Addressing Data Challenges for Bicycle Crash Analysis

Eleni Christofa, PhD Civil and Environmental Engineering University of Massachusetts Amherst

TREC Friday Transportation Seminar Series Portland, OR

March 10, 2017

Addressing Data Challenges for Bicycle Crash Analysis

2 University of Massachusetts Transportation Center

Motivation

https://www.flickr.com/photos/infomatique/6210901187


Motivation

Bikesharing Station Demand Data

Mobile Bicycle Data

Fixed Location Demand Data


Data Challenges


Background

R =1, 000, 000A

365V

R: crash rate in crashes per million vehicles A: average number of crashes per year V: average volume of vehicles per day, or average annual daily vehicles (AADT)


Background

RIVER STREET

MAIN STREET

MASSACHUSETTS AVENUEBROADWAY

HAMPSHIRE STREET

JOHN F KENNEDY STREET

VASSAR STREET

ALEWIFE BROOK PARKWAY

CAMBRIDGE STREET

MOUNT AUBURN STREETBRATTLE STREET

GARDEN STREET

WESTERN AVENUEBROOKLINE STREET

HURON AVENUE

QUINCY STREET

0

250

500

750

1,000

1,250

1,500

1,750

2,000

2,250

2,500

2,750

3,000

0 2,500 5,000 7,500 10,000 12,500 15,000 17,500 20,000 22,500AADT

AA

DB


Background

0

25

50

75

100

125

150

175

200

0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000AADB

Bicy

cle

Cras

h Fr

eque

ncy

0

25

50

75

100

125

150

175

200

0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 20,000 22,000 24,000 26,000 28,000 30,000AADT

Bicy

cle

Cras

h Fr

eque

ncy


Objective

§  To assess bicycle crash risk accounting for: §  double exposure to both cars and other bicycles

§  data challenges

§  seasonality in bicycle demands

§  lack of continuous counts in multiple locations


“Double Exposure” Crash Rate

Rdual: crash rate in crashes per million vehicles A: average number of crashes per year Vauto: average volume of automobile traffic per day, or average annual

daily vehicles (AADT) Vbike: average volume of bicycle traffic per day, or average annual daily

bicycles (AADB)

Rdual

=

✓1, 000, 000

365

◆2

· A

Vauto

Vbike


Research Approach: Framework

AADB Estimation

Corridor Assignment

Crash Analysis


Count data

Automobile:

AADT from 2012 (MassDOT)

Bicycle:

Manual peak hour count data at 28 locations City of Cambridge: 2-hr AM and PM peaks (3 days in Sept. 2012) Boston MPO: 1-4 hrs (2009-2014)

Continuous count data at 2 locations Broadway Avenue (November 2013 – June 2014) Hampshire Street & Cardinal Medeiros Avenue (July 2015 – 2016)

Source: City of Cambridge, MA; Boston MPO


Bicycle Crash Data

Location: Cambridge, MA

Time Interval: 2011-2014

Number of crashes: 622 bicycle-vehicle crashes

Source:

UMass Safety Data Warehouse

Source: https://velosurance.com


1. Annual Average Daily Volume Estimation: A Sinusoidal Bicycle Demand Model


Bicycle Demand Estimation: Research Approach

1.  Bicycle counts and bike-share data

2.  Data analysis

3.  Model calibration

4.  Model validation


1. Bicycle Counts

City Count Locations Ottawa, ON 12 Cambridge, MA 1 Arlington, VA 21 Portland, OR 6 Vancouver, BC 4 Seattle, WA 3


1. Bike-Share Data

City Bike-Share Name Available Data Boston, MA Hubway Bike-Share 2011-2013 Washington D.C. Capital Bike-Share 2013-2015 New York City, NY Citi Bike-Share 2010-2015 Saint Paul, MN Nice Ride Bike-Share 2010-2015


2. Data Analysis

0

500

1,000

1,500

2,000

Jun-11

Sep-11

Dec-11

Mar-12

Jun-12

Sep-12

Dec-12

Mar-13

Jun-13

Sep-13

Dec-13

Mar-14

Jun-14

Sep-14

Dec-14

Mar-15

Mon

thly

AD

B

2011 2012 2013 2014 Estimated ADB

Laurier Ave. and Metcalfe St., Ottawa, ON


3. Model Calibration

Portland, OR


3. Model Calibration

Seattle, WA


3. Model Calibration: Sinusoidal Function

Month, 𝑡

0.0

0.2

0.4

0.6

0.8

1.0

1 2 3 4 5 6 7 8 9 10 11 12

𝑊𝑎𝑣𝑒𝑙𝑒𝑛𝑔𝑡ℎ, 𝜔= π⁄6 

Amplitude, A

Centerline Average, AADB

𝑊𝑎𝑣𝑒 𝐶𝑟𝑒𝑠𝑡, ADBMax

𝑊𝑎𝑣𝑒 𝑇𝑟𝑜𝑢𝑔ℎ, ADBMin

Nor

mal

ized

Mon

thly

ADB

Phase�Shift, φ



MADBt = AADB +A · sin(! · (t� �))

MADBt: Monthly Average Daily Bicycle Count for month t, [bicycles/day] AADB: sinusoidal centerline [bicycles/day] t: time value [months] A: amplitude of MADB sinusoid, or the average seasonal change

[bicycles/day] ω =2πf: wavelength of MADB sinusoid [months] f=1/12: frequency of MADB sinusoid φ: phase of sinusoid



A =MADB

Max

�MADBMin

2

AADB =

P365i=1 ADBt

365

AADB =MADB

Max

+MADBMin

2

or


3. Model Calibration: Seasonal Change

α



a =MADB

Max

�MADBMin

MADBMax

+MADBMin

City Count Locations Temp. Difference (oF)

(High-Low)

α

Ottawa, ON 11 56.6 (70.2-13.6) 0.96 Cambridge, MA 1 44.4 (73.4-29.0) 0.58 Arlington, VA 20 43.8 (79.8-36.0) 0.61 Portland, OR 6 29.1 (69.5-40.4) 0.45 Vancouver, BC 4 25.9 (64.4-38.5) 0.62 Seattle, WA 3 24.1 (66.1-42.0) 0.55



AADB =MADBt

↵ · sin(⇡6 (t� �)) + 1

MADBt = AADBh↵ · sin

⇣⇡6(t� �)

⌘+ 1

i


Bike-share demand data

alpha = 0.52

alpha = 0.37

alpha = 0.39

alpha = 0.45

alpha = 0.44

2011

2012

2013

2014

2015

0

2,500

5,000

7,500

10,000

12,500

15,000

Jun-10

Sep-10

Dec-10

Mar-11

Jun-11

Sep-11

Dec-11

Mar-12

Jun-12

Sep-12

Dec-12

Mar-13

Jun-13

Sep-13

Dec-13

Mar-14

Jun-14

Sep-14

Dec-14

Mar-15

Jun-15

Sep-15

Dec-15

Mar-16

Mon

thly

AD

B

2010

2011

2012

2013

2014

2015

AADBEstimatedADB

Capital Bike-share Data, Washington D.C.


Bike-share demand data

Bike-share System α Hubway (Boston, MA) 0.69 Citi Bike (NY, NY) 0.63 Capital Bike (Washington, DC) 0.49 Nice Ride (Saint Paul, MN) 0.78


4. Model Validation

Ottawa, Canada, α = 0.99

Month

MA

DB


4. Model Validation

Arlington, VA, α = 0.57

Month

MA

DB


Estimation Accuracy

•  AADB errors varied between 8.64% (August) and 28.11%

(January) for all cities

•  MADB errors varied between 5.36% (July) and 41.54%

(January) for all cities


2. Corridor Assignment


Map of Bicycle-Vehicle Collisions, Cambridge, MA

9

8

7

6

5

4

3

2

1

0

17

16

1514

13

12 11

10

0 0.5 1 1.5 20.25Miles

Cambridge Bicycle-Vehicle Collision Map, 2011-2014

LegendStudy Intersection

Bicycle Collision

Bicycle/Pedestrian priority

Cycle track

Bike lane; Hybrid

Marked shared lane

On-Road/No Lane

0 - Inman Square1 - Massachusetts Ave & Vassar St2 - Broadway & Hampshire St3 - Massachusetts Ave & Memorial Drive4 - Lafayette Square5 - JFK St & Memorial Drive6 - Porter Square7 - Brattle St at Sparks St & Craigie St8 - Western Ave & Memorial Drive9 - Massachusetts Ave & Linear Park10 - Brookline St & Granite St11 - Quincy Square12 - Brattle St & Mason St13 - Fresh Pond Parkway & Concord Ave14 - Arsenal Square15 - Huron Ave & Fayerweather St16 - River St & Putnam St17 - Hampshire St & Cardinal Mederios Ave


Bicycle Corridors for Cambridge, MA

!

!!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

! !

!!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

BROOKLINE

BELMONT

ARLINGTON

SOMERVILLE

WATERTOWN

BOSTON

MEDFORD

BROADWAY

CAMBRIDGE STREETBRATTLE STREET

VASSAR STREET

MASSACHUSETTS AVENUE

RIVER STREET

GARDEN STREET

HAMPSHIRE STREET

BROOKLINE

STREE

T MAIN STREET

ALEWIFE BROOK PARKWAY

HURON AVENUE

MOUNT AUBURN STREETMASSACHUSETTS AVENUE! Bicycle Crash Location

Assumed Bicycle CorridorStreet without Bicycle FacilitiesStreet with Bicycle FacilitiesBicycle Count Location


3. Crash Risk Analysis


“Double exposure” vs Conventional Crash Rates

R =1, 000, 000A

365VR

dual

=

✓1, 000, 000

365

◆2

· A

Vauto

Vbike


“Double Exposure” Crash Rate & Crash Frequency

Combined Rate2468

Crash Frequency123

Crash frequencies along corridors Crash rates along corridors


Conclusions: Bicycle Crash Rate

Innovative components:

1.  accounts for bicycle exposure to both

automobiles and bicycles

2.  addresses data challenges through:

•  a seasonal bicycle demand model

•  corridor-based analysis

Limitations: •  Uncertainty in the impact of automobiles vs

bicycles on bicycle crash risk


Conclusions: Seasonal Demand Model

Advantages:

•  Can estimate MADB and AADB using only two short-term

counts

Considerations:

•  Lack of seasonality

•  Cyclist type

•  Low counts


Other Ongoing Projects



Acknowledgments

Nick Fournier and Mike Knodler

Funding: •  SAFER-SIM UTC & New England UTC

•  Eisenhower Graduate Transportation Fellowship

Data: •  UMass Safe Traffic Safety Research Program

•  Portland, OR; Arlington, VA, Seattle, WA; Ottawa, ON; Vancouver, BC;

Cambridge, MA


References

1.  Fournier, N., Christofa, E., and Knodler, M.A. 2017. A Mixed Method Investigation of Bicycle Exposure in Crash Rates, Accident Analysis & Prevention. [in press]

2.  Fournier, N., Christofa, E., and Knodler, M. A. 2017. A sinusoidal model for seasonal bicycle demand estimation. Transportation Research Part D: Transport and Environment, 50, 154-169.


Questions?

Eleni Christofa [email protected]

Education

Addressing Data Challenges for Bicycle Crash Analysis