U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-05-137
Date: July 2006

Evaluation of Safety, Design, and Operation of Shared-Use Paths

Final Report

PDF Version (1172 KB)

PDF files can be viewed with the Acrobat® Reader®




Shared–use paths are paved, off–street travel ways designed to serve nonmotorized travelers. Shared–use paths are gaining popularity in two different ways in recent years in the United States. First, there are many more paths and many more miles of paths being created. Second, shared–use paths are attracting a increasingly greater amount of use. Some urban trails attract thousands of users per hour during peak periods and some are experiencing rush hours and traffic jams. Trail managers are becoming increasingly concerned about user conflicts and injuries, and some are also concerned that potential users are deciding not to come out and use a trail because of crowding.

The design of a new path or a path to be rebuilt is thus an increasingly important activity. During the design of every shared–use path, someone eventually asks, "How wide should this pathway be?" That question nearly always raises even more questions: What types of users can we reasonably expect? When will we need to widen the path? Do we need to separate different types of users from each other? These are very difficult questions for designers who face that classic design dilemma of overbuilding versus obsolescence. If the designer specifies a trail to be wider than future use justifies, there is a waste of money that could otherwise have been used to construct more miles of trail elsewhere. If the designer specifies a trail that proves to be too narrow for the future volume and mix of users, there will be more user conflicts and collisions, greater unhappiness among users, and the need to consider expensive trail widening.

At the present time, conventional design manuals do little to help designers resolve their dilemmas. The classic procedure first specified by Hein Botma in The Netherlands, which bases LOS on the estimated number of meetings and passings for bicyclists, is an attractive framework that could help designers. There can be little debate that, in general, paths where bicyclists must make more meetings and passings should be less desirable than trails with fewer meetings and passings. However, the LOS procedure in the 2000 HCM (4), adapted from Botma's work, has a number of serious limitations that make it difficult for designers to use in order to resolve their path design dilemmas.

The overall project objective was the production of a tool that professionals can use to evaluate the operational effectiveness of a shared–use path, given a traffic forecast or observation at an existing path along with some geometric parameters. The project adopted Botma's method as the basic framework for the LOS procedure. In particular, the objective was to produce a tool that overcomes the major limitations in the current LOS procedure noted above. The desired procedure emerging from this project would:

  • Be calibrated and validated.
  • Be based on U.S. data.
  • Have LOS criteria based on user input.
  • Include more modes.
  • Include the ability to change key parameters such as mean speeds.
  • Account for delayed passing.
  • Analyze the full range of existing and possible path widths.
  • Be in a form ready for use by path designers.


Our four major achievements during the project included:

  • Development of the additional theoretical framework necessary to overcome the limitations to the existing procedure noted above.

  • Collection of field data on path operations to calibrate and validate the theoretical equations for U.S. conditions.

  • Collection of path user perception data to establish LOS criteria.

  • Development of an LOS estimation tool that professionals working with shared–use paths can use and a plan to provide them with that tool.

During the project, we made two significant advances in the theory of traffic flow on shared–use paths. First, we extended the previous method to other modes, other speed distributions, and passive passings. As noted, a passive passing is an occasion when the test bicyclist is passed by a faster path user. Second, the team developed a way to calculate the number of delayed passings. These are times when the test bicyclist would arrive behind a slower path user and not be able to pass because of the lack of an adequate gap in the next lane to the left (oncoming or same direction). Obviously, delayed passings are undesirable for bicyclists since they then have to slow down and then probably expend energy accelerating when an adequate gap does appear. Delayed passings are also critical because they are so highly related to path width. Prior to this project, there were some delayed passing calculations in the literature related to two–lane highway operation, etc.; however, there was nothing in the literature related to shared–path operation. Our new theory will estimate delayed passings of various modes for two–lane, three–lane, and four–lane paths.

The objective of the operational data collection portion of this project was to collect the field data needed to calibrate and validate the LOS model for shared–use paths. To calibrate and validate an LOS model, the main variables that we needed to collect were meetings and desired and actual passings by path users, and the speeds and volumes of the path user groups. The team chose to use the moving–bicycle method to collect these variables. We collected a database of more than 700 runs of (mostly) 0.80 km (0.5 mi) each by our test bicyclist on 15 shared–use paths in 10 cities across the United States. The database included a wide array of volumes and speeds for five different major user groups. The major analysis of the database was to use the theory described above to predict of the number of meetings and passings for each run, and to compare that prediction to the number of meetings and passings counted in the field. Our results showed that the prediction matched the field count fairly well at most sites. We also were able to use the operational database to find the default values that we needed for the LOS procedure, and to use the videotapes for our perception data collection.

The third major part of this effort was to collect data on user perceptions of multi–use trail designs and operations to help set the LOS criteria. We did this by showing more than 100 volunteer respondents thirty–six 1–min videotapes of paths recorded using the helmet camera during the operational data collection effort. We selected the videotapes to represent a wide range of path designs and operations. The respondents were asked to state, on a five–point scale, how much they would enjoy using the path in four different ways: longitudinal separation, latitudinal separation, ability to pass, and overall. An analysis of the responses showed that there was a strong relationship between the number of meetings and passings and enjoyment, and that path width was the primary geometric variable that affected user perception. The user perception model that we developed from the data was the backbone of our LOS procedure.

The most important accomplishment of the project team was the production of a new method to estimate the LOS for a shared–use path, based on the results from the other parts of the project. The procedure uses simple inputs that should be readily available to path designers, including one–way user volume, mode split, path width, and the presence or absence of a centerline. The procedure uses the theory that we developed and validated to predict the number of meetings and passings that will occur; it uses a number of default values based on data we collected if local values were not available. The LOS is based primarily on the model of user satisfaction that emerged from our perception study. Our model of delayed passing also plays a role in some LOS estimations. The output from the LOS procedure is a traditional A through F grade with which users can judge the performance of an existing path or a design alternative. The project team has also produced software to calculate the LOS automatically, greatly easing the burden on future users.



The new LOS estimation method produced during this project should be widely adopted by path designers because it provides many advantages over current methods, including:

  • The new method is based on sound theory. The theory (development of which is shown in chapter 3) began with the work of Hein Botma in The Netherlands, and was extended here to include more modes, more speed distributions, passive passing, and delayed passing.

  • The procedure to estimate meetings and passings has been validated for the first time in the United States on a wide variety of trails.

  • The LOS method and scale has now been calibrated based on the perceptions of a sample of U.S. users viewing videotapes of actual operations on actual U.S. paths.

  • The method only provides the LOS for bicycles, but does so while accounting for four other modes, including child bicyclists, inline skaters, pedestrians, and joggers.

  • The method works for and was based on path widths from 2.44 to 6.1 m (8 to 20 ft).

  • The method works for and was based on a full range of volumes, from virtually zero users in an hour to several thousand users per hour.

  • The inputs to the procedure are still easy for a user to assemble (as were the inputs to the previous procedures), and sound default values are provided for most variables.

  • The research team has provided easy–to–use software to make the LOS calculation.

Marketing the New LOS Procedure

Unlike the roadway environment, which is almost exclusively the domain of civil engineers, shared–use paths are designed by a wide variety of practitioners. Some of the most creative and unique trails in the country are the direct result of the diverse skills that these designers bring to the table. We identified three main target audiences for the marketing of our shared–use path LOS model: transportation professionals; trail designers/coordinators; and walking, bicycle, and trail advocates and organizations.

Reaching each of these three groups with the results from this research will be challenging. To reach transportation professionals, the Highway Capacity and Quality of Service Committee should consider the new LOS method for inclusion in the next edition of the HCM,or as an interim release until the next edition is ready. Presentations and papers at conferences and in journals of the Transportation Research Board and the Institute of Transportation Engineers will also help. To reach trail designers and trail advocates, the LOS method and User’s Guide should be posted on FHWA’s pedestrian and bicycle Web site and on the sites of other pedestrian and bicycle information clearinghouses. The new LOS method should also be presented to the AASHTO Task Force on Geometric Design for their consideration. If they were to include it in an update of the Guide for the Development of Bicycle Facilities,(1) there would be a great likelihood that it would be widely used and accepted. In addition, presentations and papers at conferences and in the journals of the major parks and recreation societies will also be helpful.

Future Research Needs

The scope of the project, and therefore of the products emerging from the project, was limited in several important ways, as explained below. We recommend that future research on shared–use paths focus on erasing those limitations and otherwise extending the work conducted here.

First, we recommend research to extend this work to estimate LOS from other points of view besides bicyclists. While one can easily assume that wider paths with fewer events lead to better levels of service for pedestrians and for other path users, as they do for bicyclists, the precise nature of that relationship is in doubt. We collected some user perceptions from the pedestrian point of view; however, the sample was not large and the videos that the respondents were rating, were not from the pedestrian point of view. It would be quite feasible to conduct similar user perception surveys from groups of pedestrians, inline skaters, joggers, and others by showing each group video clips and learning their reactions. It would also be possible to extend the LOS procedure to consider other modes emerging as important users on shared–use paths, such as tandem bicyclists, Segways, and wheelchairs. The data on emerging path users compiled in a recent FHWA project (44) provide a good start.

Second, we recommend collecting more data on high–volume and four–lane paths to put the LOS method on a more firm footing in these situations. The database collected in this project had one such path– the Lakefront Trail in Chicago, IL– and we recorded extremely heavy volumes on that path. However, there are other paths of that width in the United States, and there will be more in the future. In addition, if the current growth in path volumes continues across the United States in the next few years, there will be many more paths with Chicago–like volumes. We noted in chapter 7 that the perception data from Chicago was almost an outlier compared to the rest of the data set; we need more data from high–volume paths to fully calibrate the LOS method under those conditions.

Third, we recommend a research effort to validate the delayed passing theory that we developed during this project. The number of delayed passing attempts would seem to be a natural factor in an LOS procedure, and our theory appears sound. However, we developed the theory relatively late in the project, and did not have the opportunity to show that the theory correctly estimates the number of delayed passings versus field data. In addition, none of our perception video clips showed bicyclists that had delayed passings. We need more direct evidence of how bicyclists weigh delayed passings against undelayed active passings, meetings, and other factors critical to the LOS.

Fourth, we recommend a project to validate the video method of obtaining user perceptions. This could be done by comparing a sample of perceptions from viewers of videos to a sample of perceptions from people actually using the trails.

Next, we heartily recommend a project to find a better LOS procedure for intersections on shared paths. The current intersection analysis procedure in the HCM(4) is like the current LOS procedure for shared–use path segments in that it is based on good logic, but is not validated against field data from the United States. This project, of course, only provided data and analysis for segments between intersections. Given the difficulty we had during this project in finding high–volume path segments of at least 0.80 km (0.5 mi) long, intersections are probably a major concern for path users across the United States.

Finally, the theory estimating the number of delayed passings we developed in this project could be applied to improve the LOS procedure for two–lane highways. The current procedure in the HCM(4) is based on a series of microscopic simulation runs. It would require some field data collection; however, with some effort, our theory could be applied to two–lane highways, and a comparison to the simulation could be revealing.



Previous | Table of Contents | Next

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101