pedestrian and bicycle accidents with motor vehicles gudmundur f. ulfarsson, ph.d. assistant...

44
Pedestrian and Bicycle Accidents with Motor Vehicles Gudmundur F. Ulfarsson, Ph.D. Assistant Professor Director, Transportation Systems Engineering Program Department of Civil Engineering

Upload: teresa-snow

Post on 18-Dec-2015

215 views

Category:

Documents


1 download

TRANSCRIPT

Pedestrian and Bicycle Accidents with Motor

VehiclesGudmundur F. Ulfarsson, Ph.D.Assistant ProfessorDirector, Transportation Systems Engineering ProgramDepartment of Civil Engineering

Acknowledgement

University of North Carolina Highway Safety Research Center provided the data in this study

Mr. Joon-Ki Kim (doctoral candidate, Washington University)

Dr. Sungyop Kim (University of Missouri-Kansas City)

Dr. Venky N. Shankar (Pennsylvania State University)

Background and motivation Objectives Factors affecting bicycle and

pedestrian safety Data description Methodology Results Conclusions

Background and Motivation

Bicyclists and pedestrian Two particularly vulnerable groups 2% and 11% of all fatalities by traffic

accidents, respectively 46,000 bicyclists were injured in 2003 68,000 pedestrians were injured in

accidents in 2004

Background and Motivation

Previous studies Accident rates Accident frequency associated with a

certain type of injury Aggregate data

There is a need for studying the vulnerable groups’ safety based on disaggregate individual accident data

Research Objective

To build probability models of injury severity for bicyclists and pedestrians By employing econometric analysis and

behavioral models as a statistical method

For examining the factors affecting injury severity in bicycle-vehicle accidents and pedestrian-vehicle accidents

Research Objective

The primary goals Develop discrete probability models of

injury severity by using disaggregate accident data

Consider heteroscedasticity (non-identical variance of error terms) across individuals

Examine the factors affecting bicyclist injury severity in bicycle-motor vehicle accidents

Pedal-Cyclist Fatalities in U.S.

The number of pedal-cyclist fatalities 833 fatalities in 1995 629 fatalities in 2003 784 fatalities in 2005

Pedal-cyclist Fatalities

500

600

700

800

900

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

Year

Factors Affecting Bicycle Safety

Alcohol Correlated with casualties Increases the risk of injury

Higher Speed Limits Modify driver’s scanning pattern

Pay attention to the most relevant direction Ignore the less relevant direction

The most frequent type of bicycle-motor vehicle accident

A driver turning right and a bicycle coming from the driver’s right

Factors Affecting Bicycle Safety

Age Children and the elderly are the primary groups

that suffer from injuries

Gender Males are overrepresented compared to females

at almost all ages

Obeying the law Bicyclists are more likely than drivers to violate

traffic laws

Lighting, weather, annual daily traffic, road design, and so on.

Factors Affecting Pedestrian Safety

Age The older pedestrians suffer from more serious

injuries than the other age groups in pedestrian-vehicle accidents

Vehicle speed Increases the risk of pedestrian-vehicle accidents Increases the pedestrian's injury severity

Alcohol Increases the risk of pedestrian-vehicle accidents Increases the pedestrian's injury severity

Factors Affecting Pedestrian Safety

Visibility The main causes of pedestrian-vehicle accidents

Crosswalk Most pedestrian accidents occur when

pedestrians cross roadways Only 13.5 % of all pedestrians look left and right

when they enter the crosswalk No traffic control device increases the risk of

collision involving an older pedestrian Traffic volume, educational achievement,

gross national income per capita, and so on

Bicycle Data Description

Police-reported accident data from North Carolina for the years 1997 through 2000 (The University of North Carolina Highway Safety Research Center)

Analyze only accidents that involve a single motorist and a bicyclist

Bicyclist injury severity categories: Fatal injury, incapacitating injury, non-

incapacitating injury, possible or no injury Sample size = 2,934

Bicycle Data Description

Bicyclist injury severity distributionN %

Fatal 104 4

Incapacitating 363 12

Non-incapacitating

1,323 45

Possible or no injury

1,144 39

Total 2,934 100%

Pedestrian Data Description

Police-reported accident data from North Carolina for the years 1997 through 2000

Analyze only accidents that involve a single motorist and a pedestrian

Sample size = 5,808

Pedestrian Data Description

Pedestrian injury severity distributionN %

Fatal 591 10

Incapacitating 1,094 19

Non-incapacitating

1,976 34

Possible or no injury

2,147 37

Total 5,808 100%

Methodology

Multinomial logit (MNL) model Heteroscedastic generalized extreme

value (HET GEV) model

Methodology: Multinomial Logit

Discrete Outcome Multinomial Logit (MNL) model: 1) Fatal Injury, 2) Incapacitating Injury, 3) Non-Incapacitating Injury, and 4) Possible or No Injury

Estimate propensity towards an injury severity Categories of explanatory variables:

Bicyclist or Pedestrian, Driver, Vehicle, Accident, Control, Geometry, Land development, and Environmental/Temporal

Tt

S

S

nttn

nt

e

eP

Heteroscedastic Extreme Value (HEV) Model

Heteroscedasticity across individuals Propensity function

The probability of being in the injury severity

ni i ni niU β x ni n i ni n ni nU β x

Individual-specific scaling parameter

1

Pr{ }

Pr{ }

e, where 0

e

i ni n

j nj n

ni ni nj

ni n nj n

nI

j

P U U j i

U U j i

β x

β x

2

2and Var( )

6nin

Heteroscedastic Extreme Value (HEV) Model

Individual-specific scaling parameter

Estimation

where is a vector of observed individual-specific variables is a vector of estimable coefficients

nenγznzγ

where is the total number of individual is the total number of injury severities is 1 if individual suffers from injury severity , 0 otherwise

N nI iniy n i

N

n

I

inini PyL

1 1

,ln

Explanatory Variables

Bicyclist Characteristics Bicyclist age and gender, Intoxication, Helmet Use

Pedestrian Characteristics Pedestrian age and gender, Intoxication

Driver Characteristics: Driver age and gender, Intoxication

Vehicle Characteristics: Estimated vehicle speed Vehicle Type

Land Characteristics Urban Area Land development type (e.g., commercial area)

Explanatory Variables (Cont.)

Accident Characteristics Bicycle direction (facing traffic or with traffic) Type of accident Party at fault Speeding involved Road defects involved Accident location

Control Characteristics Signal, sign, other control, or no control

present

Explanatory Variables (Cont.)

Geometry Characteristics Intersection, Asphalt road, Posted speed limit Road class type, Road geometry (e.g., curve) Road type (e.g., two-way divided) Number of traffic lanes

Temporal Characteristics Weekend, Time

Environmental Characteristics Weather, Light conditions, Road surface

(e.g., dry)

Model Selection

Bicyclist injury severity turned out better with the multinomial logit model

Pedestrian injury severity turned out better with the heteroscedastic extreme value model The pedestrian age forms the

heteroscedasticity Gender was explored and not found

significant

Bicyclist Model Findings

Bicyclist Characteristics Bicyclist age 55 and over increases the

probability of fatal injury (109%) Bicyclist intoxication increases the

probability of fatal injury (174%) Helmets decrease the probabilities of fatal

injury (-24%) and possible or no injury (-24%)

Driver Characteristics Driver intoxication increases the

probabilities of bicyclist fatal injury (265%) and incapacitating injury (87.7%)

Bicyclist Findings (Cont.)

Vehicle Characteristics Estimated vehicle speed beyond 32.2 km/h

(20 mph) increases the probabilities of serious injuries

Pick-up trucks and heavy trucks increase the probability of fatal injury (10% and 381%, respectively)

Speed Fatal injury Incapacitating Injury

32.2-48.3 km/h 93% 93%

48.3-64.4 km/h 303% 100%

64.4-80.5 km/h 1,159% 83%

80.5 km/h + 1,504% 134%

Bicyclist Findings (Cont.)

Accident Characteristics Head-on collisions increase the probability

of fatal injury (101%) Speeding involved increases the probability

of fatal injury (300%)

Bicyclist Findings (Cont.)

Geometry Characteristics A curved road increases the probabilities of

fatal injury (68%) and incapacitating injury (68%)

Two-way divided roadways decrease the probabilities of fatal injury (-11%) and incapacitating injury (-11%)

Temporal Characteristics Weekend increases the probability of fatal

injury (13%) Time (06:00-09:59) increases the probability

of fatal injury (85.4%)

Bicyclist Findings (Cont.)

Environmental Characteristics Inclement weather (fog, rain, snow etc.)

increases the probability of fatal injury (129%)

Darkness without streetlights increases the probabilities of fatal injury (111%) and incapacitating injury (50%)

Pedestrian Model Findings

Pedestrian Characteristics Pedestrian age is continuous and has a elasticity

of 1.85 towards fatal injury, indicating a 1.85% increase in the probability of a fatality for each 1% increase in age

Driver Characteristics Driver age is continuous with elasticity of -0.126

towards fatal injury, indicating a small reduction in fatality probability with increasing driver age

Driver intoxication increases the probabilities of pedestrian fatal injury (168%)

Pedestrian Model Findings

Environmental Characteristics Dark-lighted conditions increase the

probability of fatal injury (148%) Dark-unlighted conditions increase the

probability of fatal injury (338%) Vehicle Type

Truck increases the probability of fatal injury (265%)

Pedestrian Model Findings

Road Types Freeway increases the probability of

fatal and incapacitating injury (143%) US Route increases the probability of

fatal injury (216%) State Route increases the probability of

fatal injury (146%)

Pedestrian Model Findings

Accident Characteristics Speeding-involved increases the

probability of fatal injury (309%) Both driver and pedestrian at fault

increases the probability of fatal injury (114%)

Pedestrian solely at fault increases the probability of fatal injury (386%)

Pedestrian Model Findings

Pedestrian age shows a significant effect on heteroscedasticity

Simulation of probability of fatal injury for all pedestrians by starting each pedestrian at age 18 and increasing by one to 100

Average for all pedestrians by age Shows MNL overestimates fatality

probabilities with age

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

18 24 30 36 42 48 54 60 66 72 78 84 90 96

Age

Ave

rag

e P

rob

abili

ty o

f F

atal

Inju

ry

HET

MNL

Bicyclist Conclusions Factors that significantly increase the probability of

fatal injury Vehicle speed >20 mph (fatality 2-16x more likely) Truck involved (fatality ~5x more likely) Speeding-involved (fatality ~4x more likely) Intoxicated driver/bicyclist (fatality ~3x more likely) Inclement weather (fatality ~2x more likely) Darkness w/o streetlights (fatality ~2x more likely) Bicyclist aged 55 and over (fatality ~2x more likely) Head-on collision (fatality ~2x more likely)

Bicyclist Conclusions

Policy perspective Lower speed limit in residential area (e.g. 20

mph)

Education perspective Discourage biking against traffic due to

increased injury severity in head-on crashes Discourage drunk biking Encourage helmet usage Develop education programs for older

bicyclists, in addition to education for youth

Bicyclist Conclusions

Engineering perspective Enable separated bicycling from high-speed

traffic (e.g., when speed limit of 30 mph or over)

Design to avoid conflicts with oncoming trafficand heavy trucks

Pedestrian Conclusions Factors that significantly increase the probability of

fatal injury Pedestrian age (fatality elasticity 1.85) Pedestrian at fault (fatality ~5x more likely) Darkness w/o streetlights (fatality ~4x more likely) Speeding-involved (fatality ~4x more likely) Truck involved (fatality ~3.6x more likely) US Route (fatality ~3x more likely) Intoxicated driver (fatality ~2.7x more likely) Darkness with streetlights (fatality ~2.5x more likely) State Route (fatality ~2.5x more likely) Freeway (fatality ~2.4x more likely) Both driver and ped. fault (fatality ~2x more likely)

Pedestrian Conclusions

Policy perspective Reduce truck traffic in pedestrian areas Restrict pedestrian access to major routes if

possible

Education perspective Drunk driving Encourage reflectors worn at night Develop education programs for older

pedestrians Walking along major routes

Pedestrian Conclusions

Engineering perspective Improved street lighting Design to avoid conflicts with traffic streams

with higher percentage of trucks Improve road crossings with older

pedestrians in mind

Future Studies

Further study needs to develop a more detailed picture of land use and the built environment around accident sites to facilitate specific design and policy interventions that may reduce bicyclist and pedestrian injury severities

Contact Information

Gudmundur F. Ulfarsson [email protected] tel: +1 (314) 935-9354