Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
A. Literature Review Findings Sections A.1-A.4 describe literature related to fleet, regional-level activity, project-level activity, and other MOVES inputs, respectively. Section A.5 provides an overview of surveys conducted by other organizations on the use of MOVES. A.1 FLEET DATA INPUTS This section discusses literature related to vehicle fleet inputs, including the distribution of vehicles by age and vehicle type (source type). Age Distribution Age distribution is a series of fractions that sum to 1, representing the percentage of vehicles that fall within each age category. MOVES requires these fractions for 31 age categories (bins) that each contain one year of vehicles, except for the final category, which represents all vehicles 30 years old and older. MOVES requires a separate age distribution for each of 13 MOVES source (vehicle) types. This information is used in emissions analyses because older vehicles typically have higher emission rates due to older technology in those vehicles as well as deterioration. This input was required with MOBILE6, but had a slightly different format. [5] Chatterjee and Miller (1994). The study identifies numerous data sources that can be used for developing model inputs, including age distribution profiles for fleets. The study discusses the use of vehicle registration data and inspection and maintenance records, both of which can be used to obtain age distributions. [20] Malcolm et al. (2003). In this study, vehicle activity and vehicle fleet data were collected in the South Coast Air Basin in southern California. Average traffic speed, density, flow rates, and license plate data were captured with a digital video camera and subsequently analyzed using vehicle registration databases and VIN decoders. This method could be used in an expanded area to establish vehicle age distribution. [14] FHWA (2009). The information collected in the U.S. DOT 2009 National Household Travel Survey included vehicle model year, which may be used as an existing data source to help build age distribution profiles for light-duty vehicles. Locality-specific profiles could be developed for states and for major metropolitan areas, since records are associated with individual metropolitan areas if they are located in the 50 largest areas. It may be possible to develop locality-specific profiles for other areas in states which paid for add-on samples, but this would require working with the state NHTS data coordinator to obtain data coded with the needed geographic identifiers (e.g., county). A-1
[15] FTA (no date). The Federal Transit Administration annually collects data from transit operators to populate the National Transit Database (NTD). The database includes number of vehicles by type and age for each reporting transit operator, so age distributions could be developed for the regional public transit vehicle fleet. Source Type (Vehicle) Population Source type (vehicle) population is simply the total number of vehicles for each of the 13 MOVES source (vehicle) types, as shown in the example in Table A.1. It is a new input for MOVES that was not required under MOBILE6 and is used to calculate nonrunning emissions, such as vehicle starts and evaporative emissions. It can often be obtained from the same vehicle registration data that is used for age distributions, but there are also other methods of creating this input based on the travel activity of each vehicle type in a county. Table A.1 Example Source Type (Vehicle) Type Population Input yearID sourceType sourceTypeID sourceTypePopulation 2002 Motorcycle 11 1,622 2002 Passenger Car 21 39,398 2002 Passenger Truck 31 21,126 2002 Light Commercial Truck 32 7,058 2002 Intercity Bus 41 29 2002 Transit Bus 42 18 2002 School Bus 43 201 2002 Refuse Truck 51 27 2002 Single Unit Short-Haul Truck 52 1,413 2002 Single Unit Long-Haul Truck 53 113 2002 Motor Home 54 300 2002 Combination Short-Haul Truck 61 354 2002 Combination Long-Haul Truck 62 316 [5] Chatterjee and Miller (1994). This study suggests using vehicle counts, vehicle registration data, and inspection and maintenance records for developing vehicle population data for use in emissions inventory development. [14] FHWA (2009). The 2009 NHTS collected vehicle information, including vehicle type. This data may assist a user in supplementing vehicle population data on light-duty vehicles, in particular, factoring total light-duty vehicle population by type. However, this source does not cover commercial vehicles and also will not reflect through traffic. The NHTS sample size is probably adequate for state-level analysis and possibly for the 50 largest MSAs. Survey A-2
data might also be usable for other geographic subareas (e.g., counties or aggregations of counties) in states with oversamples. [8] Cohen and Chatterjee (2003). The purpose of this report is to use available data and information to develop an improved understanding of the magnitude and spatial/temporal distribution of different types of commercial vehicle travel. The study compared different data sources, concluding that there are significant discrepancies among the available data sources, and that some data sources are useful to answer one particular question, but other sources were needed to answer other questions. The study highlights the challenges faced with developing source type population data for use in MOVES using existing data sources. While the study is based on reviews of a multitude of sources, the following were specifically identified: ⢠Commercial vehicle survey data (Detroit, Atlanta, Denver, and the Piedmont- Triad area); ⢠California Department of Motor Vehicle data (Los Angeles, San Francisco, San Diego, and Sacramento); ⢠National Transit Database (198 cities in the United States); ⢠United Postal Service data; ⢠School bus fleet surveys (largest 100 school districts); ⢠Taxi Fact Book (all major U.S. cities); ⢠Airport Ground Access Planning Guide (27 U.S. cities); and ⢠Vehicle Inventory and Use Survey (VIUS). [19] Lindhjem and Shepard (2007). This report analyzes data compiled by FHWA from the Highway Performance Monitoring System, weigh-in-motion sensors, and other data sources (visual observation, weigh stations, and other special projects) with the objective of improving heavy-duty vehicle modeling capability. Vehicle weights and mix of vehicle classes are investigated depending upon a number of regional and temporal factors by vehicle and roadway types. The methods described in this study can be used to develop vehicle characteristics as well as weight and class fractions of the in-use heavy- duty vehicle fleet, which are directly applicable for source type population development for MOVES modeling. [20] Malcolm et al. (2003). The video data collection method used in this study is applicable for collecting source type population data. Digital cameras are used to capture license plate data, which is matched to registration data using VINs from which specific vehicle information may be obtained and classified through postprocessing. However, this approach may not be practical for all road types (e.g., off-network and rural, where traffic volumes are low). [4] Caltrans (no date). This web site describes the use of in-pavement weigh-in- motion sensors for collecting truck data, as well as an extensive existing dataset A-3
from California from which heavy-duty vehicle source type population data could be developed. Vehicle weight, length, and classification data are collected. California data may or may not be representative of other areas, but the method of data collection could offer a viable source for data development in other areas with WIM programs. However, these programs typically only sample vehicles on major highways. A.2 REGIONAL ACTIVITY INPUTS Regional VMT by Vehicle Class This input (called HPMSVTypeYear within MOVES) requires annual VMT for each of six HPMS vehicle types, as shown in Table A.2. It may be constructed from two separate factors, the annual VMT for all vehicle types and the percent distribution of VMT among the six HPMS vehicle types. Since MOBILE6 was an emission rate model and did not have an inventory mode option it did not require annual VMT (although it would be necessary information to create an emissions inventory after the MOBILE6 run). MOBILE6 had distribution of VMT among vehicle types programmed into it and did not request this information as an input. VMT by the six HPMS vehicle classes may be obtained directly from the HPMS at a county or state level. Forecasts may be developed using travel demand model estimates of total VMT (or light and heavy-duty VMT) factored by the fraction of current VMT by vehicle type from the HPMS. Other sources may also be used, especially to refine estimates for specific vehicle types, as described below. Table A.2 Example Region VMT by Vehicle Class Input HPMS Vehicle Type HPMSVtypeID yearID HPMSBaseYearVMT baseYearOffNetVMT Motorcycle 10 2002 165,354 0 Passenger Car 20 2002 35,332,808 0 Light Truck 30 2002 20,086,413 0 Buses 40 2002 383,910 0 Single Unit Truck 50 2002 1,934,527 0 Combination Truck 60 2002 4,240,177 0 Chatterjee and Miller (1994). This report (which was developed considering MOBILEâs need for VMT distributions by a set of types not consistent with the HPMS classes) describes a number of methods that can be used to develop VMT distributions by vehicle class. HPMS and other classification counts can be used to determine VMT by road type and vehicle type. Network-based travel demand models can be used to determine VMT by road type, and to a limited extent vehicle type. Vehicle registration data and inspection and maintenance records A-4
(in areas where such programs exist) can be used to determine VMT distributions by vehicle type (based on total registrations by type and possibly annual mileage driven) although the applicability of these sources to heavy-duty vehicles may be limited. [15] FTA (no date). The NTD includes annual mileage by vehicle type and age for each reporting transit operator in the U.S., so VMT could be estimated for the regional public transit vehicle fleet. [8] Cohen and Chatterjee (2003). The purpose of this report is to use available data and information to develop an improved understanding of the magnitude and spatial/temporal distribution of different types of commercial vehicle travel. The report reviews a number of data sources on commercial vehicle activity as previously noted. Some of these might be used to support the development of VMT inputs, e.g., by disaggregating heavy-duty VMT by source type. [10] Farzaneh, et al. (2011). This paper describes the use of the FHWA Freight Analysis Framework, a commodity origin-destination database, in conjunction with highway network routing procedures, truck registration data, and MOBILE6.2 emission factors to estimate the air quality impacts of freight movement in a multistate corridor. The FAF data and assignment method can be used to help estimate freight truck VMT on major highway links, for state-level studies as well as interregional corridor studies. FAF is also a useful tool for forecasting truck VMT as it provides forecasts of future freight flows. However, the FAF database is limited primarily to heavy trucks carrying freight over longer distances and is not a complete accounting of all truck sources. Temporal Adjustments Temporal adjustments include month, day, and hour VMT fractions, i.e., the fraction of total annual VMT that occurs in a given month, the fraction of total monthly VMT that occurs on weekdays versus weekends, and the fraction of total daily VMT that occurs in a given hour, respectively. In all three cases MOVES requires these fractions for each of the 13 MOVES source types. For day and hour VMT fraction it also requires them by the four MOVES road types. [19] Lindhjem and Shepard (2007). This report analyzes data compiled by FHWA from HPMS traffic monitoring and other sources on heavy-duty vehicle activity. The study developed temporal profiles by month, day of week, and time of day. These inputs are directly related to MOVES modeling. The data analysis approach may also be useful to users in developing state or locally specific temporal adjustments to MOVES activity. [20] Malcolm et al. (2003). The video data collection method used in this study would allow a user to develop reasonable temporal adjustments for the applicable roadway types and source types. However, the method is probably more costly than using existing traffic counters for developing temporal adjustments, unless video cameras already are being used to collect traffic data A-5
for other purposes across a broad enough sample of roads to be representative of regional conditions. [2] Boriboonsomsin, Sheckler, and Barth (2012). This study describes the use of wireless communication or telematics technology, increasingly adopted by the fleet management industry, to collect a variety of data on heavy-duty vehicle operations. The dataset used in this study comes from a collective fleet of more than 2,000 Class 8 trucks traveling across the U.S. for the entire year of 2010. The data were used to observe temporal patterns among other things. However, the data source is currently applicable only to a limited set of participating fleets, which may or may not be representative of the entire heavy vehicle population. The method may have broader applicability if telematics technology is employed on a wider range of vehicles. [1] Bar-Gera (2007). This study validated GPS data on traffic speeds and travel times collected using cellular phones against data collected using dual magnetic loop detectors and floating cars. The cellular phone-based data collection method has direct applicability for developing temporal profiles, since vehicle activity would be tracked at all hours of the day. However, since the data are anonymous, it would not be possible to develop temporal distributions specific to each source type. This could be done in theory of a set of cell phone users who would be willing to identify the type of vehicle they are using could be recruited. [18] Lee et al. (2011). This study used portable emissions monitoring systems (PEMS) to track activity for refuse trucks. PEMS could be used to develop temporal profiles for specific types of vehicles when PEMS are installed for emissions monitoring purposes. [17] Hatzopoulou and Miller (2010). A micro-simulation activity-based travel demand model for the Greater Toronto Area was extended with capabilities for modeling and mapping of traffic emissions and atmospheric dispersion. Hourly link-based emissions and zone-based soak emissions were estimated. This study makes use of advanced travel demand forecasting model capabilities; while the approach is useful for developing temporal adjustments for MOVES emissions modeling, considerable model enhancement work would be required in most regions which do not have similar time of day modeling capabilities. Road Type Distribution This input is a set of five fractions that sum to 1, which represents the percent of VMT on each of five road types used in MOVES. These road types are off- network, rural restricted access, rural unrestricted access, urban restricted access, and urban unrestricted access. MOVES requires this distribution for each of the 13 source (vehicle) types. Table A.4 shows an example of this input for one type of vehicle (this would be repeated for each of the 13 vehicle types). This was an input in MOBILE6, but different roads types were used (highway, arterial/ collector, local, ramp). This information is required for emissions modeling because the traffic conditions on different types of roads affect the emission rates. A-6
For example, arterials and local roads have more stop and go conditions with more acceleration/deceleration patterns than highways. This information is usually fairly easy to obtain from a travel demand model that has links coded as different roadway types or from HPMS, although travel demand models would typically only provide a single distribution (rather than one for each vehicle type), and the six HPMS vehicle classes need to be mapped to 13 MOVES classes. Table A.3 Example Road Type Distribution Input Road Type roadTypeID roadTypeVMTFraction Off-Network 1 0 Rural Restricted Access 2 0.166473 Rural Unrestricted Access 3 0.24494 Urban Restricted Access 4 0.235751 Urban Unrestricted Access 5 0.352836 [5] Chatterjee and Miller (1994). This study identifies numerous existing data sources that can be used for developing model inputs, including road type distributions for fleets. The vehicle miles traveled section in Chapter 4 describes sources of VMT on different types of roads, such as continuous counts, short- term counts, HPMS, and travel demand models. Possible sources of error for each of these are also discussed. A method for improving estimates of VMT on local roads is provided in Chapter 8. Although not discussed directly in this document, once VMT is developed for various roadway segments from one of these, it can be used to develop road type distributions since the road type of each segment is known. [2] Boriboonsomsin, Sheckler, and Barth (2012). The telematics-based dataset used in this study was used to determine road type distributions for the instrumented truck fleets, among other things. The method presented in this study offers a very accurate data collection method that already is being used in commercial fleets. However, the existing data set is currently limited to participating fleets, primarily Class 8 trucks. [19] Lindhjem and Shepard (2007). This report analyzes data compiled by FHWA from HPMS traffic monitoring and other sources on heavy-duty vehicle activity. HPMS traffic count data can be used to develop distributions of activity by vehicle type and road type, for roads that are part of the HPMS network. However, appropriate expansion factors need to be used to ensure that monitored locations are representative of the entire road network. [1] Bar-Gera (2007). This study validated GPS data on traffic speeds and travel times collected using cellular phones against data collected using dual magnetic loop detectors and floating cars. The cellular phone-based data collection method has direct applicability for developing road type distributions, by A-7
associating vehicle trajectories with road links and types. However, since the data are anonymous, it would not be possible to develop road type distributions specific to each source type. [18] Lee et al. (2011). This study used PEMS to track the activity for refuse trucks. PEMS data can be mapped, using the GPS coordinates to activity on specific roadways, allowing a user to develop a good road type distribution for vehicles that may be instrumented for emissions monitoring purposes. Average Speed Distribution This input is a set of 16 fractions that sum to 1, which represents the distribution of vehicle-hours traveled among 16 speed bins, as shown in Table A.5. MOVES requires this information for every combination of 13 source (vehicle) types, four road types, 24 hours of the day, and two types of days (weekdays/weekends). Therefore, it effectively asks for 2,496 distributions. Typical sources for this data include travel demand model output and possibly some observed data. Table A.4 Example Average Speed Distribution Input avgSpeedBinID avgBinSpeed avgSpeedBinDesc avgSpeedFraction 1 2.5 speed < 2.5 mph 0.0000000 2 5 2.5 mph ⤠speed < 7.5 mph 0.0000000 3 10 7.5 mph ⤠speed < 12.5 mph 0.0000000 4 15 12.5 mph ⤠speed < 17.5 mph 0.0000000 5 20 17.5 mph ⤠speed <22.5 mph 0.0000000 6 25 22.5 mph ⤠speed < 27.5 mph 0.0000000 7 30 27.5 mph ⤠speed < 32.5 mph 0.0212651 8 35 32.5 mph ⤠speed < 37.5 mph 0.0027255 9 40 37.5 mph ⤠speed < 42.5 mph 0.0000000 10 45 42.5 mph ⤠speed < 47.5 mph 0.0000000 11 50 47.5 mph ⤠speed < 52.5 mph 0.0000000 12 55 52.5 mph ⤠speed < 57.5 mph 0.0000000 13 60 57.5 mph ⤠speed < 62.5mph 0.3890143 14 65 62.5 mph ⤠speed < 67.5 mph 0.0590208 15 70 67.5 mph ⤠speed < 72.5 mph 0.5279743 16 75 72.5 mph ⤠speed 0.0000000 [31], [34] U.S. EPA (2010c, 2012c). These guidance documents on the use of MOVES2010 for emission inventory preparation and GHG estimation note that the recommended approach for estimating average speeds is to post-process the output from a local travel demand network model. Speed results from A-8
most travel demand models must be adjusted to properly estimate actual average speeds. However, the documents do not offer any guidance on how the postprocessing can be done. [5] Chatterjee and Miller (1994). This study identifies numerous existing data sources that can be used for developing model inputs, including speed distributions for fleets. Chapter 5 describes how travel demand models are the most common method for estimating speeds for air quality planning. It further describes how speeds developed as part of the traffic assignment step are mainly used to achieve realistic traffic flows and are really too crude to be used for air quality planning. It describes a process of using link volumes and capacities along with a standard function such as Bureau of Public Roads or Highway Capacity Manual formulas to obtain more accurate link speeds and vehicle hours traveled. This process can be applied within the travel demand model or as a post processing step. Chapter 8 discusses methods to improve speed estimation, such as improving underlying variables like traffic volume and roadway characteristics. Intelligent Transportation Systems are also mentioned as an emerging and more affordable data source for continuous speed and volume data over a wider area. Finally, modal emissions models are mentioned as a way to eliminate the need for speed inputs all together and move to what many researchers believe is a more accurate way to estimate emissions. [13] FHWA (2006). This FHWA guidance on transportation conformity notes that since emissions estimates are extremely sensitive to vehicle speed, EPA and DOT recommend that speeds be estimated in a separate step after traffic assignment, using refined speed-volume relationships and final assigned traffic volumes. Postprocessed speeds estimated in the validation year should be compared with speeds empirically observed during the peak- and off-peak periods. These comparisons may be made for typical facilities, for example, by facility class/area type category. Based on these comparisons, speed-volume relationships used for speed postprocessing should be adjusted to obtain reasonable agreement with observed speeds. Regardless of the specific analytical technique, every effort must be made to ensure that speed estimates are credible and based on a reproducible and logical analytical procedure. [26] TMIP (2008). This synthesis paper discusses data collection methods used to develop speed inputs to travel demand models, as well as the algorithms used within travel demand models to estimate and forecast travel speeds at a link level, based on volumes, capacity, etc. Collecting of speeds for different time periods, days of the week, and in different area types is discussed. Collection methods discussed include: ⢠Floating car method to collect point-to-point travel times; ⢠Freeway loop detector data; ⢠Toll transponders; A-9
⢠GPS devices; and ⢠Video technology. [27] TMIP (2009). This study provides a synthesis of the derivation of speed adjustments in travel demand forecasting models. The study discusses different approaches to volume-delay functions (VDF), including 1) applying a single volume-delay formulation for all facility types; 2) applying unique user specified VDF functions developed for each facility type and possibly area type in the network; and 3) developing unique user specified VDF functions to account for delay at signalized intersections. Speeds that are produced by the travel models need to be postprocessed and refined to produce more realistic network link- specific values for use in mobile source emission modeling. The information presented here is relevant to developing average speed inputs to MOVES based on travel demand model data. [21] Miller (2002). The purpose of this study was to identify or develop a prototype postprocessor that VDOT staff could use to determine vehicle speeds for the purposes of conducting air quality conformity analysis using the MOBILE model. The postprocessor converts 24-hour link VMT to hourly volumes within each period, divides each link volume by the linkâs capacity, uses this ratio with a simple formula to estimate a link speed for each of the three periods, and then computes VMT and VHT for each link and for each period. The postprocessor then aggregates link-specific volumes, speeds, VMT, and VHT by period and facility type. The postprocessor reports total volume, total VMT, average speed, and total VHT by facility type and within that category by time period (AM peak, PM peak, and off-peak). The authors suggest that further validation is required. This method may hold promise for future use, but until validated, other sources of data may be preferable for developing speed distribution data for MOVES modeling. [7] Choe, Skabardonis, and Varaiya (2001). The Freeway Performance Measurement System is a freeway performance measurement system (PeMS) for all of California that processes loop detector data. The PeMS system can help establish speed distribution data on California freeways. Similar methods could be applied to loop detector data in other areas to identify local speed distributions. However, the data are limited to freeways instrumented with loop detectors. [3] Boriboonsomsin, Zhu, and Barth (2011). This study presents a statistical method for estimating truck traffic speed that takes advantage of existing traffic monitoring systems: the California PeMS, California weigh-in-motion stations, and the Berkeley Highway Laboratory (BHL). PeMS provides data on total flow, truck flow, and overall traffic speed, but not truck traffic speed. WIM stations provide data for truck flow and truck traffic speed, but not total flow and overall traffic speed. BHL collects data for all four required variables, but is limited to only a few miles of freeway. No single data source is sufficient for providing truck speed data for the entire state. This study developed methods to combine A-10
the data from these three sources to produce statewide truck speed estimates on freeways within California. The analysis showed that with traffic data from these systems, traffic speed can be effectively estimated on the basis of the knowledge of the overall traffic speed alone. [1] Bar-Gera (2007). Data from GPS-equipped cell phones, if obtained for a specific geographic area and associated with network links, can be used to develop speed distributions by road type, hour of the day, and day of week. [22] NREL (no date). The Secure Transportation Data Project is assembling detailed GPS data from household travel surveys, which contains information on trajectories of the instrumented vehicles. GPS data could be used to validate speed results from speed post processors. The dataset is currently limited to survey data from five metropolitan areas. [2] Boriboonsomsin, Sheckler, and Barth (2012). Second-by-second data collection using GPS-equipped heavy vehicles creates a very accurate speed distribution profile for participating fleets. However, this method is limited to participating fleets which may or may not be representative of the entire heavy- duty vehicle population. [20] Malcolm et al. (2003). Video data collection methods were tested and compared against instrumented vehicle methods for collecting speed distribution data. The instrumented vehicle method is commonly used for average speed measurement. The video method, while capturing all vehicles on a road segment, only collects speed profiles at single points on the highway network. Multiple measurement points, and calibration from instrumented vehicles, would probably be needed to capture segment or corridor-level average speeds. [18] Lee et al. (2011). Instrumentation of vehicles with PEMS can provide a highly accurate method for collecting speed distribution data, but is limited to the vehicles with PEMS installed in them. [24] Song and Yu (2011). This study uses large samples of second-by-second floating car data, collected from expressways in Beijing, to associate the VSP distributions with the average travel speed from 0 to 20 km/hr. While the data itself may not be relevant to U.S. conditions, second-by-second data collection, as seen in this study as well as others, is a very valuable tool in establishing activity profiles, including speed distribution. A.3 PROJECT-LEVEL INPUTS Project-level inputs in MOVES must be entered for each roadway link that is used to define a project. Therefore, while some of the inputs may appear to be similar to regional inputs, the difference is that at the project-level inputs can be varied for each individual link. A-11
Project-Level Link Activity Project-level link activity may be described using operating mode distributions, drive schedules, or average speed. A number of resources that generally address link activity are discussed first, and then sources specific to each method are discussed. General Resources [29] U.S. EPA (2010a). This document provides guidance on data sources and preparation methods for all MOVES inputs required for project-level hot spot analysis for carbon monoxide. Appendix A has guidance on estimating vehicle activity via three options: 1) average congested speeds, using appropriate methods based on best practices for highway analyses (for example, as described in TMIP publications and methodologies for computing intersection control delay provided in the Highway Capacity Manual); 2) link drive schedules; and 3) operating mode distributions. The guidance notes that for both free-flow highway and intersection links, users may directly enter output from traffic simulation models in the form of second-by-second individual vehicle trajectories. The U.S. EPA guidance for PM Hot Spot Analysis [30] also identifies the same three options for developing vehicle activity. [10] Eastern Research Group (2010). This study presents a procedure for developing MOVES inputs from link-level travel demand forecasting model output (speeds and VMT). It assumes that the speeds already have been processed into MOBILE classes. The study also presents a âmappingâ of travel demand forecasting model roadway types to HPMS roadway types. [24] Song and Yu (2011). This study uses large samples of second-by-second floating car data, collected from expressways in Beijing, to associate the VSP distributions with the average travel speed from 0 to 20 km/hr. Second-by- second data collection is directly applicable to link-level data development. The data collected in this study is directed only at low-speed vehicles, and under conditions in China. However, the method used here can be used effectively to develop several inputs to MOVES on a broader application. Operating Mode Distributions (Including Drive Schedules) These two MOVES inputs rely on similar data, but provide it to MOVES in a different format. One or the other should be input into MOVES. If operating mode distributions are chosen the input would be a set of fractions, which represent the percent of time spent in each operating mode bin (a combination of VSP and speed). If drive schedules are chosen, the input would be a set of second-by-second speeds. [30] U.S. EPA (2010b). This document, which provides guidance on PM hotspot analysis for conformity, contains general recommendations for the development of operating mode distributions. Operating mode distributions may be obtained from other locations with similar geometric and operational (traffic) A-12
characteristics, or output from traffic microsimulation models. For example, chase (or floating) cars, traffic cameras, and radar guns have been used previously to collect some traffic data for use in intelligent transportation systems and other applications. Users should consider the discussion in Section 4.2 when deciding on the appropriate activity input. The MOVES model is capable of using complex activity datasets with high levels of resolution to calculate link-level emissions. EPA encourages the development of validated methods for collecting verifiable vehicle operating mode distribution data at locations and in traffic conditions representative of different projects covered by this guidance. However, the user should determine the most robust activity dataset that can be reasonably collected while still achieving the goal of determining an accurate assessment of the PM air quality impacts from a given project. The decision to populate the Links table, Link Drive Schedule, or Op- Mode Distribution should be based on the data available to the user and should reflect the vehicle activity and behavior on each link. [9] Dowling et al. (2005). This study developed a complete modeling framework for analyzing the regional impacts of traffic flow improvements. The research produced a set of relationships through microscopic simulation that determine the proportion of the time spent on a network link in a driving mode as a function of the linkâs type. The link type was based on typical link classifications employed in planning and operational studies (facility types) and on key design/operational characteristics (e.g., number of lanes, free-flow speed, and signal spacing). Thirty-three link types were selected. The relationships were developed through processing of simulated vehicle trajectories using the INTRAS (predecessor of FRESIM). These data could be reprocessed to produce default operating mode distributions for MOVES. [25] Song, Yu, and Zhang (2012). This study examined the effect of using vehicle trajectories from microscopic simulation models as a basis for establishing operating mode distributions. The study concluded that the traditional approach of integrating traffic simulation models with emission models was not applicable for vehicle emission estimations â large errors were reported between the operating mode distributions of simulated versus real-world data. The study did not directly compare actual driving conditions to a simulation of those driving conditions â rather the âreal worldâ data were collected independently, making the background traffic conditions incomparable. Even if the research findings can be validated, microscopic models may still be useful for the relative comparison of project impacts. [22] NREL (no date). The Secure Transportation Data Project is assembling detailed GPS data from household travel surveys, which contains information on trajectories of the instrumented vehicles. GPS data could be a used to validate simulation model output and/or as a direct source of vehicle trajectories for specific routes for specific times of day, although there would be concerns about how representative it would be due to small number of GPS samples in most travel surveys to date. A-13
[11] E.H. Pechan and Associates and Cambridge Systematics, Inc. (2010). This report develops MOVES operating mode distributions from microscopic simulation model outputs for various conditions of congestion, including volume to capacity (v/c) ratio and incident characteristics. It creates operating mode distributions from the trajectories of every vehicle in the simulation network for every link. The relationships could be linked to traffic demand models at either a regional or project-level scale, providing the use of trajectory-based operating mode distributions rather than average speed inputs for MOVES analysis. The linkage is provided through the estimation of v/c â each v/c level for a given link type corresponds to different operating mode distribution. [20] Malcolm, et al. (2003). Vehicle activity and vehicle fleet data were collected in the South Coast Air Basin in southern California. Vehicle activity was characterized primarily using a large second-by-second speed and acceleration data set collected from probe vehicles operated within the flow of traffic. The average speeds represented in the data were very high, even for weekdays (57.1 mph). Although the speed-acceleration distributions are different for the three locations, the primary factor influencing the differences in distribution appears to be the level of congestion. This finding must be kept in mind if default trajectories are developed for MOVES input. [23] Papson, Hartley, and Kuo (2012). This analysis estimates emissions using a time-in-mode (TIM) methodology, which allocates vehicle activity time to one of four modes: acceleration, deceleration, cruise, and idle. This is somewhat of an in-between methodology that is more advanced than providing only average speed, but less data-intensive than providing full operating mode distributions. The TIM was based on HCM methods â assumptions were used as to what percent of vehicles were in each TIM based largely on control delay. The method could be useful for project-level analysis of intersections where the available data allows an HCM analysis. [6] Chamberlin, Swanson, and Sharma (2012). This paper uses a signal optimization project at an intersection as an example project to demonstrate a quantitative PM hot spot analysis using MOVES at the project level. The study describes how calculating operating mode distributions for input into MOVES centers on calculating the VSP for each vehicle at each second using second by second speed information from a VISSIM model. A separate operating mode distribution is calculated for every link in the VISSIM model, which correspond to the links provided to MOVES. [37] Xu et al. (2012). This paper develops a model based on real world data to predict operating mode distributions based on average speed. Applicability is limited since the paper was mainly concerned with testing the average speed input approach. However, the paper is useful in describing collection of the real world data on 238 km of expressways in Beijing using 46 GPS-equipped light- duty vehicles. A-14
Average Speed This input simply asks for the average speed (in mph) on each link in the project. It would be used if operating modes distributions or drive schedules are not available. Note: A number of the references for average speed distributions for regional analysis may also be relevant to project-level analysis. For example, some EPA and FHWA guidance documents (FHWA 2006, EPA 2010b and 2012) on conformity and emissions inventories address the postprocessing of link-level speed data from travel demand models, which may be used at the project level as well as the regional level. [28] U.S. EPA (1999). This guidance for the development of facility type VMT and speed distributions is geared to MOBILE inputs, but is still a good starting point for the current NCHRP research. Speed estimation procedures are based on BPR equations and HCM methods using data from traffic counts or travel demand forecasting models. [30] U.S. EPA (2010b). For average speeds, this guidance on PM hot spot analysis recommends that project sponsors determine average congested speeds by using appropriate methods based on best practices used for highway analysis. The document notes that resources are available through FHWAâs TMIP program and that methodologies for computing intersection control delay are provided in the Highway Capacity Manual. [21] Miller et al. (2002). This report provides a useful exploration of several methods for developing postprocessed speeds from travel demand forecasting model output. [36] Vallamsundar and Lin (2012). This study provides insight to the process with respect to data input preparation, sensitivity testing, and MOVES model setup for PM hot spot analysis. Average speed was used as the input for vehicle activity â emission factors were developed for 5 mph increments but were reduced to 1 mph increments for some speed ranges, based on sensitivity tests (21 speed ranges in all were used). Applicability is limited, but the guidance on speed ranges is appropriate if emissions factors are to be developed in this way. Off-Network Data Off-network data includes start fraction, extended idle fraction, and parked fraction. [2] Boriboonsomsin, Sheckler, and Barth (2012). The GPS-based method presented in this study offers a very accurate data collection method that already is being used in commercial fleets. The second-by-second data collection allows a user to analyze multiple variables applicable to MOVES modeling, including start fractions, extended idle fractions, and parked fractions. However, this method is limited to participating fleets and data may or may not be representative of the entire heavy-duty vehicle population. A-15
[18] Lee et al. (2011). Second-by-second data collection using PEMS- instrumented vehicles allows a user to develop very detailed and accurate activity profiles for the instrumented fleet. Inputs can be developed at a very fine level of granularity, making it directly applicable to the development of link- level data, start and extended idle fractions, parked fractions, and off-network activity development, among others. However, a special data collection effort is required with instrumented vehicles that are representative of the fleet as a whole. Link Characteristics Link characteristics include traffic volume, length, and grade. [29], [30] U.S. EPA (2010a, 2010b). These guidance documents for hot spot analysis recommend that the links specified for a project include segments with similar traffic/activity conditions and characteristics. For example, at an intersection EPA recommends defining cruise, queuing/deceleration, and acceleration links. The link length will depend on roadway geometry, how the geometry impacts activity, and the level of detail available in the activity data. Appendices D and E to [29] provide examples of determining link lengths for intersection and highway projects respectively. Traffic volumes and road grade are assumed to be available from the project sponsor and no other guidance is provided by EPA on these inputs other than to give the units for volume as vehicles/hour and for grade as percent grade. [6] Chamberlin, Swanson, and Sharma (2012). This study describes examples of project-level MOVES analyses, which provide insights on data sources and methods for link characteristics. The links defined in simulation modeling should consider the needs in MOVES, i.e., shorter links closer to the intersection center will capture the greater emissions generated. Volumes were obtained from a simulation model of an intersection and the grade was assumed to be zero based on observation of the intersection. [36] Vallamsundar and Lin (2012). This study conducted an analysis of a freeway interchange where the authors assumed that links are defined as individual ramps and freeway links before and after intersecting ramps. The authors obtained traffic volumes for these links from the Illinois DOT. Some ramps are described in the paper as âinclined ramps,â but no information was provided on data sources or methods for determining the road grade. Link Source Types This input asks for a set of 13 fractions that sum to 1, which represents the percent of vehicle hours traveled by each of the 13 source (vehicle) types on a particular link. [30] U.S. EPA (2010b). This guidance on PM hotspot analysis notes that the user needs to ensure that the source types selected in the MOVES Vehicles/ Equipment panel match the source types defined through the Link Source Type. A-16
For projects that will have an entirely different source type distribution than that of the regional fleet, the preferred option is for the user to collect project-specific data. If the project traffic data suggests that the source type distribution for the project can be represented by the distribution of the regional fleet for a given road type, the user can provide a source type distribution consistent with the road type used in the latest regional emissions analysis. [20] Malcolm et al. (2003). Vehicle activity and vehicle fleet data were collected in the South Coast Air Basin in southern California. Results showed significant spatial and temporal differences in vehicle activity patterns and vehicle fleet characteristics in different areas. For example, the report found that the average percentage of heavy-duty trucks in the freeway vehicle fleet varies from 2.35 percent in Riverside to 7.45 percent in Yorba Linda. [35] University of California at Riverside (pending publication). This report describes a survey of five MPOs to identify the state of practice in data collection for mobile source inventories. One chapter describes a license plate survey in Las Vegas. For regional emissions, it shows that in-state registration data alone may not be sufficient in certain areas, such as those with high tourist activity (e.g., Las Vegas), transportation hubs, and near state borders. The report provides some details on the methods (automated versus manual) and costs involved in collecting license plate data. Other chapters will address heavy-duty truck activity and emissions data. The study recommends using the vehicle license plate survey technique in conjunction with a vehicle registration database and VIN decoder to obtain highly resolved and area-specific vehicle fleet data. A.4 OTHER INPUTS Meteorology Inputs Meteorology inputs include temperature and humidity for each of the 24 hours of the day. MOVES requests a set of temperatures and humidity for each month being modeled in that particular MOVES run. For example, ozone runs would usually be for the month of July or for the summer months of June, July, and August. This input is similar to the meteorology inputs required with MOBILE6. [32] USEPA (2012a). The MOVES User Guide describes how to use the Meteorology Data Importer to import temperature and humidity data for months, zones, counties, and hours. The MOVES model contains 30-year average temperature and humidity data for each county, month, and hour. [33] USEPA (2012b). Local temperature and humidity data are required inputs for SIP and regional conformity analyses using MOVES. Ambient temperature affects most pollutant processes and relative humidity is important for estimating NOx emissions. EPA has provided tools to help develop the meteorological inputs to MOVES. The guidance cites the National Climatic Data Center as a good source for meteorological data. The document provides A-17
guidance for using MOVES in emission inventory mode versus emission rate mode. [16] Fincher et al. (2010). This study presents a road map for developing emissions inventories using MOVES. The study describes the options for developing meteorological inputs for MOVES, including using the existing template provided by MOVES. MOVES requires hourly temperature and relative humidity data. In this study, these inputs were taken from the MOBILE6 input files. Inspection and Maintenance Programs The I/M program input in MOVES is similar to the one found with MOBILE6 except that the information is provided in a tabular format instead of through a text input file, as shown in Table A.6. The table requests information on testing standards that impact different pollutants and emissions processes for various model years of different vehicle types. MOVES provides a default table for every county in the country, but EPA states that this should be used as a starting point only and that local information on individual programs is required for official SIP and conformity analyses. Table A.5 IM Program Inputs po lP ro ce ss ID st at eID co un ty ID ye ar ID so ur ce Ty pe ID fu elT yp eID IM Pr og ra m ID in sp ec tF re q te st St an da rd sID be gM od elY ea rID en dM od elY ea rID us eIM yn co m pl ian ce Fa ct or 101 47 47065 2015 21 1 1 1 11 1975 1995 Y 98.01 102 47 47065 2015 21 1 1 1 11 1975 1995 Y 98.01 201 47 47065 2015 21 1 1 1 11 1975 1995 Y 98.01 202 47 47065 2015 21 1 1 1 11 1975 1995 Y 98.01 [32] U.S. EPA (2012a). As described in the User Guide, MOVES allows a user to specify the level of compliance and general effectiveness of an I/M program. Currently, MOVES does not include I/M emissions effects for diesel fuel or continuous I/M inspection frequency. The compliance factor reflects the program performance metrics such as waiver rates, exemptions, special training programs, and general effectiveness. The compliance factor is a function of pollutant-process, location, source type, model year range, fuel type, and specific I/M test types. MOVES allows the user to turn certain aspects of an I/M program on and off to aid in what-if analyses. Only one I/M program description record can apply to each pollutant-process, source type, fuel type, and model year combination, and program choices are restricted to applicable model years only. A-18
[33] U.S. EPA (2012b). The guidance for using I/M programs for the purposes of SIPs and conformity analyses follows the guidance discussed in the MOVES User Guide. [16] Fincher et al. (2010). This report discusses the methodology for formatting MOBILE6 inputs for I/M programs for use within the MOVES framework. The report reiterates EPAâs recommended approach of starting with the default MOVES I/M files and making modifications as necessary. The study also discusses the challenges of mapping MOBILE6 vehicle types to MOVES source types. Fuel Formulation and Supply The fuel supply table defines the market share of various fuel formulations for each county and month found in the MOVES run as shown in Table A.7. Separate market share distributions should also be defined for each fuel year, which is the same as the calendar year for years 2012 and earlier, but is set equal to 2012 for all subsequent years. The market shares for all gasoline formulations should sum to one just as it should for all diesel formulations. The fuel formulation input defines all of the fuel formulations used in the fuel supply table by using 16 characteristics, such as Reid Vapor Pressure (RVP) and sulfur content. MOVES defaults for both tables are available for every county in the country, but local knowledge and/or fuel surveys should be used to verify and correct these defaults. Table A.6 Fuel Supply countyID fuelYearID monthGroupID fuelFormulationID marketShare marketShareCV 47065 2012 4 3850 0.333333 0.5 47065 2012 4 3869 0.166667 0.5 47065 2012 4 3870 0.083333 0.5 47065 2012 4 3872 0.416667 0.5 47065 2012 4 20011 1 0.5 [32] U.S. EPA (2012a). The MOVES User Guide notes that the fuel formulation and fuel supply importers should be used together to import the appropriate fuel data. The User Guide explains that fuel formulation data contains specific fuel properties and the fuel supply data assigns fuel formulations to specific counties and defines the market share within that county. At present, a user must edit existing data and may not enter new records until a bug associated with fuels is corrected. [33] U.S. EPA (2012b). The guidance for the fuel supply and fuel formulation MOVES inputs for the purposes of SIPs and conformity analyses follows the guidance discussed in the User Guide. The guidance outlines how the MOVES A-19
default gasoline and diesel fuel formulation and supply data was derived and cautions users that based on this approach, some fuel properties in the default database may not match actual local data. Users should take care to review the default fuel characteristics to accurately reflect real-world data for a specified modeling domain. The guidance cites sources of publicly available fuel information outside of reformulated gasoline (RFG) areas, including the National Institute for Petroleum Energy Research (NIPER) and the Alliance of Automobile Manufacturers (AAM) North American Fuel Survey. The guidance cautions users against averaging fuel properties from multiple fuel sample data to create a generic fuel profile. [16] Fincher et al. (2010). This study discusses how the fuel formulation and supply data were developed for use within the MOVES framework using information from the MOBILE6 input files in conjunction with ozone episode modeling emissions inventories created by the appropriate state agency. The study also revealed bugs associated with MOVES when updating fuel formulation inputs. The interim solution is to simply modify existing fuel formulations instead of creating new ones. Another bug concerned the expected ethanol percentages associated with a specific fuel ID (i.e., the fuel used in the specified area is identified as E10 with an ethanol percentage of 9.38 percent; MOVES expects E10 fuel to have an ethanol percentage of 10 to 20 percent). The solution suggested by EPA was to use an alternative existing fuel ID (for E8) to accommodate the known ethanol percentage. This section discusses three prior surveys on transitioning to the MOVES model, as follows: ⢠AMPO 2011 MOVES Survey; ⢠FHWA 2011 Survey of Five MPOs; and ⢠Tennessee DOT Strategic Plan for Transition to MOVES. A brief description of each survey and when it was conducted is provided along with a short list of findings, if available, followed by brief interpretation applicable to the planned survey for this project. One issue that stands out in this review is that agencies that have never had to address transportation conformity or SIP development have generally not been represented. This is a critical point as these input guidelines need to help agencies that are new to tools like MOVES or its predecessor, MOBILE6. The existing knowledge base about user capabilities does not include information about these agencies. In late 2012 and early 2013, FHWA was also undertaking a survey to help determine technical gaps/capabilities of MPOs and DOTs in using MOVES. The survey was to include an experiment that aims to sort out which factors are most important or critical to MOVES users. Results from this survey were not available at the time of this writing. A-20
A.5 OTHER MOVES-RELATED SURVEYS AMPO 2011 MOVES Survey The most significant previous MOVES-related survey was conducted by the Association of Metropolitan Planning Organizations during June-July 2011.17 This survey consisted of 33 questions compiled by AMPO with input from the FHWA, EPA, and the North Central Texas Council of Governments. The survey was sent to all state DOTs and about 90 MPOs in nonattainment and maintenance areas. Responses were received from 68 agencies, including 23 state DOTs (48 percent), 43 MPOs (46 percent), and 2 air agencies. The survey asked 33 questions on topics, including nonattainment status, level of MOVES experience, development of MOVES data inputs, meeting MOBILE6- based budgets, and MOVES training. The questions from the AMPO survey most relevant to the NCHRP 25-38 survey are Nos. 14-16, pertaining to the level of MOVES experience, and Nos. 17-19, related to MOVES inputs. From these questions, it was found that: ⢠Approximately 26 percent of the DOT responses indicated that they were the lead agency for conformity; presumably this addresses rural nonattainment areas. Approximately 81 percent of the MPO responses indicated that they were the lead agency for conformity. ⢠Approximately 70 percent of the agencies surveyed (48 out of 69 agencies) have either not begun to use MOVES for emissions analysis or have begun, but not developed all of the inputs required to run MOVES for a SIP or conformity analysis. This is illustrated in Figure A.1. ⢠Approximately 67 percent of the agencies surveyed that are using MOVES (24 out of 36 agencies) perform the model runs in-house and 22 percent (8 out of 36 agencies) use some combination of in-house and contractor support. Contractors complete all portions of the MOVES runs in the remaining cases (11 percent). ⢠Approximately 75 percent of the agencies surveyed that are using MOVES (27 out of 36 agencies) rate themselves as being 1, 2, or 3 in terms of proficiency with MOVES when rated on a scale of 1-5, with 1 being a beginner and 5 being proficient. ⢠Approximately 92 percent of the agencies surveyed that are using MOVES (33 out of 36 agencies) said they are using some combination of locally derived and national default data to prepare MOVES inputs. 17 http://www.ampo.org/assets/1480_ampomovessurveysept2011.pdf. A-21
Figure A.1 Status of Agencies Using MOVES and Developing MOVES Inputs 32 Agencies (47%) 16 Agencies (24%) 20 Agencies (29%) Not using MOVES yet Using MOVES, inputs not ready Using MOVES, developed all inputs In general, this AMPO survey provides the best existing snapshot of who is using MOVES and how it is being applied. It is important to note that the listed percentages in the AMPO survey are not adjusted to account for self-selection bias in the responders (i.e., nonresponders may have a tendency to be behind the responders in terms of their transition to MOVES). Keeping this in mind, it can still be concluded from these survey results that the large majority of agencies need help developing MOVES inputs, are performing this work in-house, are not proficient with MOVES, and are at least partially dependent on national default data. This confirms that the objective of NCHRP 25-38 to produce guidance on methods and procedures for obtaining and preparing input data for MOVES is in line with the needs of the MOVES user community. The AMPO survey results also indicate a large array of experience and proficiency in developing MOVES inputs and running the MOVES model. It will be important to capture responses from agencies at all of these levels during the survey conducted by this study to appropriately characterize the state of practice. The new survey also needs to consider how to address potential response bias, notably an apparent weighting towards more proficient MOVES users that is observed in the AMPO survey, for reasons discussed later in this memorandum. A sampling plan is presented later on which attempts to address this. The survey conducted by this study needs to avoid asking the same questions as the AMPO survey. However, some of the questions to characterize the agency and their general level of MOVES experience may be necessary to include again since we are not able to obtain individual answers from the AMPO survey (AMPO is unable to release the raw data due to privacy concerns). The AMPO survey did not ask about specific MOVES inputs and their sources of data, which will be an important characteristic of the NCHRP 25-38 survey. A-22
FHWA 2011 Survey of Five MPOs For a project entitled âImproving Vehicle Fleet, Activity, and Emissions Data for On-Road Mobile Sources Emissions Inventories,â The University of California (UC) at Riverside conducted a survey in 2011 of five MPOs on behalf of FHWA to characterize the state of practice for preparing certain MOVES inputs. MPOs were surveyed in the following areas: ⢠Southern California; ⢠The San Joaquin Valley of California; ⢠New York, New York; ⢠Maricopa County (Phoenix), Arizona; and ⢠Denver, Colorado. Since the report was not published at the time of this review, the nature of the questions and results were not known. Tennessee DOT Strategic Plan for Transition to MOVES The Tennessee DOT completed a strategic plan for transitioning Tennessee to the MOVES model. As part of this plan, the agency conducted telephone interviews with all eight MPOs in Tennessee, as well as the Tennessee DOT and Tennessee Department of Environment and Conservation (TDEC). These interviews covered a range of MOVES-related topics, including current practices and capabilities with respect to travel demand and emissions modeling; and MOVES- specific development and data preparations underway or planned. The report provides rich details on the data sources that each MPO is using for each of the 13 MOVES inputs when running MOVES for regional conformity. The MOVES capabilities and status of their transition to MOVES varies greatly among the eight MPOs in Tennessee. Most of the larger MPOs (Nashville, Memphis, Knoxville, Chattanooga) have begun working with MOVES at some level, although Chattanooga appears to be further along than the others. Most of the smaller MPOs had not begun working with MOVES at the time of the interviews (summer 2011) often because they were waiting to see if a new ozone standard would impact them. The report notes that several MOVES inputs, such as age distribution and source type population, are currently being prepared by the University of Tennessee (UT) using registration data. Similarly, it also notes that other MOVES inputs, such as I/M program and fuel data, are being prepared by TDEC. The MOVES inputs from both UT and TDEC are being prepared for all areas of the State and the report recommends that all MOVES users in Tennessee use these for consistency. The report recommends that MPOs enhance their travel demand models to include a time of day component and to add trucks as a vehicle type, if they have not already done so. A-23
