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.
D-1 This appendix describes additional reviews of long-distance and rural travel O/D data sources that already are collected or could be effectively collected both with traditional and emerging intercept technologies. This appendix describes data limitations and the minimum amount of local data required to assess the reasonableness of travel parameters derived from data collected elsewhere with traditional roadside intercept and advanced technologies. The advantage of employing traditional household survey methods for long-distance and rural trip data collection is that these surveys can gather most information required for travel analysis and modeling. But this approach also has obvious disadvantages, including very high costs (in order to have a sufficiently large sample), significant respondent burden, difficulty in gathering accurate longitudinal information, potential biases in the sampling frame, difficulty of data expansion, and data reporting and measurement errors. Several new and emerging technol- ogies have been proposed and/or tested in recent studies, which have the potential of replacing or supplementing traditional household and intercept travel surveys, including the following: ⢠GPS-based longitudinal survey with in-vehicle data loggers or tracking systems installed by vehicle manufacturers (e.g., OnStar system); ⢠GPS-based longitudinal survey with on-person data loggers; ⢠Travel survey based on GPS-enabled smartphone technology; ⢠Automatic license-plate capture and re-identification technology; ⢠Bluetooth technology; ⢠Cell phone wireless network location technology; ⢠Other wireless locationing technologies such as Radio Frequency Identification (RFID) and Wireless Internet (Wi-Fi); ⢠Web-based surveys; and ⢠Web-based surveys integrated into social networking applications and using crowd sourcing. D.1 Traditional Roadside Intercept Travel Surveys An important category of supplemental survey data is those efforts where details can be gleaned through data mining and aggregation efforts using existing intercept surveys. It should be noted though that most intercept surveys are performed at model area boundary locations for the pur- poses of estimating external-internal and through travel. It also should be noted that in some states intercepting vehicles for survey purposes is illegal. At a minimum, where available, these secondary sources can help with reasonable tests of the results. For example, in Texas, the DOT has extensive documentation on travel patterns into and through metropolitan areas, obtained through external station surveys. TxDOT has also conducted border crossing surveys. Relevant statistics (as readily available in reports) are presented here. A p p e n d i x d Other Demographic and Origin-Destination Data
D-2 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models Texas External Travel Surveys Since 1990, TxDOT has funded travel surveys on a continuous, rotating basis across Texas MPO areas. These include household, commercial vehicle, workplace, special generator, and external station surveys. Since 2000, TxDOT has sponsored 56 surveys (http://www.travel surveymethods.org/), including 17 household surveys, 8 workplace surveys (each of which includes several special generator surveys), 10 commercial vehicle surveys, and 21 external station surveys. The external station surveys are straightforward in terms of the data elements collected: origin, destination, home location, trip purpose, travel party details, and vehicle details. From this, analysts estimate the volume of through versus local travel in a region. A summary of travel statistics obtained from across the state is shown in Table D.1. As indicated therein, there are significant variations in commercial and through trip traffic, based on multiple factors. Through trips are more likely to be long-distance or rural-generated trips than internal- external trips. Texas Border Crossing Surveys In 2001 and in 2007, TxDOT sponsored statewide surveys to document the level of traffic entering and leaving the state. The 2001 effort surveyed approximately 17,000 vehicles enter- ing, exiting, and passing through Texas at 46 of the 115 highway border crossings in Texas (Texas Transportation Institute, 2001). For the 2001 effort, the survey results were used to derive an estimate of 746,000 vehicles and 1.29 million persons in vehicles cross the Texas border on a daily basis. Of these, 43 percent of the noncommercial vehicles transported nonresidents across the state line. About 10 percent of these nonresidents remained overnight in Texas, with each person staying an average of three nights per trip. The geographic distribution of these trips is shown in Figure D.1. Table D.1. Texas external station survey details. Study Area Year Population Daily Vehicles Number of Persons Percent Commercial Percent Through Trips Abilene 2005 116,000 80,000 86,300 83 16 Rio Grande Valley 2004 1,030,000 145,000 174,000 87 4 El Paso 2002 N/A 85,000 145,000a 67 7 Laredo 2002 193,117 70,000 82,400 68 7 Sherman-Denison 2005 117,000 118,600 135,000 82 17 Wichita Falls 2005 104,200 84,000 92,200 89 10 San Antonio 2005 1,145,000 290,000 313,000 82 11 Amarillo 2005 174,000 78,000 84,500 73 12 Killeen-Temple 2006 141,400 178,000 182,000 82 23 San Angelo 2004 88,000 49,000 55,400 85 9 Longview 2004 25,000 193,000 197,000 79 18 Source: Technical memos for the urban areas listed. aExcludes pedestrian crossings.
Other demographic and Origin-destination data D-3 Other relevant statistics from the 2001 effort include the following: ⢠An estimated 83 percent of the border crossing traffic was noncommercial vehicles; ⢠The average trip length for noncommercial vehicles was 60 miles, for commercial vehicles the average trip length increased to 101 miles; and ⢠The primary purpose of noncommercial vehicle trips was for work or work-related activities. TxDOT repeated the effort in 2007. For that effort, 21,000 surveys were administered, includ- ing 17,900 surveys of noncommercial vehicles and 3,800 surveys of commercial vehicles (Texas Transportation Institute, 2008). These surveys were conducted at 54 of 115 highway or bridge border crossing locations around the stateâs perimeter. The 2007 survey results suggest that an estimated 787,000 vehicles and 1.13 million persons in vehicles cross the border on a typical weekday, which shows a slight increase from the 2001 survey results. The geographic distribution of border crossings is shown in Figure D.2. Estimates suggest that 84 percent of vehicle traffic is noncommercial. Average trip length for noncommercial vehicles was 61 miles, and for commercial was 101 miles. This study enhances our understanding of long-distance travel in several respects. First, it pro- vides estimates of interstate travel based on direction of travel. It also shows the influence of the border crossings, useful for estimating long-distance travel in New Mexico, Arizona, and California. Finally, given the relative stability between 2001 and 2007, it supports the use of the older survey results in our estimations, particularly the Ohio statewide and long-distance survey efforts. Other Origin-Destination Intercept Surveys Origin-destination intercept surveys have traditionally been conducted largely to get a handle on the split between through and internal-external trips at a study area/model boundary. With 18% 24% 15% 43% East Texas Mexico Okalahoma New Mexico Source: Texas Transportation Institute (2001), page 12. Figure D.1. Texas border crossings, 2001. Source: Texas Transportation Institute (2008), page 15. 20% 19% 28% 33% M East Texas exico Okalahoma New Mexico Figure D.2. Texas border crossings, 2007.
D-4 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models external splits being largely the focus of such surveys, minimal emphasis was placed on where trips were originating from, or destined to, on the outside of the study area boundary where the rural and long-distance characteristics could be tabulated. As such, few roadside intercept surveys provide information of benefit to understanding rural and long-distance trip-making. In 2007â2008, TxDOT conducted a series of O/D intercept surveys along the I-35 corridor (Wilbur Smith Associates, Inc., 2008) for use in updating the Texas Statewide Analysis Model (SAM) for use in studying future needs along the proposed Trans-Texas Corridor. These sur- veys included a combination of roadside interviews, mailout/mailback, and license tag match- ing approaches, and addressed both passenger and commercial vehicle travel. With the focus of these studies largely on long-distance travel, there is some potential that these survey data could provide useful information, should the survey files be obtained. Figure D.3 depicts the survey intercept locations for this study. Source: Wilbur Smith Associates, Inc. Figure D.3. Texas Interstate 35 intercept survey locations.
Other demographic and Origin-destination data D-5 Another, albeit older, roadside intercept survey with a focus on long-distance and rural trip- making was conducted by the Florida DOT in 1992 (Transportation Consulting Group, 1992) for the purposes of understanding the travel patterns of visitors to Florida traveling by automo- bile. Data collected using laptop computers at survey stations along the state line and several screenline locations crossing major highways in the Florida peninsula were later used to validate the statewide model. Unfortunately, documentation of the surveys is focused on data collection and processing rather than on survey findings. Roadside intercept surveys were traditionally conducted by pulling over random platoons of vehicles to the roadside for a brief face-to-face survey about the current trip being made. Early data collection efforts were conducted using clipboards and paper surveys, which later gave way to the use of laptops and palm devices. As traffic congestion increased and privacy issues came to the forefront, use of high-speed video cameras became a common approach to locating survey respondents and asking questions about their trips. This high-speed video approach has led to other complications in linking tags to drivers, impacts of these delays on trip recall, and inability to reach travelers due to extensive use of rental cars. The next section describes the use of new and emerging technologies to enhance the collection of both O/D and household travel surveys. D.2 Surveys Using New and Emerging Technologies It is clear that no single technology can overcome all shortcomings of the traditional household- based survey method, as evidenced by a summary of technology capabilities depicted in Table D.2. Web-based surveys share the same advantages and disadvantages of traditional household-based surveys, including limitations of self-reporting, memory recall, and distance estimation (which is even more important when trying to identify long-distance and rural travel). GPS, smartphone, and license plate technologies, when used alone, can directly provide a wealth of information on long-distance and rural passenger travel, but not all the required information (such as travel purpose). Bluetooth technology does not allow re-identification of long-distance and rural trav- elers and is limited to instances where Bluetooth has been activated on devices within vehicles. Data-collection-based social networking sites have sampling bias issues at this time (which is expected to become a less limiting issue in the future) and limitations on data availability. It appears that the most promising future research directions regarding these new/emerging technologies for the purpose of long-distance and rural travel analysis include the following: ⢠Testing of the feasibility of these technological options for long-distance and rural travel data collection, including estimating required sample sizes; ⢠Exploring software applications for event-prompted recall surveys and advanced data impu- tation algorithms to gather and/or generate information on trip purpose, mode, and traveler characteristics and to supplement data directly collected from GPS, smartphone, cell phone, license plate, and/or Bluetooth technologies, while considering the limitations that may be imposed to protect privacy of the users of those technologies; and ⢠Development of methods and procedures that employ advanced technologies to identify long- distance and rural trips, then re-identify the long-distance and rural travelers, and finally conduct follow-up household long-distance and rural travel surveys. GPS-Based Travel Surveys GPS technology has been used, to some extent, in household travel surveys for more than a decade. One of the main uses of GPS technology in travel surveys has been for logging second- by-second GPS point data that can be processed into a robust dataset that includes detailed trip information, such as start and end times and locations, route/link details, and travel speeds. Early
D-6 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models studies used vehicle-based GPS data loggers that were deployed in tandem with travel diaries to collect 1 day of passive GPS data. These data were processed into trips and then compared to reported diary-based household vehicle trips. Over time, as GPS logger technology improved with respect to both power and storage capacity, and decreased in size, this method was modi- fied to include person-based approaches. More than 20 GPS-enhanced travel surveys have been conducted in the United States to date, with a full complement of household travel survey data and GPS traces for a subsample of households (typically 5 to 10 percent of all households). Since the intent of both the vehicle and person-based approach has been to compare diary- reported trips with GPS-measured trips for the same travel day, extended deployment periods for the GPS devices were not justified given the need to get the devices back quickly to redeploy to other households. Given this short deployment duration, the likelihood of capturing long- distance or rural travel in a single GPS day was limited to households that habitually made long-distance or rural trips. However, as GPS device costs continued to decline, more recent household travel surveys have deployed GPS devices for longer durations, sometimes up to 1 week. The advantage of longer duration GPS deployments is that it is more probable that less common or more infrequent long-distance and rural travel could be collected. Currently in the United States, passenger vehicle GPS data are collected mainly for two pur- poses: supplementing/enhancing traditional paper-/telephone-based household travel surveys and providing data for research and pilot tests on VMT-based revenue collection systems. Several Information Needs T ra d it io n al M et h od s G P S S m ar tp h on e L ic en se P la te B lu et oo th C el l P h on e W eb -B as ed S oc ia l M ed ia / N et w or k in g Origin-Destination Y Y Y Y Y Y Y Y Mode: Main/Access/Egress Y M Y Y-auto Y-auto Y Y Y Trip Purpose Y M Y M N M Y Y Routes Y Y Y Y Y Y M M Trip Frequency Y Y Y Y Y Y Y Y Travel Season Y Y Y Y Y Y Y Y Trip Duration Y Y Y Y Y Y Y Y Itinerary and Side Tours Y Y Y M M M Y Y Trip Cost and LOS Y M M M M M Y Y Travel Party Size Y M M M M M Y Y Traveler Characteristics Y M M M N M Y Y Domestic or International Y M M N N N Y Y Other Considerations Passive Data Collection N Y M M Y Y N M Major Privacy Concern M M M N N M Y Y High-Respondent Burden Y M M N N N Y Y Sampling Bias M M M N Y M Y Y Sufficient Sample Size M M Y Y M Y Y Y Note: âMâ (maybe) implies that although the information cannot be directly collected with a specific technology, it may be estimated based on other data sources and/or data post-processing algorithms. Table D.2. Capabilities of various new and emerging technologies.
Other demographic and Origin-destination data D-7 jurisdictions in the United States have recently collected GPS data on passenger travel, which are summarized in this appendix. In general, the sample sizes of these GPS-based surveys are not very large. In addition, most of the GPS surveys target urban travelers and the collected trip samples typically contain few long-distance and rural travel records. Nevertheless, these GPS data collection efforts may provide valuable lessons for possible future surveys that focus on long-distance and rural travel behavior. In the previously mentioned Front Range Travel Counts Survey, conducted in four MPO regions in Colorado, more than 1,000 households were recruited into the GPS augment, which included both a 1-week vehicle GPS component and a 4-day wearable GPS compo- nent in two of the regionsâDenver and Colorado Springs. The preliminary GPS datasets for this study reveal 176 trips out of 27,515 (0.6 percent) with distances greater than 50 miles (made by 61 households), 214 external-external trips (0.8 percent), and 15 external-internal or internal-external trips (0.01 percent). These percentages confirm the rarity of these behaviors, especially in small samples. The latest trend in GPS-enhanced travel surveys is a 100-percent GPS approach, in which households are recruited and report traditional sociodemographic information and then receive person-based GPS devices for 3 or more days. The GPS devices are then retrieved, and the GPS data downloaded and processed into trips and trip details. Sophisticated algorithms are applied to impute travel details such as travel mode, trip purpose, and household travel companions (Wolf, Guensler, and Bachman, 2001). GPS-based prompted recall techniques are applied to a subsample of these households to validate and/or calibrate imputation algorithms (Schuessler and Axhausen, 2009). The appeal of this approach is that other than carrying and charging a GPS data logger for a few days, there is minimal respondent burden once the recruit interview is over. Both long-distance and rural travel can be easily identified within the GPS dataset, with highly accurate trip lengths and travel routes/locations readily available. The benefit of 100-percent GPS travel surveys is that much larger GPS sample sizes (and datasets) are available for data mining. The Greater Cincinnati Household Travel Survey (2009â2011) is the first of its kind and size to adopt a 100-percent GPS sampling methodology (http://www.oki.org/departments/dataservices/faq.html). GPS-Enabled Smartphone Travel Surveys Many smartphones, such as iPhone, Droid, Windows Mobile, and Blackberry phones, have embedded GPS tracking capabilities. It is feasible to collect GPS location data at very frequent time intervals (e.g., several seconds) from smartphone users with or without them being aware of the data collection efforts. For long-distance and rural travel analysis, the value of smart- phones relies, most likely, on third-party applications (i.e., apps) that users choose to install onto their devices either voluntarily for information provided by the apps or based on financial and other incentives. Members of this research team from the University of Maryland have tested several apps that enable researchers to collect GPS location information from smartphone users. Figure D.4 shows these apps running in the foreground of an iPhone and a Droid phone, respectively, and the col- lected GPS location data of a trip originating from Frederick, Maryland, in the early afternoon of March 24, 2011. It is certainly feasible to conduct travel surveys with third-party applications like the one tested by the research team. At this time, there could be sampling biases among smartphone users who are likely to be younger and more affluent than the average traveler. However, this issue is expected to become less serious with smartphones almost becoming must- haves for younger generations. Compared to traditional GPS-based surveys, major advantages of travel surveys based on smartphones include zero device cost (users have already purchased
D-8 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models the phones), zero or very low data transmission/collection cost (usersâ own data plans may cover the location data transmission especially when the data logging frequency is set at a low level), and potentially large sample sizes. There has been recent interest in, and discussion about, GPS-enabled smartphone travel surveys among members of the travel survey research community. The technology certainly offers promise in the sense of convenience to study participants in that an application could be downloaded onto a GPS-enabled smartphone that logs GPS data while the user travels and then prompts the user for details about the trips made. However, there are several key issues to be addressed before this approach is feasible in a large-scale survey effort. These issues include the following: ⢠Technology SupportâThere currently exist several different types of smartphone platforms with varying levels of functionality. Any solution attempting a large sample size would need to support these different models and systems. At present five platforms control over 99 percent of the market. ⢠Incomplete InformationâSome smartphones will require the user to start the survey/GPS logging application at the start of each trip and to close it at the end of each trip. Some older smartphone models (e.g., iPhone 3G or versions of iOS that are older than 4) do not support GPS logging while the phone is in use for other purposes (such as when used for a phone call or to play music). ⢠Power ConsumptionâContinuous GPS data logging significantly increases power draw and may require some users to recharge more frequently than once a day. ⢠Data Plans/CostsâMany smartphone data plans are limited; unless only basic trips details are transferred (rather than complete GPS traces), users are likely to get hit with extra cell phone costs, which could be significant. Users may also not understand this clearly when volunteering for surveys. ⢠Travel Surveys are Typically Conducted at the Household LevelâNot all members (even all adult members) of sampled households are likely to have GPS-enabled smartphones. This would result in incomplete household information. More importantly, many households may not have any GPS-enabled smartphones, causing significant biases in the survey results. Figure D.4. GPS location data collected from smartphones.
Other demographic and Origin-destination data D-9 Web-Based Surveys Surveys implemented on the Web have become prevalent across many industries and top- ics; travel surveys are no exception. Most household travel surveys conducted today include an option for survey participants to report sociodemographic and travel information on-line. Web-based mapping interfaces, implemented using toolkits such as the Google Maps API, allow for real-time geocoding of trip ends by participants as well as provision of actual route traveled. These two features, in turn, enable the automatic identification of both long-distance and rural travel. Limitations of this approach include Internet access to all targeted survey populations and technology expertise in using interactive maps (the latter of which could be circumvented if participants are asked to provide address information only, with geocoding occurring in a post-processing step). Web-Based Surveys Integrated with Social Networking Social networks can be used to increase the reach of traditional Web-survey recruitment efforts by exposing them to a wider audience. For example, one could add social-network-like features to existing Web-survey efforts through integration with platforms such as Twitter for notifying participants of changes or updates to the survey and Facebook for allowing participants to invite friends and family to complete the survey (âsnowball samplingâ). The resulting bias from such an approach must be addressed in analyzing survey response. Location-based features built into social networks can also be leveraged to facilitate the collection of origin and destination information. Bluetooth Technology Bluetooth Traffic Monitoring (BTM) has emerged since 2007 as an anonymous vehicle re- identification technology that has proved to be an effective tool for collecting travel time and trip O/D information. Bluetooth subsystems in consumer electronic devices utilize a unique identifier known as a MAC address to facilitate communications. This unique identifier can be used similarly to toll tags or license plates to identify vehicles at different locations and to assess travel time as well as the O/D distribution of trips through the network. In the past 2 years, several studies have used Bluetooth to sample O/D movements on a small scale to determine turning movements within a freeway or arterial corridor or for distribution of traffic around a major attractor such as a subway stop. The University of Maryland has also recently deployed Bluetooth sensors along the I-95 corridor for travel time, traffic diversion under real-time traveler information, and O/D studies. Theoretically, nothing prevents the use of BTM on long-distance and rural travel O/D data collection, though no such attempt is known to have occurred. The unique electronic IDs are just as applicable to long distances as they are to short or medium distances. The primary limiting factor is the BTM sampling rate, which is known to be generally in the 5-percent range for most areas in the United States. For travel sur- veys and O/D analysis, the anonymous nature of BTM (i.e., no way to retrieve user information from MAC addresses) is both an advantage in that there should be no privacy concerns and a disadvantage in that it is impossible to conduct any follow-up surveys. In summary, Bluetooth Traffic Monitoring ⢠Can anonymously re-identify consumer electronic devices emitting Bluetooth signals (e.g., Bluetooth calling devices installed as vehicle parts, Bluetooth devices such as cell phones carried by drivers and/or passengers in vehicles); ⢠Can be deployed in a temporary, portable format for short-term studies, or permanently for long-term continuous travel time and O/D pattern monitoring; ⢠Can achieve approximately a 5-percent sampling rate;
D-10 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models ⢠Is an emerging technology with ongoing parallel experiments for various transportation oper- ations and planning purposes; and ⢠Does not allow the retrieval of any user information and, therefore, any follow-up surveys. Automatic License-Plate Capture When applied for travel surveys and O/D analysis, automatic license-plate capture (ALC) technology is very similar to Bluetooth technology because both enable re-identification of vehi- cles at multiple sensor locations, which makes it possible to easily single out long-distance and rural trips from other trips. ALC usually is based on high-definition video sensors and machine vision technologies for post-processing video streams. The advantage of ALC, compared to Blue- tooth, is that ALC at least theoretically allows the retrieval of vehicle ownership information. In practice, obtaining vehicle owner information from license plate readings for applications other than law enforcement can be very challenging. On the other hand, Bluetooth is more anony- mous, and Bluetooth sensors can be encased in a protective box and locked to various roadside features. ALC sensors typically require human monitoring for sensor security, unless they are permanently installed above ground. D.3 Emerging Sources of Data from Private Companies There are a variety of new technologies that are being refined and marketed that may have value in the passive collection of long-distance and/or rural travel data. Most of these approaches have not been publicly marketed for these uses, but have been leveraged for passive measure- ment of similar O/D or travel information. In general, these technologies rely on the fact that the majority of the travelers leave âbreadcrumbsâ wherever they go due to their use of mobile electronic devices or credit cards. Privacy concerns and existing regulations have prevented any significant development of this capability in the past. Recently, however, smartphones and personal navigation devices (PND) have circumvented privacy issues through the use of license agreements of software or data services. When a user agrees to the licensing terms of an application or device, there can be clauses in that agreement that allow the licensing organization to access these data for other purposes. These data can then be used as a source of traffic data for real-time or historical traffic data applications and as a source for exploring location-based services. In other cases where there is no relevant license agreement, or existing agreements preclude such derived uses, there are attempts to use detailed information while still protecting personal data. This is usually handled through data aggregation and anonymization techniques that sepa- rate the useful data from anything that can be directly linked to an individual or individual device. Selected examples of these emerging technologies that could be used for long-distance travel or rural travel identification follow, although all of these are missing key household- and person- level sociodemographic information as well as household-level travel details. AirSage Cell Phone Data (www.airsage.com) (Airsage Inc., 2011). Private-sector companies routinely accumulate, anonymize, and analyze cell phone signal data from individual handsets and determine accurate location information and convert it into real-time anonymous location data. AirSage markets cell phone data as a source of real-time traffic information and is now providing their data as a source of O/D data for public agencies. The data generated by their tech- nology comes from the triangulation or translation of phone signals from cell phone towers that have known fixed locations. A device location is not polled at regular intervals, but is generated by some action (an active phone call and the switching between different towers). The resulting
Other demographic and Origin-destination data D-11 dataset may be a little spotty for retracing entire trip details, but it can be significant for reviewing overall long-distance travel patterns. This information does come from a biased sample (users of certain cell phone companies), but has a very large sample of continuous data. Detection of repetitive patterns, in conjunction with land use and demographic data, can be used to classify trips as work, shop, school, etc. TomTom Personal Navigation Device (PND) Data (trafficstats.tomtom.com). TomTom sells PND devices and supporting software. They are now marketing processed results derived from GPS data collected by these devices to transportation planners and engineers that are inter- ested in detailed traffic data. Their marketed solution offers travel time and speed data for almost any route in the United States and Europe. Although the details of their capabilities and methods are not known publicly, it can be assumed that they have access to full data traces from their users, resulting in massive amounts of detailed travel data. Origin-destination products derived from these data have not been marketed to date. GPS-Tracking Systems Installed by Vehicle Manufacturers. GPS-tracking systems installed by vehicle manufacturers to protect the safety and security of vehicles and their owners (e.g., the OnStar system) also have the capability of tracking vehicle locations over time. The GPS data collected from these systems are from a very large sample for long periods of time and include long-distance and rural trips, which make them desirable for long-distance and rural travel analysis. However, these data products are usually proprietary and may be subject to privacy- related scrutiny. A major issue with GPS travel surveys is attributed to the lack of trip purpose and travel mode information in the collected data. However, it is possible to develop statistical and artificial intelligence algorithms (Stopher, Clifford, Zhang, and FitzGerald) to estimate trip purpose and travel mode information with GPS location data, GIS land-use data, andâif availableâsociodemographic information. Smartphone Track Logs via User-Installed Applications. There have been a range of studies around the world that involve GPS-enabled smartphone users installing a GPS logging applica- tion that transfers the GPS point information to a central location for other uses. For exam- ple, the Mobile Millennium Project (traffic.berkeley.edu) was a cooperative research endeavor between UC Berkeley, Nokia Research Center, and NAVTEQ to investigate the provision of a traffic monitoring application that was based on GPS data provided by the users of the applica- tion. These databases could be mined for long-distance and rural trips but would face the same limitations as other datasets in this group; namely, lack of information about the travel charac- teristics, lack of household-level information and trips, and bias in available samples. Wireless Network Locationing Technologies. Compared to GPS-tracking, person/vehicle tracking based on wireless network locationing technologies provides distinct advantages. A very large percentage of U.S. residents already owns cell phones and/or other devices (e.g., RFID, Wi-Fi) that can be tracked by wireless network towers/receivers. This can provide a huge sam- pling frame for long-distance and rural travel studies. Recently, cell phone tracking for traffic monitoring/management and O/D information gathering has become a common practice in other countries. Wireless networks also tend to function much better than GPS underground, inside buildings, and in areas with high-rise/dense structures. The anonymity of trip informa- tion collected via wireless network locationing technology is a major advantage. Since location information from this technology also comes with time stamps, it is relatively straightforward to impute travel modes based on how fast the cell phones move from location to location. Trip purpose information, which is also important for long-distance and rural travel analysis, may be estimated based on collected information such as land use, trip duration, etc. Social Network-Based Location Tracking. Emerging online social network services and fea- tures, such as Foursquare and Facebookâs Check In, allow users to publish their active location
D-12 Long-distance and Rural Travel Transferable parameters for Statewide Travel Forecasting Models information to their network; they also allow users to know who else is at the same place. Par- ticipating merchants can use this information for marketing purposes such as tracking customer loyalty and implementing advertising campaigns. In the context of O/D surveys, this technology could be used to recruit participants to participate in a visitor survey and also assist in the col- lection of personal movement. However, it is very likely that existing end-user agreements would limit how previously collected data can be used to derive travel. D.4 Emerging Sources of Data from Research Organizations Since 2002, the American Transportation Research Institute (ATRI) has assembled and main- tained a substantial database of truck GPS position data, with information coming from more than 600,000 trucks in North America. The interval of GPS reads can be as frequent as 1 to 3 min- utes or as long as 60 minutes depending on location and individual trucks. Each GPS track posi- tion data point is also assigned a unique truck ID and time stamp, which theoretically enables the tracking of the positions of any individual truck and its total vehicle miles traveled. The trucks included in the ATRI GPS database are primarily multi-unit semitrailers owned by large trucking companies, which would not be representative of the entire truck fleet in the nation. Between January and March 2011, ATRI and University of Maryland researchers, who are also part of the NCHRP Project 8-36 team, worked together to develop a preliminary methodology for linking GPS truck position data to TIGER GIS transportation network data. A 250-foot buffer distance was selected, and any truck GPS position points located within this buffer distance of a transportation network polyline were considered points on that polyline. Figure D.5 shows the number of recorded truck position points per mile on individual roadway segments in the City of Baltimore for the whole year of 2010. Overall, the recorded truck travel patterns show higher truck counts on higher-level roads such as Interstate highway and arterial streets. This type of volume graph can be plotted for all regions of the United States with the ATRI data. More detailed analysis taking advantage of the trucking ID and time stamp information can also be conducted. While certain sampling and methodological issues with the ATRI truck position data must be addressed before it can be widely used for long-distance and rural truck travel behavior/ performance analysis, its continuous national coverage is impressive and appealing for national, regional, and state-level analysis.
Other demographic and Origin-destination data D-13 Source: FHWA Office of Freight Management, American Transportation Research Institute and University of Maryland. Figure D.5. Geocoded 2010 ATRI truck GPS data for City of Baltimore.
