Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Office of Planning, Environment, & Realty (HEP)
PlanningEnvironmentReal Estate

HEP Events Guidance Publications Awards Contacts

Handbook for Estimating Transportation Greenhouse Gases for Integration into the Planning Process

Chapter 5 - VMT-based Methods

PDF files can be viewed with the Acrobat® Reader®

This section discusses VMT-based inventory and forecasting approaches. All of these approaches involve two main components:

  1. Developing VMT estimates - which tend to rely upon travel and land use forecasting tools. Section 5.1 describes relatively simple options relying on vehicle, household, and land use data where a network-based travel forecasting model is not available. These methods may be most applicable for areas seeking to develop a GHG inventory in a relatively quick manner. Section 5.2 describes expanded options based on HPMS data and the use of a network-based travel forecasting model, which tend to be more robust and allow for more sophisticated analyses of speeds and other factors
  2. Estimating emissions- which can range from applying a simple emissions factor (in grams per mile) to the VMT estimate, or may involve use of sophisticated emissions models in order to calculate emissions from travel (options are described in Section 5.3).

5.1. Estimating VMT Relying on Vehicle, Household, and Land Use Data

Using Simplified Methods

Each of the methods described in this section are relatively simple and may be defined as "sketch planning" approaches. In general, it would be better to use calibrated and validated travel forecasting and emissions models. These simplified methods can be used when more sophisticated tools or the resources to apply those tools are not available. As such, it is important for the analyst to recognize the limitations of these approaches.

VMT is a key factor that influences transportation GHG emissions because the level of travel activity is a determinant of fuel consumption. While there are many sources of VMT data available, this section focuses on relatively simple methods of obtaining VMT data based on odometer data from vehicles, household travel surveys, and land use information. These VMT methods are generally intended for calculating passenger GHG emissions -- not freight. They also are largely intended for developing inventories, although extrapolations of historical trends can be made to develop forecasts, recognizing a high degree of uncertainty in these results. A brief description of each of these methods is provided.

Table 8. Selection Criteria for Methods of Estimating VMT that Rely on Vehicle, Household, and Land Use Data

Selection Criteria


Odometer Data

Household Travel Data

Land Use Data

Analysis Type

Inventory; forecast possible

Inventory; forecast possible

Inventory; forecast possible

Geographic Scope

State, regional, or county

State, regional, or county

Regional or county (particularly suited for smaller geographic areas)

Analysis Precision

Approximate - appropriate for simple calculation, largely for personal vehicles; may lack data for certain vehicle types and speeds.

Approximate, due to limitations in sample size - appropriate for simple calculation, largely for personal vehicles; typically does not account for speed and other factors.

Approximate, based on trip generation factors that may not be applicable in all areas, and lack of application of speeds and other factors.

Data Needed

Odometer data, vehicle stock data

Results from household travel survey - e.g., miles traveled, trip purposes

Land uses (including square footage and/or employment levels), trip generation rates (from Institute of Transportation Engineers, or local study)

Necessary Analytical Capabilities

Moderate - although the calculation is relatively simple, there may be complexities in analyzing the odometer data. Analysis is more complex for forecasts, to account for changes in vehicles/fuels.

Limited for inventories - relatively simple calculation. More complex for forecasts, to estimate changes in travel patterns.

Limited to moderate - generally need fairly substantial amount of land use data and calculations across each land use type.

Level of Resources Required (i.e., staff/budget)

Limited to moderate for inventories. Higher for forecasts as knowledge about future vehicle/fuel trends is required.

Limited, assuming existing survey data; conducting a new survey would require additional resources.

Limited to moderate - depending on quality of existing land use databases, and extent of geographic area.

Capable of Addressing Vehicle Technology/ Fuels Changes**

Yes; odometer data should be matched with vehicle information

No; needs supporting vehicle/fuel data

No; needs supporting vehicle/fuel data

Capable of Addressing Changes in Travel Demand

Limited - depends on extent to which changes are captured in odometer data but generally not designed to assess.

Limited - depends on the extent to which the effects of these changes are captured in household travel data but generally not designed to assess.

Limited - depends on the extent to which the effects of these changes are captured in land use data

Capable of Addressing Changes in Vehicle Speeds and Operations


Limited - depends on extent to which survey data account for time of day and speeds.

Limited - depends on whether trip generation data are applied by time of day and associated with speeds.

** Requires combination with other models and methods such as MOVES to address vehicle technology or fuels changes.


Vehicle Inventory/Odometer Data
One way to determine VMT for inventories is to directly observe the number of miles driven through periodic odometer readings. In some areas, odometer data are collected as part of vehicle safety inspections, air pollution vehicle inspection and maintenance (I&M) programs1, or as part of the vehicle registration process. One important value of these data are that they can typically match information on miles traveled with specific types of vehicles (e.g., make and model), which when combined with fuel economy information, can be used to calculate fuel consumption and GHG emissions.

A sample website from the State of Delaware at the link below shows how odometer data are collected and used by consumers: Delaware is one of the many states that contribute to the National Motor Vehicle Title Information Service, which also collects and stores odometer data for consumer protection: (

Some insurance companies also collect odometer data from drivers, which can be either self-reported or verified by a certified third party. Despite the collection of this data, its availability is often a limitation to using this method to estimate GHG emissions. Unlike many other government data sources, odometer data cannot simply be accessed from a government website. In addition, information on heavy-duty trucks is limited since these vehicles often operate outside of the state where they may be registered and few states individually track heavy-duty truck odometer data.

Where available, vehicle odometer data must be requested from state licensing departments, there may be a fee to access this information, and confidentiality agreements may also be necessary. The reliability of odometer data may also pose an issue. Newer cars and some very old cars may not be included in state I&M program emissions checks and therefore no data are collected. Based on information from the National Household Travel Survey (NHTS), newer cars are driven more than older cars. Additionally, the odometer data may not be available for all geographic areas. Small counties may be excluded because of confidentiality reasons, and many rural counties do not have air pollution inspection and maintenance programs. Lastly, vehicle odometer data are typically recorded at annual or biennial intervals. While this type of data could be useful for estimating long-term vehicle-related GHG emissions, the data can be challenging to use to estimate daily VMT or GHG emissions.

Household Travel Survey Data
Household travel surveys represent another source of VMT data. The most commonly available types of household travel surveys are the NHTS, statewide household travel surveys, and MPO household travel surveys. For the most part, these travel surveys recruit a socioeconomically and geographically diverse range of volunteers to have their travel activities monitored. As part of the travel survey, respondents are asked to report information such as the age of their car(s), odometer readings, and to estimate their annual mileage driven. In addition, daily mileage driven is estimated for the survey days by either directly estimating mileage using a GPS device or by using self-reported mileage from the respondents. It may be possible to estimate VMT if physical addresses or parcels associated with trip ends are recorded.

Household surveys are often viewed as one of the most reliable sources of data for daily travel and VMT estimates; however, it is important for the analyst to understand limitations of the data as it relates to key issues such as sample size, selection bias, and limited time period. Travel surveys may have underreporting errors due to discrepancies between self-reported daily VMT and actual VMT and may focus on travel time instead of distance unless GPS tracking is used.

While household surveys are fairly accurate for daily travel and VMT estimates, their ability to generate annual VMT estimates has limitations. The potential inaccuracy relates to the short (typically one or two-day) survey periods, the reliance on self-reported data, and small sample sizes. The average person does not have a good sense about their annual VMT, yet this is a common question on household travel surveys. In addition, reporting errors can occur in households with multiple drivers, since they are more likely to report the annual mileage driven by the car as opposed to the mileage driven by each driver (which can lead to overestimates).

MPOs often take steps within their travel model development processes to correct for these potential errors. If these data sources are used to estimate VMT for GHG emissions reporting purposes, it may be prudent to validate the household travel survey results against other data sources such as fuel consumption, or HPMS data. For calculating GHG emissions, it will be important for surveys to classify information on vehicles, so that VMT by vehicle type (e.g., automobile, light-duty truck) is produced.

Land Use Data
Methods that rely on land use data to estimate VMT typically use land use-based trip generation factors to estimate vehicle trips and then multiply the trips by average trip lengths to calculate VMT. These approaches are often used at a small scale, such as for a municipality or to report GHG emissions at a small geographic scale across a metropolitan region where accurate and complete land use data are available. Many local jurisdictions have complete and relatively accurate inventories of land uses based on comprehensive plans, building permit data, or for local tax purposes. These land use inventories can be used to estimate passenger vehicle and truck VMT if adequate information is known about the number and length of vehicle trips generated. Some newer land use inventory and planning tools have travel embedded in the programs to provide for a direct estimate of VMT.

Of the three methods highlighted in this section, land use data are probably the most problematic for generating large-area (regional or statewide) VMT and GHG emissions data, and tend to be geared toward lower levels of analysis (e.g., a city, county, or sub-area of a county). The geographic area issue relates to the complex nature of travel between different land uses. For example, it is fairly simple to develop rough estimates of vehicle trip generation for a given land use based on trip rates from the Institute of Transportation Engineers (ITE) Trip Generation report. Combining the information on generated trips with trip lengths from a household survey, odometer readings, or data from the regional travel demand model can be used to estimate VMT for a given land use type.

The challenge is how to combine the information for a large area. Trips go between land uses and if one is not careful, it is easy to double count VMT and therefore GHG emissions. To help illustrate this point further, consider a spreadsheet tool developed by King County, Washington to help estimate GHG emissions. The tool has VMT and transportation GHG estimates for a wide variety of land uses from residential, to retail, to office and manufacturing. If one were to apply the King County tool on a citywide basis, there would be no way to account for the travel between workplaces and homes. Since both types of land use have a VMT and GHG estimate, both land uses are counting the same trip (and therefore VMT/GHG emissions) twice. More complex tools like travel demand forecasting models were developed to help untangle the mix of trips between different land uses to provide accurate estimates and forecasts of performance measures such as VMT. While there are significant drawbacks to using land use data for large-area planning efforts, these tools can be useful at the parcel level. In fact, the King County tool mentioned above was developed for just this purpose and is effective for isolating the VMT generated from individual land uses.

Strengths and Limitations

The main strengths of these VMT-based methods relying on vehicle, household, and land use data are quickness and low cost because they rely on existing and available data. As mentioned earlier, these methods are primarily applicable where a network-based travel demand forecasting model is not available. The following table highlights some of the strengths and limitations of the different methods.

Table 9. Strengths and Limitations of Methods of Estimating VMT that Rely on Vehicle, Household, and Land Use Data




Using vehicle odometer data

  • For purposes of GHG inventories, data on travel can be matched directly with information on vehicle type to develop estimates of fuel consumption by each vehicle type, if needed.
  • Odometer data are not readily available in many locations.
  • Only provides data on the mileage traveled by vehicles, no information on where the travel occurred or under what conditions.
  • Data are often limited to light-duty vehicles so no freight vehicle data are available.
  • Measures current travel only. Forecasts rely on extrapolation that are unlikely to be sensitive to changes in the transportation network, fuel type, travel cost, or other important variables. This insensitivity would be magnified the further out the extrapolation goes.

Using household travel survey data

  • Survey data can provide more detail on travel behavior, trip purposes, travel times, and other characteristics that are useful for more detailed analysis.
  • Household surveys do not address freight traffic.
  • Measures current travel only. Forecasts rely on extrapolation that are unlikely to be sensitive to changes in the transportation network, fuel type, travel cost, or other important variables. This insensitivity would be magnified the further out the extrapolation goes.
  • Surveys typically do not provide information on operating conditions (speeds, congestion).

Using land use data

  • Land use-based approaches can be useful for analyzing GHG emissions at small geographic scales or for distributing GHG emissions within a region to origins and destinations.
  • If future land use forecasts are available, then this method can also be used to forecast VMT and GHG emissions.
  • Land use data at regional or state levels may be incomplete.
  • Double counting of trips and VMT may occur when including residential and non-residential land uses in a VMT estimate or forecast.
  • Trip generation rates from ITE are based on a limited sample and may not be reflective of actual travel behavior in all areas, so the estimates need to be calibrated to study area conditions/sources.
  • Forecasts are unlikely to be sensitive to changes in the transportation network and travel cost.

The different methods for estimating VMT may be useful for different purposes. For instance, while HPMS-based VMT estimates (which are described in a subsequent sections) are very detailed, and provide information on where travel is occurring (e.g., on specific roadway links), the data lack information on the origin and destination of trips, which may be useful for certain types of analysis. For some states and regions, pass-through travel can account for a large portion of VMT. Since transportation agencies have limited ability to influence travel generally, and intercity travel in particular, some may wish to report intercity travel or "through" trips separately in their GHG inventory, or exclude this travel altogether from the analysis. This is the current practice in California, which focuses on the emissions that a DOT or MPO can most directly influence through its transportation and land use planning efforts. Some MPOs have taken the approach of using HPMS data to establish a regional total VMT and then subtracting through-trip VMT, using regional cordon point license plate surveys to estimate through trip VMT.

Moreover, the odometer-based, household survey, and land use based methods only capture travel for a defined population of vehicles or households: generally passenger travel, not freight. Odometer-based estimates in particular, are not able to provide information on where the vehicles travel, and so these estimates are less useful in developing a detailed inventory accounting for travel speeds.

Key Steps and Data Options

This section describes how to obtain the VMT data and calculate GHG emissions from each of the three data sources above.

Option 1: Use Vehicle Data
As described above, directly reported vehicle odometer data is collected by most states through their motor vehicle departments. This data is often available for individual vehicles during sales or registration transactions. Odometer data usually contains the annual miles driven for passenger vehicles but the data are not readily available, particularly to non-government parties. While there are many limitations (as described above), vehicle odometer data are one of the few directly measured indicators of vehicle travel. These data may be particularly useful as a way to check VMT estimates from other sources or methodologies.

When these data are available, the following steps would typically be followed to estimate GHG emissions.

Step 1: Collect vehicle odometer data. The data may be available from vehicle registration or emissions inspection checks, and can be tied to the Vehicle Identification Number (VIN) to estimate annual mileage for different types of vehicles. This step can involve reconciling some complexities in the data, such as different registration and inspection dates for different vehicles.

Output: Vehicle odometer data

Step 2: Collect vehicle stock data. Ideally, the odometer data would be summarized by vehicle type to aid in a more accurate assessment of fuel consumption and GHG emissions. If the odometer data is not available by vehicle type, then data from vehicle registration programs or air pollution emissions models could be used to estimate vehicle type classes. Care should be taken if age-based emissions profiles are used since odometer data may not include a comprehensive inventory of older and newer cars.

Output: Vehicle odometer data by vehicle type

Step 3: Multiply the annual mileage per vehicle by the number of vehicles of each type. This calculation will result in estimates of the total VMT annually by vehicle type.

Output: Total VMT annually by vehicle type

Step 4: Apply Emissions Factors. Per mile emissions factors can be applied to estimate total GHG emissions. Emissions factors can be extracted from air quality analysis software such as MOVES and EMFAC. For more information on approaches, see Section 5.3.

Output: Total GHG emissions

In addition to odometer data serving as a check for other VMT estimates, it can also be used to compare similar households in different geographic locations to understand how variables such as proximity to urban centers or high quality transit may affect VMT generation. An example based on California data is shown below. Note, however, that these data are per household (and households tend to be larger in low density areas than in urban centers or transit villages) and they are for selected areas, not necessarily representative for all regions.

Figure 7. Daily Passenger VMT per Household by Selected Area Land Use Patterns

A bar graph comparing daily vehicle miles of travel in various land use patterns. The 'Y' axis shows VMT generated per household, and the 'X' axis shows three locations/areas: 'Low Density' in San Ramon, CA; 'Transit Village' in Rockridge, Oakland, CA; and 'Urban Center' in North Beach, San Francisco, CA. The low density area has the highest VMT per household at 76, the transit village is in the middle with 35 VMT per household, and the urban center has the lowest at 15 VMT per household.

Source: Developed using data from Holtzclaw et al., "Location Efficiency: Neighborhood and Socio-Economic Characteristics Determine Auto Ownership and Use - Studies in Chicago, Los Angeles, and San Francisco." Transportation Planning and Technology, 25:1 (2002).

Option 2: Use Household Travel Survey Data
As described above, there are a variety of household survey data sources. Data for the NHTS was last collected in 2009. The data for this widely used survey are located at While the NHTS has good national coverage, its statistical validity drops below multi-county or MPO geographic levels. Therefore, other sources should be considered for smaller geographies.

Several states (e.g., Michigan, Ohio, New York, Utah, Massachusetts, Oregon, Idaho, and California) have developed their own statewide household travel surveys. Sometimes these surveys are collected at the same time as the NHTS since the states can take advantage of the national survey effort. Typically, statewide surveys have a sample size that is large enough to provide statistical reliability for all but the most sparsely populated counties. These surveys serve as the backbone data source for many travel modeling efforts. One disadvantage of statewide household travel surveys is that they are often not updated as frequently as other survey sources, and therefore, may not be as reliable in terms of estimating current VMT patterns or GHG emissions.

The most widely conducted travel surveys are at the MPO level. MPOs rely on household travel survey data as inputs to their travel models. MPO surveys tend to contain the most detail and can be used to provide city-wide or even sub-area specific VMT and GHG emissions data. As described earlier, trip underreporting is a significant issue for all travel surveys; however, given their smaller size and lower overall survey budgets, some MPO models have fewer error correction techniques. Practitioners should consult with the MPO prior to using the data to determine if there are any known errors or corrective measures. This website from the University of Minnesota provides links to many MPO travel surveys, although the list is far from complete:

Depending on the specific household travel survey, a number of different methods and techniques are available to analyze VMT and ultimately GHG emissions. For household level estimates, the following steps would be followed.

Step 1: Obtain Household Travel Data. Data can be obtained from surveys such as the NHTS, regional surveys, or other sources.

Output: Household travel data

Step 2: Expand Survey Sample to Universe. The next step is to expand the survey sample estimates to the entire population. For example, the NHTS includes estimates of the average VMT generated per driver by age (see table below). This information can be used to estimate VMT for a state or MPO, using the number of registered drivers by age to expand the survey data. This type of estimate will exclude some travel on the state or MPO network (e.g., pass through trips) but that may be acceptable if the focus of the inventory and any associated GHG reduction strategies is on the residents of the area.

Table 10. Average Annual VMT per Licensed Driver by Age, 2009

Driver's Age


16 to 19


20 to 34


35 to 54


55 to 64




Source: FHWA, "Summary of Travel Trends: 2009 National Household Travel Survey," Prepared by A. Santos, et al, June 2011, FHWA-PL-11-022, Table 42, available at:

Output: VMT for State or MPO

Step 3: Apply Emissions Factors. Per mile emissions factors can be applied to estimate total GHG emissions. For more information on approaches, see Section 5.3.

Output: Total GHG emissions for State or MPO

Option 3: Use Land Use Data
A general approach using land use data would involve the four basic steps described below.

Step 1: Collect Land Use Data. Land use-based trip generation methods typically rely on an inventory of existing land uses associated with other planning efforts. In most cases, the inventories at a state or MPO level use residential data (i.e., housing units) only and accept the limitation that the method does not address commercial, visitor, and some employment trips.2 Regional and local agencies are more likely to have complete land use inventories including both residential and non-residential land uses.

Output: Land use data

Step 2: Estimate Vehicle Trips. The land use amounts are multiplied by vehicle trip rates (or person trip rates and then converted to vehicle trips) using trip generation rate sources such as the Institute of Transportation Engineers (ITE) informational report, Trip Generation or NCHRP Report 365, Transportation Research Board, 1998. Freight trips may be estimated using truck trip generation rates based on land use.

Output: Vehicle trips by land use type

Step 3: Estimate VMT. Once trips are estimated, they are multiplied by average trip lengths, which can be obtained from a variety of sources such as the NHTS or NCHRP Report 365. The trips may also be disaggregated into common purposes such as home-based work (HBW), home-based other (HBO), and non-home-based (NHB) since trip lengths are often available by purpose. An important clarification to consider when estimating household generated VMT is whether to include NHB trips generated by residents. Many trip rate sources, such as the ITE report noted above, only include HBW and HBO trips because they only measure trips that crossed the driveway of the home. A full accounting of household generated VMT would track all the trips made by residents throughout the day. Approximately 25 percent of daily vehicle trips are NHB so excluding them could result in an underestimate of VMT and GHG emissions. VMT from freight trips may also be generated if data on average truck trip length are available.

Output: VMT

Step 4: Apply Emissions Factors. Per mile emissions factors can then be applied to estimate GHG emissions from the vehicle types for which VMT data have been estimated. These factors may or may not account for the effects of speeds on emissions. For more information about developing emissions factors see section 5.3.

Output: GHG emissions


The table below contains an estimate of household generated VMT and GHGs for an average weekday using NHTS data.

Table 11. Weekday Household GHG Emissions Estimate Using NHTS VMT Data


2010 Households (1)

Average VMT Generated Per Household Per Weekday (2)

Total Weekday Household Generated VMT

CO2 Equivalent Emissions Factor (for gasoline) (lbs per mile) (3)

CO2 Equivalent Emissions per Weekday
(metric tons)








5.2. Estimating VMT Relying on HPMS Data and/or a Network-based Travel Model

Another way of developing estimates and forecasts of GHG emissions relies on VMT data derived from models. Two key sources of these model-based VMT forecasts are the Federal Highway Administration's HPMS and network-based travel forecasting models, both of which assign VMT to the roadway network (in contrast to methods described in Section 5.1, where VMT is estimated based on sources, such as vehicle population, households, or land uses). This section provides additional information about using HPMS data and network-based travel forecasting models, describes how to extract relevant information, and presents the strengths and weaknesses of these approaches.

Table 12. Selection Criteria for Estimating VMT with HPMS or Network-based Travel Models

Selection Criteria

HPMS or Network-based Travel Model


Network Model

Analysis Type

Inventory or Forecast

Inventory or Forecast

Geographic Scope

State, regional, or county

State, regional, or county

Analysis Precision

Moderate for inventory
Low for forecasts

Relatively high, but depends on sophistication of network model.

Data Needed

HPMS VMT Data, VMT by vehicle type

Network Model Output

Necessary Analytical Capabilities

Limited - HPMS data are readily available

Requires a travel demand model*

Level of Resources Required (i.e., staff/budget)

Depends on level of adjustments required

Depends on existing modeling capability*

Capable of Addressing Vehicle Technology/ Fuels Changes

Not directly addressed in travel modeling, but can be addressed through emissions modeling.**

Not directly addressed in travel modeling, but can be addressed through emissions modeling.

Capable of Addressing Changes in Travel Demand

Limited; requires additional analysis

Yes, designed to address transportation system investments but model sophistication varies in terms of ability to address land use factors, bicycle/pedestrian investments, etc.

Capable of Addressing Changes in Vehicle Speeds and Operations

Yes, to the extent that travel speeds are incorporated into the analysis.

Yes, accounts for congestion, but typically does not address system management strategies or eco-driving.

*Note: The level of effort required may not be significant for agencies with network model already well developed.

**Can be combined with other models and methods, such as MOVES, to incorporate changes in vehicle and fuel technology.


Highway Performance Monitoring System (HPMS)
The HPMS is a program administered by the FHWA,3 which requires that all State DOTs submit annual traffic count, highway inventory, revenue generation, and safety information as a condition of receiving Federal funding. Since it is impractical to count traffic or evaluate the pavement quality of every roadway segment in a state, models are used to translate a sample of data into the regional and statewide data required by FHWA. Related to GHG emissions estimation, the traffic count data are typically aggregated into VMT by vehicle class and roadway functional class at a variety of geographic levels.

Because all states collect HPMS data that must conform to FHWA requirements, these data are available for all states and metropolitan areas over 50,000 in population. FHWA reports VMT by Federal Aid Urbanized Area in Highway Statistics, which is the annual report that summarizes the HPMS data. Typically, Federal Aid Urbanized Area boundaries do not match with MPO boundaries although state HPMS programs often provide VMT by county.

By their nature, HPMS data are backward looking and can provide a good review of historic trends. FHWA's Publication Highway Statistics is available back to 1945 and the HPMS was established in 1978. HPMS reporting requirements include providing estimates of future VMT, which should be a 20-year forecast of annual average daily traffic (AADT). Since the data collected for HPMS are based on observed conditions, there are limitations to the forecasts, which may be developed from state procedures or MPOs or other sources. Other approaches could involve extrapolating trends by functional class, using regression to correlate changes in the population to VMT, or other types of statistical analyses. However, for all VMT-based methodologies, extrapolation of VMT and vehicle trends needs to account for demographic and economic changes. These (and other) factors would require analytical assumptions, which are subject to significant degrees of uncertainty and need to be thoroughly "vetted" and disclosed.

Another limitation of HPMS data is that it does not account for time of day variation in volumes and speeds. Moreover, since the data are associated with travel on the roadway network, this can create challenges if trying to assign emissions to trip generating sources within the state or region. For example, some states like California require VMT estimates and forecasts to account for trips that enter and exit MPO regions when analyzing GHG emissions, but exclude "through trips", those traveling through the MPO without a stop. This is different than conventional air pollution analysis that focuses on travel within a non-attainment or maintenance area, and HPMS does not provide information related to the origins or destinations of trips.

Network-Based Travel Forecasting Models

Understanding the Sophistication of Travel Forecasting Models
Travel demand forecasting models are commonly used by MPOs, and several State DOTs also have statewide travel demand forecasting models. These tools vary in their sophistication, and the extent to which travel models account for different factors (e.g., land use, transit, bicycle and pedestrian activity) will affect the accuracy of VMT forecasts and their ability to address different types of strategies.

While this Handbook is not designed as a resource on travel forecasting, it is important for those who wish to analyze GHG emissions to understand the strengths and limitations of their travel models. The following resources may be consulted for support.

  • Travel Model Validation and Reasonableness Checking Manual, FHWA/TMIP, 2010
  • NCHRP Report 716, Travel Demand Forecasting: Parameters and Techniques, TRB, 2012
  • Metropolitan Travel Forecasting, Special Report 288, TRB, 2007
  • 2010 California Regional Transportation Plan Guidelines, CTC, 2010

Other useful resources are available through the Travel Model Improvement Program (TMIP) -

Network-based travel forecasting models are computer programs that are developed to estimate future travel patterns in a given area based on variables that influence both transportation supply and demand. Key structural and input variables for these models often include land use, socio-demographic characteristics, travel modes, transportation network, and travel costs. The models can be simpler or more complex, 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 that is not contemplating transit over the planning horizon may leave out the mode choice step. A number of the largest MPOs are replacing their existing models with advanced tour- or activity-based models and some are testing advanced dynamic traffic assignment (DTA) to improve sensitivity to traffic operating conditions and provide more accurate forecasts of peak period conditions. As with all models, network models have various limitations, and it is important to consider these limitations.

The most common network models are trip-based models where trip generation is usually a function of land use data. These models are often called three or four-step models after the number of key submodels (trip generation, trip distribution, mode choice, and traffic assignment) they include. Network-based travel models are fairly ubiquitous at the MPO level, which are urbanized areas (or portions of an urbanized area) with a population greater than 50,000. However, outside of MPOs, network-based travel models are less common. For example, many smaller communities do not have a network-based travel model and only a handful of states have a statewide model available.

In contrast to the HPMS data, network-based travel models are forward looking and generally produce a reasonably reliable forecast of future travel patterns. Therefore, these models can generally be relied on to forecast future GHG emissions (provided that they are well calibrated and validated, and provided that analysts can reliably predict future VMT by vehicle type and the GHG emissions characteristics of future vehicles). Unfortunately, while these types of models have been available since the 1960's they are not a good source of historic data. More often than not, when jurisdictions periodically update their models to remain current with the latest land use and transportation system changes, the prior version of the model is discarded. As time progresses, many jurisdictions cannot retrieve either input or output data from these older versions of the model. Thus, estimating historic VMT and GHG emissions can be difficult, which may be important, since several state GHG emissions reduction targets are based on 1990 GHG emissions levels (consistent with the Kyoto Protocol).4

A limitation of network-based models similar to that noted for HPMS data is that these models also have physical boundaries, which can make full accounting of VMT difficult for those trips that cross the boundaries. Further, some GHG mitigation strategies are targeted at VMT generators (i.e., land use development) so knowledge about both ends of a trip is often required to understand the effectiveness of the strategy. Quantifying VMT changes from mitigation strategies is subject to uncertainties. For GHG quantification, these uncertainties may be compounded by uncertainties in future vehicle technologies and fuels, and difficulty in distributing VMT among many different vehicles/fuel types/models/years expected in the future. GHG quantification by all models, including network-based VMT models, is also limited by the inability to accurately reflect non-recurring (for instance, incident-based) congestion and eco-driving behavior.

Strengths and Limitations

Of the tools described up to this point, network models have the advantage of being able to capture the GHG emissions related to population and employment growth as well as transportation network or system changes. Network models also allow for testing of both transportation demand and supply in an integrated model. The accuracy of network models is better than many of the other methods or tools discussed depending on the level of detail in the model. It is also important to consider that network models have been used for a long time, their modeling framework is well understood, and the inputs they require are generally available. The main limitation of these models is their complexity and cost. Smaller agencies may lack the staff expertise to develop a network model as such models require a considerable amount of resources to develop, calibrate, and validate. These models also require fairly expensive software and regular maintenance for land use and transportation databases to produce accurate results. As such, HPMS may be an effective alternative method because of data availability and limited effort necessary to obtain VMT data and then calculate GHG emissions. Key strengths and limitations of each method are summarized in the table below.

Table 13. Strengths and Limitations of Estimating VMT with HPMS and Network-based Travel Models





  • Simplest method for estimating statewide VMT by roadway functional type
  • County, district, or regional VMT may also be directly available
  • Information on VMT by vehicle type is available
  • Reliable historic data available
  • Given limited traffic count sample sizes, data quality may be an issue at small geographic scales or in rural areas
  • No explicit data on operating speeds or congestion, only speed limits on roadways are available
  • VMT forecasts rely on extrapolation that may not be sensitive to changes in land use, travel cost, or other important variables
  • Data is based on travel on a network and cannot isolate origins and destinations (which may be useful for some analyses)
  • GHG results are subject to uncertainty and inaccuracy in out years due to uncertainties in vehicle technologies and fuels, and distribution of VMT among different vehicle types and fuels

Network-based model

  • Explicitly includes data on speed and traffic congestion (assuming model is calibrated to such factors)
  • Compared to HPMS, much more reliable forecasts for future conditions
  • Some ability to account for sources or contributors to VMT
  • Data can be difficult to extract without properly trained staff or consultant assistance
  • Because network-models are custom tools, the VMT results may not be directly comparable between regions
  • Can be difficult to generate historic VMT estimates
  • Estimates or forecasts are based on fixed boundaries and may not capture the full length of trips entering or leaving the MPO or state boundary.
  • GHG results are subject to uncertainty and inaccuracy in out years due to uncertainties in vehicle technologies and fuels, and distribution of VMT among different vehicle types and fuels

Key Steps and Data Options

This section provides step-by-step instructions on how to extract the data needed to estimate GHG emissions from the two methods described above.

Step 1 - Estimate VMT
The first step in developing a GHG emissions estimate is the extraction of VMT data. Because the data format of the HPMS and network-based travel models are so different, there are two distinct methods from extracting the data from each data source.

Option 1: Use HPMS data
HPMS VMT data are fairly easy to extract. At the statewide level, the FHWA's Highway Statistics publication ( provides VMT data organized by the following:

The data from Highway Statistics are available as both webpage data and Microsoft Excel files, which make the data easy to post-process.

In addition to the FHWA data sources, all State DOTs also have HPMS data available; although the organization of this data varies considerably from state-to-state. For example, Ohio DOT summarizes HPMS-type VMT data by county organized by Federal functional classification ( On the other hand, the Alaska DOT organizes VMT data into three districts of the state, with distinctions made for urban and rural areas and Federal functional classification.

If conducting a base year estimate of GHG emissions, the HPMS data can be used directly. For future years, the basic steps involved in forecasting VMT in areas without a network model is to 'grow' or 'adjust' HPMS base year estimates based on mathematical relationships between VMT and commonly forecasted variables such as population and employment; or in more complex models, built environment variables (i.e., the Ds) such as land use density, land use diversity, regional accessibility, and distance to transit service. The main steps involved are dependent on the level of sophistication of the method or model. Some examples are listed below, which range from a very simple linear trend-line projection to more complex regression analyses that base VMT forecasts on a range of demographic and economic factors.

More information on these methods and their use is available in the FHWA report: Sample Methodologies for Regional Emissions Analysis in Small Urban and Rural Areas, available at:

Option 2: Use a network-based travel model
Obtaining VMT data from a network-based travel model is fairly straightforward for base year or future year conditions. Nearly every network-based model that is likely to be in common use will have the ability to output link-by-link average weekday traffic volumes. Some network-based models may only have a PM peak hour component, and in this case, only average PM peak hour weekday traffic volumes may be available. Many models may automatically generate a VMT report, which can be used to estimate VMT directly, however, as will become apparent later in this section, the disaggregate link volumes are more useful for ultimately estimating GHG emissions.

If annual GHG emissions estimates are desired, the average weekday or PM peak hour data will need to be factored up to generate annual link volumes or VMT. There are several methods to perform this factoring, but the most common is to develop an AADT factor, which will (as the name suggests) convert the average weekday traffic volume into an annual average daily traffic volume. Most State DOTs have a method to convert daily traffic data into annual average daily traffic data. Additionally, most MPOs have regular traffic data collection programs that can account for the weekend/weekday, PM peak hour/daily, and seasonal variations in traffic to develop a more localized AADT factor. Once AADT is known, it is straightforward to develop annual traffic volumes and VMT by multiplying traffic volumes by the length of road segments.

Output: VMT

Step 2: Estimate Speeds/Disaggregate VMT into Speed Bins (optional)
With the VMT data extracted, the next step in developing GHG emissions is to estimate the speed at which VMT was accumulated. While there are generic GHG emissions factors that do not explicitly consider speed, the figure below shows that GHG emissions rates for light-duty vehicles are highly variable based on speed - with nearly two times as much CO2 per mile at low speeds as for mid-level speeds. Moreover, CO2 emissions per mile are higher during transient driving (stop and start conditions) than during smooth driving at the same overall average speed.

Figure 8. CO2 Emissions with respect to Speed, Light-duty Vehicles

A line graph showing grams per mile of CO2 generated for various speed bins ranging from 0 mph to 80 mph. For transient driving, gCO2/mi decreases from 700g to 350g between 0 and 60 mph and increases to 400g after 60 mph. For smooth driving, gCO2/mi decreases from 700g to 275g between 0 and 40 mph and increases to 380g after 40 mph.
Source: U.S. EPA, analysis using MOVES for all light-duty vehicles for 2010.

Given the importance of speed when calculating GHG emissions from VMT, the HPMS is at a disadvantage compared to network-based models since no future speed data is explicitly included as part of the HPMS. However, if only rough approximations are required, using functional class and geographic information from the HPMS can help to relate some speed information to the VMT data, based on reported speed limits.6

For example, the 2008 HPMS database showed the following information for the state of Arkansas:

Table 14. Arkansas 2008 Annual VMT (millions)

Functional Class






Other Principal Arterial



Minor Arterial



Major Collector



Minor Collector








Source: FHWA, HPMS 2008.

Given these data from the HPMS, a similar table of average speeds for each functional classification could be developed. An example table is below. These data could be developed based on speed limits, but ideally, these data would be developed from actual speed survey data from locations across the state.

Table 15. Arkansas Estimated Travel Speed by Functional Class

Functional Class

Rural VMT

Urban VMT




Other Principal Arterial



Minor Arterial



Major Collector



Minor Collector






Source: Sample table based on common speeds.

Alternatively, an analyst could use a sample of speed data across several roads in each functional class to develop an average speed for that functional class, and apply that across the entire functional class. In some cases, an analysis could also be based on formulas relating the volume to capacity (V/C) ratio and the free-flow speed to estimate speeds on individual road links. Grouping links with similar parameters and analyzing together, such as at a functional class basis, could simplify this analysis.

Extracting speed data from a network-based model is typically a simple process so long as the entire link database from the model run was extracted7. Within the output link database, information on traffic volumes, link length, "congested" link speed, and "free-flow" link speed are included. It is important to distinguish between congested link speed (which is the speed the model predicts on the link given the forecasted traffic congestion) and free-flow link speed, which is typically the speed limit or prevailing free-flow speed. Depending on the type of model, both congested and free-flow speed may be needed to accurately estimate GHG emissions. For example, many simple network-based travel models have a daily traffic model that will produce an estimate of congested speed. However, this congested speed is used as part of the traffic assignment portion of the model and is typically more representative of congested peak travel periods. Using congested travel speeds with a daily or annual VMT estimate would overstate potential GHG emissions. Therefore, if the congested speed represents conditions affecting 20 percent of the VMT and the remainder of the VMT occurs at or near free-flow speed, then the analysis should separately account for VMT data at the different speed levels. Most large MPO models have more explicit treatment of off-peak and peak travel times. Compiling the speed data is typically done in a spreadsheet where the total VMT occurring in a given speed "bin" (a speed range) is totaled. The table below is an example of VMT data organized into 16 different speed bins.

Table 16. Example of Network-based Model VMT by Speed Bin Data








speed < 2.5mph





2.5mph <= speed < 7.5mph





7.5mph <= speed < 12.5mph





12.5mph <= speed < 17.5mph





17.5mph <= speed <22.5mph





22.5mph <= speed < 27.5mph





27.5mph <= speed < 32.5mph





32.5mph <= speed < 37.5mph





37.5mph <= speed < 42.5mph





42.5mph <= speed < 47.5mph





47.5mph <= speed < 52.5mph





52.5mph <= speed < 57.5mph





57.5mph <= speed < 62.5mph





62.5mph <= speed < 67.5mph





67.5mph <= speed < 72.5mph





72.5mph <= speed



Source: Example values from MOVES2010, using the 2010 Lake County example from the MOVES demonstration training files, for HPMS vehicle type 20, source type 21,road type 3.

While it is fairly straightforward to extract speed information from a network-based travel model it is important to understand how reliable the model's speed data are. For example, many models are not calibrated to match observed travel speeds. Therefore, before network-based travel model data are used, the practitioner should evaluate to make sure that the speeds make sense for roads of different functional classifications and for the model as a whole. In addition, the practitioner should perform some sensitivity tests to ensure that the model responds reasonably to changes in anticipated traffic flow. Guidance on these types of dynamic validation tests are provided in the following resource documents.

If a model is not appropriately sensitive, increases in traffic volumes may not cause appropriate reductions in congested speeds, which would affect the GHG emissions because the VMT stratification across speed bins would not be correct. If the network-based model's speed estimates are questionable, then the same procedures outlined for the HPMS data may be applicable, including relying on the built-in speed distributions in air quality emissions modeling software. It is also possible to post-process the output speeds from travel models, using volume to capacity relationships, to generate more reasonable estimates of congested speeds.

Output: VMT by speed bin

Step 3: Estimate Vehicle Fleet Mix
GHG emissions vary widely based on the type of vehicle and type of fuel used. The table below provides some example emissions factors in grams of CO2 equivalent (CO2e) per mile for several types of vehicles:

Table 17. Example of Composite CO2e Emissions by MOVES Vehicle Type



Source Type ID

Source Type

CO2 Equivalent (g/mi)





Denver, CO







Passenger Car





Passenger Truck





Light Commercial Truck





Intercity Bus





Transit Bus





School Bus





Refuse Truck





Single Unit Short-haul Truck





Single Unit Long-haul Truck





Motor Home





Combination Short-haul Truck





Combination Long-haul Truck




Source: Sample outputs predicted with MOVES2010b, for calendar year 2010 for Denver County, Colorado at the National scale. CO2e emissions include running, start, and extended idle (as appropriate for each vehicle type) exhaust emissions of CO2, N2O, and CH4, each pollutant normalized according to its Global Warming Potential (GWP).

Given this variability in emissions rates by vehicle class, it is important to determine the amount of VMT generated by different vehicle types. The FHWA's Highway Statistics publication provides national VMT by vehicle type, but not state-level detail. Travel information by vehicle type is more difficult to obtain than basic VMT data, and in many cases it may be necessary to contact a goods movement or freight division of the DOT to obtain the data. Often the data are split only into autos/light trucks and commercial trucks.8 Below is an example data from Denver County, Colorado for 2010 travel by vehicle class.

Table 18. Example of VMT disaggregated by HPMS Vehicle Type












Passenger Cars




Other 2 axle 4-tire vehicles








Single Unit Trucks




Combination Trucks

Source: Sample outputs predicted with MOVES2010b, run for calendar year 2010 for Denver County, Colorado at the National scale.

Network-based travel models may also have limited information about vehicle type. In general, only the largest MPOs have a freight component to the travel model and there is often very little data by which to calibrate these freight models. If direct model output is used to estimate VMT by vehicle type, steps should be taken to ensure that the data are reasonable. Moreover, it should be noted that vehicle type groupings encompass a large variety of vehicles, with a wide range of variation in GHG emissions/mile. Issues such as vehicle age can be accounted for through the application of appropriate emissions factors or through use of an emissions model.

Given the limitations for both HPMS and network-based travel models when it comes to estimating VMT by vehicle type and the lack of finer-grained vehicle data related to vehicle age and other characteristics, it may be most advantageous to use a VMT distribution from emissions modeling software like MOVES or EMFAC. Note that the current version of MOVES does not have default VMT by vehicle type distributions, except at the national scale (it does produce estimates of VMT by vehicle type, which can then be converted to a distribution if needed). However, if MOBILE6 data are available, there is a documented procedure from the U.S. EPA to convert MOBILE6 files to MOVES format. Otherwise, vehicle type VMT (at the HPMS-class level) may be obtained from state agencies, HPMS reports, or other sources.

However, even these approaches have limitations. In many cases the VMT by vehicle type distribution is based on vehicle registrations within a given area (state, county, region). While the registration distribution across different vehicle classifications is probably reasonable for light and medium duty vehicles, heavy duty trucks are often registered far from where they are operated. In addition, some fleet operators register many of their vehicles in one location even if they are not operated in that location. These factors can lead to difficulty in getting an accurate distribution of VMT by vehicle type, particularly for trucks.

Output: VMT by vehicle type

Step 4: Develop Information for Other Factors Influencing Emissions (optional)
VMT and vehicle emissions factors are the two most important pieces of information for developing GHG emissions from HMPS or network-based travel mode data. However there are other factors that could be influential for developing GHG emissions estimates. Many of these factors are not common, particularly across large geographies, but may be important. Examples include the carbon content of the fuel (e.g., based on whether there is a renewable biofuel blended with the standard fuel), penetration rate of hybrid electric or full electric vehicles, eco-driving programs, and steep grades or other unusual geography. Other common factors that influence criteria air pollutants such as meteorological data do not have a substantial impact on GHG emissions.

Most of the common emissions modeling tools have the ability to account for some of these factors explicitly. Other factors will have to be accounted for after the fact by either scaling emissions factors (e.g., speed management strategies) or scaling preliminary emissions totals. Any post-processed adjustments should be transparent and accompanied by substantial evidence that justifies the modifications. Given the variability of other factors that can influence GHG emissions, there is no standard protocol that can be recommended. Rather a reasonable estimate of the change in GHG emissions rates or total emissions must be made and applied accordingly.

Output: Estimate of the change in GHG emissions rates or total emissions due to other influencing factors

Step 5: Develop Emissions Factors and Estimate Emissions
Based on the VMT and fleet data developed using the procedures described above, GHG emissions can be calculated by applying an emissions factor. Or, MOVES can be run in Inventory Mode if all of the model-related inputs are available. For more information about developing emissions factors see Section 5.3.

Output: GHG emissions

Step 6: Conduct Additional Strategy Analysis (optional). To consider strategies that are not well accounted for in HPMS or travel model forecasts (e.g., travel demand management strategies, truck stop electrification), additional off-model analyses should be conducted. For more information about specific transportation strategy analysis methods, see Section 7.


Delaware Valley Regional Planning Commission (DVRPC) GHG Inventory

One approach to developing a GHG inventory using a transportation demand model can be seen in the inventory developed by DVRPC. DVRPC developed a regional GHG emissions inventory that relies on travel demand model outputs to allocate GHG emissions to different traffic analysis zones.

HPMS data was used to determine a VMT total. Through traffic was estimated based on the travel demand model trip table that shows trips with origins and destinations outside the region. VMT from through traffic was subtracted from total VMT to focus the analysis on travel within the region. VMT was then apportioned to municipalities based on trip origins, destinations and trip length. Emissions were mapped per acre, per population and per employee.

Figure 9. DVRPC Maps Showing GHG Emissions by Geographic Area

Left Side: A map showing the DVRPC region's greenhouse gas emissions per acre by municipality. The area with the highest emissions per acre is in the center of the map. Right side: A map showing DVRPC's greenhouse gas emissions per population and employment by municipality. The area with the lowest emissions per population and employment is the center of the map.
Source: Delaware Valley Regional Planning Commission.

The map above to the left shows emissions per acre, which indicates that GHG emissions are higher in Philadelphia's urban core. If emissions for trips are allocated 50 percent to the trip origin and 50 percent to the trip destination, the map on the right shows that emissions are higher on a per population and per worker basis in the suburban and exurban areas around Philadelphia. The DVRPC inventory helps make the case for the role of smart growth in reducing the GHG emissions intensity of development in the region. DVRPC's inventory is available at:

Capital District Transportation Committee (CDTC) GHG Forecast

CDTC, the MPO for the Albany metropolitan area, conducted an analysis of the GHG effects of its long-range transportation plan using guidelines from the New York State DOT, which include a set of lookup tables and adjustment factors to estimate fuel use per vehicle mile by average speed group. CDTC used its 3-step network-based travel forecasting model, and made an off-model adjustment that reduced VMT to account for land use, transit, and demand management policies. It then used a set of lookup tables of fuel economy by speed and adjustments to account for future fuel economy improvements and multiplied the VMT in each speed bin by the fuel consumption rate at each speed. Finally, CDTC converted its fuel consumption into CO2 using a set of equations provided by the state.

Atlanta Regional Commission (ARC) GHG Forecast

In 2007, ARC staff began to prepare a study to inform local policy makers and citizens about the impact of transportation investment decisions on GHG emissions. ARC performed a scenario analysis using its 4-step travel demand model and MOBILE6 to model the emissions impacts of various land use scenarios describing different types of possible growth. Changes in land use and the transportation network were used as inputs in the travel demand model, which fed into the MOBILE6 calculations. The result allowed ARC to demonstrate the impact on GHGs of a variety of strategies, including Federal fuel efficiency standards, land use policies encouraging density, as well as the previous regional plan, Envision6, as shown below. ARC has since conducted additional analyses using MOVES.

Figure 10. ARC's Emissions Forecast under Multiple Scenarios

A graph showing the potential on-road CO2 emissions in Atlanta through 2030 given a variety of different scenarios. The given scenarios are future local plans (trend), which projects the highest number of on-road emissions at about 158 thousand tons per day. This is followed by Envision6, which predicts about 149 thousand tons; trend + EISA, which predicts about 110 thousand tons per day; Envision6 + EISA, which predicts about 101 thousand tons per day; density land use + EISA, which predicts about 95 thousand tons per day; TPB Concept 3 + transit focused land use + EISA, which predicts about 90 thousand tons per day; and C3 +TFLU + 2009 CAFE, which also projects about 90 thousand tons per day.
(1) Envision6 is ARC's previous Regional Transportation Plan; it has now been replace by PLAN 2040.
(2) Density Land Use is a schema that increases regional density into key activity centers and curbs sprawl.
(3) EISA is the Energy Independence and Security Act CAFE standard
(4) TPB Concept 3 is the Transit Planning Board concept for regional transit buildout
(5) Transit Focused Land Use focuses on nodes identified in CONCEPT3
Source: Atlanta Regional Commission. See

Metropolitan Washington Council of Governments (MWCOG) GHG Forecast

The National Capital Region Transportation Planning Board (TPB), the Federally designated MPO for the region, conducted a "what would it take" scenario study to determine how the MWCOG goals would be met in the transportation sector. TPB determined that cumulative emissions would need to be reduced by 33.5 percent from 2010 to 2030 to meet MWCOG's goals.

Figure 11. MWCOG's "What Would it Take" Scenario Study

A graph answering the question, 'What if we had to meet these MWCOG multi-sector goals in the transportation sector?' The graph shows Annual Emissions on the 'Y' axis and years/emission reduction targets on the 'X' axis with a negative sloping line called 'COG Goals' and a positive sloping line called 'CAFE 25 (pre EISA 2007).' The area above the COG Goals line but below the CAFE 25 line is shaded pink and represents a 33.5% cumulative reduction in annual emissions.
Source: "What Would it Take? Transportation and Climate Change in the National Capital Region: Final Results," Presentation to the Climate Energy, and Environment Policy Committee by Ronald Kirby, National Capital Region Transportation Planning Board, May 26, 2010.

As part of this analysis, TPB conducted an inventory and forecast for years 2005, 2010, 2020, and 2030, and applied different scenarios to determine various ways of reducing emissions to meet the target. TPB used its travel demand model to forecast VMT using assumptions about projects and the network from its 2009 Constrained Long Range plan and its 2010-2015 Transportation Improvement Plan. TPB then applied CO2 emissions factors generated by MOBILE6.2, and used software developed by a consultant to generate the CO2 emissions totals (like ARC, TPB now uses MOVES rather than MOBILE6.2). Once the CO2 emissions forecasts were generated, it was possible to apply newly adopted CAFE standards and other possible emissions reduction measures using spreadsheet tools to better determine what it would take to meet the goals.

Maryland Department of Transportation, Draft 2012 Implementation Plan for the Maryland Climate Action Plan Analyses

Maryland DOT (MDOT) developed a 2006 baseline inventory, a 2020 baseline forecast, and a 2020 action scenario reflecting application of investments and strategies. The procedures were fairly intensive and involved use of HPMS data and the MOVES2010a (MOVES) model. The state does not as yet have a statewide travel demand forecasting model, however several alternatives were available to determine forecast growth rates. MDOT developed a statewide link level traffic database using Maryland State Highway Administration's (SHA) HPMS data and used forecasts based on historic trends of 1990-2006 HPMS VMT growth for the 2020 baseline (business as usual - BAU) scenario.

Annual GHG values were calculated based on 12 monthly runs of MOVES, each using traffic volumes, speeds, temperatures, and fuel values specific to an average day in each month to arrive at an annual total value. For the 2020 BAU scenario, the procedures for emissions analysis involved the following steps: 1) adjust traffic data to an average day in each month; 2) run MOVES for each of 12 months; 3) multiply VMT and emissions by the number of days in each month; and 4) aggregate to an annual total. Since MOVES includes the effects of post-2006 national vehicle programs (Model Year 2008-2011 CAFE standards and Model Year 2012-2016 program), these technology programs were removed by revising the MOVES default emission rate database.9

The traffic database has several link level attributes including number of lanes, AADT volumes, truck percentage, speed limit, area types and functional class. Maryland uses pre and post-processing software called PPSUITE to process the traffic data, develop MOVES input files, and process the results. Since the traffic data for roadway segments did not include congested speeds and hourly detail needed by MOVES, PPSUITE calculates the hourly congested speeds for each link, applies vehicle type fractions, aggregates VMT and vehicle hours traveled (VHT), and prepares and formats the required MOVES input files. PPSUITE relies on data from the SHA Traffic Trends System Report Module to seasonally adjust AADT in order to estimate average daily traffic each month, and to disaggregate volumes to each hour of the day. MDOT used MOVES defaults for the miles per vehicle by source type.

For the 2020 action scenario, MDOT conducted analyses by modeling a range of strategies, including 2020 vehicle technology emissions reduction programs (including the CAFE standards 2008-2011 MY, the national program 2012-2016 MY, the Maryland Clean Car Program (2011MY), National Fuel Economy Standards (2017-2025 MY), and National 2014-2018 Medium and HDV Standards), by making adjustments to the emission rate default database in MOVES. MDOT also examined transportation fuels, including renewable fuels and a low carbon fuel standard, as well as implemented and adopted transportation plans and programs. In order to analyze the impacts of the committed and funded state, regional, and local transportation and land use plans, MPO VMT forecast data were used to adjust VMT growth rates on a county level, resulting in an estimate of a 1.4 percent annual rate of VMT growth between 2006 and 2020, compared to the HPMS historical baseline of 1.8 percent statewide. MDOT also separately analyzed a range of transportation emission reduction measures (TERMs) that have GHG benefits.

See documents related to the MDOT climate action plan at:

5.3. Developing Emissions Factors & Emissions Inventories

This section describes how to estimate emissions either by developing and applying emissions factors from several common sources or using existing models.

Table 19. Selection Criteria for Approaches to Estimating Emissions Once Travel Activity Has Been Determined

Selection Criteria

Emissions Factors and Inventories

Simple Factor

Look-up Table Only Accounting for Fleet Characteristics

Look-up Table Accounting for Fleet Characteristics and Speeds

MOVES(1) Inventory

MOVES(1) Emission Rates

Analysis Type

Inventory or forecast

Inventory or forecast

Inventory or forecast

Inventory or forecast

Inventory or forecast

Geographic Scope

State or regional

State or regional

State or regional

Nation, state, region, county, or project

Nation, state, region, county, or project

Analysis Precision




High, with precision highest using local factors rather than defaults

High, with precision highest using local factors rather than defaults

Data Needed


VMT, fleet characteristics

VMT, speed bins, fleet characteristics

VMT and Population by vehicle type(2)

VMT and Population by vehicle type(2)

Necessary Analytical Capabilities

Limited - factors easily available from EPA/EIA sources

Moderate - needs disaggregation of VMT by vehicle type

Moderate - needs disaggregation of VMT by vehicle type and speed bin

Moderate - requires use and postprocessing of MOVES model inputs and outputs, although with use of National scale defaults, inputs are minimized.

High - requires use and postprocessing of MOVES model inputs and outputs, although with use of National scale defaults, inputs are minimized.

Level of Resources Required (i.e., staff/budget)

Limited - relatively easy

Limited -requires using additional factors

Moderate - requires additional analyses of speeds

Requires staff trained in the use of MOVES

Requires staff trained in the use of MOVES

Capable of Addressing Vehicle Technology/ Fuels Changes






Capable of Addressing Changes in Travel Demand

Yes, in aggregate; not by vehicle type





Capable of Addressing Changes in Vehicle Speeds and Operations






(1) EMFAC can be applied in California instead of MOVES.
(2) MOVES may require significant input data if not relying on national default values. Increasing accuracy is achieved when more locally specific data are used. At a minimum, VMT by HPMS vehicle type is required for running emissions and population for other emission processes. Both have defaults available at the National scale but not at finer scales. More information is available from EPA.1010
(3) If different
VMT scenarios are considered then travel and land use changes may be addressed.


There are several options for estimating emissions, once travel activity data (VMT) are estimated. These range from applying a simple composite emissions factor that reflects all on-road vehicles, to using tables of emissions factors for different vehicle types, or tables of emissions factors that account for vehicle speeds, to using the detailed emissions models, such as MOVES or EMFAC (in California). Generally, simpler approaches will lead to less accurate analytical results.

MOVES is the preferred tool for estimating GHG emissions

For more information about EPA's MOVES model, including appropriate guidance documents, see:

  • Resources on MOVES, including user manual:

Simple CO2 emission factors obtained from published sources can be multiplied with estimated VMT to produce an estimate of CO2 emissions. Such factors are typically not sensitive to aspects like vehicle speed and fleet mix. For instance, the U.S. EPA has a simple GHG emissions factor of 460.2 grams of CO2 equivalent per light duty vehicle mile traveled -

However, the accuracy of the CO2 estimates can be improved if either the proportions of key vehicle types in the fleet or average speeds are known. To more accurately account for different emissions from the range of vehicle types in the fleet more accurately and efficiently, a look up table that provides simple emission factors by vehicle and fuel type may be used to calculate emissions from VMT data. These emission factors are also available from sources published by EPA.

Currently, the best tool available to produce estimates of on-road transportation GHG (and other) emissions is EPA's MOVES model. In California, the EMFAC model may be used. The MOVES model estimates energy consumption and emissions, including atmospheric CO2, CH4, N2O, and CO2e. MOVES can estimate emissions at the national, county (or custom, multi-county), or project scales and for annual or shorter periods of time.11

The model itself requires many inputs. Although defaults are available for most factors, locally specific inputs produce more accurate results. Inputs to MOVES include data on vehicle population, fuel type, and VMT. The model works by simulating actual vehicle drive cycles, including the effect of travel at different speeds and vehicle power loads. Although the current version of MOVES does not calculate emissions from off-road sources, a future version of the model will. EPA released a series of guidance documents for use of the MOVES model in 2010, updated and released guidance on use of the model for GHG analysis in November 2012.12

Sensitivity to local driving conditions is valuable when examining transportation plans and policies, such as new highway capacity investments, congestion pricing, and other strategies that affect vehicle speeds and operating conditions. To capitalize on the additional capabilities of MOVES, though, a significant volume of input data may be required, along with significantly greater amount of user (and processing) time, as shown in the table below.

Table 20. MOVES Data Inputs for GHG Analysis


Default Data

Typical Data Source

Source (Vehicle) Type Population

Only for national default analyses (which can be used to analyze one or more states or counties) 13

Local registration data - national defaults for heavy trucks and some other classes

Vehicle Type VMT

Only for national default analyses (which can be used to analyze one or more states or counties)

Travel model or HPMS

Month, Day, Hour VMT Fractions

Default data available, but local data preferred

MOVES, HPMS, count stations

Average Speed Distribution

Default data available, but local data preferred

Travel model

Road Type Distribution

Default data available, but local data preferred

Travel model

Age Distribution

Default data available through EPA's website, but local data preferred

Local registration data for light-duty and MOVES national data for heavy vehicles (if no better local source)

Ramp Fraction


Travel model

Meteorological Data

Yes, but local data preferred

MOVES or National Weather Service

Fuel Supply/ Formulation


MOVES or local data

I/M Program


MOVES or local data

The level of effort needed to use MOVES depends on the type of analysis and the existing capabilities of the organization. Using MOVES at the National scale (where the model relies mostly or entirely on national default input data) is relatively simple. Likewise, if an area is already using MOVES for transportation conformity analysis or for development of emissions inventories or forecasts for air quality planning purposes (i.e., state implementation plans), adding GHGs to the list of pollutants being modeled in the analysis involves almost no extra effort. On the other hand, if an area is starting from scratch with MOVES, and wishes to perform an analysis involving extensive use of local data, more effort will be required.

There are two calculation types in the MOVES model: to produce emissions inventories or emission rates. Each of these calculation types is described below.

Strengths and Limitations

Table 21. Strengths and Limitations of Approaches to Estimating Emissions, Once Travel Activity Data Estimated




Simple Factor

  • Easily applied
  • Readily available
  • Does not capture effects of policies affecting vehicle types and operation
  • Lack of precision.

Look-up table only accounting for fleet characteristics

  • Easily applied
  • Captures regional fleet characteristics (albeit, not by model and year)
  • Does not capture effects of strategies that affect vehicle operation, such as congestion reduction strategies
  • Less precise than an analysis using MOVES

Look-up table accounting for fleet characteristics and speeds (vehicle operating characteristics)

  • Easily applied
  • Captures regional fleet characteristics and operating conditions
  • Less precise than an analysis using MOVES

MOVES in Inventory Mode

  • The inventory mode in MOVES can reduce human error, because it eliminates the need for the user to conduct a separate post-processing analysis to apply emissions factors.
  • Using MOVES in inventory mode requires running MOVES every time the travel demand model is modified or a new VMT scenario is prepared. This can be time consuming.

MOVES in Emissions Rate Mode

  • May significantly reduce frequency of MOVES runs. Emissions rates calculated by MOVES are used along with outputs of the travel demand modeling process (speed, etc.) to determine emissions. Thus, a new MOVES run is not required for each transportation demand model run.
  • Emission rates can be complex to apply, as separate rates are produced for running and non-running processes (starts and extended idle), all of which are needed for a complete GHG inventory.
  • Individual model runs can be very time consuming. These rates often then need to be recombined for the analysis.
  • MOVES runs can take longer in emission rate mode

Key Data Steps and Options

Option 1: Use a simple emissions factor
The simplest way to estimate GHG emissions from VMT would be to apply a single emissions factor that is either not sensitive to fleet mix and speed or is blended to at least consider fleet mix and average speeds and/or loads. A web search will reveal many emissions factors for different vehicle types from a variety of sources.

For example, the U.S. EPA has several emissions factors directly published, including a default rate of 423 grams of CO2 per VMT for a passenger car (

Once an emissions factor is selected, calculating GHG emissions is simple:
VMT x Emissions Factor = CO2 Emissions

To provide a slightly more refined answer, other GHG emissions (CH4, N2O, etc.) can be accounted for and the total reported in units of grams of CO2 equivalent (CO2e). The U.S. EPA document cited above provides guidance on this topic . As noted, CO2 emissions constitute up 95 to 99 percent of the global warming potential emissions from a typical vehicle so assuming a conservative five percent of the global warming potential from light vehicle emissions come from other GHGs, the initial CO2 calculation could be multiplied by (1/.95) = 1.053 to estimate the grams of CO2e.

Using Appropriate Emissions Factors

Transportation agencies may be able to collect emissions factors or tables of factors for use in GHG analysis by working with their state air quality or environmental agency. These factors could be developed based on national defaults or reflect location-specific fleet information, and may be developed using MOVES. Regardless of how the emissions factors are developed, the analyst should be careful to apply the appropriate emissions factors. Specifically:

  • Make sure the emissions factor matches with the types of vehicles being examined. For instance, use a factor that reflects only light-duty vehicles if analyzing a travel demand management strategy that reduces commute trips, and use a factor that reflects all vehicles (including heavy-duty trucks) for a strategy like a traffic improvement that affects all vehicles. Even so, this is an extremely crude approach, in that a vehicle type encompasses a wide range of different models and model years with significant variations in GHG (as much as a factor of 4), and the variation would be enormous if all vehicle types are aggregated together.
  • Recognize that emissions factors may be produced for running emissions (per mile) and non-running emissions (per vehicle, to reflect emissions from engine starts and extended idling of heavy-duty trucks), or may be developed in a composite form (accounting for running and non-running emissions). Use a factor reflecting only running emissions to evaluate a strategy like ramp metering that affects only running emissions, and use appropriate factors reflecting total emissions to evaluate a strategy like ridesharing that reduces full trips.

While this approach is very simple to develop and apply, it does not account for several important factors such as the effects of variations in speed or non-running emissions processes, such as vehicle starting or idling. Additionally, if the fleet mix is fairly complex, this approach can be more cumbersome than the other approaches listed below that provide a better way of accounting for variations in the vehicle fleet. These approaches also cannot account for changes in fleet emissions due to changes in fuel economy rules and emission standards, so they would not be useful for projecting future GHG emissions.

Option 2: Use look-up table only accounting for fleet characteristics
This option is similar to the Option 1 but utilizes different emissions factors for different vehicle types. Common types include the following:

Separate emissions factors may be developed for gasoline, diesel, or fuel types, and may be weighted to reflect the appropriate shares of the vehicle fleet. These factors would ideally be created using MOVES to reflect locally specific information but could be developed based on national data or by estimating vehicle fuel economy for each type of vehicle and multiplying by appropriate carbon coefficients.

Option 3: Use look-up table accounting for fleet characteristics and speeds (vehicle operating characteristics)
This option is similar to Option 2 but goes further and includes look-up tables that show different emissions factors for different speeds, which can be used in combination with VMT data that is broken out by speed bin. The key advantage of this approach over the simpler methods is that speed is an important factor that affects GHG emissions. Consequently, using different emissions factors at different speeds is necessary in order to show the emissions effects of changes in traffic congestion and other strategies that affect vehicle speeds.

Option 4: Apply MOVES in Inventory Mode
VMT estimates are a required input to the MOVES model, 14 which can then provide estimates of total quantity of emissions for a given location and time.

Step 1: Convert transportation model output (or other VMT estimates) into MOVES data input format. VMT by vehicle and road types as well as the temporal distribution is characterized in five components:

EPA provides several spreadsheet tools to develop these inputs for county or multi-county scale modeling.15

MOVES Conversion Factors
MOVES requires annual VMT by vehicle type as an input. U.S. EPA provides a series of tools to convert inputs into the form needed by MOVES, including a spreadsheet tool that converts Annual Average Weekday Vehicles Miles Travelled (AADVMT) at the HPMS level into yearly VMT in MOVES' required input format, including by type of day, by month, and by MOVES source types. Default or user-supplied adjustment factors can be used.

Users would also need to map the transportation demand forecast model links to the MOVES road types.

Step 2: Generate other MOVES data requirements. Other data that would need to be collected for the model (if not relying on defaults) would include:

Step 3: Import data into the MOVES database. Entry of non-default data would be handled by the MOVES' County Data Manager tool at the county scale or the Data Importer at the national scale, similar to the process followed by nonattainment or maintenance areas when conducting SIP or conformity analyses. Much of both this step and step 2 can be avoided by running MOVES at the National scale, relying on national defaults, as described in EPA's MOVES GHG analysis guidance, albeit with decreased accuracy.

Step 4: Run MOVES model and estimate emissions inventory. Once the data inputs are prepared, the model can be run. The model estimates GHG emissions according to a variety of classifications, including vehicle type/class, fuel type, model year, pollutant, and others, as well as by process (running, start, refueling, and extended idle).

Option 5: Apply MOVES in Emissions Rate Mode
MOVES can also be used to estimate emissions rates: emissions per unit of activity. Use of MOVES in rates mode is generally more complex than using MOVES in inventory mode, but it may be more convenient for agencies that have already developed processes for using MOVES rates to generate estimates of emissions for other pollutants.

The basic activities included in MOVES are distance traveled and vehicle population. In the MOVES model, the variety of emissions processes (running exhaust, start exhaust, evaporative emissions, etc.) included are associated with one of these two types of activities for its reporting of emission rates. The resulting rates have units such as grams per mile (for running exhaust, for example) or grams per vehicle (for starting exhaust, for example). The emission rates produced by the MOVES model can be then applied to estimates of the associated activities to calculate total emissions. A typical instance would be coupling VMT estimates, such as from travel demand modeling, with running emission rates, produced with MOVES, to predict total running emissions. The same approach would be used to combine vehicle population estimates (the total number of vehicles) with the applicable emissions rates for start or extended idling to predict total emissions for these processes.

This analysis of emission rates, and subsequent calculations of total emissions, could be done using either post processors that may be specifically available for travel demand models or manual analyses using spreadsheet or database tools. More information on these technical and operational issues is available in the MOVES User Guide, and in the MOVES training materials found on EPA's website at:

Step 1: Collect MOVES Inputs. This step involves all the components listed above in Option 4, Steps 1 - 3,

Step 2: Run MOVES and estimate emission rates. The MOVES model produces running emission rates by road type, emissions process, and speed bin. In addition, the user can select to have MOVES produce rates that are disaggregated further, depending on available activity data. For example, most users would have VMT by vehicle type, and therefore should request output of rates by vehicle type. It is less likely that users would have VMT by fuel type or model year, but those choices are also available. Similar emission rates are also estimated for non-running emissions processes, generally replacing variations in speed with variations in temperature or hour.

Step 3: Disaggregate vehicle activity as needed and match road types between the models. This step first involves disaggregating vehicle activity, if necessary, so that the appropriate amount of activity can be multiplied by the MOVES-predicted emission factors. The second step is to match road types between the travel demand model and the MOVES outputs. For example, a typical approach could be to determine a total emissions factor for methane from all heavy-duty vehicles on highway links with an average speed of 65 mph. Proper disaggregation and matching ensures the MOVES emission factors represent the activity on the road type or for each link on the network.

Step 4: Estimate GHG emissions. Properly matched emission rates are then multiplied by the applicable VMT estimate to produce total running emissions. Similarly, non-running emission rates are multiplied by the appropriate vehicle population. This calculation of GHG emissions may be performed manually or the emissions factors can be loaded into a post processor of a travel model.


The examples in sections 5.1 and 5.2 show how different agencies have applied emissions factors. A variety of states and MPOs have used MOVES to generate emission factors that are then used to post process the output of their travel demand model. For example, Hillsborough County MPO (Tampa) has used MOVES GHG emission rates by speed to calculate emissions for long range transportation plan scenarios. The analysis was implemented in a postprocessor. A regional inventory of GHG emissions from transportation was developed by MWCOG. GHG estimates from mobile sources were calculated using data and forecasts of VMT by vehicle type from the air quality conformity analysis and by applying MOVES in inventory mode.

1 Emissions testing was first done in California in 1966 and now there are currently 32 states that require some form of vehicle emissions testing. Information on state I&M programs is available at:

2 The lack of non-residential data for statewide or MPO levels may be appropriate to avoid double-counting VMT between residential and non-residential uses.

3The HPMS data are helpful in that the data reveal travel patterns over time, though accuracy depends heavily on the resources of each state. More information about HPMS data is available at:

4 It may therefore be worthwhile for states to set GHG reduction targets/ policy in a way that future GHG reductions are tied to readily available starting points or model base years.

5 FHWA, "Sample Methodologies for Regional Emissions Analysis in Small Urban and Rural Areas," prepared by ICF International, October 2004.

6 Note that if a more complex emissions factor program such as MOVES is used, HPMS speed data are not required since these programs have built-in default speed distribution data. However, even though defaults are included in the models, local data are recommended if they are available.

7 As described above, many network-based travel models can output model-wide or district-wide VMT totals. However, these VMT totals do not typically have any speed information associated with them.

8 As was the case with speed data, if a practitioner plans to use an air quality emissions modeling software like MOVES, these programs typically contain a built-in estimate of VMT by vehicle classification and additional detail may not be necessary.

9 To remove the benefits of the 2008-2011 CAFE standards and the 2012-2016 National Program, the database was revised so that all energy rates beyond 2007 were the same for each vehicle type, model year and fuel type.

10 See EPA's GHG Guidance (Using MOVES for Estimating State and Local Inventories of On-Road Greenhouse Gas Emissions and Energy Consumption, EPA-420-D-12-001, January 2012).

11 MOVES models emissions at the national, county, and project scales. Statewide estimates may be developed with national-level inputs and default downscaling, although EPA does not recommend this method. Instead, state-level estimates could be developed by aggregating counties of interest. See EPA's guidance on this topic at:

12 See EPA's web page at: for the latest version of this guidance.

13 As noted above, defaults may be used for national scale runs at the state or county level; however EPA cautions users about using the national scale for reasons of accuracy in the downscaling factors.

14 For all but the national scale, at which they are a suggested input. Note that the national scale can be used to estimate emissions at smaller geographic levels such as a state (or states) or a county (or counties).

15 See for a list and additional information on the available tools.

Updated: 04/07/2014
HEP Home Planning Environment Real Estate
Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000