Assessing Mechanisms for Integrating Transportation-Related Greenhouse Gas Reduction Objectives into Transportation Decision Making (2010)

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.

NCHRP Project 20-24 (64) Final Report 59 5. ANALYSIS TOOLS AND TECHNIQUES TO SUPPORT IMPLEMENTATION OF POLICY MECHANISMS In order for transportation agencies to comply with policy mechanisms such GHG or VMT budgets or performance standards, they may need analytical methods that are not widely used today. In some cases, the tools and techniques necessary to implement new GHG policy mechanisms are available and sufficient; in other cases they are not. This section reviews these tools and techniques and discusses their limitations and capabilities. Policy mechanism implementation could require two types of analysis that are closely related but distinct. 1) First, a DOT or MPO would need analysis to determine if it meets a target. This would require quantifying VMT or GHGs for the current year and potentially for multiple future years, based on how the target is set. The ability to do this accurately, in the face of changing policies, technologies, and transportation systems, varies widely. 2) Second, a DOT or MPO will need tools and techniques to determine the effects of GHG reduction strategies. In many cases, these tools and techniques will be the same as used by the DOT or MPO to determine if it meets a target in the first place. In other cases, additional tools and techniques will be needed to analyze strategies. This section focuses on three basic analysis functions:  Measuring and estimating VMT  Forecasting VMT  Estimating and forecasting GHGs For each of these functions, we first describe the analytical options and then discuss their capabilities and limitations. For the purpose of this section, we use the term VMT or GHG “estimate” to refer to a current year (or recent historic year) calculation of VMT or GHGs. A VMT or GHG “forecast” is a calculation for a future year. In addition to these three analysis functions, Section 5.5 reviews the non-model tools and techniques for analyzing GHG reduction strategies. Key findings from this section are as follows. 1. The tools and techniques for implementing a state-level target are more limited than for an MPO-level target. Most states do not have tools to produce accurate strategy-sensitive forecasts of statewide VMT. Except for a handful of small states that function essentially like MPOs and a few others that have invested in complex statewide forecasting models, state DOTs use relatively simplistic VMT forecasting methods for the purpose of fuel tax projections. Fuel-based state GHG inventories in many state climate action plans are available but differ from MPO estimates and do not provide information to select strategies. For many larger states, creation of a new network-based travel demand model to forecast VMT would not be feasible. Instead, econometric models can be used to forecast statewide VMT, as is done in California and several other states. But approaches that do not involve travel modeling do not lend themselves to analysis of many types of transportation GHG reduction strategies. 2. MPOs are generally better positioned than DOTs to implement a VMT or GHG target. Most use a travel demand model for long range planning analyses, and some have applied regional models to analyze emission reduction strategies or comply with air quality conformity regulations. Many

NCHRP Project 20-24 (64) Final Report 60 MPOs also have experience with emissions modeling through the conformity process. But there is a wide variation in technical capabilities among MPOs, often related to the size and resources of the agency. Many small MPOs do not maintain travel demand models, and a few have no VMT forecasting capabilities at all. 3. Current travel demand models are not suitable to analyze many types of demand management strategies that can reduce VMT and GHGs. This is in part because zone-based models are simply not sensitive to fine-grained changes in land use, bicycle/pedestrian systems, and transit service. Several off-model tools and techniques can be used to analyze these types of strategies, but they are inherently limited in their predictive ability because they rely on generalized relationships (i.e., they are not customized for a specific region). Traffic simulation models are being developed but their current use is limited. 4. Current travel demand models and emissions models are not suitable to analyze most types of operations strategies. These strategies are often of greatest interest to DOTs because they can be directly controlled by DOTs and often can be implemented relatively quickly. Analysis of operations strategies typically must be done using sketch planning tools or traffic simulation models. Even if the transportation impacts of operations strategies can be accurately predicted, previous transportation and emissions models have not been suitable for analyzing strategies that affect congestion and traffic flow. Traffic simulation models and EPA’s new MOVES2010 model help to remedy this shortcoming. 5. Freight transportation is not as well represented in MPO travel models as passenger transportation. While most MPOs include truck forecasts in their models, relatively few independently model truck travel. Moreover, the traffic counts used to calibrate travel demand models may not accurately reflect truck volumes. Most MPOs do not estimate travel or emissions from rail, marine, or other non-road freight transportation. The tools and techniques for analyzing freight strategies are limited. 5.1 Measuring and Estimating VMT A policy mechanism that involves a per capita VMT target would require DOTs or MPOs to develop accurate estimates of current VMT within their boundaries. VMT estimates are also essential inputs to a GHG inventory, and thus are necessary for implementation of a GHG target. Description of Options Data used for estimating VMT is typically based on traffic counts collected by states and localities. FHWA requires that states report annual statewide VMT by 12 functional roadway classes and 4 vehicle types through the Highway Performance and Monitoring System (HPMS). A number of different procedures are used to collect this data. All states rely on samples of traffic counts on state, county, and city roadways to estimate the current VMT on state highways in the HPMS database. Continuous traffic counters can collect data year-round at fixed locations. States also employ portable traffic counters to conduct counts for shorter times on a larger sample of roadway segments. Statistical methods that account for seasonal and weekday variations in traffic are then employed to expand these counts into annual estimates of traffic volumes on all segments of the road system. Volumes are multiplied by the length of the roadway segment and summed to get total annual VMT. Data on traffic counts collected by the states are used by the federal government. Two separate but related federal traffic data programs are the HPMS and the Traffic Volume Trends (TVT) program.

NCHRP Project 20-24 (64) Final Report 61  HPMS provides limited data for the entire universe of public roads and more detailed information for a sample of these segments. The sample data can be expanded to represent the universe of public roads. The sample data includes information on the physical characteristics, condition, performance, use, and operations on sampled roads, including estimates of average annual daily traffic (AADT) on each segment. The database contains information on peak and off-peak volumes. In addition, VMT can be calculated for truck and other vehicles separately. Due to lag times in collecting and processing the data, published HPMS VMT estimates are often two years old.  The TVT program is intended to track travel activity by month and report results quickly. TVT estimates of travel activity (including annual month-on-month data) lag current conditions by only two or three months. TVT is based on continuous traffic counts that are voluntarily reported by the states, using information collected from approximately 4,000 continuous traffic counters. Data for missing functional classes and states are imputed based on available functional classes, surrounding state data, or national estimates. Other traffic counts There are other sources of traffic count data besides the information collected by state DOTs to satisfy federal requirements. Many local governments perform traffic counts as part of their planning and congestion management processes. Regional traffic management centers may maintain archived real-time traffic data, although this data is available only for select roadways, usually the most heavily traveled. Because of the lack of standardization in these traffic counts (conducted for a variety of purposes by multiple state, regional, and municipal agencies), they are generally not a reliable source of VMT estimates. Travel demand models Travel demand models can be used to estimate current-year VMT at a regional (or county) scale. Many MPOs estimate baseline or other past years in the conformity process, to either meet baseline year conformity tests or validate models. In some states, a statewide travel model can be used to estimate VMT at the state level. These models (discussed in the following section) are created for the purpose of forecasting VMT. The models are calibrated to match current-year volumes on selected roadways and screenlines. The models produce estimates of traffic volumes on every link represented in the model network – typically collector and higher functional classes. Local roadways are not included in these models, so local roadway VMT is typically estimated as a fixed fraction of total VMT. Capabilities and Limitations HPMS is designed to produce VMT estimates at the state level. In addition, special sampling is conducted in HPMS to allow accurate estimation of VMT in air quality nonattainment areas and selected other regions. HPMS sample sizes are typically not large enough to allow accurate estimation of VMT at the county level or smaller geographic areas, although some states collect sufficient sample data to allow county-level estimates with sufficient accuracy. The HPMS sample segment data includes peak and average daily volume fractions for single unit trucks and combination trucks, so VMT for these vehicle types can be estimated at the state level. However, the vehicle classification counts used to estimate truck fractions are not updated every year, so this data may be less reliable than total vehicle counts.

NCHRP Project 20-24 (64) Final Report 62 HPMS is available free from FHWA, and summary statistics are reported in Highway Statistics. Since it is a large database, there are costs associated with manipulating and analyzing the raw data. A limitation of HPMS is that there are relatively few samples available for urban arterials, resulting in more variability over time in these estimates and greater uncertainties in the estimates. HPMS does not include traffic volumes on local roads, so although state-level estimates for local road VMT are available in Highway Statistics, the quality of these estimates is considered low. States use a variety of methods to estimate VMT for local roadways. Using HPMS data to track urban-area VMT over time can sometimes be problematic when roadways are reclassified from rural to urban categories. Like HPMS, TVT data is available free from FHWA. An important limitation of TVT is that it does not provide data below the state level. Since TVT is based on using trends in continuous traffic count data to grow HPMS VMT estimates, the TVT estimates are subject to the same uncertainties associated with HPMS. In addition, TVT data is subject to additional uncertainties because missing data for functional classes and even whole states are imputed. TVT estimates at the state level are typically within five percent of HPMS data, but VMT estimates for particular functional classes can vary by 25 percent or more.33 The capabilities and limitations of travel demand models are discussed in the following section. 5.2 Forecasting VMT VMT is the key variable in estimating and forecasting GHG emissions from on-road transportation. CO2 emissions are a function of VMT, vehicle efficiency, and carbon content of fuels; but the latter two are relatively constant and predictable. The ability to accurately forecast VMT is therefore essential for MPOs or DOTs to implement a policy mechanism like a VMT or GHG target. Transportation agencies would need VMT forecasts in order to demonstrate compliance of their long range plans with a target and to evaluate the effects of many GHG reduction strategies. Approaches to forecasting VMT can be distinguished in two ways: MPO-level vs. state-level, and approaches using network-based travel models vs. non-travel model (or “off-model”) approaches. Network-based travel models include not only traditional four step models, but newer “activity-based” travel models using network simulation (discussed below). MPO Travel Models MPOs are required by federal law to develop transportation plans and programs for their regions. To develop these plans, most MPOs rely on travel demand forecasting models to estimate future vehicle travel patterns and to analyze the impacts of alternative transportation investment scenarios. In regions required to conduct air quality conformity analysis under the Clean Air Act, the MPO travel demand model typically provides the inputs used to estimate emissions using emissions models.34 Nearly all MPO travel modeling involves a sequential four-step approach. The basic four-step model (1) estimates the number of daily trips generated by analysis zone, (2) distributes trips from origin zones to destination zones, (3) divides the trips according to mode of travel, and (4) assigns the trips to the roadway (and sometimes transit) network. The trip assignment step produces an estimate of volumes on each roadway link in the network; link volume multiplied by link length give link VMT, which can then aggregated to develop regional VMT estimates. This basic structure has been used since the 1950s. MPO 33 ICF Consulting, “Estimation of Monthly State Level Highway Travel: Final Report,” Prepared for the Federal Highway Administration, November 2005. 34 40 CFR 93.122(b).

NCHRP Project 20-24 (64) Final Report 63 travel models can be more complex or simpler, depending on the resources and needs of the region. For example, some larger MPOs have models that include separate components for forecasting truck travel or automobile ownership, or models designed to be responsive to changes in the pedestrian environment. Conversely, a region without transit or with limited transit may eliminate the mode choice step. A small number of MPOs are replacing their existing models with advanced tour- or activity-based models. Other Regional-Level VMT Forecasting Methods Many small urban and rural areas are not covered by a travel demand forecasting model. Some of these areas may be in nonattainment for federal air quality standards, and therefore require VMT forecasts in order to demonstrate transportation conformity.35 Areas without travel demand forecasting models generally rely on calculations that involve spreadsheets to forecast future VMT. The methodologies range from very simple linear trend lines to more complex non-linear regression analyses.36 State Travel Models Statewide travel models have been developed and are operational in about half the states. Statewide models are used for different purposes, and they range widely in their form and capabilities. In small states (such as Rhode Island, Massachusetts, New Hampshire, and Delaware), the statewide travel model is often an extended MPO travel model, with the same use and capabilities. Other states have models that use a coarse zone system for modeling travel in inter-regional corridors or for modeling freight movement. A handful of mid-size states have created statewide travel models with extensive networks and relatively fine-grained zonal systems. These models are capable of estimating VMT for the purpose of demonstrating transportation conformity in areas not covered by an MPO model, and likely produce relatively accurate statewide VMT forecasts. Florida’s Statewide Urban Transportation Modeling System (FSUTMS) joins together MPO models with county models for non-MPO areas to provide coverage for the entire state.37 The following states appear to have comprehensive statewide travel models that could produce accurate VMT forecasts.  District of Columbia (MPO model)  Florida  Indiana  Kentucky  Massachusetts  Michigan  New Hampshire  Oregon  Rhode Island (MPO model)  Wisconsin  Vermont  Delaware Aside from these examples, most states do not maintain travel models capable of forecasting statewide VMT.38 35 40 CFR 93.122(d). 36 ICF International, Sample Methodologies for Regional Emissions Analysis in Small Urban and Rural Areas: Final Report, Prepared for the Federal Highway Administration, October 18, 2004. 37 Robert McCullough. “Evolution of Statewide Modeling in Florida.” Florida Department of Transportation. September 1999. 38 NCHRP Synthesis 358, Statewide Travel Forecasting Models, TRB, 2006.

NCHRP Project 20-24 (64) Final Report 64 Other State-Level VMT Forecasting Methods Many states develop VMT forecasts without the use of network-based travel demand models. These forecasts are often done for the purpose of estimating future fuel tax revenues. For example, historical trends in traffic volumes can be used to project statewide VMT by functional class and vehicle type. Colorado, for example, develops traffic volume and VMT projections using 20-year growth factors based on a statistical analysis of trends. The trend method is sufficiently accurate for projecting federal-aid funding for a state, but is of little value in exploring the effects of policy options or infrastructure improvements on VMT. Other states have more sophisticated tools for forecasting VMT without travel modeling. California’s Motor Vehicle Stock, Travel and Fuel Forecast (MVSTAFF) model uses a macroeconomic approach to forecast statewide vehicle mix, VMT, and total fuel consumption. VMT forecasts are based on variables of population, income, vehicle ownership and fuel cost per mile. Capabilities and Limitations MPO Level Regional travel demand models maintained by MPOs are designed to forecast traffic on the entire roadway network, and most are capable of producing regional VMT forecasts with acceptable levels of accuracy. However, some smaller MPOs maintain models that forecast peak-period travel only. These models are not intended to estimate daily or annual VMT as would be needed for GHG estimation; factoring of peak-period volumes to estimate AADT can potentially introduce significant levels of error. Because the model networks do not include most local roads, VMT on these facilities must be estimated outside the model framework. Local roadway VMT typically ranges from 5 to 20 percent of total metropolitan VMT.39 Nearly all large and mid-size MPOs (those with population greater than 200,000) use a travel model; some smaller MPOs (those with population less than 200,000) may not. In a 2004 survey, 15 percent of small MPOs reported no modeling capabilities at all.40 According to a recent GAO survey, about half of the MPOs do their own travel modeling, while the rest rely on consultants or their state DOT.41 In general, larger MPOs are more likely to develop and operate models in-house, and smaller MPOs, if they use a model, are more likely to require outside technical assistance. If MPO travel models are to be used to demonstrate compliance with a regional VMT or GHG target, a critical question is to what extent can they capture the impacts of strategies that transportation agencies might use to reduce VMT and GHG emissions. There is extensive literature on the capabilities and shortcomings of four-step travel demand models. As noted in a recent TRB report: “The demands on forecasting models have grown significantly in recent years as a result of new policy concerns. Existing 39 Based on Federal Highway Administration, Highway Statistics, Table HM-71. 40 Transportation Research Board, “TRB Special Report 288: Metropolitan Travel Forecasting – Current Practice and Future Direction.” October 2007. 41 U.S. Government Accountability Office, “Metropolitan Planning Organizations: Options Exist to Enhance Transportation Planning Capacity and Federal Oversight,” Report GAO-09-868, September 2009.

NCHRP Project 20-24 (64) Final Report 65 models are inadequate to address many of these new concerns.”42 Below are some key points on this subject.  Sensitivity to Land Use Changes. All travel demand models rely on socioeconomic data (population, households, employment) as an input at the zone level, and therefore the models are sensitive to land use changes that shift population or employment among zones. However, because they operate at the zonal level, most models are not sensitive to small-scale land use changes, such as mixing of residential and commercial uses, that occur within zones.  Sensitivity to Transit Improvements. Models that include a mode choice element will be able to capture at least some of the mode shift effects of strategies that make transit use more attractive. However, many small and mid-size MPOs do not incorporate mode choice in their models, or do so in ways that do not sensitive to transit improvements.  Sensitivity to Pedestrian and Bicycle Improvements. Some MPOs include bicycling and walking modes within their mode choice models, and some even assign these trips to a non- motorized facility network. However, most MPO models do not account for these trips. Even models that do include bicycling and walking modes will have difficulty capturing changes to the bicycle and pedestrian network, since these changes typically occur within zones.  Sensitivity to Road Pricing. Most travel models can simulate the effects of static road pricing by changing the composite travel cost associated with a roadway link. However, the static nature of the traffic assignment process in a four-step model does not allow analysis of variable pricing.  Sensitivity to Operations Strategies. Four-step travel demand models do not account for non-recurrent (incident) delay. Therefore, without post processing, they are not capable of capturing the effects of operational improvements designed to address non- recurrent congestion, including many ITS strategies.  Sensitivity to Time-of-Day Shifts. Travel models vary greatly in the time periods they cover. Small MPOs may only model peak periods. Large MPOs typically model the AM peak, mid-day, PM peak, and evening time periods separately; the sum of these provides total daily volumes. Because they model fixed time periods, the models may not accurately capture effects of strategies that shift travel times, such as peak- period pricing or changes in freight delivery times.  Sensitivity to Fuel Prices. Models are typically not sensitive to changes in fuel 42 Transportation Research Board. “TRB Special Report 288: Metropolitan Travel Forecasting – Current Practice and Future Direction.” October, 2007. Model Improvements for SB 375 California has identified model improvements necessary for MPOs to accurately quantify and meet the GHG emission reduction targets under SB 375. The state is awarding $12 million in voter-approved bond funds to MPOs to accelerate and implement improved modeling capabilities. The priorities for model improvement are: 1. A 4-step model with post-processing capabilities to include density, diversity, design, destinations, etc. This minimum level of modeling is considered critical to effectively evaluate all strategies relating to SB 375. 2. A tour/activity-based model with post-processing capability. These models illustrate that trips made by a household are not independent of each other but are often connected or chained together for efficiency or convenience. 3. An inter-regional/regional integrated tour/activity- based transportation model with land use and economic modeling components that support a healthy quality of life.

NCHRP Project 20-24 (64) Final Report 66 prices, with the exception that models containing mode choice components would see a shift in some trips from automobile modes to other modes as the relative cost of driving increases.  Sensitivity to Freight Strategies. A majority of MPOs include truck forecasts in their models, but many estimate truck trips simply as a fraction of passenger vehicle trips. Relatively few MPO models independently model truck travel. Thus, most MPO models cannot accurately forecast the effects of investments or policies directed at the freight sector. For MPOs that do not use a travel model to forecast VMT, their ability to capture the effects of VMT or GHG reduction strategies rests entirely on off-model techniques, some of which are discussed in Section 5.5. It is worth noting that a small number of MPOs are using or developing more advanced “activity-based” travel models in order to address some of these shortcomings (see Exhibit 5-1). While the traditional four- step model estimates an aggregate number of trips from each zone, activity-based models estimate the travel patterns of individual household members, using the specific socioeconomic details of the household along with time of day constraints, accessibility indicators, available modes of travel, and other factors. And rather than modeling discrete trips between a single origin and destination, activity-based models analyze travel in “tours” that can have multiple stops. The activity-based modeling framework is therefore better suited to understanding the impacts of transportation programs and policies on underlying traveler behavior, which determines travel patterns in a region.43 Exhibit 5-1 lists agencies currently known to be using or developing activity-based travel models. Exhibit 5-1: Agencies Currently Using or Developing Activity-Based Travel Models Currently using advanced activity-based travel models Current developing advanced activity-based travel models Mid-Ohio Regional Planning Commission (Columbus) New York Metropolitan Transportation Council (New York City) San Francisco County Transportation Authority Sacramento Area Council of Governments Tahoe Regional Planning Agency (Lake Tahoe, CA and NV) Puget Sound Regional Council (Seattle) Atlanta Regional Commission Denver Regional Council of Governments Metropolitan Transportation Commission (San Francisco Bay Area) North Central Texas Council of Governments (Dallas–Fort Worth) Portland Metro (Oregon) St. Louis East-West Gateway Council of Governments San Diego Association of Governments Southern California Association of Governments Sources: TRB Special Report 288, Metropolitan Travel Forecasting: Current Practice and Future Direction, Transportation Research Board, 2007; Cambridge Systematics, A Snapshot of Travel Modeling Activities, Prepared for the Federal Highway Administration, August 8, 2008; Bowman, John, “How is an Activity-Based Model Set Developed?”, Presentation at 12th TRB Conference on Transportation Planning Applications, May 17-21, 2009. To be fully responsive to policy questions, activity-based models need to use network simulation or “dynamic traffic assignment,” rather than the traditional static network assignment process. Dynamic traffic assignment intends to account for congestion effects that evolve over time by using a much more detailed representation of network characteristics, including turn lane capacities, intersection controls, and 43 Cambridge Systematics, A Snapshot of Travel Modeling Activities, Prepared for the Federal Highway Administration, August 8, 2008.

NCHRP Project 20-24 (64) Final Report 67 time-dependent demand.44 However, use of dynamic traffic assignment for a full region is not yet a feasible option for MPOs because of the required computational time and resources.45 State Level Most DOTs produce statewide VMT forecasts, but few do so using travel models or other methods that are sensitive to policies and strategies that affect VMT. For the limited number of states that have statewide travel models with detailed road network representation, the capabilities and limitations of the models are similar to those described above for MPOs. In these states, statewide VMT forecasts can be developed with acceptable levels of accuracy, and the forecasts will be sensitive to some of the strategies that influence travel demand. Many other types of GHG reduction strategies of interest to transportation agencies will not be captured in these statewide models. In the remainder of states, statewide VMT forecasts are generally not developed for planning purposes. Many of these states forecast VMT for fuel tax estimating only, using statewide variables such as population and economic forecasts. The VMT forecasts in these states would often not be suitable for determining compliance with a statewide VMT or GHG target. A related shortcoming with statewide VMT forecasts is their inconsistency with MPO forecasts. While a few states (such as Florida) have statewide modeling systems that are built off MPO models, most do not make use of MPO models. Thus, a state VMT forecast from a DOT is likely to be inconsistent with the forecasts from the MPOs in that state. 5.3 Estimating and Forecasting GHG Emissions The estimation and forecasting of GHG emissions is a new exercise for many transportation agencies. A majority of states have developed an inventory and forecast of GHG emissions, often as part of a state climate change planning process led by the state environmental agency. DOTs usually participate in these exercises, but do not lead the GHG forecasting. A small number of MPOs have estimated GHGs resulting from transportation plans, and many more are now in the process of doing so.46 Vehicle emissions for a state or region are not “measured;” rather, emissions must be modeled using related parameters. Conceptually, there are two approaches to estimate on-road transportation GHG emissions – a fuel-based top-down approach and VMT-based bottom-up approach. The top-down approach, used in national and state GHG inventories, relies on fuel consumption by fuel type to determine emissions. The bottom-up approach, typically applied at the regional or municipal level, relies on estimates of VMT data and fleet fuel efficiency or emission factors to calculate GHG emissions.47 44 Virginia DOT, Implementing Activity-Based Models in Virginia, VTM Research Paper 09-01, July 2009. 45 Transportation Research Board. “TRB Special Report 288: Metropolitan Travel Forecasting – Current Practice and Future Direction.” October, 2007. 46 ICF International, Integrating Climate Change into the Transportation Planning Process, Prepared for FHWA, July 2008. Available at: http://www.fhwa.dot.gov/hep/climatechange/index.htm. 47 Gallivan, Frank, Michael Grant, and John Davies, “Improving the Transportation Component of State Greenhouse Gas Inventories.” Paper presented at the 17th Annual International Emission Inventory Conference, Portland, Oregon, June 2008.

NCHRP Project 20-24 (64) Final Report 68 Scenario Planning A number of MPOs are using scenario planning to evaluate the impacts of alternative regional growth scenarios. MPOs began developing and comparing regional growth scenarios in the 1990s, often as part of the long-range planning process, in part to minimize congestion and air quality impacts. More recently, scenario planning has been used to evaluate the effects of transportation investments and land development patterns in GHG emissions. A variety of software tools are available to assist with scenario planning, such as: INDEX, CorPlan, UPlan, Places, CommunityViz, MetroQuest, PlanMaster, and WhatIf? These tools can be used to facilitate a process in which stakeholders (the general public, business leaders, and elected officials) strive to agree on a preferred scenario. This scenario becomes the long- term policy framework for the community’s evolution, is used to guide decision making, and can be embodied in the long-range transportation plan. MPO Level Regional-level transportation GHG estimates and forecasts can start with VMT by vehicle and fuel type. Current and future year VMT is estimated using the techniques described in Section 5.3 – typically a four- step travel demand model. CO2 emissions can be calculated using per-mile emissions factors, or combined with fleet fuel efficiency information to estimate fuel use, which in turn can be used to estimate CO2 emissions based on the carbon content of the fuel. Emissions of the other two significant transportation GHGs, CH4 and N2O, are calculated using per-mile emission factors. GHG emission factors typically come from EPA’s emission factor models – MOBILE6 or the next generation model, MOVES2010, which was released in December 2009, and provides improved estimates. Agencies in California use the Air Resources Board’s EMFAC model. Vehicle fuel efficiency, and hence CO2 emission rates, vary with vehicle speed and operating conditions (i.e., congestion levels). These effects may or may not be reflected in a regional-level GHG estimate. More simplistic estimates ignore speed and congestion effects. VMT across all roadway types is simply summed by vehicle type, and translated into emissions using fleet-average fuel efficiency data or emission factors. Use of the MOBILE6 model necessitates this approach, because CO2 emission factors in MOBILE6 do not vary with speed. More sophisticated regional GHG estimates attempt to capture speed and congestion effects. California’s EMFAC model provides CO2 emission factors that vary with speed. Using travel demand model output, each roadway link is assigned an average speed, and VMT is then grouped into speed “bins,” to be multiplied by the appropriate CO2 emission factor for that speed. The MOVES2010 model allows for much more sophisticated and complex analysis of GHGs that attempts to capture effects of both speed and vehicle operating conditions (acceleration patterns). MOVES2010 defines vehicle types on the basis of HPMS vehicle classifications (as opposed to EPA’s weight-based emission classifications used in MOBILE6) to avoid the need for transportation practitioners to map their HPMS data to EPA categories.48 State Level Approximately 35 states have developed statewide GHG inventories, which report emissions by source category for a recent year. Most states follow a process developed by EPA called the State Inventory Tool (SIT) to develop these estimates. The key input variable is the amount of each fuel consumed in the state by fuel type. Fuel consumption data comes from fuel tax records, state energy estimates, or U.S. 48For more information on MOVES, see: U.S. Environmental Protection Agency, MOVES (Motor Vehicle Emission Simulator) web site at: http://www.epa.gov/otaq/models/moves/index.htm.

NCHRP Project 20-24 (64) Final Report 69 Energy Information Administration data. The inventory process typically splits fuel use into modal categories for reporting purposes (e.g., automobile vs. aviation gasoline, marine vs. railroad diesel). CO2 emissions are calculated based on the carbon content of the fuel; CH4 and N2O emissions are calculated by applying emission factors to VMT (for on-road vehicles) or as a fraction of CO2 emissions (for non- road sources). Most of these states have also developed GHG forecasts, typically to 2020 or 2025. GHG forecasts are calculated by multiplying base year emissions by growth factors for each mode. For on-road vehicles, the growth factors usually come from state DOT forecasts for light- and heavy-duty VMT. For non-road sources, growth factors are typically based on past trends. 5.4 Agency Forecasting Capabilities and Limitations The tools and techniques for creating regional-level GHG estimates and forecasts are relatively new and not fully developed. In general, because regional-level forecasts are based on travel demand model output, they are far more sensitive to transportation policy and investment changes than top-down, fuel- based approaches. For example, regional-level GHG forecasts would capture the effects of shifts in regional land use patterns or increases in transit use, but only to the extent the region’s travel model captures the VMT effects of these changes. MPO Level Regional-level GHG estimates reflect the emissions that occur within the political boundaries as defined by the MPO modeling area. In this way, the emissions estimate matches well with the portion of the roadway system under the jurisdiction of the MPO. This contrasts with a fuel-based approach, which can suffer from “leakage” – vehicles purchasing more or less fuel in a geographic area than is consumed in that area. Regional-level GHG estimates contain a fine level of detail that is useful for analyzing many local and regional GHG reduction strategies. Because they are built from VMT and other local data (e.g., local fleets) for every roadway link, GHG emissions can be accurately estimated for individual roadway types (e.g., freeways), or for individual cities or counties in the region. And many regional GHG estimates contain more detailed information on vehicle types than fuel-based inventories, which can be useful for analyzing some GHG reduction strategies. Region-level GHG forecasts suffer from the same limitations as regional-level VMT forecasts. They do not directly model emissions on local roads, since these facilities are typically not included in travel model networks. Most importantly, they cannot fully capture the effects of some types of strategies due to the modeling limitations outlined in Section 5.3. For example, in the absence of post-processing or other off-model techniques, region-level GHG forecasts will not be able to capture the effects of many fine- grained land use and urban design changes, bicycle and pedestrian improvements, or operational strategies. Another potential shortcoming of a regional VMT-based emissions estimate is the treatment of emissions caused by alternative fuels. For example, a proper regional transportation GHG estimate should include the electricity generation emissions attributable to electric rail or trolley bus service. Activity data for these sources might need to be calculated outside of the travel demand model. Similarly, off-model adjustments may be necessary to properly capture emissions from buses running on natural gas or other alternative fuels. If battery electric or plug-in hybrid electric vehicles establish significant market share, their electricity generation emissions will also need to be included.

NCHRP Project 20-24 (64) Final Report 70 A critical issue for transportation agencies is the extent to which regional-level GHG forecasts can capture the effects of system efficiency improvements that seek to improve traffic flow. As noted above, the use of vehicle fuel consumption rates or emission factors that do not vary with speed (e.g., those from MOBILE6) prevents almost any analysis of system efficiency improvements. The use of average link speeds and emission factors that vary with average speed (e.g., those from EMFAC, or similar to the way MOBILE6 criteria pollutant emission factors are applied) will capture some system efficiency impacts. However, this approach also has limitations, in that it will not reflect differences in operating conditions and congestion levels. For example, emissions would look identical for a roadway segment with traffic moving at a steady-state 30 mph as compared to a congested segment on which speed averages 30 mph. (The latter would have higher emissions in reality, because of the acceleration effects.) EPA’s MOVES model is intended to better capture the congestion effects on emissions, provided it can be paired with detailed traffic flow data. The average vehicle speed by roadway link is available from metropolitan travel demand models, but with several caveats. The main caveat is that these models do not simulate traffic in order to determine specific delays caused by intersections, bottlenecks, signal timing, lane configurations, or other detailed factors that influence travel speed. Instead, the models typically use a lookup table based on certain basic roadway features, such as the number of lanes and the functional classification of the roadway, to determine a default “free flow” speed for each road in the model, which is then factored using a formula to reflect the “congested” travel speed accounting for traffic. The default speeds and formula factors are typically adjusted to aid in model calibration such that they may not accurately represent ground conditions. Some MPOs use post-processing tools to adjust the speeds taken from the model output for use in air quality conformity analyses.49 These issues can also be addressed by using speed outputs from traffic microsimulation models such as SimTraffic, Paramics, or VISSIM. Microsimulation models typically reflect on-the-ground conditions much better than standard travel demand models, but they are typically not available at a regional scale.50 State Level The top-down fuel-based approach to estimating transportation emissions has the advantage of its basis on a relatively accurate metric that captures all activity – fuel sales. Unlike VMT estimates, which must be modeled based on data samples, fuel sales reflect the activity of all vehicles on all roadways. At a large geographic scale with minimal cross-border leakage, this likely makes a fuel-based GHG estimate more accurate than a VMT-based estimate. A fuel-based approach also has the advantage of being simple to calculate. The state-level GHG forecast is subject to the same limitations as state-level VMT forecasts described in Section 5.3. In most states, the state-level GHG forecast is simply the product of the base year emissions estimate and a growth rate for that fuel type. The growth rate for on-road gasoline is usually based on the state’s automobile VMT projection, which may not be reflective of adopted transportation policies and investments. In many cases, the state’s VMT growth forecast has no relationship to the VMT forecasts in the state’s metropolitan areas. 49 Gong H, Chen M, Mayes J, and Bostrom R. “Speed Estimation for Air Quality Analysis.” Journal of Transportation and Statistics, Vol. 9, No. 1, 2006. 50 Federal Highway Administration. “Traffic Analysis Toolbox, Volume II: Decision Support Methodology for Selecting Traffic Analysis Tools.” July, 2004.

NCHRP Project 20-24 (64) Final Report 71 For transportation agencies, the main limitation of the fuel-based approach is the lack of sufficient detail to assess emission reduction strategies. State-level transportation emissions are typically reported only by the following categories:  On-road Gasoline  On-road Diesel  Aviation Gasoline  Jet Fuel  Rail  Boats and Ships  Other In order to analyze most emission reduction strategies, these totals need to be disaggregated. For example, to analyze strategies affecting automobiles, on-road gasoline emissions must be split into light-duty vehicles and heavy-duty trucks (some of which burn gasoline). Analysis of statewide land use strategies may require splitting vehicle emissions into urban and rural components. This type of disaggregation can be done using state or national VMT factors, but reduces the accuracy of the estimates, particularly in forecast years. Most GHG reduction strategies for transportation would be applied in the state’s urban areas. The analysis of these strategies using the statewide GHG forecast is challenging because the emissions are not reported by urban area. Moreover, as noted above, the statewide GHG forecast may bear no relationship to the GHG forecasts for the metropolitan areas in the state, since they are developed independently using fundamentally different approaches. Both State and MPO Level Several analytical limitations apply to both state and MPO level approaches to GHG estimation and forecasting. One relates to the accuracy of fleet fuel efficiency data, a key component in performing GHG analysis without emissions models. Fuel efficiency is reliable on a national level, but less so at the state and MPO levels. The vehicle fleet mix varies from state to state based on factors such as the physical environment and personal vehicle preferences. Fuel efficiency varies based on the type of vehicles in the fleet, the age of vehicles, and driving conditions. For example, urban areas are likely to have more small vehicles; rural areas may have more large utilitarian vehicles. The fleet mix and fuel efficiency of vehicles in any given state may vary significantly from national averages. However, few states and regions (with the notable exception of California) have data on the fuel efficiency of their fleets. EPA’s MOVES2010 model accounts for fleet efficiency information to calculate GHG emissions factors, based on inputs of state or local vehicle registration information. These inputs are similar to what is used for criteria pollutant analysis in conformity, but states and MPOs that have not already conducted emissions analyses will need to collect appropriate vehicle registration data. 5.5 Off-Model Tools and Techniques for GHG Reduction Strategy Analysis Some types of GHG reduction strategies cannot be analyzed using conventional travel demand models and the GHG forecasting methods described in Section 4.4. For example, a number of transportation demand management (TDM) strategies are not easily incorporated into MPO travel demand forecasting models.51 Similarly, the travel impacts of many operations strategies cannot be analyzed using current travel forecasting tools. A variety of other tools and techniques are available for these analyses. Some techniques simply involve spreadsheet calculations that rely on the effectiveness of a strategy as reported in the literature. Other strategies are best analyzed using software tools designed to estimate the travel and 51 Regional Targets Advisory Committee (California). “MPO Self-Assessment of Current Modeling Capacity and Data Collection Programs.” May 5, 2009.

NCHRP Project 20-24 (64) Final Report 72 emissions impacts of specific types of transportation strategies. Most of the analytical strength of these tools is in the estimation of travel impacts; the user does not need to calculate a change in VMT or speeds, since the model performs that analysis. TDM Strategies TDM strategies include a wide variety of measures to reduce VMT, including transit improvements, ridesharing programs, and bicycle and pedestrian improvements. In general, if a TDM strategy affects transportation between zones throughout a region, a travel demand model is the best approach to analyze strategy VMT impacts. Some tools are designed to process trip tables that are the output of regional travel models – allowing for region-wide analysis without the need to re-run the full travel model. If a strategy affects only regional sub-areas or affects transportation within a zone (such as employer-based strategies, parking, and bicycle/pedestrian transportation), then an off model tool such as sketch planning or spreadsheet analysis is the best approach. Some examples of tools to analyze TDM strategies are summarized below.  The COMMUTER model, developed by the U.S. EPA, is designed to analyze the impacts of transportation control measures such as transit employer-based transportation demand management programs and transit improvements, on VMT, criteria pollutant emissions, and CO2. By default, the COMMUTER model uses national average MOBILE6 emission factors for criteria pollutants and CO2, but the user has the option of importing locally specific MOBILE6 emission factors for more locally accurate emissions estimates.  TRIMMS (Trip Reduction Impacts for Mobility Management Strategies) has recently been developed by the University of South Florida. It is conceptually similar to the COMMUTER model, but tends to report lower VMT reductions than the COMMUTER model.  TCM Analyst is a spreadsheet based tool developed by the Texas Transportation Institute in 1994-95. It is now considered somewhat dated.  The Transportation Emissions Guidebook was created by the Center for Clean Air Policy; it allows for relatively simplistic analysis of a large number of emission reduction strategies.  STEAM (Surface Transportation Efficiency Analysis Model) is essentially a benefit-cost analysis tool that can also be used to analyze travel activity and emissions changes. STEAM requires extensive inputs from regional agencies in the form of baseline and improvement case trip tables for each type of strategy. Operations Strategies Operations strategies include measures such as traffic surveillance, work zone management, electronic toll collection, traffic incident management, road weather management, emergency management, and traveler information services. These strategies reduce vehicle delays associated with incidents and other non-recurring events; they typically reduce emissions by reducing idling and delay, and allowing for smoother traffic flow. Most travel demand models do not capture non-recurrent delay. In addition, most regional travel models are not capable of simulating traffic flows to account for factors such as intersections, bottlenecks, signal timing, lane configurations, etc. For these reasons, a regional travel model cannot, by itself, be used to analyze most operations strategies.

NCHRP Project 20-24 (64) Final Report 73 The tools and techniques for analyzing operations strategies is an area of on-going research, led in part by FHWA.52 Some of the most common current methods for analyzing operations strategies are outlined below.53  Sketch planning tools produce general order-of-magnitude estimates of travel demand and traffic operations in response to transportation improvements. These approaches are typically the simplest and least costly of the traffic analysis techniques, but are usually limited in scope, analytical robustness, and presentation capabilities. Three examples are: the ITS Deployment Analysis System (IDAS), Screening for ITS (SCRITS), and STEAM.  Deterministic tools typically implement the procedures of the Highway Capacity Manual (HCM) to quickly predict capacity, density, speed, delay, and queuing on a variety of roadway types. They are good for analyzing the performance of isolated or small-scale transportation facilities, but limited in their ability to analyze network or system effects. Two examples of deterministic models are Traffix and Highway Capacity Software (HCS).  Traffic simulation tools perform detailed representations of traffic flow in real-world locations. These tools require a large amount of detailed input data, including detailed roadway geometric, signal timing, and trip generation/distribution data, and extensive validation and quality control. Because of their data and computer processing requirements, simulation tools are generally not appropriate for use at a regional scale.54 Simulation tools can be combined with travel demand models to examine freeway performance in individual corridors. Simulation tools include macroscopic simulation models such as FREQ, PASSER, and TRANSYT-7F, mesoscopic simulation models such as SYNASMART-P and TRANSIMS, and microscopic simulation models such as CORSIM/TSIS, Paramics, and VISSIM. Freight and Non-Road Strategies The options for analyzing freight and non-road transportation emission reduction strategies are generally less developed than the tools and techniques for automobile-focused strategies. These strategies can include:  Strategies to reduce idling/berthing emissions by trucks, locomotives, or ships  Strategies to reduce truck travel to/from a port or rail intermodal facility  Strategies to shift freight to more fuel efficient modes  Strategies to enable smoother, more fuel efficient movement of trains or ships  Strategies to reduce emission from transportation refrigeration units (i.e., refrigerated trailers and shipping containers) The US EPA SmartWay Transport Partnership offers information to analyze the benefits of these types of strategies. For example, the SmartWay web site (www.epa.gov/smartway) includes several calculators and models that provide fuel consumption rates of idling trucks and of idle reduction solutions; guidance for states that want to incorporate idle reduction projects in their air quality plans; current and prior idle reduction projects funded by SmartWay and others, and the environmental and related benefits of these projects; and other key tools and information on the effectiveness and benefits of reducing idling from 52 See http://ops.fhwa.dot.gov/travel/plan2op.htm 53 FHWA, Applying Analysis Tools in Planning for Operations, (unpublished). 54 Federal Highway Administration. “Traffic Analysis Toolbox, Volume II: Decision Support Methodology for Selecting Traffic Analysis Tools.” July, 2004.

NCHRP Project 20-24 (64) Final Report 74 trucks and locomotives. EPA also has created the SmartWay Transport Partnership FLEET Performance model, which can also be found on EPA’s website. 5.6 Summary of Findings on Analysis Tools and Techniques As this section illustrates, new policy mechanisms for integrating GHGs into transportation planning would place analytical demands on DOT or MPOs that these agencies are not necessarily prepared to meet. The exhibit below summarizes the main capabilities and limitations for each of the three basic MPO and DOT functions reviewed in this section. Exhibit 5-2: Summary of Findings on Analysis Tools and Techniques Function Level Capabilities Limitations Measuring and Estimating VMT MPO level  Many MPOs used travel models to estimate VMT  Some MPOs estimate VMT based on counts rather than model  A few smaller MPOs lack capability to estimate VMT  No local roads in models  Truck fraction may not be reliable State level  HPMS (and TVT) programs provide current statewide VMT by major vehicle and facility type  No speed information  Truck fraction may not be reliable  HPMS data has 2-3 year time lag Forecasting VMT MPO level  Most MPOs have models that forecast VMT  Some smaller MPOs don’t use travel models for VMT forecasts  A few smaller MPOs do not have VMT forecasting capability  MPO models not sensitive to some TDM strategies (bike/ped, transit/land use changes, etc.)  Most MPO models not sensitive to operations strategies State level  A few states have travel models that can forecast statewide VMT  Most states use spreadsheet analysis for VMT projections  Most states: VMT forecasts not sensitive to VMT reduction strategies  Most states: VMT forecasts not sensitive to changes in fuel prices, fleet fuel efficiency Estimating and Forecasting GHGs MPO level  GHG inventory reflects emissions within MPO boundaries  MPOs doing conformity have experience that could be used in GHG emissions modeling  VMT-based GHG forecast more useful for strategy analysis than fuel- based forecast  MPO models not sensitive to some TDM strategies and operation strategies  Fleet fuel efficiency data is limited (but emissions models help address this limitation) State level  Fuel-based GHG inventory is highly accurate (if minimal cross-border travel)  Easy to account for statewide technology and fuels changes  If cross-border travel is extensive, GHG inventory may not reflect emissions within state boundary  Forecast of statewide GHGs are based on statewide VMT forecasts, which may not be accurate  Not sufficiently disaggregated for analysis of many strategies  Fleet fuel efficiency data is limited (but emissions models help address this limitation)