Guidebook on Pedestrian and Bicycle Volume Data Collection (2014)

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.

1 C H A P T E R 1 1.1 About This Guidebook This Guidebook on Pedestrian and Bicycle Volume Data Collection is a resource for practi- tioners who are, or would like to be, involved in collecting non-motorized count data. The guidebook describes methods and technologies for counting pedestrians and bicyclists, with a focus on automated technologies. The guidebook also provides guidance on developing a data collection program, selecting appropriate counting technologies, and using the data to help fulfill an organization’s mission. 1.1.1 Guidebook Objectives This guidebook is designed to help practitioners • Understand the value of collecting pedestrian and bicycle volume data; • Develop a pedestrian and bicycle data collection plan, including identifying count locations, determining count frequencies and durations, and establishing data management and sharing standards; Introduction Chapter 1 Topics • Guidebook objectives • Guidebook organization • Guidebook development • Topics covered in the guidebook • Related topics not covered in the guidebook • Relationship of the guidebook to the FHWA Traffic Monitoring Guide • Important differences between motorized and non-motorized counting • Components of a non-motorized volume counter Sidebars Sidebars such as this are provided throughout the guidebook to provide more detail on particular topics and to provide examples of real-world applications.

2 Guidebook on Pedestrian and Bicycle Volume Data Collection • Identify and recommend data collection methods that will meet their project needs, while considering the organization’s available resources; and • Correct raw count data to account for systematic over- or undercounting resulting from the use of a particular counting technology. 1.1.2 Guidebook Organization This guidebook contains five main chapters. It is not necessary to read the entire guidebook cover to cover to make use of it. In particular, Chapters 4 and 5 provide reference information that applies to various counting technologies; readers only need to look at the portions of these chapters that apply to the specific technologies they are using or considering using. The guidebook’s chapters are organized as follows: 1. Introduction: This chapter provides an overview of the guidebook’s purpose and describes related topics outside the guidebook’s scope. This chapter also introduces the components of a counting system. Finally, the chapter provides a brief summary of the research behind this guidebook. 2. Non-Motorized Count Data Applications: This chapter describes and illustrates potential uses of pedestrian and bicycle data. 3. Data Collection Planning and Implementation: This chapter is the heart of the guidebook— it provides guidance on the various steps of the process involved with starting, implementing, maintaining, and expanding a non-motorized volume data collection program. 4. Adjusting Count Data: This chapter provides adjustment factors for correcting raw count data to account for systematic over- or undercounting associated with the use of a particular counting technology. 5. Sensor Technology Toolbox: This chapter provides a toolbox of the existing and emerg- ing sensor technologies available for non-motorized counting, including information on each technology’s strengths and limitations, accuracy, costs, and availability and usage in the United States as of the time of writing. 1.1.3 Guidebook Development and Research Objective This guidebook is a product of NCHRP Project 07-19, “Methods and Technologies for Pedestrian and Bicycle Volume Data Collection.” The objective of this research was to assess existing, new, and innovative technologies and methods and to provide guidance for trans- portation practitioners on how to best collect pedestrian and bicycle volume data. This assessment was to consider, among other factors, the feasibility, availability, quality, reliability, cost, and compatibility of various counting technologies. A summary of the research activities conducted as part of this project can be found in the project’s final report (Ryus et al. 2014). 1.2 Guidebook Scope 1.2.1 Topics Covered in the Guidebook This guidebook focuses on methods and technologies for collecting non-motorized volume data. It addresses both manual and automated methods, although it emphasizes automated methods, as these have been covered less comprehensively in the literature to date. A mature non-motorized counting program typically employs a mix of manual and automated counts. The guidebook addresses both intersection and screenline counts, but focuses on screenline counts, as most of the automated counting technologies on the market are used for such counts.

Introduction 3 Screenline counts are counts of the number of pedestrians or bicyclists crossing an imaginary line. Intersection counts include counts of pedestrians crossing each roadway leg or counts of bicyclists turning left, turning right, or going straight. Intersection counts have a number of uses, but the technologies available at present to collect these counts are limited to (1) manual counts in the field, (2) manual counts from video, and (3) automated counts from video. (The same technologies are also used for collecting motorized vehicle counts at intersections.) As previously mentioned, the guidebook also discusses how to develop a non-motorized count program and how to adjust raw counts to account for systematic errors associated with a particular counting technology. 1.2.2 Related Topics This guidebook covers methods for counting the actual number of pedestrians or bicyclists pass- ing a given point or screenline or crossing an intersection. Three topics related to non-motorized volume counting not covered in this guidebook are trip sampling techniques, presence detection, and trip generation estimation. Trip Sampling Techniques As the name suggests, sampling techniques count a sample of the total volume passing a loca- tion or traveling between two points. These techniques are better suited for evaluating origin– destination travel patterns, investigating traveler route choice decisions, and estimating overall mode split, than they are for estimating volumes at a given location. Examples of sampling techniques include the following. Bluetooth and WiFi Detection Electronic devices with actively engaged Bluetooth and/or WiFi communication capabilities regularly transmit “here I am”–type messages. Included in these messages is a unique identifier (Media Access Control, or MAC, address) associated with the device’s Bluetooth or WiFi trans- mitter. Bluetooth or WiFi readers can detect and record these MAC addresses; by comparing the times and locations when a particular MAC address was recorded by different readers, a possible route and travel time can be estimated. It is not possible to differentiate between modes (e.g., motor vehicles, bicyclists, pedestrians) by this means; therefore, its application to pedes- trian and bicycle studies is limited to isolated non-motorized environments, such as trails, malls, and stadiums (Liebig and Wagoum 2012). Estimating total pedestrian or bicycle volumes from these data samples is problematic even in these isolated locations, due to the need for location- specific adjustment factors, such as • Percentage of users with Bluetooth-enabled devices, • Percentage of Bluetooth-enabled devices turned on, and • Percentage of users with multiple Bluetooth devices (e.g., phone and earpiece). GPS Data Collection Multiple efforts have used standalone GPS units or smartphone applications that use a phone’s GPS functionality to collect non-motorized trip data (Hood et al. 2011). These applications have been used primarily to evaluate route choice, but have also been used to compare demand at different locations. The sample data collected through this method can be used to establish minimum volumes at a location, but cannot be adjusted to estimate total pedestrian or bicycle volumes. Sample bias is also an issue with these technologies, as those being counted have to opt-in to the program and—with smartphone apps—have to own a smartphone and remember to use the app on each trip.

4 Guidebook on Pedestrian and Bicycle Volume Data Collection Radio Frequency ID (RFID) Tags RFID tags are commonly used in the logistics industry for tracking individual packages and containers. The tags can be read at a distance of 5 to 10 meters, depending on the antenna power and particular radio frequency used (Andersen 2011). As with GPS-based methods, sample bias is an issue with this technology, as people have to “opt-in” to the program by placing tags on their bicycles. Unlike GPS-based methods, a bicycle’s position is known only at specific locations where an RFID reader has been placed. Fredericia, Denmark, uses these tags as part of a program to encourage residents and commuters to bike more often, with each “check-in” at a location counting as an entry to a prize drawing (www.cykelscore.dk). Bike Sharing Data Bike sharing stations can record the identification number of a bike when it is checked in or out at a particular station. Some bike sharing programs also equip their bicycles with GPS devices. These data can be used to estimate origin–destination patterns and possible routes and travel times. Because the data are reflective of bike share users only, rather than the bicycling popula- tion as a whole, and because the bicycle’s location may only be known with certainty at the bike share stations, it is not a practical method for determining actual bicycle volumes at a given time and place. Pedestrian Signal Actuation Buttons At some traffic signals, pedestrians have to push a button to activate the walk signal for the pedestrian crossing. The number of requests can be stored, and researchers have found that signal activation rates can be a reasonable proxy for determining relative rates of pedestrian demand (Day et al. 2011). However, these researchers also stated that observing these rates is not an effective method for collecting total pedestrian counts. At the time of writing, Portland, Oregon, counted and stored pedestrian button activations at 14 locations, with more locations planned, and was investigating the possibility of developing relationships between actuations and demand, based on-site characteristics (Kothuri et al. 2012b). Surveys Surveys can be used to collect such pedestrian- and bicyclist-related data as mode share and origin–destination patterns. Mode shares can then be extrapolated to determine total pedestrian volumes for a larger area, such as within a neighborhood or traffic analysis zone (TAZ). However, estimates made this way are not suitable for collecting count data, due to the relatively small sample size in contrast with a relatively large sample area with complex land use patterns. Presence Detection Some of the sensor technologies discussed in this guidebook can be used to detect pedestrians and bicycles. For example, they may be used to detect the presence of bicyclists or pedestrians at a traffic signal, so the traffic signal can adjust its timing to serve those users. They can also be used to detect the presence of a pedestrian or bicyclist in an unauthorized area, such as a tunnel. Because these applications focus on whether or not a pedestrian or bicyclist is present, rather than determining the number of people present, these applications are not a substitute for counting. However, at the time of writing, vendors that specialized in presence-detection applications were working on expanding their functionality to include counting applications. Trip Generation Transportation engineers have long used the Institute of Transportation Engineers’ Trip Gen- eration Manual (ITE 2012) to estimate the number of motorized vehicle trips that would be generated by a particular land use. A recent NCHRP Project 08-78, “Estimating Bicycling and Walking for Planning and Project Development,” has developed a guidebook (published as NCHRP

Introduction 5 Report 770) with guidance on estimating pedestrian and bicycle trip generation associated with particular types of trips and demand to use particular pedestrian and bicycle facilities (Kuzmyak et al. 2014). A fundamental difference between trip generation estimation and volume count- ing is that the former forecasts future demand, while volume counting measures current usage. However, improved and more widespread volume counting can aid the development of better trip generation estimates. 1.2.3 Relationship to the Traffic Monitoring Guide The FHWA’s Traffic Monitoring Guide (TMG) provides “up to date guidance to State high- way agencies in the policies, standards, procedures, and equipment typically used in a traffic monitoring program” (FHWA 2014). The 2013 edition of the guide includes information on non-motorized traffic counting; Chapter 4 specifically addresses traffic monitoring for non-motorized traffic. This NCHRP research project and guidebook complement the FHWA guide by providing more recent data on the accuracy of various counting technologies and providing real-world examples of non-motorized traffic counting applications. The TMG is available at http://www.fhwa.dot.gov/policyinformation/tmguide/. 1.3 Non-Motorized Counting Concepts 1.3.1 Differences between Motorized and Non-Motorized Traffic Counting Developing a non-motorized traffic count program presents unique challenges in comparison to motorized vehicle counting. Most transportation agencies have standard practices for collect- ing vehicular counts and have historical data available for assessing daily, seasonal, and annual trends. Agencies have a history of using “rules-of-thumb” to expand short-term motorized counts based on a roadway’s classification and character, as well as the time of day and year when the short-term count was taken. In contrast, there is limited U.S. guidance available on bicycle and pedestrian data collection—mostly developed since 2004 (e.g., the National Bicycle and Pedestrian Documentation Program [NBPD, Alta Planning + Design 2012], the Traffic Monitoring Guide [FHWA 2013])—and few agencies have developed a standard local practice. Differences in Demand Variability One key difference between non-motorized and motorized volume counting that must always be kept in mind is that non-motorized volumes are much more sensitive to environmental conditions—precipitation, temperature, darkness, etc.—than are motorized vehicle volumes. Figure 1-1 compares observed hourly bicycle volumes on a multi-use path in Minneapolis with observed hourly automobile volumes on a parallel freeway a couple of miles away, for 1 week in October 2013. Although the auto volumes are fairly similar, with the p.m. peak-hour volume Hyperlinks In the PDF version of this guidebook, click on web addresses to go directly to that website or document. Website addresses are subject to change; if a link no longer works, use an Internet search engine to find the document or site.

6 Guidebook on Pedestrian and Bicycle Volume Data Collection varying only 5% from the lowest to the highest volume day, the bicycle volumes show 200% variability in the p.m. peak hour. According to Weather Underground’s archived weather data for Minneapolis, it rained about 1 inch on Tuesday, ½ inch on Monday and Thursday, 0.1 inch on Friday, and 0.01 inch on Wednesday. Another critical difference is that hourly bicycle or pedestrian volumes at a given count site tend to be relatively low, compared to the volumes observed at typical motorized vehicle count sites; the lower volumes also contribute to higher day-to-day variability. The greater variability present in non-motorized volumes means that factoring techniques used to estimate long-term (e.g., annual) motorized volumes based on short-term (i.e., 24-hour or less) counts are not necessarily appropriate for non-motorized counting. For example, a 12-hour motorized vehicle count could be converted to a daily count by dividing the counted Sources: Auto: Minnesota DOT automatic traffic recorder #326 (I-394); Bicycle: NCHRP Project 07-19 testing, Midtown Greenway. 0 2,000 4,000 6,000 8,000 10,000 12,000 0 6 12 18 24 H ou rl y Vo lu m e (v eh /h ) Hour Star ng Monday Tuesday Wednesday Thursday Friday Annual Weekday Average Figure 1-1. Comparative variability of automobile (top) and bicycle (bottom) volumes, October 14–18, 2013.

Introduction 7 volume by the proportion of daily traffic occurring on average during the count period (based on previous 24-hour or longer counts), and the result could then be adjusted for monthly variations in traffic, as determined from data from a permanent counting station, to reach an estimate of average annual daily traffic (AADT). In comparison, a study of sites in Boulder, Colorado, found that estimates of average annual bicycle traffic (AABT) based on 12 hours of counts from a mid- week day had an average error of 40% from the true value (Nordback et al. 2013). Scandinavian research also confirms the difficulty of estimating AABT (Danish Road Directorate 2004) or year-to-year change in bicycle volumes (Niska et al. 2012), even when 1 week of count data was available. The longer timeframes required for accurate non-motorized volume estimates require greater usage of automated counting techniques. Ease of Detection Pedestrians and bicyclists are more challenging to detect than motor vehicles because they are smaller, move in less regular patterns, and are not confined to fixed lanes, as are vehicles. Pedes- trians and bicyclists may travel outside designated walkways and bikeways, take unmarked short- cuts, or stop unexpectedly. In addition, pedestrians and bicyclists may travel in groups, which some types of counting technologies have difficulties distinguishing. In contrast, motor vehicles are large, metal objects that move with relatively sizable gaps between each vehicle, which make them easier to detect (FHWA 2013). Experience with Counting Technology Another key difference between motorized and non-motorized traffic counts relates to the level of experience with motor vehicle counts versus non-motorized counts. The technologies commonly used to count motor vehicles are well established, counting errors associated with particular technologies are understood, and methods for addressing errors are fairly well devel- oped (FHWA 2013). In contrast, some of the counting technologies used for non-motorized counting are different from those commonly used for motorized vehicle counting, and new technologies are emerging. Even when counting technologies are similar (e.g., pneumatic tubes, inductive loops), the counting errors associated with these technologies can be different for non-motorized users than for motorized users. Therefore, one of the key objectives of this guidebook is to expand the available knowledge about the accuracy of non-motorized count- ing technologies. 1.3.2 Counting System Components Although it is easy to focus on the technology used to detect pedestrians or bicyclists when evaluating counting devices, it is important to remember that the sensor technology is just one piece of the overall counting system and that all of the pieces need to work well together for a successful application. This section briefly describes the elements constituting a count- ing device. Sensor The sensor is the portion of the counting device that detects pedestrians or bicyclists. Various sensor technologies are available for non-motorized counting, including • The human eye (for manual counting in the field); • Video cameras (for use with manual counting in the office or as an input to an automated method for identifying pedestrians or bicyclists); • Pneumatic tubes, where a pulse of air is generated when a bicyclist rides over the tube; • Piezoelectric strips, which emit an electrical signal when deformed by bicycle wheels passing over them;

8 Guidebook on Pedestrian and Bicycle Volume Data Collection • Fiberoptic pressure sensors, which detect changes in the amount of light transmitted through buried fiberoptic cable based on the amount of pressure (weight) applied to the cable; • Inductive loops, where the magnetic field produced by an electrical current running through buried metal loops is changed when the metal parts of a bicycle pass over the loops; • Magnetometers, which detect changes in the Earth’s magnetic field when the metal parts of a bicycle pass over the detector; • Pressure and acoustic sensors, which are placed under a pathway and detect the weight (pressure) or footsteps (acoustic) of passers-by; • Passive infrared and thermal sensors, which detect the infrared radiation (heat) emitted by pedestrians and bicyclists; • Active infrared and radio beam sensors, which send infrared light or radio beams from a trans- mitter to a receiver and detect when the beam is broken; and • Laser scanning, in which laser pulses are emitted and the characteristics of the reflected pulses are used to detect pedestrians and bicyclists. Counter (Processor) The counter receives the sensor output and determines whether or not a detection should be recorded. In many cases, the counter’s sensitivity can be adjusted to reduce the number of false- positive or false-negative detections. In some implementations, this is a component of the data collection unit, while in others this is a processing step undertaken on a personal computer after the data have been collected. Data Logger When a pedestrian or bicyclist is detected, this information is sent to the device’s data logger. Some devices record every detection, along with the time of detection (timestamping), while others keep a running count of detections and store the totals in 15- or 60-minute bins. Typi- cally, when the data logger’s memory becomes full, the data logger begins to overwrite data, starting with the oldest data. The amount of available memory and the way it is used helps determine how often data must be downloaded from the device. Power Supply The counting device requires power to work, which can be provided by a replaceable battery, a permanent electrical connection, a solar panel, or other means. Battery life helps determine how often a maintenance trip must be made to the counter to change the battery. Communications A method is needed to transfer count data from the device to the user’s database. Possible methods include communications ports for physically attaching a laptop computer or memory card, wireless (e.g., Bluetooth or WiFi) connections, or a cellular modem. The communications method has an associated cost (e.g., staff time to visit the counter to download data or sub- scription fees for cellular data service). The method may also influence how often data can be obtained from the device and how quickly problems with the device are detected. Securement A counting device usually needs some sort of physical housing (e.g., a box or a utility cabinet) to protect it from the elements and vandalism and a method of fastening it in place to prevent undesired movement or vibration and to discourage theft. During the research done for this project, it was sometimes difficult to obtain permission to install a device in a desired location if the method of securing the device was deemed to be visually unattractive. Additionally, one unit with a less-robust securement mechanism was stolen during the course of testing.

Introduction 9 Data Management Some counting products come with their own data management software; others require end users to manage the count data themselves (e.g., by storing it in a spreadsheet). The Counting Device as a Whole The complete counting device includes all of the above components. Different vendors may incorporate the same sensor technology, but use different components for other aspects of the device, which can affect the suitability of a particular product for a particular location. Some vendors offer the opportunity for the end user to customize certain components (e.g., the power supply). Finally, the customer service and training provided by the vendor plays an important role in a successful counting device deployment.