Elsevier

Landscape and Urban Planning

Volume 107, Issue 3, 15 September 2012, Pages 307-316
Landscape and Urban Planning

Estimating use of non-motorized infrastructure: Models of bicycle and pedestrian traffic in Minneapolis, MN

https://doi.org/10.1016/j.landurbplan.2012.06.005Get rights and content

Abstract

Traffic counts and models for describing use of non-motorized facilities such as sidewalks, trails, and bike lanes are generally unavailable. Because officials lack the data and tools needed to estimate use of facilities, their ability to make evidence-based choices among investment alternatives is limited. This paper (1) summarizes counts of cyclists and pedestrians between 2007 and 2010 at 259 locations in the city of Minneapolis, MN, (2) develops scaling factors for estimating 12-h (6:30 am–6:30 pm) “daily” counts from hourly counts, (3) presents models for estimating non-motorized traffic using ordinary least squares and negative binomial regressions, (4) validates each model using bicycle and pedestrian counts for 85 locations, and (5) estimates non-motorized traffic for every street in Minneapolis, MN. Across all locations, mean pedestrian traffic (51/h) exceeded mean bicycle traffic (38/h) by 35%. One-hour counts were highly correlated with 12-h “daily” counts suggesting that planners may focus on short time scales without compromising data quality. Significant correlates of non-motorized traffic vary by mode and model and include weather (temperature, precipitation), neighborhood socio-demographics (household income, education), built environment characteristics (land use mix), and street (or bicycle facility) type. When controlling for these factors, bicycle traffic increased over time and was higher on streets with bicycle facilities than without (and highest on off-street facilities). Our models can be used by policy-makers to estimate non-motorized traffic for streets where counts are unavailable or to estimate changes in non-motorized traffic associated with other changes in the built environment (e.g., adding bicycle lanes or changing land use).

Highlights

► 1- or 2-h bicycle and pedestrian counts can predict reasonable estimates of ‘daily’ (12-h) counts. ► On-street and off-street bicycle facilities are associated with higher volumes of bicycle traffic. ► Bicycle traffic volumes in Minneapolis, MN are significantly increasing over time. ► Negative binomial models yield reasonable relative estimates of non-motorized traffic volumes.

Introduction

Interest in providing infrastructure for non-motorized modes of transportation – bicycling and walking – are growing at federal, state, and local levels of government. Traffic counts and other basic information about use of existing non-motorized infrastructure, however, are generally unavailable for planning and managing transportation systems. Because officials lack both the data and tools needed to forecast use of facilities, their ability to make evidence-based choices among investment alternatives and to optimize management of non-motorized transportation systems is limited. Transportation officials need better data and new tools to meet the needs of cyclists and pedestrians (Porter & Suhrbier, 1999).

Researchers have worked on methods of forecasting non-motorized traffic for at least forty years. Two early examples from the 1970s are: (1) use of aerial photography to count pedestrians and develop regression models to estimate pedestrian traffic as a function of built environment variables (Pushkarev & Zupan, 1971) and (2) estimating pedestrian traffic/h for blocks in Milwaukee, Wisconsin as a function of land use and other variables (Behnam & Patel, 1977). More than 20 years after these exploratory studies, Hunter and Huang (1995) completed a comprehensive review of reports on the use of bicycle lanes and off-street trails and found wide variation in the level of detail and quality. They concluded that many studies had not collected representative information which could be generalized and used in other applications. In a national assessment, the Bureau of Transportation Statistics concluded that data about “the number of bicyclists and pedestrians by facility or geographic area” is “poor” and the “priority for better data is high” (BTS, 2000). At about the same time, the Federal Highway Administration published its Guidebook on methods to estimate non-motorized travel (US DOT, 1999) which summarized historic efforts to measure and model non-motorized traffic.

Demand for information about non-motorized traffic since has grown. However, the data and tools needed for effective management still are not generally available (although some progress is being made [Handy et al., 2009, Jones, 2009, US DOT, 2010]). The National Bicycle and Pedestrian Documentation Project (NBPDP; Jones, 2009), cited by Handy et al. (2009) as a “promising beginning”, is designed to establish consistent count and survey methodologies, create a national database of counts, and support analyses of factors that influence non-motorized traffic. Results include assessments of automated count technologies and protocols for counts and surveys. More recently, the U.S. Department of Transportation (Jones, 2009) issued a policy statement on bicycle and pedestrian accommodation that recommends transportation agencies and local communities collect data on biking and walking, observing, “the best way to improve transportation networks for any mode is to collect and analyze trip data to optimize investment.”

Although the scope of studies remains insufficient, researchers have added new insights in several key areas thought to impact non-motorized travel such as built environment characteristics (Dill, 2009, Guo et al., 2007Haynes and Andrzejewski, 2010, Jones et al., 2010), infrastructure design characteristics (Lindsey et al., 2006, Reynolds et al., 2007), neighborhood socio-economics (Adams, 2010), and weather (Aultman-Hall, Lane, & Lambert, 2009). Furthermore, researchers have made incremental steps toward developing traditional traffic models (e.g., gravity models) for non-motorized travel by developing impedance functions for cycling and walking (Iacono, Krizek, & El-Geneidy, 2010), modeling mode share near bicycle facilities (Krizek, Barnes, & Thompson, 2009), and building route-choice models (Hood, Sall, & Charlton, 2011). For example, researchers explained more than 80% of observed variation in traffic at 30 locations on five multiuse trails in Indianapolis, Indiana by modeling daily counts as a function of weather, day of week (and month of year), neighborhood socio-demographics, neighborhood form, and trail characteristics (Lindsey et al., 2006, Lindsey et al., 2007, Lindsey et al., 2008).

There may be regional differences in patterns of non-motorized traffic. Jones (2009) illustrated that the patterns of trail use documented in Indianapolis vary from those in other regions of the nation and concluded that “unlike vehicle use patterns, there appear to be significant regional differences in seasonal patterns” and that, for non-motorized facilities, analysts may need to “accept variation as part of the normal estimating process.” A review by Pucher, Dill, and Handy (2009) highlights this point for bicycle infrastructure interventions stating that there is “considerable variation in estimated impacts, both by type of intervention and by study design, location, and timing”.

Our research focuses on developing models that can be used by practitioners to make evidence based choices on where and how to build non-motorized infrastructure. This paper summarizes counts (2007–2010) of cyclists (n = 436) and pedestrians (n = 431) in the city of Minneapolis, MN, and presents models that (1) explore correlates of pedestrian and bicycle traffic volumes, (2) estimate non-motorized traffic in locations without counts, and (3) test for trends in non-motorized traffic over time. Scaling factors for adjusting hourly counts to 12-h ‘daily’ counts are derived and two types of regression models (ordinary least squares and negative binomial) for estimating 12-h bicycle and pedestrian traffic are presented and validated. We then discuss how these data and models can inform non-motorized transportation planning, the limitations of the models, and research needed to improve modeling non-motorized traffic.

Section snippets

Data and methods

Our general research approach involved assembling non-motorized traffic counts, constructing independent, explanatory variables, computing descriptive statistics, developing scaling factors, and building regression models of cycling and pedestrian traffic (Borah et al., 2010).

Scaling factors and time of day traffic patterns

Time of day proportions and scaling factors for bicycle and pedestrian traffic are presented in Table 2. Peak-hour traffic for both bicycles and pedestrians occurs between 5:00 pm and 6:00 pm. However, the mid-day hours account for a larger proportion of traffic for pedestrians (8.7–9.7%) than for cyclists (5.2–7.2%).

The bivariate regression results (using the 43 12-h observations; see Table 2 and Supplemental material) show that the 12-h counts correlate best with peak-hour counts (although

Discussion and conclusions

We summarized traffic counts of bicycles and pedestrians in Minneapolis, MN and developed statistical models to estimate non-motorized traffic on roads and sidewalks without counts. Several important observations with implications for counting, modeling, and estimating bicycle and pedestrian traffic can be drawn from our results. However, our analyses also confirm the need for additional research to improve the tools available to planners and managers responsible for non-motorized

Acknowledgements

We thank Shaun Murphy (Minneapolis Department of Public Works) and Tony Hull (Transit for Livable Communities) for providing the original data set as well as offering helpful comments.

References (44)

  • J. Borah et al.

    Non-motorized transportation in the city of Minneapolis

    (2010)
  • BTS (Bureau of Transportation Statistics)

    Bicycle and pedestrian data: Sources, needs, and gaps

    (2000)
  • A. Cameron et al.

    Econometric models based on count data: Comparisons and application of some estimators and tests

    Journal of Applied Econometrics

    (1986)
  • X. Cao et al.

    The influences of the built environment and residential self-selection on pedestrian behavior: Evidence from Austin, TX

    Transportation

    (2006)
  • City of Minneapolis. (2010). Crime statistics and reports. www.minneapolismn.gov/police/crime-statistics Accessed May...
  • City of Minneapolis, Department of Public Works. Road and bicycle facility types. Data available upon...
  • J. Dill

    Bicycling for transportation and health: The role of infrastructure

    Journal of Public Health Policy

    (2009)
  • DPW (Department of Public Works). (2011). Non-motorized traffic counts operations and estimation methodology. City of...
  • Y. Fan et al.

    Impact of light rail implementation on labor market accessibility: A transportation equity perspective

    Journal of Transport and Land Use

    (2012)
  • J. Guo et al.

    Effect of the built environment on motorized and non-motorized trip making: Substitutive, complementary, or synergistic?

    Transportation Research Record

    (2007)
  • S. Handy et al.

    The regional response to federal funding for bicycle and pedestrian projects

    (2009)
  • S. Hankey et al.

    Health impacts of the built environment: Within-urban variability in physical inactivity, air pollution, and ischemic heart disease mortality

    Environmental Health Perspectives

    (2012)
  • Cited by (93)

    • A review of bike volume prediction studies

      2024, Transportation Letters
    • Assessing the effects of the built environment and microclimate on cycling volume

      2023, Transportation Research Part D: Transport and Environment
    View all citing articles on Scopus
    View full text