Guide to Truck Activity Data for Emissions Modeling (2019)

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.

49 7.1 Overview “Hotelling” is defined as any long period of time that drivers spend in their vehicles during mandated down times during long-distance deliveries by tractor-trailer combination heavy- duty trucks. Only the long-haul combination truck source use type (sourceTypeID 62) is assumed to have any hotelling activity. At the national and county scales, total hotelling hours are based on U.S. DOT driver rest rules, which specify 8 hours of rest per 10 hours of driving. MOVES assumes that trucks are hotelling for the entire rest period, of which the majority of time is idling. Hotelling activity is linked specifically to operation on rural highways using an activity rate of 0.0554 hour of truck hotelling per mile traveled on rural restricted roads. MOVES2014 defines four operating modes for hotelling periods: (1) extended idling of the main engine, (2) using an on-board APU, (3) using power from an external source such as truck stop electrification, and (4) having all engines and accessories off. In MOVES2014, all of the hotelling hours for long-haul trucks of model years before 2010 are assumed to use extended idle to power accessories. Starting with the 2010 model year, the trucks are assumed to use extended idle 70 percent of the time and APUs 30 percent of the time. Other types of trucks, including single-unit trucks and short-haul trucks, also may engage in idling activities and produce emissions that are not accounted for in MOVES2014. Emissions from idling while stopped at traffic signals or being stuck in traffic congestion are part of running exhaust emissions, but emissions from idling while loading/unloading cargo, queuing at a port terminal, being on a restroom or meal break, and so forth, are not captured by any emission pro- cesses in MOVES. This issue is investigated further in Case Study #6 (see Appendix F, available in NCHRP Web-Only Document 261). The EPA also is conducting research to examine idling time beyond what currently is captured in MOVES and is likely to provide a new “off-network idle” input option with the next release of MOVES (Brzezinski, 2017). Future versions of MOVES may include other revisions to idle and hotelling user-input data structures and default values consistent with the latest available data on and understanding of idling processes. 7.1.1 Project-Scale Analysis For project-scale runs, the user is expected to provide data for total hotelling activity and OMDs. Total hotelling activity is specified in the OffNetworkLink table via the OffNetwork table of the Project Data Manager (the same table that is used to input starts information). Activity is specified as the fraction of the total vehicle population that is engaged in hotelling (extended idling) during the project period being modeled (typically 1 hour). MOVES only supports extended idle for long-haul combination trucks (source type 62), so only input for this source type is relevant. This is shown in Table 7.1; in this case, there would be 100 total hours S E C T I O N 7 Hotelling

50 Guide to Truck Activity Data for Emissions Modeling of hotelling (since one-half of the long-haul combination truck population is idling during the 1-hour project period). Idling events of less than 1 hour would be accounted for by scaling the extendedIdleFraction. Operating modes are the same as for national and county scales. These are found in the HotellingActivityDistribution table, which can be populated through the Hotelling tab in the Project Data Manager. 7.1.2 County-Scale Analysis In MOVES2014, total hotelling hours can be provided by the user as a function of month, day type, hour, and vehicle age. These inputs are provided in the HotellingHours table avail- able via the Hotelling table of the Data Importer, as shown in Table 7.2. Currently, there is not a practical method for directly observing total truck hotelling hours in a modeling domain. Therefore, estimating a measure of total activity would typically require preprocessing by the user to scale up metrics that are available. For example, instrumented vehicles or telematics sources can provide hotelling hours on a per-truck basis, as well as distributions of hotelling activity by month, day, or hour. These parameters must be scaled up to total hotelling hours based on estimates of the total long-haul truck population in the modeling domain. Truck population is most frequently derived from VMT estimates based on traffic counts or demand models (see NCHRP Web-Only Document 210, Volume 1, Section 4.2) (Porter et al., 2014a). An alternative preprocessing approach would be to scale up total hotelling hours based on parking spaces in the area being modeled and the fraction of time a space is filled with an idling truck (idle rate) from field studies. A simpler alternative to providing total hotelling hours is to provide a rate of hotelling hours per mile traveled on rural, restricted-access roads, which can be found in the table Hotelling CalendarYear, available in the Generic tab of the Data Importer. The national rate of hotelling zoneID sourceTypeID vehiclePopulation startFraction extendedIdleFraction parkedVehicleFraction 120110 21 200 0 0 0 120110 31 200 0 0 0 120110 32 200 0 0 0 120110 42 200 0 0 0 120110 61 200 0 0 0 120110 62 200 0 0.5 0 Table 7.1. Sample OffNetworkLink table. hourDayID monthID yearID ageID zoneID sourceTypeID hotellingHours 15 3 2018 0 90140 62 15 3 2018 1 90140 62 15 3 2018 2 90140 62 15 3 2018 3 90140 62 …a … … … … … aEllipsis points indicate that the selected data shown are a sample and there may be many more rows of data in the table. Table 7.2. Sample HotellingHours table.

Hotelling 51 hours per mile of rural, restricted-access, roadway VMT is stored in the HotellingCalendarYear table for each calendar year. The value calculated for 2011 is used as the default for all calen- dar years. The default rate of 0.0554 hour per mile is estimated based on a federal law limiting long-haul truck drivers to 10 hours of driving, followed by a mandatory 8-hour rest period (U.S. EPA, 2016).10 However, analysis of recent truck telematics data suggests a somewhat lower hotelling rate of 0.016 to 0.033 hour per mile for stops of at least 8 hours and a much lower level of extended idle activity during hotelling, with only about 10 to 15 percent of hotelling time (all stops between 8 and 16 hours) spent in extended idle mode, compared to the MOVES default of 70 percent. The result is an extended idle rate that is perhaps only 6 to 8 percent of the MOVES default rate (about 0.0025 to 0.0030 hour per VMT), as discussed in Case Study #4 (see Appendix D, available in NCHRP Web-Only Document 261). Operating modes for hotelling are used to distinguish technologies that would reduce idle emissions. The operating modes available in MOVES2014 are shown in Table 7.3. In default MOVES2014, only modes 200 and 201 are populated. Note that these are a function of model year, as APU usage is phased in according to provisions in federal heavy-duty GHG standards. The Phase 2 heavy-duty GHG rule (U.S. EPA and NHTSA, 2016) includes revised assumptions that vary by model year, as shown in Table 7.3; however, these assumptions were developed after MOVES2014a was released. Total hotelling hours are assigned these operating modes based on user input of the HotellingActivityDistribution table, available via the Hotelling tab of the Data Importer. Table 7.4 is an example table showing default entries from the EPA’s MOVES documentation. 7.2 MOVES Embedded Data In MOVES2014, the total hours of hotelling are estimated by using the national estimate of VMT by long-haul combination trucks divided by an estimated average speed to calculate total hours of driving. The total hours of driving divided by 10 gives the number of 8-hour rest periods needed and, thus, the national total of hotelling hours. To allocate these total hotelling hours to locations, a hotelling rate is calculated as the national total hours of hotelling divided by 10 Note that the 2011 rule, amended in 2017, actually specifies a limit of 11 hours of daily driving after 10 consecutive hours off duty for property-carrying drivers or 10 hours of daily driving after 8 hours off duty for passenger-carrying drivers. See Hours of Service of Drivers Final Rule (FMCSA, 2011) and “Summary of Hours of Service Regulations” (FMCSA, 2017). The rule contains additional details that could affect the actual ratio of hotelling to driving activity. OpModeID Description MOVES2014 Default Phase 2 Heavy-Duty GHG Final Rulea Model Year: <2010 2010+ <2010 2010– 2020 2021– 2023 2024– 2026 2027+ 200 Extended idling of main engine 100% 70% 100% 91% 60% 50% 45% 201 Hotelling diesel APU 30% 9% 30% 40% 40% 203 Hotelling battery or AC (plug in) 10% 10% 15% 204 Hotelling all engines and accessories off aU.S. EPA and NHTSA, 2016. Table 7.3. MOVES operating modes for hotelling.

52 Guide to Truck Activity Data for Emissions Modeling the national total miles driven by long-haul trucks on rural, restricted-access (freeways) roads. The hotelling rate (hotelling hours per mile of rural, restricted-access travel by long-haul combi- nation trucks) is applied to the estimate of rural, restricted-access VMT by long-haul combina- tion trucks to estimate the default hotelling hours for any location, month, or day. The hours of hotelling activity in each hour of the day are not proportional to VMT. For each hour of the day, the number of trips that would end in that hour were estimated, based on the number of trips that started 10 hours earlier. The MOVES2014 data are based on a 120-truck sample collected by Battelle in California from 1997 to 1998. An estimate of the distribution of truck hotelling duration times is derived from a 2004 CRC paper based on a survey of 365 truck drivers at six different locations (U.S. EPA, 2016). 7.3 Sensitivity/Importance Emissions from extended idling can make up a significant fraction of emissions from long- haul trucks, and total truck emissions are highly sensitive to the amount of hotelling activity. Hotelling emissions are limited to combination long-haul trucks in MOVES, but over time are projected to contribute an increasing share of motor vehicle emissions, especially NOx and PM. Contributing factors include the assumption that the 2010 heavy-duty NOx standards will not reduce idle NOx emissions as much as over-the-road emissions and the assumption built into MOVES that 70 percent of time spent in mandatory rest periods (8 hours rest per 10 hours on the road) will be in extended idle mode. On the basis of these assumptions, long-haul hotelling emissions are estimated by MOVES2014a to contribute 30 percent of combination long-haul truck NOx emissions and 13 percent of total motor vehicle NOx emissions in 2018, with this share growing over time. No prior studies to assess the sensitivity of hotelling emissions to idle activity were found in the literature, so the NCHRP Project 08-101 team conducted an assessment using data from Vnomics that is described in detail in Case Study #4 (see Appendix D, available in NCHRP Web-Only Document 261). Considering stopping periods of 8 to 16 hours in dura- tion, the hotelling activity idle rate is estimated to be 0.00224 hour per mile traveled on rural, restricted-access roads. This rate is about 20 times lower than the MOVES default rate of 0.03878 hour per mile traveled on rural, restricted-access roads (with the default assump- tion that 70 percent of hotelling activity is in extended idle mode). Even if the threshold of “hotelling” is lowered to 2 hours, the Vnomics extended idle rate is still 16 times lower than the MOVES default rate. beginModelYearID endModelYearID opModeID opMode Fraction 1960 2009 200 1 1960 2009 201 0 1960 2009 203 0 1960 2009 204 0 2010 2050 200 0.7 2010 2050 201 0.3 2010 2050 203 0 2010 2050 204 0 Table 7.4. Sample HotellingActivityDistribution table.

Hotelling 53 Table 7.5 shows the change in emissions using the idle rate observed in the Vnomics data versus the MOVES2014a default when run at the national scale. The lower hotelling rate reduced extended idle emissions by 96 percent. The effect on the national emissions inven- tory was a 28 percent reduction in NOx emissions and 11 percent reduction in PM2.5 emissions for the source type 62 category (combination long-haul trucks), and an overall reduction (all source types) of 12 percent NOx and 5 percent PM2.5. Further effort may be needed to quantify real-world truck hotelling activity and examine the effect of this activity on mobile source emissions inventories. To assess the sensitivity of emissions to idling inputs at a project scale, the NCHRP Project 08-101 team performed a sensitivity analysis using a project-level MOVES run (starts, soak, and extended idling) composed of (1) two 0.5-mile, rural, restricted-access, on-network links and (2) one off-network link. The vehicle population was split evenly between gaso- line light commercial trucks and diesel combination long-haul trucks, and the off-network vehicle populations were set at 10 percent of the on-network populations. In a calendar year 2020 scenario, the majority of emissions are from the heavy-duty diesel trucks (91 percent of HC, 98 percent of NOx, and 99 percent of PM2.5), and most emissions are from off-network operation (88 percent HC, 72 percent NOx, and 35 percent PM2.5). Increasing the off-network hotelling hours by a factor of three increased total emissions (on and off network) by 42 to 175 percent. These results are also shown in Table 7.5. 7.4 Generating Local Data Options for generating local data on truck hotelling may include • Conducting field surveys that involve observations of trucks parked at truck stops, rest areas, and other overnight parking locations or instrumentation of trucks serving a specific facility. • Purchasing GPS/ECU data collected from truck fleets. Field surveys allow the analyst to identify quantities of idling vehicles, as well as the idle oper- ating modes, for specific local areas where emissions are of concern. However, field surveys are resource intensive to conduct and evaluate and may miss extended idle events that occur at dis- persed locations. Purchased GPS/ECU data on idle locations and durations on a per-truck basis may be more comprehensive, but cannot distinguish whether an APU or plug-in unit is being used, only whether the engine is on or off. Also, the size of the data purchase required to obtain Parameter Varied Scenario Change Delta Composite HC Delta Composite NOx Delta Composite PM2.5 Project-Scale Scenario, Calendar Year: 2020; Source Type 62 Extended idle (hotelling) hours 62.5 187.5 175% 114% 42% National Default Run, Calendar Year: 2018; Source Type 62 Hotelling ratio (HotellingCalendarYear table) Replace default 0.0554 hours/ VMT with 0.00224 hours/VMT −28% −11% National Default Run, Calendar Year: 2018; All Source Types Hotelling ratio (HotellingCalendarYear table) Replace default 0.0554 hours/ VMT with 0.00224 hours/VMT −12% −5% Source: Analysis by Eastern Research Group, Inc. “Hotelling ratio” is the ratio of hours of hotelling activity to source type 62 VMT on rural, restricted-access roadways. Table 7.5. Hotelling sensitivity analysis.

54 Guide to Truck Activity Data for Emissions Modeling sufficient data to represent a geographic area (county or even state) may make such a purchase prohibitively expensive. 7.4.1 Field Data Collection Source, Availability, and Cost The purpose of field data collection for hotelling is to gather data that enable an areawide estimate of hotelling activity, which MOVES requires if users wish to replace default data. Instrumented truck data will not capture all trucks idling in an area, and for quantifying (and projecting) the amount and location of idling, there needs to be an inventory of idle locations (preferably by parking space) and their usage rate (the frequency with which spaces are filled with idling trucks). Steps describing a large-scale (e.g., statewide) field study to gather this information are detailed below. Examples of real-world field surveys focused on truck stops and other rest areas also are documented in a variety of studies, mainly in Texas (Zietsman and Perkinson, 2003; Eastern Research Group, Inc., Cambridge Systematics, and Alliance Transpor- tation Group, 2004; Miller et al., 2007; and Hoekzema, 2015). A heavy-duty vehicle idle study will consist of three main components. The primary com- ponent will focus on identifying and characterizing the source generator categories across the study domain (mainly truck stops). The goals during the primary stages will be to identify the universe of truck stops, rest areas, and intermodal locations (e.g., ports and railyards) and to obtain reliable information on the location and number of parking spaces. This information can then be used to identify sites for data collection, as well as to extrapolate field activity to nonsurveyed sites. The second component of a heavy-duty vehicle idle study will involve developing and imple- menting a data collection plan to collect field observations of truck occupancies and idling rates. Additional information, such as time of visit, truck class, and truck stop amenities, also will need to be collected to refine the idling and/or emissions models as necessary. The third component of a heavy-duty vehicle idle study will involve analysis of the field data and development of the idling and emissions model as dictated by the end-use of the idling study. Identifying the sources for the generator categories is straightforward and will rely on publicly available data. Some guidelines for locating information on generator categories such as truck stops, rest areas, and intermodal locations are the following: • Truck stops—Truck stops are potentially the largest of the generator categories, and infor- mation on truck stop locations can be found on the Internet. The websites of industry trade associations, such as the National Association of Truck Stop Operators (NATSO), can serve as a starting point. Other “trucker-oriented” websites and truck stop operator websites11 may contain information on truck stop locations. Also, FHWA studies may contain information on truck stops within the study area. It is likely that a complete list cannot be compiled from public sources on the Internet, as some truck stop operators may not advertise using standard media methods. Information on the truck stop such as location, operating hours, estimate of parking spaces, and available amenities should be collected. If this information is not available online, a preliminary telephone interview could potentially provide this information. 11 Examples of online sources are the following: dieselboss.com, gocomcheck.com, natsn.com, otrprotrucker.com, road staronline.com, trucker.com, truckerslink.com, truckingnetwork.com, truckrealm.com, Yahoo Yellow pages (search for “Truck Stops”), flyingj.com (Flying J), loves.com (Love’s), petrotruckstops.com (Petro Stopping Centers), tatravelcenters.com (Travel Centers of America), and pilotcorp.com (Pilot Travel Centers).

Hotelling 55 • Rest areas—Information on rest area locations may be found on state DOT websites and/or through Internet search engines. • Intermodal locations—The term “intermodal” refers to facilities that handle truck-to-ship, truck-to-rail, and truck-to-aircraft goods, as opposed to operations that only involve high- way trucks. Information on these generator categories may be found on the websites of the following organizations: – Commercial cargo airports—U.S. DOT Bureau of Transportation Statistics. – Intermodal rail yards—Rail Intermodal Terminal Directory. – Commercial Marine Ports—American Association of Port Authorities. Once a list of potential data collection locations has been developed, the next step will be to focus on collection of real-world truck-idling data. A data collection plan that contains the pro- cedures for field data collection should be developed. The plan should indicate the various data elements that need to be collected, provide data collection guidelines, and provide data collec- tion templates for recording field observations. The plan should also account for spatial coverage (i.e., coverage of sampling locations within the study area) and temporal coverage (seasonal and day of week coverage for sampling) of field observations. Additionally, the plan should contain provisions for situations in which data collection at any location is not feasible. The field data collection effort should be designed to collect two types of data. The first type of data includes quantifiable information, directly observable to field staff, such as the number of spaces with idling trucks for truck stops and rest areas, and length of queue and idle duration for intermodal facilities. The second kind of data are qualitative data that should be collected from facility managers and truckers via interviews for use in validating direct observations and extrapolating results to nonsurveyed locations and over different time periods (i.e., annual level, or ozone season level, and daily). The observable data collected at truck stops and rest areas should include, but are not limited to: • Unique identifier for surveyed site. • Address, descriptive location (adjacent highway and mile marker), and GPS coordinates. • Owner/operator information. • Functional class of adjacent roadway. • Time of day, date, and day of week of observation. • Number of parking spaces and basis (count or estimate). • Occupancy and idling fractions. • Total number of trucks. • Number of diesel trucks. • Number of long-haul trucks. • Amenity characterization (truck stops only). • Lot status (paved or unpaved). • Surveyor comments (e.g., problems with access, presence of truck stop electrification systems, and reliability of interviews). Data also should be collected on heavy-duty truck idling occurring at intermodal facilities such as marine ports, airports, and rail terminals. As a facility reaches its capacity, it is constrained by (1) the acreage of the cargo-loading marshalling area and (2) the number of gates serving the facility. It is anticipated that most of the idling activity would occur at the entry gate where manifests are checked. In addition to applicable Example: Truck Stop Surveys in Texas Researchers in Texas surveyed occupancy and idling rates and APU use at three truck stops in the Port of Beaumont area. The weighted average of the idling rates (or ratio of idling trucks to total trucks) was found to be approximately 70 percent. Interviews with carriers and truck drivers shed light on how carrier policies and other factors might encour- age or discourage extended idling (Zietsman and Perkinson, 2003). Researchers in the greater Austin region performed surveys to estimate the total amount of extended idling, as well as APU use in the 11-county region. A total of 42 facilities was surveyed, with an estimated 9.7 hours per day of idling per space and an average idling rate of 55 percent for parked vehicles. A survey of trucks parked on frontage roads also found an idle rate of about 50 percent, although the sample size was small (Hoekzema, 2015).

56 Guide to Truck Activity Data for Emissions Modeling data elements listed above for truck stops and rest areas, the following data elements need to be collected at intermodal facilities: number of trucks in the idling queue, idle duration, and peak periods of activity. Due to Department of Homeland Security measures and time constraints, it might not be possible to inspect the onsite load-out areas of every terminal, and, in many cases, field observa- tions will only be collected at the entry gate. Normally, intermodal facilities maintain detailed records noting the time when a truck entered the queue and when the same truck proceeded through the terminal gate. Such records or summaries could be obtained by contacting the intermodal facility. Additionally, collecting information on the intermodal facility’s idling policy could be very useful if field observations at the facility are not possible. Level of Data-Processing Effort Conducting a comprehensive field study as described is very time and resource intensive. Data-Processing Steps—Truck Stops and Rest Areas A methodology for estimating idling rates should be developed to analyze the data collected during field observations and onsite interviews at truck stops and rest areas. Statistical tests can be used to determine which independent variables are statistically significant in terms of correlation with idling rates. Then, the idling data collected in the field should be expanded to estimate the idling under annual, daily, and hourly scenarios, as needed. Similarly, the idling data should be extrapolated to nonsurveyed truck stops and rest areas. The following correlations can be useful for developing an idle activity data model for the study area: • Occupancy, truck-idling rates, and parking space idling rates by hour. • Idling rates by geographic location. • Idling rates by facility size and facility type (truck stop versus rest areas). • Weekday versus weekend day idling rates. • Idling rates versus local area population. Statistical analyses can be performed to identify the factors that are significantly influencing average idle rates per parking space. Two potential methods are tests of means and regression analysis. Tests of means can be performed under scenarios where multiple independent variables will be tested or for categorical attributes (e.g., amenities at truck stops). Regression analysis can be used when there is a singular and quantitative independent variable to test. The idling data collected through field observations and interviews represent real-world activ- ity during the time the observations were made. Also, not all truck stops and rest areas in the study domain will be included in the survey. Therefore, extrapolation profiles must be developed from the collected data to estimate idling activity for all locations and for different time periods. Inter- views of truck stop managers can be used to estimate occupancy rates during different seasons. Therefore, an analysis of the accuracy of the manager survey responses will provide an indication of the accuracy of the seasonal extrapolation data developed from these interviews. Data elements such as number of truck parking spaces and occupancy for a truck stop can be evaluated based on both observed data and manager interview responses. However, additional data collected in alternative seasons for each of the survey locations would provide a more robust final model. Data-Processing Steps—Intermodal Facilities Intermodal facilities are expected to have restricted space, thereby causing queue lines that could result in substantial idling. In general, most airport terminals are open 24 hours a day. Most marine and rail terminal operations are conducted during daylight hours (usually 7:00 a.m. to

Hotelling 57 7:00 p.m.). Unlike airport freight operations, which mainly run overnight, most containerized facilities are anticipated to have three peak periods: in the early morning, at noon, and in the early evening. In any given study area, there will typically be fewer intermodal facilities than truck stops and rest areas. It might be possible to attempt to collect field data from all the intermodal facilities within the study area and thereby avoid the need to extrapolate idling rates to nonsurveyed intermodal facilities. However, data will have to be extrapolated to cover time periods that were not surveyed. In certain cases, there could be extended idling due to weather emergencies or the holiday season rush (e.g., flights delayed due to weather conditions and/or holiday travel). Such instances should be treated as exceptions and should not be accounted for when extrapolating idling rates to other seasons. Additionally, the survey team should collect information on the idling policy in effect at intermodal facilities in order to use that information for extrapolating data to nonsurveyed time periods. Applicability and Limitations Conducting a comprehensive field study as described is very time and resource intensive. Online sources may reduce the effort required in the field—for example identifying idle locations—but use of these sources will not entirely eliminate the need for fieldwork to deter- mine the ongoing usage rate of individual locations. Temporal distribution of idle and activity at nonsanctioned idling spots (on the side of the road, etc.) will be better captured by instrumented trucks (GPS/ECU). 7.4.2 Third-Party Data Logging/Telematics (GPS/ECU) Source, Availability, and Cost It is unlikely that adequate information on hotelling could be obtained only from GPS readings; long-duration stopped events could be identified, but not the critical information of whether the engine is on. ECU data can be used to identify times when a vehicle is stopped and the engine is running, although not the use of an APU or plug-in power. The ECU data also must be linked with GPS coordinates if the activity is to be linked with a specific study area. The NCHRP Project 08-101 team identified only one provider, Vnomics, which was able to provide ECU data linked with GPS data. While this may change in the future, analysis of sample data from this provider gives insights into the opportunities and potential limitations of this data source. As described in Case Study #4 (see Appendix D, available in NCHRP Web-Only Docu- ment 261), the project team was able to purchase a sample of 300 trucks observed for 1 month nationwide. These were trucks that performed long-haul operation and were likely to engage in hotelling activities due to the FMCSA Hours-of-Service rule (FMCSA, 2011). The data were pro- vided by vehicle, rather than for a specific geographic area. Since many trucks travel across several geographic areas in their daily course of operation, the sample’s hotelling patterns occurred over a broad area. A larger sample size would be required to develop statistics that are meaningful for a localized geography, such as a county, metropolitan area, or state. The NCHRP Project 08-101 team was able to acquire GPS and GPS/ECU data sets from multiple providers for a total cost of less than $50,000. The data purchases included GPS-based data for three counties, GPS/ECU data for three county-months, and a national sample of GPS/ECU data for 300 trucks for one month. The NCHRP Project 08-101 team also attempted to match hotelling information with the locations of truck stops identified in databases available to the public for a nominal cost. The goal was to see if a relationship between truck stop parking capacity and hotelling events could

58 Guide to Truck Activity Data for Emissions Modeling be established, so that truck stop data could be used as a proxy for hotelling. However, this effort was ultimately deemed unsuccessful because (1) most of the hotelling events occurred at dispersed locations not captured in truck stop databases (about two-thirds on average, but this ratio could vary locally) and (2) the observations in the GPS database were too sparse to draw a clear relationship to truck stop locations and capacity. Level of Data-Processing Effort Analysis of a local sample of GPS/ECU data is likely to require significant effort due to the large data files that are involved and the multiple processing steps needed. Data-Processing Steps The raw data in the sample Vnomics data set included a number of data files containing second-by-second time stamp, latitude, longitude, vehicle speed, accelerator pedal position, fuel rate, engine RPM, total distance (vehicle odometer), and reference torque, as well as the make, model, and year of the vehicle. A further description of the Vnomics data is provided in Section 6.4.2. The following steps can be undertaken to analyze this data set to support local MOVES inputs: 1. Flag extended idling events and duration. Extended idling events can be defined as events where the truck was stationary for a period of at least 2 hours with the engine running. Note that an extended idling event may occur at the beginning, at the end, or in the middle of a trip. Also note that a trip may include multiple extended idling events. 2. Calculate the VMT driven by heavy-duty trucks in the data set for each vehicle age. 3. To develop inputs for the HotellingHours table: a. Calculate the total hours of extended idle for each vehicle age group in the data set. b. Divide by VMT for each age group to obtain idle hours per VMT by age group. c. Multiply by an external estimate of analysis domain VMT by age group (e.g., total long- haul heavy-duty truck [source type 62] VMT on rural, restricted-access roads calculated elsewhere for MOVES inputs, multiplied by an age distribution for the same vehicle class) to obtain an estimate of total extended idling hours by age group. d. Divide by the value for OMD 200 (extended diesel idling) in the HotellingActivity Distribution table to obtain estimates of total hotelling hours by age group (including idling and APU operation). 4. To develop an updated hotelling rate for the HotellingCalendarYear table: a. Using the data prepared for Step 3, compute the total idle hours in the data set across all age groups, divide by the total VMT in the data set across all age groups, and multiply by 0.248 to obtain average idle hours per VMT on rural, restricted-access roads. 5. To compare other hotelling parameters against MOVES default values: a. Analyze the distribution of idle times, including the number of extended idling events by 2-hour increment up to 16 hours. b. Analyze the percent of extended idle events that begin in each hour of the day. 6. (Optional) Plot the locations of long-duration idling events on aerial imagery to illustrate the types of locations where the events occur (e.g., truck stop, warehousing facility, roadside). Applicability and Limitations A substantial difference was found between observed extended idle rates and the default hotelling rate contained in MOVES. The sample data sets analyzed for this project, as well as any similar samples obtained by local analysts, will only include trucks in the fleets that are customers of the data set vendor. These trucks may or may not be representative of the entire fleet of heavy-duty vehicles either nationwide or in a localized geographic area of analysis. It also is possible that seasonal variations would be observed in hotelling patterns.

Hotelling 59 Differences also were found in the temporal distribution of hotelling activity. Figure 7.1 shows the hourly distribution of trip starts and trip ends for the 383 trips in the Vnomics data set that include extended idling events, as well as the hourly distribution of truck trips used to calculate hotelling hours in MOVES. The Vnomics distributions contrast with the MOVES distributions, where the Vnomics distributions show trip starts peaking in the mid afternoon and trip ends peaking in the morning, indicating a preference for nighttime driving. Figure 7.2 shows the distribution of extended idling activity by hour of the day in the Vnom- ics data set, as compared to the MOVES defaults. The shape of the distributions is similar, with Source of MOVES data: U.S. EPA, 2016, Table 12-9. 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Percent of Truck Trips Hour of Day (1 = midnight - 00:00:59) MOVES - Trip Starts MOVES - Trip Ends Vnomics - Trip Starts Vnomics - Trip Ends Vnomics - %starts, all trips Figure 7.1. Hourly distribution of truck trips with extended idling event(s). Source of MOVES data: U.S. EPA, 2016, Table 12-11. 0% 1% 2% 3% 4% 5% 6% 7% 8% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Percent of Fleet Hotelling Hour of Day (1 = midnight - 00:00:59) MOVES2014 Vnomics Data Set Figure 7.2. Percent of hotelling activity by hour of the day.

60 Guide to Truck Activity Data for Emissions Modeling a high of about 6 to 7 percent in overnight hours in both distributions, and a low of around 2 to 2.5 percent around midday to early afternoon. However, the Vnomics distribution shows substantially less extended idling activity between 6:00 p.m. and midnight, while showing significantly more activity between 3:00 and 9:00 a.m. The total daily hotelling activity is most important for estimating total daily emissions. Hourly distributions are mainly important for hourly emissions outputs that may be needed for air qual- ity modeling, or in some cases, peak-period hot-spot analysis. However, if a project hot spot is a major source of hotelling activity, local field surveys may be needed to identify project-specific hotelling activity levels and temporal patterns. 7.4.3 Third-Party Data Logging/Telematics (GPS Only) Extended idling could potentially be inferred from GPS-only data sets as described for the starts inputs in Section 6.0. Processing methods would be similar to those described above. In the sample Vnomics national truck data evaluated in this study, the average ratio of engine-on to key-on duration was about 0.83. This factor could be used to estimate engine-on duration and the amount of extended idling based on the “key-on” duration as obtained from other truck activity data sets with only GPS data.12 7.4.4 Other Resources Case Study #4 (see Appendix D, available in NCHRP Web-Only Document 261) describes the estimation of hotelling activity using Vnomics GPS/ECU data from a nationwide sample of trucks. One of the Excel files (NCHRP08-101_Data_CS3+4+6_Starts-Idle-Hotelling.xlsx) that sup- plements this guidebook contains summaries of the Vnomics data referenced in Case Study #4, as well as comparison MOVES data. This file is available on the NCHRP Research Report 909 web page (http://www.trb.org/Main/Blurbs/178921.aspx). 12 The ratio of engine-on to key-on duration during off-network parking for the local Vnomics data sets collected in two counties was 0.62 to 0.65. These data sets appear to represent mainly trucks in short-haul use; the 0.83 factor is deemed more appropriate for long-haul trucks for which extended idling is an input to MOVES2014.