New Tools for Estimating Walking and Bicycling Demand

New Tools for Estimating Walking and

Bicycling Demand

9 September 2014

Practical Motivations for Estimating Demand

• How many people are on our trails? Ray Irvin, Indy Parks Greenways, 1996

• Quality of data about “number of bicyclists

and pedestrian by facility … is “poor” and

the “priority for better data is “high” Bureau of Transportation Statistics, 2000

• What traffic controls are needed at this

intersection to protect cyclists and

walkers? Minneapolis Department of Public Works, 2010

What is Demand?

• “Pedestrian and bicycle activity” (NCHRP 770, p. 7)

• Willingness of people to walk or bike

– Expression of choice – presumed to

maximize well-being

– Contingent on multiple factors (accessibility)

– Difficult to measure

• Measures – estimates – of pedestrian or

bicycle volumes important for planning,

investment, and other decisions

NCHRP 770 Tools for Estimating Demand

• Tour-Generation and Mode-Split Models

• GIS-Based Walk-Accessibility Models

• Enhancements to Trip Based Models

• Walk-Trip Generation and Flow Models

• Portland Pedestrian Model

• Facility Demand Models

– Route choice models

– Direct demand models*

Modeling Demand from Counts

• Bikes on streets; pedestrians on sidewalks (Hankey et al. 2012)

• Mixed-mode traffic on multiuse trails (Wang et al. 2013)

• Traffic volume is function of:

– neighborhood socio-demographics

– built environment (e.g., land use, jobs)

– transportation infrastructure

– weather

Bike & Ped Counts in Minneapolis

TLC and City of Minneapolis Count Locations, 2007-2009

0 1 2 3 40.5Miles

5

On-Street Bicycle Facility

Bike Lane, One-Way

Bike Lanes

Shared Lane

Off-Street Trail

Off-Street Bicycle Facility

None

Count Locations

Count Description

Method of observation

Manual

Traffic observed Cyclist - separate

Pedestrian - separate

Locations in Minneapolis

On /off-street bike facilities and no bike facilities

(n=259)

Period of observation 2007-2010

Number of observations

436

Length of observations

12-hour (n=43) 2-hour peak period

(n=352) Other

Limitations Human error

Correlates of Bike & Pedestrian Traffic Hankey et al. 2012

Bicycle models* Pedestrian models*

• % non-white (+)

• % college (+)

• HH income (-)

• LU mix (+)

• Distance to CBD (-)

• Precipitation (-)

• Arterial (+)

• On/Off-street (+)

• Year (+)

• % non-white (+)

• % college (+)

• Distance to water (-)

• Distance to CBD (-)

• Precipitation (-)

• Arterial (+)

• Collector (+)

*negative binomial regression; bold is significant at p=0.05

Estimated 12-hour pedestrian traffic

Model Validation

Modeling Mixed-Mode Trail Traffic Wang et al. 2013

Example Monitoring Site:

Midtown Greenway

Correlates of Mixed Mode Trail Traffic

Variables Expected Sign

Neighborhood Socio-demographic Characteristics

African American residents (%) -

Residents with college degrees (%) +

Population over 64 or below 6 (%) -

Median household income. (1,000 dollars) +

Neighborhood Built Environment

Population density (per square kilometer). +

Weather Conditions

Recorded high temperature.(in Celsius) +

Deviation from the 30-year normal temperature +/-

Precipitation.(centimeters) -

Average wind speed. (kph) -

Temporal Dummies

Saturday or Sunday (equals 1, otherwise 0) +

Validation of Trail Traffic Models (Wang et al. 2013)

Using Factoring to Estimate Daily Traffic

from Short Duration Counts

Estimating Performance Measures: AADT and Trail Miles Traveled in Minneapolis

Segment AADT

Mean 954

Median 750

Max 3,728

Min 39

• 6 reference sites • 7 day short duration counts

on each segment

> 28 million

miles traveled

on 80 mile trail

network in

2013:

16

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

0:00 6:00 12:00 18:00 0:00

% o

f d

ail

y t

raff

ic

Weekdays Weekends

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

0:00 6:00 12:00 18:00 0:00

% o

f d

ail

y t

raff

ic

Weekdays Weekends

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

0:00 6:00 12:00 18:00 0:00

% o

f d

ail

y t

raff

ic

Weekdays Weekends

Short-duration monitoring

identified three different

traffic patterns (factor

groups). Need new

reference monitoring sites.

Utilitarian (weekday)

Mixed Recreational – Utilitarian

(all current reference locations) Recreational

Facility Demand Models

• Require counts or other measures as inputs

• Useful for planning, understanding system

• Do not explain causation

• Have limitations (NCHRP 770): – Need to include variables of interest

– Need to be calibrated

– Need to be validated

– Should not be not transferred

• Can be strengthened (NCHRP 770): – Potential to cross-validate with choice models

Questions?

Acknowledgements: MnDOT: Lisa Austin, Jasna Hadzic

Minneapolis DPW: Simon Blenski

Minneapolis Park Board: Ginger Cannon, Jennifer Ringold

Virginia Tech: Steve Hankey

USC: Xize Wang

For more information contact: Greg Lindsey (linds301@umn.edu)