center, and (d) 10,000 square feet health club/fitness center. The esti- mated trip generation impacts and the postprocessing adjust- ments for both internal capture and induced demand effects could proceed as follows. Step 1: Trip Generation Calculation for Each Land Use On the basis of the 2008 Institute of Transportation Engi- neers (ITE) Trip Generation manual rates in Table A.3, the sum-totals of trips generated by these four land uses, ignor- ing possible trip-reducing benefits from their co-presence, are 7,219 daily trips and 669 trips during the p.m. peak hour. Step 2: Internal Capture Adjustment Based on the recent findings of Ewing et al. (2011) that around 18% of total vehicle trips generated by such mixed-use develop- ments are captured internally, the second step involves simply adjusting these estimates down by 18%, assuming the same internal capture rate applies in the weekday and p.m. peak trips: • Weekday trips: 7,219 ∗ (1 - 0.18) = 5,920 • p.m. peak trips: 669 ∗ (1 - 0.18) = 549 Step 3: Induced Demand Adjustment Based on the findings of Sperry et al. (2010) that around 26% of trips that are internally captured for such mixed-used developments are newly generated or induced trips, a third step adjustment could be • Weekday trips: 7,219 ∗ {1 - [(0.18) ∗ (0.26)]} = 6,881 • p.m. peak trips: 669 ∗ {1 - [(0.18) ∗ (0.26)]} = 638 In sum, the initial estimate using ITE unadjusted rates is 7,219 weekday and 669 p.m. peak trips. Accounting for inter- nal capture lowers the estimates to 5,920 weekday and 549 p.m. peak trips. A third round of adjustments that accounts for pos- sible induced demand impacts brings these figures up slightly to 6,257 weekday and 580 p.m. peak trips. One could argue for even further refinements to reflect the traffic impacts of mixed-use development. Some of the traffic going to the shopping center might be pass-by trips, such as motorists pulling over on a whim to pick up a few items. The ITE manual recommends a pass-by adjustment of 34% for shopping centers (ITE Code 820). Thus a reasonable adjust- ment would be to take 34% of generated trips off the top of estimates for shopping centers—that is, 2,832 trips = [(4,292) ∗ (1 - .34)], though caution should be exercised because ITE’s pass-by adjustment rates were derived from a small number of observations. Also, from the Ewing et al. (2011) study, 11.5% of trips produced by mixed-use centers were external trips made by walking or public transit. Mode split adjustments might reduce some of the generated trip estimates by this fig- ure as well, particularly among trips made by residents of the 300 apartment units. State DOT Strategies Through various methods, state DOTs have attempted to measure induced travel and induced growth. The Utah DOT employed an approach for measuring induced demand in Table A.3. ITE Trip Generation Rates by Land Use Code Land Use (Code) Land Use Proposal ITE Vehicle Trip Generation Rates Total (Unadjusted) Generated Trips Weekday p.m. Peak Weekday p.m. Peak Apartments (220) 300 DU 6.65/DU 0.62/DU 1,995 186 General office (710) 50 KSF 11.01/KSF 1.49/KSF 551 75 Shopping center (820) 100 KSF 42.94/KSF 3.73/KSF 4,294 373 Health/fitness club (492) 10 KSF 37.93/KSF 3.53/KSF 379 35 Total 7,219 669 Source: ITE Trip Generation manual (2008). Note: KSF = thousand square feet; DU = dwelling unit.

108 response to a legal challenge from an environmental group regarding the suitability of the Wasatch Front Regional Council (WFRC) travel demand model for analyzing high- way expansion (Schiffer et al. 2005). Sensitivity tests were conducted that held the following constant between future base and future base with the highway: land use, auto owner- ship, trip generation, trip distribution, mode choice, and traf- fic assignment. The highway network was the only component of the WFRC travel demand model that was changed. The sensitivity test produced performance metrics and helped derive elasticity by region and by facility. The study concluded that the WFRC model was sensitive to changes in the highway network. The addition of highway capacity lead to higher VMT, lower VHT, increased driving speeds, and lower transit ridership. Elasticities were more influenced by trip distribu- tion than mode choice or highway assignment, and elasticity values fell within the range found in the literature review. The Florida DOT provides guidance on determining induced growth in Community Impact Assessment: A Hand- book for Transportation Professionals (Florida DOT 2000). Three categories of induced growth related to transportation are identified: (1) “projects serving specific land development,” (2) “projects that would likely stimulate complementary land development,” and (3) “projects that would likely influence regional land development location decisions” (7-5). The handbook observes that the first two categories are easily pre- dictable. For the third category, a checklist approach is favored over a land use modeling approach, which would be more data intensive and costly. The checklist “provides guidance toward a general conclusion on growth inducement potential through systematic consideration of common market factors applied by real estate investors when making a development or purchase decision” (7-5). This tool is based on NCHRP Report 403: Guidance for Estimating the Indirect Effects of Proposed Trans- portation Projects (Louis Berger and Associates 1998). Travel Demand Models Travel demand models are commonly used to predict the demand for transportation services, as described above. More sophisticated models will include some form of feedback loop to provide traveler reaction to the state of the network and will redistribute trips based on the feedback outputs. Advanced travel demand models include feedback loops to take into account the effects of corridor capacity, congestion and bottle- necks on mode choice, induced demand, travel speed and emissions (California Transportation Commission 2010). Wegener’s land use-transport feedback cycle is one represen- tation of these interactions based on activities and accessibilities (TRB 2010). According to this representation, land use, which accounts for population and employment, drives activities, activities rely on the transportation system, the transportation system determines accessibility, and accessibility influences land use. Simulating feedback loops between transportation and land use improves the logical consistency of model forecasts. TRB Special Report 288, Metropolitan Travel Forecasting: Current Practice and Future Direction, summarized the limita- tions of many current travel demand models with regard to induced traffic and induced growth. Since four-step models are not behavioral in nature, they cannot evaluate time shifting of travel in congested networks (TRB 2007). Four-step models are also limited in their ability to represent land use allocation, trip generation, and traffic assignment (Schiffer et al. 2005). Land use allocation methods do not consistently account for acces- sibility effects. Latent demand is not typically considered as part of trip generation. Traffic assignment routing may not be sensi- tive to the impact of queuing. Furthermore, the shortcomings of four-step models are often amplified under congested traffic conditions. When static models use base-year travel behavior parameters for future horizon scenarios, they do not account for the tendency of traffic congestion to shift the share of daily trips occurring during the peak (Schiffer et al. 2005). The Spreadsheet Model for Induced Travel Estimation (SMITE) is a sketch planning model designed by FHWA that uses travel demand model outputs to compare the costs of induced travel with the net societal benefits of highway capac- ity expansion (DeCorla-Souza and Cohen 1998). After esti- mating a diversion of traffic from arterials to the freeway, SMITE applies elasticities that relate decreases in travel time to increases in travel demand. User benefits are estimated based on conventional FHWA cost-benefit analysis proce- dures. External environmental and social costs per VMT are based on user-provided estimates. The Surface Transportation Efficiency Analysis Model (STEAM) is another FHWA model that uses outputs from travel demand models (FHWA 1997). STEAM was developed to estimate the effect of regional transportation projects on mobility and safety at both corridor and system-wide levels. STEAM allows users to produce metrics by user-defined dis- tricts. It also addresses the benefits of increased accessibility resulting from transportation investments by estimating the effect of decreased travel time on employment availability. Relationship Between Smart Growth and Congestion Performance Metrics Evaluating the effectiveness of smart growth design on traffic congestion is a multi-step process, as illustrated previously in Phoenix, Arizona, and Prince George’s County, Maryland. One must first examine the vehicle traffic stream and ascer- tain the degree to which a subject development (or collection of developments) is contributing to that traffic stream. This cannot be credibly done by simply measuring traffic levels on

109 links or at intersections in the immediate proximity of the developments, but requires methods and metrics that can attribute the impacts to source. Methods and metrics that can serve this purpose are • Traffic volumes on individual network links or inter sections by time of day and direction; and • Proportion of those volumes comprising trips with a rela- tionship to the study area (both origin and destination, or either origin or destination within the study area) versus the proportion that are entirely pass through. The through traffic share is an important indicator of the subject area’s impact on traffic. If a traffic level of service stan- dard is violated, it is important to ascertain the portion of the volume leading to the violation that is outside the control of the subject area. Short of expensive travel surveys, the only practical way to estimate these proportions is through “select link” analyses with the regional travel model. By attempting to associate the traffic volumes on a given link with the traffic analysis zone-to-traffic analysis zone (TAZ) trip movements that have been assigned to that link, it is possible to estimate the proportions of internal versus through traffic. As traffic assignment routines in travel forecasting models have become more complex, with many iterations before achieving an equilibrium assignment, this has led some practitioners to question the accuracy by which origin of these trips can be identified. Still, through traffic identification is a critical vari- able, and a select link approach is arguably better than any other available technique (other than origin–destination type studies, which are generally cost infeasible). The second set of performance metrics correspond to the structure and performance of the subject area itself. The mea- sures in this group include the following: • Rates of internal trip capture; • Mode split; • Average trip lengths; and • VMT production. A useful framework for approaching this assessment is simi- lar to the approach described above to attribute traffic con- tributions on identified roadway segments. The framework offers important insight from analyzing a breakdown of key trip market segments. This can be done by manipulating trip table data by trip purpose from the local travel model into the simple construct pictured in Figure A.1. If this compilation is done for each of the primary trip purposes shown, the following useful metrics can be obtained: • First, the proportion of total trips of each type that are retained within the area (Internal–Internal), versus those made to external destinations (Internal–External). If the area has strong smart growth characteristics, it should retain a high proportion of its trips, particularly for non- work travel. • The modal share for each trip purpose for those trips origi- nating in the subject area. If the area has good smart growth characteristics, a high percentage of the Internal–Internal trips should be made by walking, biking or local transit; for trips made outside the area, a high percentage should be made by transit, multi-passenger vehicle (reflected in vehi- cle occupancy), or bicycling. • The average trip length for trips that originate in the sub- ject area should be shorter than average, reflecting that more trips are made locally because of attractive opportu- nities and good connectivity. Combined with less auto use, this should result in lower household and per capita VMT rates for these areas. • For trips made to the area (External–Internal), the indica- tors should show a high percentage of trips arriving by transit, multi-passenger vehicle (occupancy higher), or bicycle/walk. The compact, well-designed nature of the receiving area should make alternative modes attractive Destinations O ri gi ns Internal External Internal External By Trip Purpose Home-Based Work Home-Based Shop Home-Based Other Non-Home Based Figure A.1. Framework of trip market segments.

110 and efficient, and also lead to a high percentage of inter- nally captured non-home-based trips. Application Tools It is acknowledged that conventional TAZ-based travel fore- casting models are poorly suited to estimate the effects of smart growth land patterns on travel behavior. The structure is simply too coarse to capture the effects of density, diversity and design on household and individual travel decisions, which operate at the “walking scale” of the traveler’s environment. These charac- teristics strongly affect choice of destination, mode, linking of trips, number of vehicles owned, and the like, but are outside the resolution of the TAZ. To get at these characteristics, it is necessary to engage other tools that incorporate the character- istics directly (e.g., the Ds models such as I-PLACE3S, INDEX, and Envision Tomorrow) or to look forward to the new genera- tion of activity-based or tour-based models that operate at a much finer level of resolution (parcels or points). It is also nec- essary to use tools that incorporate or are sensitive to 4 Ds mea- sures of built environment in order to evaluate or optimize the overall efficiency of a smart growth design. Nevertheless, for many of the broad measures of impact described above, a great deal of useful information can be derived from analysis of trip table data and traffic assignment results. In many cases it is more about asking the right ques- tions and properly massaging the data than having the exact right tool, per se. The Prince George’s County and Phoenix examples illus- trate how conventional tools and data can be used more effec- tively to address the smart growth versus traffic congestion question. An illustration of what such an analysis can convey is in Figure A.2 used in the Prince George’s County’s study. This setup is for the US-1 North Corridor, one of the six case study sites described earlier. To portray travel flows within the county and in connection with the broader Washington, D.C., region, the county was subdivided into 16 internal districts (not including the six case study areas) and 10 external districts rep- resenting surrounding counties and the District of Columbia. Individual TAZs were then aggregated into these districts, and trip tables reflecting person trips and trips by mode for four primary purposes (work, shopping, other home-based, non- home-based) were for the system of six activity centers plus 16 internal districts plus 10 external districts, or a 32 × 32 ana- lysis universe. The internal districts are denoted as I-1, I-2, and so forth, while the external districts are denoted as E-1, E-2, and so forth. Pulling data from the respective trip tables for this district-level setup, it can be seen that only 18% of trips that originate in the study zone remain within the zone, meaning that 82% travel outside, the largest shares to Montgomery County, Maryland (E-2), and northern Prince George’s County (I-1). Since this is much more of an employment area than a residential area, only 40,700 trips originate within the study area, while 104,300 come to the area from the outside. This is not a particularly transit-oriented area. It does not have a Metrorail station, though there is a MARC commuter rail station, and there is limited walkability in the area. Thus we see that the primary transit use is for home-based work (HBW) travel, which accounts for 23.5% of the 9% of daily trips that originate in the area, and 10.4% of the 33% of HBW trips which are made to the area. Transit use for all other purposes is less than 2%. Walk/bike data were not available for this analy- sis, though given the design, few trips would be expected. Figure A.3 provides additional insight on the nature of trips made by residents in relation to the presumed smart growth design. It shows that only 10% of resident work trips are made to destinations within the study area, which is not particularly uncommon except that this is a jobs-rich setting where a higher live-work rate might be expected. A high percentage of shop- ping trips are made internally, which is a desirable result of smart growth design, and attributable to the rich retail envi- ronment, with a study area ratio of 1.51 retail jobs per house- hold (compared with 0.32 countywide). However, only 19.6% of other home-based trips and 16.4% of non-home-based trips are made within the study area, suggesting that the pur- poses associated with these types of trips are not well served by the design of the corridor. The relative lack of large concentra- tions of identifiable locations for these trips suggests that they are scattered widely about the surrounding region. Such an analysis clearly tells a story that this particular development area is well short of what would be considered adequate smart growth performance: Too few trips retained internally, far too few trips by transit from or to the area, and certainly very little use of transit for nonwork travel or work travel that is not downtown-oriented. While the diagrams and performance indicators shown were generated manually, it would probably not be difficult to create software that would extract these relationships and create the visual elements automatically. GIS tools can be programmed to portray relationships in this manner, and some modeling soft- ware packages (such as TransCAD) actually incorporate such features in their structure and can be programmed for other custom output functions. This includes showing actual traffic conditions and congestion levels on network facilities. New tools are emerging that will contain much more of the desired capability to address land use impacts in the local and regional context. A major shortcoming among even the conventional 4 D models has been the ability to accurately account for pedestrian and bicycle travel. This is due both to the issue of modeling scale, but also reflects not having the functional relationships that are necessary to estimate non- motorized travel demand. The reason this is important is that the ultimate measure of efficiency of a smart growth designed community is in how much it encourages walking and biking

111 for basic travel. If walking and biking are viable alternatives, they can serve as a substitute for auto trips, provide improved access to and from transit, and allow both residents and visi- tors to travel between non-home-based locations without relying on a car. NCHRP Project 08-78 is focused on develop- ing such a modeling capability, which can be used to estimate bicycle and pedestrian demand at the community or corridor levels, for regional planning and policy analysis, and for local bike/pedestrian network design and prioritization (Renais- sance Planning Group et al. 2011). The proposed tools should be capable of not only guiding the development of effective smart growth designs but also accounting for the subsequent effect on traffic levels on local and regional facilities. Smart Growth and Freight Traffic Performance Metrics As used in Lemp and Kockelman (2009a), Zhou, Kockelman, and Lemp (2009), Tirumalachetty and Kockelman (2010), Kakaraparthi and Kockelman (2010), and other papers and reports, the most common method for regional-scale modeling is simulation, at one point in time or over 20+ year horizons (after including land use models), across various policy scenar- ios (e.g., congestion pricing, highway expansions, urban growth boundaries, higher gas prices, and purposeful shifting of job and household locations). Simulations can be disaggregate—at the level of individual households and businesses, for example, or in aggregate (at the level of TAZs). Zone counts generally number 1,000 or more, and link counts of more than 10,000 for regions of 1 million-plus population. Network assignment of traffic in such model almost exclu- sively relies on static assignment (where a link’s congestion cannot impact upstream links), since dynamic user equilib- rium applications require far more detail and longer run times (and stronger assumptions about route choices and the evolv- ing nature of trip tables over the course of a day). Models are estimated based on disaggregate travel records (by households and businesses), and sometimes calibrated based on observed network data. Inventories of job, population, and land use patterns are significant activities for planners that support such models, with data generally applied at the zone level. Metrics for such regional-scale models include regional VMT, VHT, and tons of emission (by type) per modeled travel day (typically a weekday). They regularly include average volume-to-capacity ratios and speeds (by broad time-of-day categories) for the network (though such values are generated at the link level). Kockelman and teammates also regularly provide measures of welfare (using monetized differences in logsums between the base case and alternative scenarios), in order to provide more substantive information than simple travel metrics. For example, travel time savings are not always a good indicator of social benefits. Land use patterns and access can be key to meeting traveler needs. Examples of this include Lemp and Kockelman (2009a) and Gulipalli and Kockelman (2008), who described spatial and demographic relationships in welfare changes under road pricing and other scenarios for Texas regions. Lemp and Kockelman (2009b) offer a detailed examination of how such values can be com- puted, using rigorous nested logit examples. Of course, modelers can also examine particular origin– destination pairs in detail: their travel times and costs before and after a system change. See, for example, Gulipalli and Kockelman (2008). They can seek to quantify the effects of system changes on travel time reliability and crash counts, and value these changes (along with traveler welfare, emissions, and policy costs) using engineering accounting (e.g., net present valuation versus base case values to produce benefit- cost ratios), as in Fagnant et al. (2011). Kockelman and team- mates are finalizing a project evaluation toolkit (PET) that quickly anticipates travel patterns by using constrained maxi- mum entropy techniques and existing or anticipated link- flow inputs, and then pivoting (via incremental logit functions and elastic trip-making equations for all origin–destination pairs) to each scenario’s estimated trip table. The PET pro- vides a variety of comprehensive project impact scores (e.g., internal rates of return and benefit–cost ratios, including their distributions over a series of random simulations, to reflect uncertainty in model parameters and inputs). But PET does so without detailed link systems (e.g., 300 links) or land use information. Coming versions may allow for planners to input their own, more detailed models’ outputs, for PET esti- mation of project values and overall scores. Such details would allow for PET evaluation of multiple land use scenar- ios, once paired with an appropriate travel demand model. In a study of Seattle freight, PSRC (2009) staff identified the following performance metrics for characterizing com- mercial vehicle activities: value of travel time savings and reliability, vehicle and facility operating and capital costs, rev- enues and jobs, access to freight-trip generators (e.g., ports and businesses), emissions rates and costs per ton of pollut- ant, accident rates and costs, and value of network redun- dancy (in case of emergency, resurfacing, or other incidents that impact access times). Many of these are already included in the PET described above, though the toolkit generally assigns generic values to all truck types, rather than allowing for industry- and/or firm-specific variations. Other metrics of interest to this work are inputs to the mod- eling process, particularly those characterizing the transport network, land use patterns, and system behavior. They include free-flow and modeled speeds, link-performance functions (travel time versus demand parameters), signal phasing, and delays. They also include the balance and mix of land uses,

112 Figure A.2. 2030 daily traffic flows in US-1 North Corridor. HH  household, HBO  home- based other trips, HBS  home-based shop, HBW  home-based work, and NHB  non– home-based other trips. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

113 Figure A.3. Internal capture analysis for US-1 North Corridor. HH  household, HBO  home- based other trips, HBS  home-based shop, HBW  home-based work, and NHB  non– home-based other trips. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

114 using simple or sophisticated accessibility indices, entropy equations, and other functions, around points of interest (e.g., homes and businesses), routes of interest, and/or zones. Application Tools The Regional Freight Plan developed by Portland’s Metro in 2010 includes a chapter on developing a freight strategy tool- kit. Freight planning goal categories include system planning for efficient freight mobility and access, system management to increase network efficiency, better public understanding of freight issues, freight-sensitive land use planning, and strate- gic transportation investments. Decision-making tools in the Washington State 2010–2030 Freight Rail Plan released by the Washington State DOT lists the following tools that can be used in modal selection of freight infrastructure: a benefit–cost calculator, a legislative priority matrix, a project management assessment matrix, a user benefit levels matrix, project evaluations, and decision documentation. The DVRPC has published freight planning guidelines as part of its Municipal Implementation Tool series. The 2010 document Freight Transportation articulates a goal of focus- ing goods movement in designated corridors. To achieve this goal, the DVRPC makes several recommendations to cities: improve the links between freight-related transportation and land use, concentrate freight growth in industrial centers, cre- ate freight villages with contiguous freight land uses, and advance access management. The New Jersey Comprehensive Statewide Freight Plan from 2007 concludes that more data and tools are needed for a proper analysis of the freight system. The summary recom- mendations state that “The development of improved data and analysis tools could help determine where it is best to target infrastructure improvement to mitigate current and forecast congestion” (12-14). It also recommends the development of a multi-modal tool that would be used “to gain a better under- standing of the relationship among improvements in capacity, travel times, and reliability at points, corridors, and Interstate routes (or freight lanes) and the impacts on freight movements as part of the overall logistics supply-chain” (12-14). Key Findings and Recommendations Performance Metrics Our research on performance measures proven most effective in comprehensive smart growth and transportation system planning include metrics designed to operate at three impor- tant levels: (1) transportation-specific indicators, (2) metrics that indicate the effectiveness of the regional and local integra- tion of transportation and land use, and (3) higher-level met- rics that capture the effects of land use and transportation decisions on a “triple bottom line” of economic, environmen- tal, and societal impact. Higher-order metrics are particularly noteworthy when eval- uating smart growth benefits. Compared with uncontrolled growth, smart growth development patterns would produce the following savings nationally (Burchell et al. 2002): • 188,305 reduction in local road lane miles, and related sav- ings of $109.7 billion; • Lower local fiscal impact of $4.2 billion; • Reduced property development cost of $420 billion or 6.6%; and • Personal savings related to reduced VMT (auto plus bus) of 4.9 million VMT or $24 billion. The authors identify the following as key metrics that address the effects of smart growth on transportation capacity needs as measured in terms of pure engineering assessment of traffic volume-to-capacity relationships and resulting congestion. The authors also identify the higher-level objectives states and regions are now using to envision and plan their future bal- ance of infrastructure and land use with respect to economic, environmental, and societal return on investment: Transportation Metrics • Daily vehicle trips and VMT; • Daily transit trips or share; • Vehicles by purpose, peak periods; • VHT, VHD, emissions, energy; • Adequate crossing time and intersections; • Right-of-way allocation to all modes (e.g., complete streets); and • Multi-modal level of service. Integrated Transportation/Land Use Metrics • Traveler cost; • Development cost; • Transportation system/service cost; • Location efficiency; • Economy, property values, jobs; • Environment and equity; and • Livability, community character. Higher-Order Metrics • Economic and social value of induced traffic over short and long terms; • Public health impacts and costs; • Local infrastructure costs (capital, operations and maintenance); • Building energy use and emissions; • Building water use and emissions;

115 • Local/jurisdictional revenues; • Land consumption; • Fiscal impact; • Resource usage and waste generation; • Housing affordability; and • Storm water management. The next section addresses whether each of the available application tools is capable of producing the above list of metrics. Application Tools Current Modeling Practice Most MPOs and state DOTs use sophisticated modeling tools to forecast the effects of land use and transportation systems and policies on future traffic levels and the need for roadway capacity expansion. All of the modeling processes contain the following basic elements: • Socioeconomic and land use forecast—projected future population and employment and land use for every sub- area of the region; • Trip generation estimate—the number and purposes of trips that will occur as a result of the future land use; • Trip distribution—the destinations and lengths of each generated trip; • Mode choice—whether each trip will occur by single- occupant automobile, carpool, transit, walking, or biking; • Route assignment—what paths will the auto and transit trips follow to reach their destinations and what volumes of traffic will result on each street and highway segment and what ridership on each transit line; • Capacity analysis—the resulting levels of congestion throughout the roadway and transit networks and result- ing travel speeds and delays; • Travel performance measures—the levels of travel, regional mobility, transportation system performance expressed, for example, as vehicle miles traveled, vehicle hours of delay, congestion levels, and air-quality emissions; and • Multi-dimensional performance—the effects of the land use patterns and transportation system conditions on an array of socioeconomic and environmental indicators specified to reflect regional, state, and federal objectives, such as livability, cost benefit, and return on investment. Within this basic analysis framework, the degree of modeling sophistication varies depending on the size, complexity, and resources of the region. Smaller MPOs often use simpler four- step models that perform basic trip generation, distribution, mode choice, and route assignment to prepare information for the evaluation of travel performance and multi-dimensional regional objectives. Larger MPOs are beginning to adopt more sophisticated activity-based models to perform forecasting at a more refined and policy-oriented level. Some of the most advanced of the activity-based models are reaching the level of specific- ity to adequately address transportation and land use inter- actions at the localized level needed to capture the effects of smart growth on travel demand. However, these models are very complex and resource intensive and even the largest and most advanced MPOs find it challenging to respond to grow- ing demands from decision makers and the public on the sub- ject of smart growth and its effects. The demand for more responsive models emerges from the desire of planners and decision makers to perform interactive scenario evaluations in a public setting and the desire to cap- ture the effects of both regional and community-level smart growth concepts on a diverse set of regional goals and con- cerns. These demands require models that are highly respon- sive, transparent, stable, and sufficiently fine-tuned to capture the effects of both local and regional land use and transporta- tion decisions on levels of travel and accessibility and conse- quential economic, environmental, and societal effects. Models employed by MPOs for evaluating regional transportation investments are, for the most part, too slow and macro scale to address these needs. Standard regional models and even advanced regional models take many hours of processing time to produce results and/or operate at a macro regional scale, too insensitive to capture the critical effects of local land use patterns and transportation choices. Smart Growth Evaluation Tools At least 12 options have emerged to address the need for tools that are responsive to smart growth policies and interactive enough to inform planning processes that involve high levels of engagement with decision makers and the public. They include • Simple spreadsheets to address a subset of planning factors and performance measures; • Sophisticated GIS tools that allow scenario planning at the land use parcel level and produce a large variety of perfor- mance indicators; and • Tools that provide a visual interface dashboard for present- ing the results of a set of analyses performed on the full MPO models in advance of the planning sessions. Of the comprehensive, multi-issue land use transportation planning tools, the most well known and commonly used (and shown in Tables A.1 and A.2) are: • CommunityViz; • Envision Tomorrow; • INDEX; • iPLACE3S; • MetroQuest;

116 • Rapid Fire; • Urban Footprint; and • TREDIS. Each of these tools has been used by at least a handful of MPOs and/or at a state level to perform interactive smart growth scenario evaluations of a broad array of social, economic and environmental indicators. Many of the tools perform ana lysis of transportation and other effects, while several tools (MetroQuest, TREDIS, CommunityViz) serve primarily as visualization platforms for standard transportation modeling. These tools may also be distinguished from one another by the scale at which they operate, the specific data they require, and the performance indicators they produce. In terms of scale, the different tools operate at one or more of the following levels: • Development project or transit station area TOD (micro); • Corridor/community (meso); and • County or regional (macro). The table also identifies a set of simpler evaluation tools that can be used to selectively produce quick-response indicators of the effects of land use and transportation strategies at various scales on specialized subsets of performance metrics. Those tools are MXD-P (project/plan), MXD-V (vision/region), DRM, BMP, and SCAG TDM Tool. These land use and transportation interactive effect tools are primarily spreadsheets, some with interactive dashboards, which have been used in local and regional smart growth analysis in various parts of the United States. In some cases these tools pivot from baseline analyses produced by more sophisticated analysis models. Their data requirements are more limited than those of the multi-issue land use transpor- tation planning tools described above. With respect to the primary purpose of the SHRP 2 C16 research and capacity building effort, a most critical question in tool selection is the question of which tools are capable of addressing the underlying relationships that measure the effects of smart growth on transportation system capacity needs. Table A.1 also indicates which of the core relationships each of the available application tools address. While most of the application tools address the effects of built environment on daily travel demand and about half address the effects of travel demand management on amounts of travel, a critical finding of this first-phase C16 analysis is that few of the avail- able tools address the effects of: • The relationship between peak travel demand and network supply (capacity) on congestion; • Congestion and accessibility on induced growth or induced travel; and • Freight demand and urban form on system capacity needs. In addition, no single application tool addresses all three fac- tors at any analysis scale. Information Gaps and Limitations of Current Practice Performance measures and metrics to evaluate the effects of smart growth on transportation system capacity needs should be compatible with and integrated with the metrics used for the broad range of regional and local transportation plan- ning, such as MPO regional transportation plans. Metrics should operate at three basic levels: (1) transportation-specific indicators, (2) metrics that indicate the effectiveness of the regional and local integration of transportation and land use, and (3) higher-level metrics that capture the effects of land use and transportation decisions on a triple bottom line of economic, environmental, and societal impact. Examples of transportation-specific indicators include VMT and VHD. Integrated land use and transportation metrics include loca- tion efficiency and induced travel impacts, livability and community character. Higher-order metrics include public health impacts, housing affordability, and fiscal impacts. Models used by MPOs and DOTs are too macro scale to fully address the effects of smart growth on trip reduction and the complexities of location-specific congestion and needed reme- diation. Regions with sufficient resources can fine-tune their models and add policy sensitivities through activity-based formulations and can analyze congestion and infrastructure needs through more detailed and sophisticated tools such as dynamic traffic assignment and simulation. However, most regions lack the resources to achieve these goals in the short or medium term. Furthermore, the resulting highly sophisticated models would not achieve the other goals cited by the agency representatives as important for smart growth scenario plan- ning: (a) the capability to perform quick-response visioning and scenario analysis and (b) the ability to scale effectively between the local, corridor and regional levels of analysis for effective communication with local governments and subregional agencies and the public. While there are at least 12 application tools that have been successfully used as stand-alone or to supplement regional travel models for scenario planning and production of travel, socioeconomic, and environmental indicators, few of the avail- able tools address the effects listed in the section above. Again, no single application tool addresses all three factors at any analysis scale. In conclusion, subsequent tasks of the Capacity Project C16 work effort will need to address the means through which to overcome the lack of sound and transferable knowledge on the phenomenon of induced travel, the effects of smart growth on peak travel generation, and the effects of network connec- tivity on infrastructure capacity needs. Subsequent work will

117 also need to investigate the lack of application tools equipped to address these issues. References Burchell, R., G. Lowenstein, W. Dolphin, C. Galley, A. Downs, S. Seskin, K. Still, and T. Moore. 2002. TCRP Report 74: Costs of Sprawl—2000. TRB, National Research Council, Washington, D.C. California Transportation Commission. 2010. 2010 California Regional Transportation Plan Guidelines. Sacramento. Cervero, R. 2002. Induced Travel Demand: Research Design, Empirical Evidence, and Normative Policies. Journal of Planning Literature, Vol. 17, No. 1, pp. 3–20. Cervero, R. 2003. Road Expansion, Urban Growth, and Induced Travel: A Path Analysis. Journal of the American Planning Association, Vol. 69, No. 2, pp. 145–163. Cervero, R. 2006. Alternative Approaches to Modeling the Travel-Demand Impacts of Smart Growth. Journal of the American Planning Associa- tion, Vol. 72, No. 3, pp. 285–295. Cervero, R., and M. Hansen. 2002. Induced Travel Demand and Induced Road Investment: A Simultaneous Equation Analysis. Journal of Transport Economics and Policy, Vol. 36, No. 3, pp. 469–490. Cervero, R., and K. Kockelman. 1997. Travel Demand and the 3Ds: Density, Diversity, and Design. Transportation Research Part D: Transport and Environment, Vol. 2, No. 3, pp. 199–219. DeCorla-Souza, P., and H. Cohen. 1998. Accounting for Induced Travel in Evaluation of Urban Highway Expansion. Available at http:// www.fhwa.dot.gov/steam/doc.htm. Delaware Valley Regional Planning Commission, 2009. Connections: The Regional Plan for a Sustainable Future. Philadelphia, Pa. Denver Regional Transportation District. 2008. Quality of Life Study. Denver, Colo. DKS Associates et al. 2007. Assessment of Local Models and Tools for Analyzing Smart-Growth Strategies. State of California Business, Transportation and Housing Agency, California Department of Transportation, Sacramento. Donnelly, R., G. D. Erhardt, R. Moeckel, and W. W. Davidson. 2010. NCHRP Synthesis of Highway Practice 406: Advanced Practices in Travel Forecasting. Transportation Research Board of the National Academies, Washington, D.C. Ewing, R., and R. Cervero. 2001. Travel and the Built Environment: A Synthesis. In Transportation Research Record: Journal of the Trans- portation Research Board, No. 1780, TRB, National Research Coun- cil, Washington, D.C., pp. 87–114. Ewing, R., and R. Cervero. 2010. Travel and the Built Environment: A Meta-Analysis. Journal of the American Planning Association, Vol. 76, No. 2, pp. 265–294. Ewing, R., E. Dumbaugh, and M. Brown. 2001. Internalizing Travel by Mixing Land Uses: Study of Master-Planned Communities in South Florida. In Transportation Research Record: Journal of the Transpor- tation Research Board, No. 1780, TRB, National Research Council, Washington, D.C., pp. 115–120. Ewing, R., M. Greenwald, M. Zhang, J. Walters, M. Feldman, R. Cervero, L. Frank, and J. Thomas. 2011. Traffic Generated by Mixed-Use Developments: A Six-Region Study Using Consistent Built Environ- mental Measures. Journal of Urban Planning and Development, Vol. 137, No. 3, pp. 248–261. Fagnant, D. J., K. M. Kockelman, and C. Xie. 2011. Anticipating Roadway Expansion and Tolling Impacts: A Toolkit for Abstracted Networks. Proceedings, 90th Annual Meeting of the Transportation Research Board, Washington, D.C. Fehr & Peers. 2004. Legacy Parkway Supplemental Environmental Impact Statement: Draft Technical Memorandum on Integration of Highways and Transit in the North Corridor. Salt Lake City, Utah, Federal Highway Administration regional office. Fehr & Peers. 2005. SLOCOG Travel Demand Model–Development Report. San Luis Obispo, Calif. FHWA. 1997. Surface Transportation Efficiency Analysis Model (STEAM). Federal Highway Administration. http://www.fhwa.dot.gov/steam/ Florida Department of Transportation. 2000. Community Impact Assess- ment: A Handbook for Transportation Professionals. Tallahassee. Florida Department of Transportation. 2008. Strategic Investment Tool. Tallahassee. Florida Department of Transportation. 2009. 2009 Quality/Level of Ser- vice Handbook. Tallahassee. Fulton, L., D. Meszler, R. Noland, and J. Thomas. 2000. A Statistical Analysis of Induced Travel Effects in the U.S. Mid-Atlantic Region. Journal of Transportation and Statistics, Vol. 3, No. 1, pp. 1–14. Gulipalli, P., and K. Kockelman. 2008. Credit-Based Congestion Pricing: A Dallas–Fort Worth Application. Transport Policy 15 (1), pp. 23–32. Hansen, M., and Y. Huang. 1997. Road Supply and Traffic in California Urban Areas. Transportation Research Part A: Policy and Practice, Vol. 31, pp. 205–218. Highway Capacity Manual. 2000. TRB, National Research Council, Washington, D.C. Highway Capacity Manual 2010. 2010. Transportation Research Board of the National Academies, Washington, D.C. Institute of Transportation Engineers. 2008. Trip Generation, 8th Edi- tion. ITE Washington, D.C. Kakaraparthi, S., and K. Kockelman. 2010. An Application of UrbanSim to the Austin, Texas Region: Integrated Model Forecasts for the Year 2030. Journal of Urban Planning and Development, Vol. 137, No. 3, pp. 238–247. Kuzmyak, R. 2006. Test Applications of New Models of Household VMT and Vehicle Ownership on Alternative Growth Scenarios for the I-95 Corridor in Howard and Anne Arundel (MD) Counties. Unpublished white paper. Baltimore Metropolitan Council, Baltimore, Md. Kuzmyak, R., R. Pratt, G. Douglas, and F. Spielberg. 2003. TCRP Report 95: Traveler Response to Transportation System Changes. Transportation Research Board of the National Academies, Washington, D.C. Lemp, J., and K. Kockelman. 2009a. Anticipating Welfare Impacts via Travel Demand Forecasting Models: Comparison of Aggregate and Activity-Based Approaches for the Austin, Texas Region. In Transpor- tation Research Record: Journal of the Transportation Research Board, No. 2133, Transportation Research Board of the National Academies, Washington, D.C., pp. 11–22. Lemp, J., and K. Kockelman. 2009b. The Financing of New Highways: Opportunities for Welfare Analysis and Credit-Based Congestion Pricing. Proc., 88th Annual Meeting of the Transportation Research Board, Washington, D.C. Litman, T. 2010. Sustainable Transport Indicator Data Quality and Availability. Paper 10-2496. 2010 Annual Meeting Compendium of Papers of the Transportation Research Board, Washington, D.C. Louis Berger and Associates. 1998. NCHRP Report 403: Guidance for Estimating the Indirect Effects of Proposed Transportation Projects. TRB, National Research Council, Washington, D.C. Metro. 2010. 2035 Regional Transportation Plan. Portland, Ore. MetroPlan Orlando. 2009. 2030 Long Range Transportation Plan. Orlando, Fla. National Cooperative Highway Research Program. 2011. NCHRP Report 684: Enhancing Internal Trip Capture Estimation for Mixed

118 Use Developments. Transportation Research Board of the National Academies, Washington, D.C. New Jersey. 2007. Comprehensive Statewide Freight Plan. New York State Department of Transportation. Smart Growth Checklist. https://www.nysdot.gov/programs/smart-planning/tools. Noland, R. B., and W. A. Cowart. 2000. Analysis of Metropolitan High- way Capacity and the Growth in Vehicle Miles of Travel. Transporta- tion, Vol. 27, No. 4, pp. 363–390. Pennsylvania Department of Transportation. 2010. Improving the Land Use-Transportation Connection through Local Improvement Tools. Harrisburg. The Pew Center on the States and the Rockefeller Foundation. 2011. Measuring Transportation Investments: The Road to Results. New York. PSRC. 2009. Planning for Freight in the Central Puget Sound Region. Puget Sound Regional Council report. Seattle, Wash. Renaissance Planning Group et al. 2011. NCHRP Project 08-78: Esti- mating Bicycling and Pedestrian Demand for Planning and Project Development. Interim Report, April 2011. Schiffer, R. G., M. W. Steinvorth, and R. T. Milam. 2005. Comparative Evaluations on the Elasticity of Travel Demand. Committee on Transportation Demand Forecasting, Transportation Research Board, Washington, D.C. Sperry, B., M. Burris, and E. Dumbaugh. 2010. A Case Study of Induced Trips at Mixed-Use Developments. Washington, D.C. Presented at 89th Annual Meeting of the Transportation Research Board, Wash- ington, D.C. Tirumalachetty, S., and K. Kockelman. 2010. Forecasting Greenhouse Gas Emissions from Urban Regions: Microsimulation of Land Use and Transport Patterns in Austin, Texas. Proc., 89th Annual Meeting of the Transportation Research Board, Washington, D.C. Transportation Research Board. 2007. Special Report 288: Metropolitan Travel Forecasting: Current Practice and Future Direction. Transpor- tation Research Board of the National Academies, Washington, D.C. Transportation Research Board. 2010. NCHRP Synthesis Report 406: Advanced Practices in Travel Forecasting. Transportation Research Board of the National Academies, Washington, D.C. Walters, G., R. Ewing, and W. Schroeer. 2000. Adjusting Computer Mod- eling Tools to Capture Effects of Smart Growth, or Poking at the Project Like a Lab Rat. Transportation Research Record 1722: Journal of the Transportation Research Board, TRB, National Research Coun- cil, Washington, D.C., pp. 17–26. Washington State Department of Transportation. 2010. The 2010 Con- gestion Report. Olympia. Zhou, B., K. Kockelman, and J. Lemp. 2009. Applications of Integrated Transport and Gravity-Based Land Use Models for Policy Analysis. Transportation Research Record No. 2133: Journal of the Transporta- tion Research Board, Transportation Research Board of the National Academies, Washington, D.C., pp. 123–132.