Using Truck Fleet Data in Combination with Other Data Sources for Freight Modeling and Planning BDK84-977-20 Final Report Prepared for: Florida Department of Transportation Mr. Frank Tabatabaee, Project Manager Prepared by: University of South Florida Dr. Abdul R. Pinjari Mr. Akbar Bakhshi Zanjani Mr. Aayush Thakur Ms. Anissa Nur Irmania Mr. Mohammadreza Kamali American Transportation Research Institute Mr. Jeffrey Short Mr. Dave Pierce Ms. Lisa Park July 2014 DISCLAIMER The opinions, findings, and conclusions expressed in this publication are those of the authors and not necessarily those of the State of Florida Department of Transportation. ii METRIC CONVERSION SYMBOL WHEN YOU KNOW MULTIPLY BY TO FIND SYMBOL LENGTH in inches 25.4 millimeters mm ft feet 0.305 meters m yd yards 0.914 meters m mi miles 1.61 kilometers km iii TECHNICAL REPORT DOCUMENTATION 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No. 4. Title and Subtitle 5. Report Date Using Truck Fleet Data in Combination with Other Data Sources for July 2014 Freight Modeling and Planning 6. Performing Organization Code USF 7. Author(s) 8. Performing Organization Report No. Abdul R. Pinjari, Akbar Bakhshi Zanjani, Aayush Thakur, Anissa Irmania, Mohammadreza Kamali, Jeffrey Short, Dave Pierce, Lisa Park 9. Performing Organization Name and Address 10. Work Unit No. (TRAIS) College of Engineering Department of Civil and Environmental Engineering University of South Florida 11. Contract or Grant No. 4202 E. Fowler Avenue, ENB118, Tampa FL 33620 BDK84-977-20 12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered Florida Department of Transportation Final Report Research Center 5/3/2011– 6/3/2014 605 Suwannee Street, MS 30 14. Sponsoring Agency Code Tallahassee, FL 32399-0450 15. Supplementary Notes 16. Abstract This project investigated the use of large streams of truck GPS data available from the American Transportation Research Institute (ATRI) for the following statewide freight modeling and planning applications in Florida: (1) Average truck speed data were developed for each (and every) mile of Florida’s Strategic Intermodal System (SIS) highway network for different time periods in the day. (2) Algorithms were developed to convert raw truck GPS data into a database of truck trips. The algorithms were applied to convert four months of raw data, comprising 145 million GPS records, into more than 1.2 million truck trips traveling within, into, and out of Florida. (3) The truck trip database developed from ATRI’s truck GPS data was used to analyze truck travel characteristics, including trip duration, trip length, trip speed, and time-of-day profiles, for different regions in the state. Further, distributions of origin-destination (OD) truck travel times were derived for more than 1,200 OD pairs in the Florida Statewide Model (FLSWM). (4) ATRI’s truck GPS data were evaluated for their coverage of truck traffic flows in Florida. At an aggregate level, the data were found to capture 10 percent of heavy truck volumes observed in Florida. (5) The data were used to develop OD tables of statewide freight truck flows within, into, and out of Florida. To do so, the truck trip database developed from ATRI’s truck GPS data was combined with observed truck traffic volumes at different locations in Florida and other states using OD matrix estimation procedures. The OD tables were developed for the spatial resolution of traffic analysis zones used in FLSWM. (6) Preliminary explorations were conducted with ATRI’s truck GPS data for analyzing truck travel routes between different locations and for analyzing truck flows out of seaports. 17. Key Word 18. Distribution Statement Freight data, truck GPS data, freight performance measures, OD flow table, statewide freight travel demand models 19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price Unclassified Unclassified 238 iv ACKNOWLEDGMENTS The authors would like to express their since appreciation to the Florida Department of Transportation (FDOT) managers on this project, Mr. Vidya Mysore and Mr. Frank Tabatabaee, for their full support, constant guidance, and valuable feedback during the project. Mr. Mysore, former manager of systems traffic modeling at FDOT, spearheaded the project and provided valuable input and feedback. Mr. Tabatabaee served as the project manager in the later stages of the project and played a key role in helping us complete the project. Thanks are due to Mr. Thomas Hill, current manager of the FDOT Systems Planning Office, for his support to the project. Mr. Frank Tabatabaee and Mr. Vladimir Majano provided the research team access to various FDOT datasets required for the project. Thanks to Mr. Daniel Lamb, Ms. Denise Bunnewith, and Mr. Larry Foutz for their input as project panel members. The authors would like to thank the Florida Model Task Force (MTF) members for their input and discussion during the project presentations made at several MTF meetings. Special thanks are due to Mr. Dan Murray of the American Transportation Research Institute (ATRI) for his input and collaboration on the project. Thanks to Dr. Vince Bernardin from Resource Systems Group and Dr. Arun Kuppam from Cambridge Systematics for sharing their experience in working with ATRI’s truck GPS data. Thanks to Mr. Vipul Modi from Citilabs for his assistance with the use of Cube software for the origin-destination matrix estimation procedure. Thanks to Dr. Ramachandran Balakrishna and Dr. Howard Slavin from Caliper Corporation for helpful insights on the theoretical and practical aspects of origin- destination matrix estimation. Thanks to Mr. Krishnan Viswanathan from CDM Smith and Mr. Colin Smith and Ms. Maren Outwater from Resource Systems Group for their input at different stages of the project. The authors would like to thank Mr. Vaibhav Rastogi and Mr. Anirban Ghosh, who implemented several algorithms developed for the project in Java programming platform. Mr. Bertho Augustin and Mr. Jonathan Koons provided valuable assistance with the preparation of different datasets and validation of outputs from the procedures developed in the project. Ms. Cherise Burton and Mr. Akshay Kancharla assisted in some of the validation activities. Finally, the authors thank Ms. Patricia Ball and Ms. Vicki Morrison for their editorial review of this report. v EXECUTIVE SUMMARY An accelerated growth in the volume of freight shipped on American highways has led to a significant increase in truck traffic, influencing traffic operations, safety, and the state of repair of highway infrastructure. Traffic congestion, in turn, has impeded the speed and reliability of freight movement on the highway system. As freight movement continues to grow within and between urban areas, appropriate planning and decision making processes are necessary to mitigate the above-mentioned impacts. However, a main challenge in establishing these processes is the lack of adequate data on freight movements such as detailed origin-destination (OD) data, truck travel times, freight tonnage distribution by OD pairs, transported commodity by OD pairs, and details about truck trip stops and paths. As traditional data sources on freight movement are either inadequate or no longer available, new sources of data must be investigated. A recently-available source of data on nationwide freight flows is based on a joint venture by the American Transportation Research Institute (ATRI) and the Federal Highway Administration (FHWA) to develop and test a national system for monitoring freight performance measures (FPM) on key corridors in the nation. These data are obtained from trucking companies that use GPS-based technologies to remotely monitor their trucks. ATRI’s truck Geographical Position System (GPS) database contains GPS traces of a large number of trucks as they traveled through the national highway system. This provides unprecedented amounts of data on freight truck movements throughout the nation (and Florida). Such truck GPS data potentially can be used to support planning, operation, and management processes associated with freight movements. Further, the data can be put to better use when used in conjunction with other freight data obtained from other sources. The overarching goal of this project is to investigate the use of ATRI’s truck GPS data for statewide freight performance measurement, statewide freight truck flow analysis, and other freight planning and modeling applications in the state. The specific objectives are to: 1) Derive freight performance measures for Florida’s Strategic Intermodal System (SIS) highways, 2) Develop algorithms to convert large streams of ATRI’s truck GPS data into a more useable truck trip format, 3) Analyze truck trip characteristics in Florida using ATRI’s truck GPS data, 4) Assess ATRI’s truck GPS data in terms of its coverage of truck traffic flows in Florida, 5) Develop OD tables of statewide freight truck flows within, into, and out of Florida for different geographic resolutions, including the Florida Statewide Model (FLSWM) traffic analysis zones (TAZs), and 6) Explore the use of ATRI’s GPS data for other applications of interest to Florida, including the analysis of truck flows out of two seaports, the re-routing patterns of trucks after a major highway incident, and the routing patterns of trucks traveling between Jacksonville and Ocala. vi Project Outcomes and Findings The outcomes and findings from the project are discussed next. Freight Performance Measures on Florida’s SIS Highway Network The project resulted in the development of average truck speed data for each (and every) mile of the Strategic Intermodal System (SIS) highway network for different time periods in the day— AM peak, PM peak, mid-day, and off-peak—using three months of ATRI’s truck GPS data in the year 2010. In doing so, it was found that the existing shape files of the SIS network available from FDOT either were not accurate enough or they lacked the details (for example, separate links by direction for divided highways) to derive performance measures using geospatial data. Therefore, a highly accurate network was developed to represent highways on the SIS network. The SIS highway network shape file and the data on average truck speeds by time-of-day were submitted in a GIS shape file that can be used in an ArcGIS environment to identify the major freight bottlenecks on Florida’s SIS highway network. In addition to the development of average speed measures, the project developed example applications of ATRI’s truck GPS data for measuring truck speed reliability and analyzing highway freight bottlenecks. Algorithms to Convert ATRI’s Raw GPS Data Streams into a Database of Truck Trips The raw GPS data streams from ATRI need to be converted into a truck trip format to realize the full potential of the data for freight planning applications. The project resulted in algorithms to convert the raw GPS data into a database of truck trips. The results from the algorithms were subjected to different validations to confirm that the algorithms can be used to extract accurate trip information from raw GPS data provided by ATRI. These algorithms were then applied to four months of raw GPS data from ATRI, comprising a total of 145 million raw GPS data records, to develop a large database of truck trips traveling within, into, and out of the state. The resulting database comprised more than 1.2 million truck trips traveling within, into, and out of the state. This database of truck trips can be used for a variety of purposes, including the development of truck travel characteristics and OD truck flow patterns for different geographical regions in Florida. The database can be used to calibrate and validate the next-generation statewide freight travel demand model being developed by FDOT. In future work, this database potentially can be used to develop data on truck trip-chaining and logistics patterns in the state. Analysis of Truck Travel Characteristics in Florida The truck trip database developed from four months of ATRI’s truck GPS data was used to analyze a variety of truck travel characteristics in the state of Florida. The truck travel characteristics analyzed include trip duration, trip length, trip speed, time-of-day profiles, and OD flows. Each of these characteristics was derived at a statewide level as well as for different regions in the state—Jacksonville, Tampa Bay, Orlando, Miami, and rest of Florida—defined based on the freight analysis framework (FAF) zoning system. In addition, the truck trips were used in conjunction with the GPS data to derive distributions of OD travel distances, travel times, and travel speeds between more than 1,200 TAZ-to-TAZ OD pairs in the FLSWM. Comparing the minimum truck travel times measured vii using GPS data for the 1,200 OD pairs with free flow travel times used as inputs to FLSWM indicated that the FLSWM travel times are systematically underestimated when compared to the truck travel times measured from ATRI data. A similar comparison with the travel times extracted from Google Maps suggested that the Google Maps travel times also underestimate (albeit not as much the FLSWM travel times) truck travel times between origins and destinations. This is most likely because the travel times used as inputs for the FLSWM and those reported by Google Maps are predominantly geared toward passenger cars that tend to have higher travel speeds and better acceleration characteristics. ATRI’s truck GPS data, on the other hand, provide an opportunity to accurately measure travel times exclusively for trucks (and for different times of the day). In addition to the measurement of OD truck travel distances, travel times, and speeds, the project team performed an exploratory analysis of truck travel routes for more than 1,600 trips between 10 OD pairs in FLSWM. A preliminary exploratory analysis suggested that a majority of trips between any OD pair tend to travel by largely similar routes (i.e., the variability in route choice is not high for the 10 OD pairs examined in this study). Specifically, considerable overlap was observed among the routes across a large number of trips between an OD pair. This observation has interesting implications for future research on understanding and modeling truck route choice. While this study did not delve further into understanding the route choice patterns of long-haul trucks, this is an important area for future research using the truck GPS data from ATRI. Assessment of ATRI’s Truck GPS Data and Its Coverage of Truck Traffic in Florida This project resulted in a better understanding of ATRI’s truck GPS data in terms of its coverage of truck traffic in the state of Florida. This includes deriving insights on (a) the types of trucks (e.g., heavy trucks and medium trucks) present in the data, (b) the geographical coverage of the data in Florida, and (c) the proportion of the truck traffic flows in the state covered by the data. Based on discussions with ATRI and anecdotal evidence, it is known that the major sources of ATRI data are freight shipping companies that own large trucking fleets, which typically comprise tractor-trailer combinations (or Federal Highway Administration [FHWA] vehicle type classes 8–13). However, a close observation of the data, from following the trucks on Google Earth and examining travel characteristics of individual trucks, suggests that the data include a small proportion of trucks that are likely to be smaller trucks that do not necessarily haul freight over long distances. The project used simple rules to divide the data into two categories: (1) long-haul trucks or heavy trucks (considered to be FHWA vehicle classes 8–13), and (2) short-haul trucks or medium trucks. Specifically, trucks that did not make at least one trip of 100-mile length in a two-week period and those that made more than 5 trips per day were considered short-haul or medium trucks. Out of a total of 169,714 unique truck IDs in the data, about 4.6 percent were labeled as short-haul trucks (or medium trucks) and separated from the remaining long-haul trucks (or heavy trucks). In future work, it will be useful to derive better definitions of heavy trucks and medium trucks. Whereas heavy trucks are of primary interest to FLSWM for updating and validating its freight truck model components (assuming these are the long-haul freight carrying trucks), medium trucks are also of potential use for updating the non- freight truck model components in FLSWM. Further, extracting sufficient data on medium trucks viii potentially can help understand truck movement within urban regions as well, because a considerable proportion of truck traffic in urban areas tends to comprise medium trucks. ATRI’s truck GPS data represent a sample of truck flows within, coming into, and going out of Florida. This sample is not a census of all trucks traveling in the state. Also, it is unknown what proportion of heavy truck flows in the state is represented by this data sample. To address this question, truck traffic flows implied by one-week of ATRI’s truck GPS data were compared with truck counts data from more than 200 Telemetered Traffic Monitoring Sites (TTMS) in the state. The results from this analysis suggest that, at an aggregate level, the ATRI data provide 10.1 percent coverage of heavy truck flows observed in Florida. When the coverage was examined separately for different highway facilities (based on functional classification), the results suggest that the data provide a representative coverage of truck flows through different types of highway facilities in the state. The coverage of ATRI data was examined for different geographical regions in Florida by examining the spatial distribution of the number of truck trips generated at TAZ-level and at county-level geography. In addition, the percentage of heavy truck traffic covered by ATRI data at different locations was examined. All these examinations suggest potential geographical differences in the extent to which ATRI data represent heavy truck traffic volumes at different locations in the state. For instance, truck trips generated from Polk County were much higher than those generated from Hillsborough and Miami-Dade counties. Further, the percentage of heavy truck traffic covered by ATRI data in the southern part of Florida (within Miami) and the southern stretch of I-75 is slightly lower compared to the coverage in the northern and central Florida regions. Such geographical differences (or spatial biases) potentially can be adjusted by combining ATRI’s truck GPS data with observed data on truck traffic flows at different locations in the state (from FDOT’s TTMS traffic counting program). OD Tables of Statewide Truck Flows An important outcome of the project was to use ATRI’s truck GPS data in combination with other available data to derive OD tables of freight truck flows within, into, and out of the state of Florida. The OD flow tables were derived at the following levels of geographic resolution: a) TAZs of the FLSWM, where Florida and the rest of the country are divided into about 6,000 TAZs, b) County-level resolution, where Florida is represented at a county-level resolution and the rest of the country is represented at a state-level resolution, and c) State-level resolution, where Florida and the rest of the country are represented at a state-level resolution. As part of this task, first, the truck trip database developed from four months of ATRI’s GPS data was converted into OD tables at the TAZ-level spatial resolution used in the FLSWM. Such an OD table derived only from the ATRI data, however, is not necessarily representative of the freight truck flows in the state. This is because the ATRI data does not comprise the census of trucks in the state. Besides, it is not necessarily a random sample and is likely to have spatial biases in its representation of truck flows in the state. To address these issues, the OD tables derived from the ATRI data were combined with observed truck traffic volumes at different ix locations in the state (and outside the state) to derive a more robust OD table that is representative of the freight truck flows within, into, and out of the state. To achieve this, a mathematical procedure called the Origin-Destination Matrix Estimation (ODME) method was employed to combine the OD flow table generated from the ATRI data with observed truck traffic volume information at different locations within and outside Florida. The OD flow table estimated from the ODME procedure is likely to better represent the heavy truck traffic volumes in the state, as it uses the observed truck traffic volumes to weigh the ATRI data-derived truck OD flow tables. Explorations of the Use of ATRI’s Truck GPS Data for Other Applications In addition to the above, this project conducted preliminary explorations of the use of ATRI’s truck GPS data for the following applications: a) Analysis of truck flows out of two ports in Florida—Port Blount Island in Jacksonville and Port Everglades in Fort Lauderdale, b) Analysis of routing patterns of trucks that used the US 301 roadway to travel between I-95 around Jacksonville and I-75 around Ocala, and c) Analysis of changes in truck routing patterns during the closure of a stretch of I-75 near Ocala due to a major multi-vehicle crash in January 2012. Note that these applications were only preliminary explorations conducted as proofs of concept. Future work can expand on these explorations to conduct full-scale applications. x