Case Studies of 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.

Guide to Truck Activity Data for Emissions Modeling D-1 Appendix D. Case Study #4: Truck Extended Idling and Starts from Fleet Telematics Data D.1 Emissions Model Inputs Supported • Truck hotelling (distribution of extended idling by hour and duration). • Truck starts (distribution of starts by hour and soak time). D.2 Level of Effort for Local Application This method requires a high level of effort for local application. D.3 Overview This case study focused on investigating the potential for using truck fleet telematics data to develop local inputs for truck hotelling (distribution of extended idling by hour and duration) and starts (distribution of starts by hour and soak time), which may be provided by users as inputs to the U.S. EPA’s MOVES model. The case study investigated the use of data purchased from a private fleet telematics provider, Vnomics. Vnomics collects data from trucks of their customer fleets on an ongoing basis. The collected data include readings from onboard GPS units that can be used to identify the location and speed of the vehicle at any given time, and readings from the vehicle’s engine control unit which can be used to identify engine starts and idling events. (The Vnomics data can only be used to identify when the engine is on or off; it does not identify the use of onboard auxiliary power units or external power.) Data were obtained from 300 trucks for a one-month period as described in more detail below. Data on hotelling characteristics were compared to MOVES embedded data. Comparisons of starts data are made separately in Case Study #3. D.3.1 MOVES Approach to Hotelling The MOVES approach to hotelling is explained in U.S. EPA (2016). “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. The total hours of hotelling are estimated by using the national estimate of VMT by long-haul combination trucks divided 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 hotelling hours. To allocate these total hotelling hours to locations a hotelling rate is calculated as the national total hours of hotelling divided by 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 combination trucks) is applied to the estimate of rural restricted access VMT by long-haul combination 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.

Guide to Truck Activity Data for Emissions Modeling D-2 During a hotelling period, the truck may be in one of the four modes: 1) extended idling of the main engine, 2) using an onboard auxiliary power unit, 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 auxiliary power units (APUs) 30 percent of the time (U.S. EPA, 2016). The EPA document references two sources of heavy-duty truck activity information. Table 12-6 references a 120-truck sample collected by Battelle in California in 1997 – 1998. The Vnomics sample is larger, more current, and more geographically dispersed. An estimate of the distribution of truck hotelling duration times is derived from a 2004 Coordinating Research Council paper based on a survey of 365 truck drivers at six different locations. This is a different sampling method than the truck-based observational method using the Vnomics data. MOVES provides user input options for entering hotelling activity: • The hotellingHours table on the Hotelling tab of the County Data Manager can be used to provide hotelling hours by hour, age, and duration for the domain’s vehicle population. • The HotellingActivityDistribution table on the Hotelling tab of the County Data Manager can be used to provide fractions of hotelling by operating mode (idle, APU, plug-in, engine/accessories off) by model year. • The national rate of hotelling hours per mile of rural restricted access roadway VMT is stored in the HotellingCalendarYear table for each calendar year. The same value calculated for 2011 is used as the default for all calendar years. This table can be edited via the Generic tab. D.3.2 Case Study Approach The following data evaluations and comparisons were made in this case study: • Locations of extended idling events were identified and compared with known truck stop locations. The objective of this comparison was to evaluate the extent to which extended idling occurs at expected locations versus other “unpredictable” locations. This information could assist MOVES users in allocating extended idling spatially (e.g., among counties for a State analysis) based on the locations of known truck stops. • The characteristics of truck hotelling activity, including duration of extended idling events and their frequency by hour of the day, were compared with MOVES defaults. • ECU and GPS data were compared to evaluate the ratio of “key on/off” duration to “engine on/off” duration, which may be useful for adjusting idling data inferred from GPS only.

Guide to Truck Activity Data for Emissions Modeling D-3 D.4 Data Sources Table D.1 Data Sources Data Source Source Agency/ Organization Availability and Cost Truck fleet telematics data (GPS and ECU) Vnomics Purchased for 300 trucks for one month Truck stop locations Truck Stops Plus Purchased in 2014 for $40, http://www.truckstopinfoplus.com/ Truck/RV fuel stations POI Factory Free, http://www.poi-factory.com/node/26860 The truck fleet telematics data were acquired from Vnomics for 300 trucks that performed long-haul operation and were likely to engage in hotelling activities due to the Hours-of-Service rule (FMCSA, 2017). The data were acquired for the month of November 2016. Figure D.1 shows the geographic coverage of the data. All of the vehicles in the dataset were model year 2010 or newer. The data were provided by vehicle, rather than for a specific geographic area. Since many trucks travel across several geographic areas in their daily course of operation (see an example for one truck in Figure D.2), this case study looks at hotelling and start patterns that occur 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. However, the statistics from this dataset can be compared to MOVES defaults and also to other data sources acquired by the project team that have greater local validity. The time period from start of negotiations to final receipt of the Vnomics data was approximately seven months. This included discussions with the vendor to determine what data items could be provided, negotiation of nondisclosure agreements, review by the project team of sample data for suitability to achieve the project’s objectives, negotiation of a purchase price and contract, execution of the contract, and data processing and transfer by the vendor. Contract negotiation was complicated by the fact that the prime contractor on the NCHRP project team made the data purchase, but the data was actually provided to a subcontractor (UC-Riverside), requiring an additional nondisclosure agreement. Subsequent data purchases from the same vendor would be much quicker now that data parameters and contract provisions have been established.

Guide to Truck Activity Data for Emissions Modeling D-4 Figure D.1 Geographic Coverage of the Data for 300 Trucks Acquired in This Study Background image: GoogleEarth. Figure D.2 Geographic Coverage of the Data for One Truck Spanning Multiple Counties and States Background image: GoogleEarth.

Guide to Truck Activity Data for Emissions Modeling D-5 In addition to the fleet telematics data, data on truck stop locations were acquired from databases available for free or at nominal cost. The Truck Stops Plus dataset was purchased for a different project in 2014 (Perez and Koupal, 2014). A review at the time of several potential sources of truck stop data concluded that Truck Stops Plus provided the most comprehensive and detailed database at a nominal cost. The database was obtained as an add-on template for Microsoft Streets and Trips. It included 7,347 truck stops including all major chain truck stops, as well as independent truck stops, and the size category of parking spaces (<20 parking spaces, 20 – 70 parking spaces, and >70 parking spaces). The project team obtained a free online source from the Points of Interest (POI) Factory to supplement the Truck Stops Plus data. D.5 Data Processing and Analysis The raw data provided included 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. An example is shown in Table D.2. A month of data in November 2016 was provided for 300 trucks. Table D.2 Sample Vnomics Data Time Latitude Longitude Speed (kph) Acc Position (Percent) Fuel Rate, (Liter/ Hour) RPM Total Distance (Kilometer) Ref Torque (N-m) Make Model Year 11/2/2016 21:34:39 29.966555 -90.15716 0 5.2 3.45 625 271646.1 2504 International LA677 2012 11/2/2016 21:34:40 29.966555 -90.15716 0 5.2 3.45 625 271646.1 2504 International LA677 2012 11/2/2016 21:34:41 29.966555 -90.15716 0 5.2 3.45 625 271646.1 2504 International LA677 2012 11/2/2016 21:35:56 29.966418 -90.15716 70.759375 27.6 20.4 1254.325 271646.1 2504 International LA677 2012 11/2/2016 21:35:57 29.966639 -90.15716 71.771874 27.6 20.4 1265 271646.1 2504 International LA677 2012 11/2/2016 21:35:58 29.966639 -90.15716 72.4656225 27.6 20.4 1285.15 271646.1 2504 International LA677 2012 11/2/2016 21:35:59 29.966639 -90.15716 73.44531 27.6 20.4 1296 271646.1 2504 International LA677 2012 11/2/2016 21:36:00 29.966639 -90.15716 74.0404265 27.6 20.4 1305.45 271646.1 2504 International LA677 2012 The project team applied the following data processing steps: 1. Verified that the dataset contained 300 trucks and correct fields. 2. Plotted data from a sample of trucks on Google Earth. Identified activity events and activities. Six “activity events” are defined as shown in Table D.3 including key on/off, engine on/off, and starting and stopping; and eight “truck activities” are defined based on these activity events, as shown in Table D.4. The data logger starts recording data after the key is switched on (“key-on” event) and it stops recording data after the key is switched off (“key-off” event). Since the recorded

Guide to Truck Activity Data for Emissions Modeling D-6 data includes both vehicle speed and engine speed, “engine-on” and “engine-off” events can also be identified and used to represent the start and the end of a trip. In other words, a “trip” in the context of this case study is the entire period of engine operation from the engine-on event to the engine-off event as shown in 3. Figure D.3. Due to possible noise in the vehicle speed and engine speed data, a vehicle speed of 1 mph is used as the threshold for vehicle not moving and an engine speed of 300 RPM is used as the threshold for engine not running. A total of 55,911 trip start and trip end events (each) were identified in the dataset, which is equivalent to about 6.2 trips per day per truck. 4. Analyzed the distribution of stopping activity, considering fraction of stopping activity by the length of the stop by hour bin and whether the state of the vehicle was engine-on, key-on/engine-off, or key-off. Various thresholds based on stopping time were used to evaluate extended idling and hotelling. 5. Analyzed the distribution of idle times for all 300 trucks, including the number of extended idling events by 2-hour increment up to 16 hours, and compared to MOVES defaults. 6. Plotted the locations of long-duration idling events (>12 hrs) on aerial imagery to illustrate the types of locations were the events occurred (e.g., truck stop, warehousing facility, roadside). A total of 16 events were plotted. Examples of these events are shown in Figure D.4 through Figure D.6. 7. Compared the locations of extended idling events of at least 8 hours (159 total events illustrated in Figure D.7) with the locations of truck stops as identified in the truck stop databases (illustrated in Figure D.8). 8. Evaluated the number and distribution of trip starts and trip ends by hour of the day and compared to MOVES defaults. 9. Evaluated the percent of fleet hotelling by hour of the day and compared to MOVES defaults. 10. Evaluated the rate of engine idling during long-duration stopped periods and compared to MOVES defaults. Table D.3 Definition of Activity Events Event ID Event Name Definition 1 Key on Start of file & v(t) < 1 mph & RPM < 300 2 Engine on v(t) < 1 mph & v(t-1) < 1 mph & RPM > 300 3 Vehicle starts v(t) > 1 mph & v(t-1) < 1 mph & RPM > 300 4 Vehicle stops v(t) < 1 mph & v(t-1) > 1 mph & RPM > 300 5 Engine off v(t) < 1 mph & v(t+1) < 1 mph & RPM < 300 6 Key off End of file & v(t) < 1 mph & RPM < 300

Guide to Truck Activity Data for Emissions Modeling D-7 Table D.4 Definition of Truck Activities Activity Definition Vehicle trip 1 --> 6 Vehicle parked 6 --> 1 Vehicle moving 3 --> 4 Vehicle stopping 4 --> 3 Hotelling x hrs < vehicle stopping < y hrs Engine trip 2 --> 5 Engine idling 2 --> 3 & 4 --> 5 & 2 --> 5 Engine soaked 5 --> 2 Figure D.3 Illustration of Key-on versus Engine-on Event Figure D.4 Long-Duration Idling (>12 hrs) at a Truck Stop Facility Image courtesy GoogleEarth.

Guide to Truck Activity Data for Emissions Modeling D-8 Figure D.5 Long-Duration Idling (>12 hrs) at a Warehousing Facility Image courtesy GoogleEarth. Figure D.6 Long-Duration Idling (>12 hrs) on a Roadside Image courtesy GoogleEarth.

Guide to Truck Activity Data for Emissions Modeling D-9 Figure D.7 159 Extended Idling Events of At Least 8 Hours Sources: Analysis of Vnomics data by UC-Riverside; image background courtesy GoogleEarth. Figure D.8 Truck Stop Facilities Sources: Truck Stops Plus (2014) and POI Factory (2017); image background courtesy GoogleEarth.

Guide to Truck Activity Data for Emissions Modeling D-10 D.6 Findings from Sample Data D.6.1 Locations of Extended Idle Events Figure D.9 shows an overlay of extended idling locations (red dots) with truck stops from the databases (blue dots). An excerpt is shown for the intermountain west region, which had a high concentration of truck activity in the provided dataset. It is apparent that many extended idling events did not take place at truck stops that are listed in these truck stop databases. The team’s analysis found that approximately 33 percent of extended idling events of at least 8 hours took place within 0.25 mile of an identified truck stop. An overlay of the locations of the other events with satellite imagery showed a diversity of locations. It was observed that about 17 percent of those other events took place at truck stops not listed in the truck stop databases, about 40 percent at off-road sites such as a warehousing facility or parking lot, and about 10 percent on roadsides. Figure D.9 Extended Idle and Truck Stop Locations Sources: Analysis of Vnomics data by UC-Riverside; image background courtesy GoogleEarth.

Guide to Truck Activity Data for Emissions Modeling D-11 D.6.2 Stopping Activity by Duration The definition of hotelling depends on the minimum and maximum thresholds of vehicle stopping period. The dataset was used to guide the selection of these thresholds. Figure D.10 shows the frequency distribution of truck stopping periods. Based on its shape, the distribution can be divided into three broad groups as follows: • Less than 8 hours—These stopping activities are shorter stops, which may include stopping at traffic lights or in congestion, meal/restroom breaks, loading/unloading, etc. Stops for less than one hour account for 98 percent of the total number of stops. • Between 8 and 16 hours—Some of these stopping activities are likely associated with the Hours of Service rules. To identify a stopping activity associated with the hours of service (HoS) rules, further analysis would need to be performed to determine whether the stop is away from the truck’s home base. • More than 16 hours—These stopping activities are likely stops after completing the trip or reaching the destination. Some of them may involve the truck not being used for a long period such as over the weekend. In such case, the truck would likely be parked at its home base. Figure D.10 Truck Stopping Activity D.6.3 Idling Events by Duration Figure D.11 shows the number of extended idling events by idling duration in the acquired dataset. There are 404 extended idling events of at least 2 hours in 383 trips. 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. In addition to the 404 events of at least 2 hours duration (extended idling), there were 663 events with a duration between 1 and 2 hours, and 36,595 events with a duration between 5 minutes and 1 hour. 0.0% 0.1% 0.2% 0.3% 0.4% 0.5% 0.6% 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48+ Percent Frequency Stopping period ("0" is 0-1 hour) Total count = 467,870 stopping events

Guide to Truck Activity Data for Emissions Modeling D-12 Figure D.11 Number of Extended Idling Events by Idling Duration Figure D.12 shows the fraction of idle events by duration, comparing the Vnomics data with MOVES data as presented in Table 12-10 of U.S. EPA (2016). Figure D.12 Fraction of Extended Idling Events by Idling Duration D.6.4 Hourly Distribution of Extended Idle Events Figure D.13 shows the hourly distribution of trip starts and trip ends for the 383 trips in the Vnomics dataset that include extended idling events as well as the hourly distribution of truck trips used to calculate hotelling 404 312 256 225 206 182 159 128 86 37 16 0 50 100 150 200 250 300 350 400 450 2 3 4 5 6 7 8 9 10 11 12 No. of Events Extended Idling Duration of At Least X Hours 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 2 4 6 8 10 12 14 16 Idle Duration (hours) MOVES2014 Vnomics Dataset Fraction of Idle Events

Guide to Truck Activity Data for Emissions Modeling D-13 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 D.13 Hourly Distribution of Truck Trips with Extended Idling Event(s) Source of MOVES data: U.S. EPA (2016), Table 12-9. Figure D.14 shows the distribution of extended idling activity by hour of the day in the Vnomics dataset as compared to the MOVES defaults. The shape of the distributions is similar, with a high of about 6 – 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 p.m. and midnight while showing significantly more activity between 3 and 9 a.m. 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

Guide to Truck Activity Data for Emissions Modeling D-14 Figure D.14 Percent of Hotelling Activity by Hour of the Day Source of MOVES data: U.S. EPA (2016), Table 12-11. D.6.5 Key-on versus Engine-on Duration The project team compared the duration of data logging from “key-on” to “key-off” events with the duration of engine operation from “engine-on” to “engine-off” events. As illustrated in Figure D.3 earlier, the data logging duration can be longer than the engine operation duration if the driver switches the key ignition on, which powers up the vehicle electrical systems and the data logger, without turning on the engine. Since many GPS-based datasets on truck activity only provide “key-on” rather than “engine-on” data, the ratio of engine- on to key-on duration could be used to adjust GPS-based estimates of idling activity. For the Vnomics dataset the project team found the average ratio of engine-on to key-on duration to be 0.8265. D.6.6 Hotelling Duration per VMT (HotellingHours Table) While not a direct user input, the hotelling duration per VMT (source type 62, rural restricted VMT only), as found in the HotellingHours table, is a key parameter in MOVES for extended idle emissions. The Vnomics dataset shows a total of 1,657 extended idle hotelling hours (only including idling events longer than 8 hours to be consistent with the calculation in MOVES). The total VMT for the dataset is 2,977,888 miles. Considerable geoprocessing work would be required to associate the Vnomics data with road types. However, using the MOVES default road type distribution of 24.8 percent of VMT on rural restricted roads, the estimated VMT on rural restricted roads is 738,516 miles. Table D.5 compares hotelling rates based on the 300-truck dataset, with various time span definitions, and the national default value in MOVES. Depending on what thresholds for stopping period are used to define hotelling, the hotelling rates from the 300-truck dataset are only 28 to 60 percent of the MOVES default. 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 Dataset

Guide to Truck Activity Data for Emissions Modeling D-15 Table D.5 Comparison of Hotelling Rates Stopping Period Key Off (hours) KeyOn_EngOff (hours) KeyOn_EngOn (hours) Total_Time (hours) Hotelling Rate (hours/mile) Fraction of MOVES 8-12 hrs 9,638 161 1,780 11,579 0.0157 28% 8-14 hrs 16,157 205 2,112 18,474 0.0250 45% 8-16 hrs 22,087 249 2,214 24,551 0.0332 60% MOVES national default 0.0554 N/A D.6.7 Hotelling Activity Distribution The HotellingActivityDistribution table is a user input table that provides the fraction of hotelling activity by mode—extended main engine idle, diesel APU, battery or plug in, and all engines and accessories off. MOVES currently contains default values of 70 percent engine idle and 30 percent APU activity. The Vnomics dataset does not indicate whether an APU, electrical hookup, or no power at all is used when the engine is turned off, so it cannot be used to produce the complete input. However, it can be used to estimate the fraction of hotelling spent in extended main engine idle (and the fractions of time in APU and battery/plug in modes could be adjusted proportionately based on MOVES defaults or another source). Figure D.15 shows the time fraction of different states of key and engine during stopping periods based on data of the 300 trucks combined. The state “KeyOn_EngOn” represents idling of main engine during the stopping periods. Key observations are: • Depending on how hotelling is defined in the 300-truck dataset, the fraction that the main engine is idled ranges from 9 to 15 percent, which is significantly lower than MOVES assumption. • Longer stopping periods are associated with a lower fraction of main engine idling. • The variation in the fraction of main engine idling within a group (i.e., < 8 hrs, 8 to 16 hrs, and > 16 hrs) is less than the variation among groups. This supports the grouping of stopping periods into these three groups.

Guide to Truck Activity Data for Emissions Modeling D-16 Figure D.15 State of Key and Engine during Stopping Periods The MOVES national default for hotelling rate with main engine idling is 0.0554*0.7 = 0.03878 hours per rural restricted VMT. Table D.6 shows compares this with the rates based on the 300-truck dataset. The rate of extended idling per rural restricted VMT observed in the Vnomics data is only 6 to 8 percent of the MOVES default rate. Table D.6 Comparison of Hotelling Rates with Main Engine Idling Stopping Period KeyOn_EngOn (hours) Hotelling Rate with Main Engine Idling (hours/mile) Fraction of MOVES 8-12 hrs 1,780 0.0024 6% 8-14 hrs 2,112 0.0029 7% 8-16 hrs 2,214 0.0030 8% MOVES national default 0.0388 N/A The MOVES user (or in future MOVES updates, U.S. EPA) also may wish to update the HotellingActivityDistribution table with the latest available information. For example, EPA’s final Phase 2 heavy-duty greenhouse gas (GHG) rule (U.S. EPA and NHTSA, 2016) includes APU assumptions as shown in Table D.7. Table D.7 Hotelling Activity Distribution in Phase 2 Heavy-Duty GHG Rule Model Years Diesel APU Penetration Battery APU Penetration 2010 – 2020 9% 0% 2021 – 2023 30% 10% 2024 – 2026 40% 10% 2027+ 40% 15% Source: U.S. EPA and NHTSA (2016), Table 5-14. 44% 49% 54% 83% 87% 90% 99%7% 6% 6% 1% 1% 1% 1% 49% 45% 41% 15% 11% 9% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% < 2 hrs < 4 hrs < 8 hrs 8-12 hrs 8-14 hrs 8-16 hrs > 16 hrs Time Fraction KeyOff KeyOn_EngOff KeyOn_EngOn

Guide to Truck Activity Data for Emissions Modeling D-17 The MOVES default assumes eight hours of hotelling for every 10 hours of operation. It is not surprising that it appears to significantly overestimate extended idling activity, since it does not account for factors such as truckers shutting off the engine, team driving, or periods of no activity that are not “hotelling” events. This is the first instance known to the project team in which real-world data could be gathered on extended idle activity per VMT. There are still significant uncertainties in the latest estimate, including the representativeness of the Vnomics data, and the distribution of VMT by road type within this dataset. D.6.8 Sensitivity of Emissions to Hotelling Rate Hotelling (extended idle) 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 for NOx and PM. A contributing factor for this is the assumption that reduction in idle NOx emissions from the 2010 heavy-duty vehicle NOx standards will not be as stringent as for over-the-road emissions, and the aggressive assumption that trucks will idle for the entire duration of mandatory rest periods (8 hours rest per 10 hours on the road). With these assumptions in place, 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, growing over time. Despite their significance, the project team is not aware of prior studies to assess the sensitivity of hotelling emissions to idle activity at the national, regional or county scale. The team therefore conducted this assessment using data from Vnomics described above. MOVES2014a was run at the national scale to produce annual 2018 emissions with the default hotelling activity, and a second case with the Vnomics idle activity. Results are shown in Table D.8—since there was only one sensitivity case, results are shown relative to hotelling emissions, all combination long-haul truck emissions, and all motor vehicle emissions. Only NOx and PM emissions were analyzed, as diesel idle contributions for HC and CO are far less significant. Table D.8 Total Hotelling Activity Sensitivity Analysis Emissions From To Delta NOx Delta PM2.5 Hotelling Only 0.0554 Hotelling hours per rural VMT 0.00224 Idle hours per rural VMT -96% -96% Combination Long-Haul Truck -28% -11% All Vehicles -12% -5% As shown, the shift in NOx and PM emissions is significant with this substantial change from default idle activity to Vnomics data, even relative to total motor vehicle emissions inventory. More focused effort may be needed to quantify real-world truck hotelling activity, particularly as mobile source NOx emissions are scrutinized by the air quality community. D.6.9 Summary of Findings Extended idling is a significant source of heavy truck emissions especially at hot-spot locations, and its relative importance is growing as hot-stabilized running emissions decline with post-2010 model years subject to increasingly stringent emission standards. The truck telematics data evaluated in this case study provided some interesting insights and comparisons to MOVES default data on extended idling. However, these sources have

Guide to Truck Activity Data for Emissions Modeling D-18 limitations for use in developing locally specific MOVES truck activity inputs. Some general observations about the data sources in this case study include: • The Vnomics data are unique among the datasets investigated for this project in that they include ECU data. ECU data is critical to identifying actual engine operation, rather than simply whether a vehicle is moving or stopped, and therefore to developing accurate estimates of extended idling activity. • The Vnomics data provide a larger and more current dataset than the data used to develop some of the embedded hotelling factors in MOVES. The dataset represents trucks in fleets recruited by the data vendor, which may or may not be representative of the entire fleet of heavy-duty vehicles in the U.S. or a specific geographic area. • As the data acquired in this case study are for only 300 trucks operated throughout the U.S., there are insufficient observations to develop total extended idling statistics for a specific geography (county or State). However, the dataset sheds some light on the distribution of extended idling by duration and by hour of day. • The available truck stop databases do not appear to be complete, somewhat limiting their utility for identifying the relative amount of truck idling by geography. • The Vnomics data do not indicate the fraction of hotelling activity by operating mode (in particular, APU and plug-in electric operation are not identified). • The Vnomics data were also used to compare total starts per day and the distribution of starts by hour of day against MOVES data as well as StreetLight data. These findings are described in Case Study #3. Case Study #6 describes the analysis of a more localized set of Vnomics data, obtained for two urban counties. Observations from the data analysis include: • The hotelling rate observed in the Vnomics dataset (any operating mode) is about one-third to two-fifths as much per truck VMT on rural restricted access roads as the MOVES default value. • The amount of extended idling per VMT by long-haul combination trucks on rural restricted access roads appears in the Vnomics dataset to be only 6 to 8 percent of the estimate contained in MOVES2014. This suggests that the current version of MOVES may substantially overestimate the contribution of emissions from extended idle activity. • About two-thirds of all extended idling events in the sample dataset appear to take place outside of truck stop locations identified in truck stop databases. This information could potentially be used to “factor up” a truck idling survey that focuses on identified locations. However, it could not be determined whether this ratio might vary depending upon the local availability of truck stops. • The average ratio of engine-on to key-on duration is about 0.83 in the sample dataset. This factor could be used to estimate the duration of engine-on and the amount of extended idling based on the “key-on” duration as obtained from other truck activity datasets with only GPS data. • The distribution of extended idling activity by hour of the day followed similar patterns to the MOVES defaults, but more activity was observed in the early morning hours while less activity was observed in

Guide to Truck Activity Data for Emissions Modeling D-19 the late evening hours. The hourly distribution is primarily of importance if MOVES outputs are to be used for peak-hour hot-spot analysis or for air quality modeling. It is less important for simply estimating total emissions inventories. • The distribution of extended idling activity by duration also shows significant differences in the Vnomics data compared to MOVES default data. For example, Vnomics shows more short (2 – 4 hours) and very long (>16 hours) events than MOVES, with much less activity in the ranges in between. D.7 Transferability MOVES defaults are estimated from truck GPS data in the late 1990s. The sample data evaluated in this data set is real-world and more up-to-date. However, both datasets are relatively small and may or may not be representative of the universe of all truck hotelling activity. Furthermore, the datasets are not large enough to evaluate whether there might be significant differences in hotelling patterns across different states or regions of the country.