Krista Nordback

Estimating Annual Average Daily Bicyclists: Error and Accuracy

Cities around the United States are investing in bicycle infrastructure, and to secure additional transportation funding, cities are reporting bicycle use and safety improvements. Data on bicyclist traffic volume is necessary for performing safety studies and reporting facility use. Meeting the need for data, available manual bicycle counting programs count cyclists for a few hours per year at designated locations. A key issue in the design of counting programs is determining the timing and frequency of counts needed to obtain a reliable estimate of annual average daily bicyclists (AADB). In particular, in which days of the week, hours of the day, and months of the year should counts be collected? And, most important to program cost, how many hours should be counted? This study used continuous bicycle counts from Boulder, Colorado, to estimate AADB and analyze the estimation errors that would be expected from various bicycle-counting scenarios. AADB average estimation errors were found to range from 15% with 4 weeks of continuous count data to 54% when only 1 h of data was collected per year. The study found that the most cost-effective length for short-term bicycle counts is one full week when automated counting devices specifically calibrated for bicycle counting are used. Seasons with higher bicycle volumes have less variation in bicycle counts and thus more accurate estimates.

Methodology to Characterize Ideal Short-Term Counting Conditions and Improve AADT Estimation Accuracy Using a Regression-Based Correcting Function

Transportation agencies’ motor vehicle count programs tend to be well established and robust with clear guidelines to collect short-term count data, to analyze data, develop annual average daily traffic (AADT) adjustment factors, and to estimate AADT volumes. In contrast, bicycle and pedestrian traffic monitoring is an area of work for most transportation agencies. In most agencies, there are a low numbers of counting sites and limited agency experience to manage a city-wide or state-wide system of collecting, processing, and using nonmotorized data. Short duration counts are used to estimate longer duration volumes such as AADT. Because bicycle or pedestrian short-term counts vary dramatically over time and significantly more than motorized vehicle counts, the direct application of motorized vehicle AADT estimation methods may be inadequate. The goal of this paper is to present a methodology that will enhance, if needed, existing AADT estimation methods widely employed for motorized vehicle counts. The proposed methodology is based on the analysis of AADT estimation errors using regression models to estimate a correcting function that accounts for weather and activity factors. The methodology can be applied to any type of traffic with high volume variability but in this research is applied to a permanent bicycle counting station in Portland, Oregon. The results indicate that the proposed methodology is simple and useful for finding ideal short-term counting conditions and improving AADT estimation accuracy.

Institutionalizing Bicycle and Pedestrian Monitoring Programs in Three States

Information about nonmotorized traffic is needed to support management of transportation systems. However, transportation officials across the United States generally have not monitored nonmotorized traffic, and most agencies lack bicycle and pedestrian counts. This paper describes current efforts by the Colorado, Minnesota, and Oregon Departments of Transportation (DOTs) to establish programs for monitoring non-motorized traffic. With FHWA principles for traffic monitoring as a framework, this case study summarizes state approaches for initiating monitoring, agency collaboration with local governments, and continuous and short-duration monitoring efforts. Agency protocols for data collection, analysis, and management, including development of factors for purposes such as estimating average daily bicyclists or bicycle miles traveled, are also compared. Agency efforts to demonstrate the effectiveness of monitoring technologies are described. This study reveals similar objectives across states, both similarities and differences in approaches, differing rates of implementation, and similar problems in implementation. The paper summarizes lessons learned and identifies challenges that DOTs will face in institutionalizing the monitoring of nonmotorized traffic.

Accuracy of Bicycle Counting with Pneumatic Tubes in Oregon

Interest in counting bicycles and establishing nonmotorized counting programs is increasing, but jurisdictions still struggle with how to integrate bicycle counting into standard practice. In this paper, the authors share findings and recommendations for how to minimize error for bicycle counting from tests conducted in conjunction with the Oregon Department of Transportation. This research studied three types of off-the-shelf pneumatic tube counters for counting bicycles, including equipment from five manufacturers: two bicycle-specific counters, three varieties of motor vehicle classification counters, and one volume-only motor vehicle counter. Tests were conducted both in a controlled environment and in on-road mixed traffic to better identify problems in accuracy. Equipment studied generally undercounted cyclists, especially those in groups. Results from the controlled test with standard bicycles showed that within 10 ft of the counter, the undercounting error ranged from 0% to −12%. In the mixed-traffic test, all the equipment tested tended to undercount with mean percent error ranging from −10% to −73%. Each counter type has pros and cons, but in general, counting accuracy decreased with increases in bicycle and motor vehicle traffic and longer tube lengths. Higher accuracy can be achieved by careful selection of equipment type, classification scheme, and tube configuration. Bicycle speeds given by off-the-shelf pneumatic counting equipment were accurate.

Creating a National Nonmotorized Traffic Count Archive: Process and Progress

Robust bicycle and pedestrian data on a national scale would help promote effective planning and engineering of walking and bicycling facilities, build the evidence-based case for funding such projects, and dispel notions that walking and cycling are not occurring. To organize and promote the collection of nonmotorized traffic data, a team of transportation professionals and computer scientists is creating a national bicycle and pedestrian count archive. This archive will enable data sharing by centralizing continuous and short-duration traffic counts in a publicly available online archive. Although other archives exist, this will be the first archive that will be national in scope and enable data to be uploaded directly to the site. This archive will include online input, data quality evaluation, data visualization functions, and the ability to download user-specified data and exchange the data with other archives and applications. This paper details the first steps in creating the archive: (a) review count types, standard formats, and existing online archives; (b) list primary functional requirements; (c) design archive architecture; and (d) develop archive data structure. The archives versatile data structure allows for both mobile counters and validation counts of the same traffic flow, an innovation in design that greatly expands the usefulness of the archive.

Measuring Traffic Reduction from Bicycle Commuting

Increased bicycle commuting is said to reduce motor vehicle traffic, but can this reduction be measured? To address one aspect of this question, Bike to Work Day events are used to explore how the volume of motor vehicle traffic changes when the volume of bicycle traffic doubles. In many communities, Bike to Work Day is an annual 1-day event that encourages commuting by bicycle. Because event participation rates were high in Boulder, Colorado, the city was chosen as an example to investigate this issue. Bicycle and motor vehicle traffic volumes measured by continuous counters on Bike to Work Day were compared with volumes from comparable workdays. Bicycle volume on Bike to Work Day was roughly double that on comparable workdays; motor vehicle volume on Bike to Work Day was lower than on comparable Wednesdays in 88% of the 16 cases studied and significantly lower at the 80% confidence level in 31% of the cases. The observed decrease in motor vehicle volume is of the order of magnitude expected for the observed increase in bicycle volume. Results of a one-tailed t-test for paired samples indicate that the decreased traffic counts on Bike to Work Day are significant at the 95% confidence level versus counts on other summer Wednesdays. These results suggest that bicycling can be a practical alternative to commuting by motor vehicle for some people and demonstrate that corresponding decreases in the volume of motor vehicle traffic can be measured.

Bicycle and pedestrian counts at signalized intersections using existing infrastructure: opportunities and challenges

Bicycling and walking have gained increased attention recently; however, systematic bicycle and pedestrian counts are still scarce. At intersections, transportation agencies are interested in counting bicycles and pedestrians and leveraging for counting purposes, if possible, existing signal detection equipment. This study evaluated four counting technologies: inductive loops and a thermal camera to count bicycles and passive infrared counters and pedestrian signal actuation data to count pedestrians. The four technologies were tested in a parking lot (controlled environment) and in an intersection (real-world environment). The findings revealed that while the inductive loops and thermal camera counted bicycles accurately in a controlled environment, the loops and cameras failed to do so at an intersection. Passive infrared counters were found to count pedestrians accurately at the intersection sidewalk, and pedestrian signal actuation data could be a cost-effective surrogate for pedestrian demand at signalized intersections.

Leveraging Signal Infrastructure for Nonmotorized Counts in a Statewide Program: Pilot Study

Transportation agencies are beginning to explore and develop non-motorized counting programs. This paper presents the results of a pilot study that tested the use of existing signal infrastructure—Model 2070 signal controllers with advanced software to log pedestrian phase actuations and detections from bicycle lane inductive loops—to count pedestrians and bicycles. The pilot study was conducted at a typical suburban signalized intersection with heavy motorized traffic that was instrumented on all four approaches with pedestrian push buttons and advance inductive loops in the bicycle lane for signal operation. One day (24 h) of video data was collected as ground truth. The data were reduced and compared with the controller logs. Results indicated that using pedestrian phases as a proxy for estimating pedestrian activity was a promising avenue for counting programs. During the pilot study day, 596 pedestrians crossed the intersection, and 482 pedestrian phases were logged (i.e., 1.24 pedestrian crossings per phase logged). However, bicycle counts were not as accurate because of a number of site-specific factors: (a) inductive loop location, (b) loop sensitivity settings, (c) loop shape, and (d) nearly half of the cyclists passing through the intersection were riding on the sidewalk. The pilot study was part of a research project to develop guidelines for a statewide bicycle and pedestrian counting program for the Oregon Department of Transportation.

Annual Average Nonmotorized Traffic Estimates from Manual Counts: Quantifying Error

Across the United States, jurisdictions are investing more in bicycle and pedestrian infrastructure, which can benefit from nonmotorized traffic volume data. The design of nonmotorized counting programs varies. Whereas some agencies use automated counters to collect continuous and short duration counts, the most common type of bicycle and pedestrian counting is manual counting either in the field or from video. The objective of this research is to identify the optimal times of day to conduct manual counts for the purposes of accurately estimating annual average daily nonmotorized traffic (AADNT). This study used continuous bicycle and pedestrian counts from six U.S. cities to estimate AADNT and analyze estimation errors for multiple short duration count scenarios. Using two permanent counters per factor group reduces error substantially (> 50%); afternoon counts seem to be best for reducing error (2:00 to 6:00 p.m.). Error on Sunday is often as good as, if not better than, Saturday, contrary to what others have found. Arlington has the lowest AADNT estimation error (mean absolute percentage error), probably because of better data quality and higher nonmotorized traffic volumes, and Mount Vernon, Washington has the highest. Average AADNT estimation errors for the studied short duration count scenarios range from 30% to 50%. Error is lower for the commute factor group, bicycle-only counts, scenarios in which more peak hours are counted, and when more than one permanent counter is available to estimate adjustment factors.