Highway Safety Analytics and Modeling: Techniques and Methods for Analyzing Crash Data

476 Pages·2021·11.551 MB·English

by  Dominique Lord, Xiao Qin, Srinivas R. Geedipally

HIGHWAY
SAFETY
ANALYTICS
AND
MODELING
D L
OMINIQUE ORD
X Q
IAO IN
S R. G
RINIVAS EEDIPALLY
Elsevier
Radarweg29,POBox211,1000AEAmsterdam,Netherlands
TheBoulevard,LangfordLane,Kidlington,OxfordOX51GB,UnitedKingdom
50HampshireStreet,5thFloor,Cambridge,MA02139,UnitedStates
Copyright©2021ElsevierInc.Allrightsreserved.
Nopartofthispublicationmaybereproducedortransmittedinanyformorbyanymeans,
electronicormechanical,includingphotocopying,recording,oranyinformationstorageand
retrieval system, without permission in writing from the publisher. Details on how to seek
permission,furtherinformationaboutthePublisher’spermissionspoliciesandourarrangements
withorganizationssuchastheCopyrightClearanceCenterandtheCopyrightLicensingAgency,can
befoundatourwebsite:www.elsevier.com/permissions.
Thisbookandtheindividualcontributionscontainedinitareprotectedundercopyrightbythe
Publisher(otherthanasmaybenotedherein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and
experiencebroadenourunderstanding,changesinresearchmethods,professionalpractices,or
medicaltreatmentmaybecomenecessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluatingandusinganyinformation,methods,compounds,orexperimentsdescribedherein.
Inusingsuchinformationormethodstheyshouldbemindfuloftheirownsafetyandthesafety
ofothers,includingpartiesforwhomtheyhaveaprofessionalresponsibility.
Tothefullestextentofthelaw,neitherthePublishernortheauthors,contributors,oreditors,
assumeanyliabilityforanyinjuryand/ordamagetopersonsorpropertyasamatterofproducts
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions,orideascontainedinthematerialherein.
LibraryofCongressCataloging-in-PublicationData
AcatalogrecordforthisbookisavailablefromtheLibraryofCongress
BritishLibraryCataloguing-in-PublicationData
AcataloguerecordforthisbookisavailablefromtheBritishLibrary
ISBN:978-0-12-816818-9
For information on all Elsevier publications visit our
websiteathttps://www.elsevier.com/books-and-journals
Publisher:JoeHayton
AcquisitionsEditor:BrianRomer
EditorialProjectManager:BarbaraMakinster
ProductionProjectManager:SwapnaSrinivasan
CoverDesigner:MarkRogers
TypesetbyTNQTechnologies
Dominique Lord:
Tomyfamily(LeahandJavier),mymother(Diane),mybrother(Se´bastien),
andmytwoformeradvisors(Dr.EzraHauerandDr.BhagwantPersaud).
Xiao Qin:
Tomyfamily(Yuchen,Ethan,andEva),myparents(XingpoandGuang-
qin), my brother(Hui),and my former advisor (Dr. John Ivan)
Srinivas R. Geedipally:
To my family (Ashwini, Akshath, and Svidha), my parents (Ram Reddy
and Laxmi), and my brother (Rajasekhar Reddy). Special thanks to my
former advisor, Dr. Lord,for involving me inthis project.
Preface
Theprimarypurposeofthistextbookistoprovidethestate-of-the-art
knowledge about how to better analyze safety data given their unique
characteristics. This textbook provides the latest tools and methods
documented in the highway safety literature, some of which have been
developed or introduced by the authors. The textbook covers all aspects
of the decision-making process, from collecting and assembling data to
making decisions based on the analysis results, and is supplemented by
real-worldexamplesandcasestudiestohelpunderstandthestateofprac-
ticeontheapplicationofmodelsandmethodsinhighwaysafety.Where
warranted,helpfulhints andsuggestions areprovidedbythe authors in
the text to support the analysisandinterpretation of safety data.
The textbook is suitable for college students, safety practitioners
(e.g., traffic engineers, highway designers, data analysts), scientists, and
researcherswhowork inhighwaysafety.Thistextbook specificallycom-
plementstheHighwaySafetyManual(HSM)publishedbyAAHSTOand
the Road Safety Manual (RSM) by the World Road Association. The
publication of the HSM, RSM, and other safety-oriented guidelines has
substantially increased the demand for training engineers and scientists
about understanding the concepts and methods outlined within. Hence,
thecontentofthistextbookhelpsfillinthisgapbydescribingthemethods
ingreaterdepthandallowsthereaderstobroadentheirknowledgeabout
the fundamental principlesand theories ofhighway safety.
Allthreeauthorsofthistextbookhavetaughtgraduate-levelcoursesin
highway safety at different institutions. The material covered had to be
usedfromvarioussources,includingchapters(orpartofthem)ofvarious
textbooks in areas within and peripheral to highway safety, published
peer-reviewed papers, class notes from the world leaders in highway
safety (e.g., Dr. Ezra Hauer), research reports, and manuals published
bynationalpublicagencies.Mostofthesematerialsdidnotcontainexer-
cises and problems that students could use to apply the knowledge ac-
quired from these documents. Throughout the years, it became clear
thatatextbookwasneededthatcouldcombinealltheseimportanttopics
into a single document. The one from which students could read and
learn about theoretical principles and apply them using observed (or
simulated) data. In this regard, the textbook includes more than nine
xi
xii
Preface
datasetsformorethan40exercises.Mostofthesedatasetshavebeenused
in peer-reviewed publications. All the datasets can be found at the lead
author’s website: https://ceprofs.civil.tamu.edu/dlord/Highway_Safety_
Analytics_and_Modeling.htm.
Thecontentofthetextbookisbasedonanaccumulationofmorethan
40yearsofresearchandapplicationsrelatedtomethodsandtoolsutilized
foranalyzingsafetydata.Thetextbookisdividedintothreegeneralareas.
Thefirstareaincludeschaptersthatdescribefundamentalandtheoretical
principles associated with safety data analyses. This area covers the na-
ture of the crash process from the human and statistical/mathematical
perspectives, as well as key crash-frequency and crash-severity models
that have been developed in the highway safety literature. The second
area groups chapters that describe how the various models described in
the first area are applied. The chapters include methods for exploring
safety data, conducting cross-sectional and before-after studies, identi-
fying hazardous sites or sites with promise as well as tools for incorpo-
rating spatial correlation and identifying crash risk on a near real-time
basis. The third area assembles alternative safety analysis tools. The
methodsincludehow tousesurrogatemeasuresofsafetyanddatamin-
ing techniques for extracting relevant information from datasets,
including those categorized as big data (e.g., naturalistic data).
It is hoped that the content will help readers to better understand the
analytical tools that have been used to analyze safety data to make
informed decisions for reducing the negative effects associated with
crashes across the globe. This is even more important given the Vision
Zero programs that have been increasingly implemented by various
agenciesinEurope,NorthAmerica,andEurasiaamongothers.Thecon-
tent should also help improveordevelopnew tools aimed at estimating
the safety performance of connected and automated vehicles, especially
when they will be deployed in mixed-driving environments (within the
next decade).
Forimplementingmethodsandtechniquesproposedinthistextbook,
the authors have provided computer codes for three advanced software
languages. Of course, the methods are not restricted to just three, but
manyothersoftwarelanguagescanbeeasilyimplementedtobeutilized
given the parameterization described in the textbook. Along the same
line, Microsoft Excel provides simple, flexible, and adequate tools that
canbeusedtoimplementvarioussimplermethods,suchasthegraphical
methods presented in Chapter 5 or before-after studies described in
Chapter 7.
This textbook would never have come to completion without the
significant help and input from numerous individuals, colleagues, and
formerandcurrentgraduatestudents:ZhiChen,SomaDhavala,Kathleen
xiii
Preface
Fitzgerald-Ellis, Ali Shirazi, Ioannis Tsapakis, Yuanchang Xie, Cheng-
cheng Xu, and Lai Zheng. After a few requests on social media, several
peoplehaveofferedinformationaboutgettingaccesstosafetydatabases
or giving us permission to use datasets. They include Jonathan Aguero-
Valverde (Costa Rica), Amir Pooyan Afgahri (Australia), David Llopis
Castello´ (Spain), Aline Chouinard (Canada), Stijn Daniels (Belgium),
Thomas Jonsson (Sweden) Neeraj Kumar (Netherlands), Pei Fen Kuo
(Taiwan), Emad Soroori (Australia), Shawn Turner (New Zealand), and
SimonWashington (Australia).
Finally,thistextbookprojectwouldnothavebeenpossiblewithoutthe
support from Elsevier. First, a large thank you to Brian Romer, who first
approached the authors several years ago and convinced us to prepare
a book on highway safety (given our reluctance about the effort needed
for such an endeavor). Thanks to the two book managers who kept us
on our toes for the duration of this project: Barbara Makinster and Ali
Afzal-Khan. Special thanks to Narmatha Mohan for helping us manage
copyright information and permission log, and Swapna Srinivasan for
handling the production of the textbook. The content of this textbook
has been partly funded by the A.P. and Florence Wiley Faculty Fellow
provided by the College of Engineering at Texas A&M University and
project 01-001 from the Safety through Disruption (Safe-D) University
Transportation Center (UTC).
Dominique Lord,
Texas A&M University
Xiao Qin,
University of WisconsindMilwaukee
Srinivas R. Geedipally,
Texas A&MTransportation Institute
C H A P T E R
1
Introduction
1.1 Motivation
Althoughalotofefforthasbeenplacedbyagenciesacrosstheworldto
reducethenumberandseverityofcrashes1viaimprovementsinhighway
design,vehicletechnology,trafficpolicy,emergencyservices,andthelike,
theeffectsofhighwaycrashesonroadtransportnetworksarestillamajor
source of morbidity (Lord and Washington, 2018). Fig. 1.1 illustrates the
historicalstatisticsinroadwayfatalitiesintheUnitedStatesbetween1913
and 2018 (similartrendshave been observed among most industrialized
countries).Thisfigureshowsthatthetrendinroadwayfatalitieshasbeen
slightly going down since early 1970s, with sharp decreases during eco-
nomic recessions (further discussed later). This figure also demonstrates
thatwhenthevaluesareanalyzedbytakingintoaccountthevehiclemiles
traveled (a measure of exposure), the rate has been going significantly
down since the beginning of official crash data collected by the federal
government.Eventhoughthecrashrateshowsagreatreduction,theraw
numbers, as a public health measure, are still the most important factor
that guides the allocation of resources. For example, although the crash
rate is generally going down, the number of injured people arriving at
various emergency rooms located within a jurisdiction, or the patient
1Inthistextbook,weusetheterm“crash”toreflectoutcomeofacollisionbetweenavehicle
andafixedobject(i.e.,aneventwhereonlyonevehicleisinvolved),oneormorevehicles,
or one or more vulnerable road users (i.e., pedestrians, cyclists, etc.). Although some
people do not like to label a crash an “accident” because the word accident could
absolve the driver of any responsibility, the word accident could still be employed as
thatwordreferstotheprobabilisticnatureoftheevent.Ifaccidentswerecomingfroma
deterministicsystem,weshouldthereforebeableto“predict”withcertaintywhenone
or more crashes would occur in the future. Obviously, in the context of this textbook,
thisisnotpossible.
HighwaySafetyAnalyticsandModeling 1
https://doi.org/10.1016/B978-0-12-816818-9.00006-8 ©2021ElsevierInc.Allrightsreserved.
2
1. Introduction
FIGURE1.1
Numberoffatalitiesandfatalitiesper100millionvehiclemilesintheUnited
Statesbetween2013and1018(NSC,2018).
arrival rate, is the primary metric that the hospital management uses to
allocate medical services. The same information is also needed, for
example, for managing first responders, such as emergency medical
services,firefighters,andnational,regional,andlocalpoliceforces.Hence,
the desired attention usually focuses on crash or injury counts for many
safetyinterventions,althoughexposureintermsofvehiculartrafficand/or
segmentlengthmaystillneedtobeincorporatedintosomeofthemethods
utilizedforassessingsafety.
AccordingtotheWorldHealthOrganization(WHO),between2000and
2016, roadway-related crashes increased from about 1.15 million to 1.35
milliondeathsglobally(WHO,2018).Onanannualbasis,about80million
nonfatal injuries warranting medical care occur on highway networks
(Word Bank, 2014). Road traffic injuries are ranked eighth as the leading
causeofdeath(2.5%)amongpeopleofallages,rightinfrontofdiarrheal
diseases and tuberculosis (WHO, 2018). Vulnerable road users (i.e.,
pedestriansandcyclists)represent26%ofroadinjurydeaths,whiledrivers
andpassengersofmotorizedtwo-wheelandthree-wheelvehiclesaccount
for another 28% worldwide (WHO, 2018). Unfortunately, while a large
proportion of high-income countries have observed either a reduction or
no change in traffic-related deaths between 2013 and 2016, a significant
numberofmiddle-andlow-incomecountrieshaveobservedanincreasein
traffic-related deaths (WHO, 2018), in large part attributed to the rapid
motorizationobservedindevelopingcountries(WorldBank,2014).
The economic burden of crashes significantly impacts the global
economy. In the United States, for instance, highway crashes are
estimatedtohavecausedmorethanUS$871billionineconomiclossand
societalharmin2010(Blincoeetal.,2015).InEurope,itisestimatedthat
3
1.1 Motivation
crashes have cost more than US$325 billion (V280 billion) in economic
harm in 2015 (this value is considered underestimated) (Wijnen et al.,
2017), while in Australia the economic burden was estimated to be US$
23.9 billion (AU$33.2) in 2016 (Litchfield, 2017). Globally, it is estimated
that 3% of gross domestic product (GDP) is lost to highway crashes (all
severities)andcanbeashighas5%formiddle-andlow-incomecountries
(WHO, 2015).In short, in addition tothe pain andsufferingthat crashes
have caused to the victims of such events, highway crashes can signifi-
cantlyimpedeacountry’seconomicgrowthorviabilityacrosstheglobe.
As described in Fig. 1.1, the relationship that economic activity is
stronglylinkedtothenumberoffatalitiesobservedonhighwayshasnow
been well established (Wijnen and Rietveld, 2015; Elvik et al., 2015;
Wegman et al., 2017; Noland and Zhou, 2017; Shimu, 2019). In times of
economic growth, the number of crashes increases, while during
economic hardship (i.e., recession), the number of crashes decreases.
Fig. 1.2 illustrates such a relationship in detail, during the “Great Reces-
sion”of2007e09intheUnitedStates(theright-handsideofFig.1.1).The
influencingfactorsincludeunemploymentlevel,especiallyamongyoung
people, mode shift for people who are unemployed and lower exposure
by high-risk drivers (e.g., drivers below 25years old) during recession
periods(Bloweretal.,2019).Therelationshipbetweeneconomicactivity
and crashriskisveryimportant tobeunderstood beforeanalytical tools
areusedforanalyzinghighwaycrashdata.Thisistoavoidthepotential
confoundingeffectswhentreatmentsareimplementedandevaluatedfor
reducing the number andseverity of crashes.
FIGURE1.2 Fatalitiestrendduringthegreatrecessionof2007e09intheUnitedStates
(NCS,2018).
4
1. Introduction
Giventhemagnitudeoftheproblemassociatedwithhighwaycrashes,
numerouspublictransportationagenciesacrosstheworld,fromnational
tolocalagencies,haveplacedalotofeffort(i.e.,labor,promotion,etc.)and
allocatedalargeamountoffundsforreducingthenumberandseverityof
crashes, especially over the last 25years. For example, in the United
States,theNationalHighwayTransportationSafetyAgency(NHTSA)has
devoted US$908 million for highway-safety initiatives related to vehicle
safety, driver safety, and traffic enforcement in 2016 (NHTSA, 2016). In
2019, the Federal Highway Administration (FHWA) allocated US$2.60
billion solely for safety projects, which include research, dissemination,
engineering, and construction projects among others (FHWA, 2019).
Similarfinancialinvestmentshavebeenplacedbyvarioustransportation
agencies in Europe, Middle East, Asia, South Asia, and Oceania. The
strong commitment to reducing the negative effects of highway crashes
by decision-makers can be seen in the Vision Zero2 movement that was
first introduced by the Swedish Government in 1997. This movement
consistsinfindingnewandinnovativeapproachesandwaysofthinking
(i.e.,shiftingtheresponsibilityfromroaduserstohighwaydesignersand
engineers for reducing crashes) for significantly reducing, if not elimi-
nating, fatal and nonfatal injuries on highways, especially on urban
highways (Kristianssen et al., 2018). Vision Zero has been assertively
implemented invarious communities acrossthe globe.
Torespondtotheincreasinginvestment insafety-related projectsand
help with the aim of reducing, if not eliminating (as per Vision Zero)
highway crashes, research into methods and tools for analyzing crash
datahasexponentiallygrownduringthesametimeperiod.Thetestament
of such increase has recently been documented in two scientometric
overviewpublicationsthatvisuallymappedtheknowledgeinthefieldof
highwaysafety(i.e.,keyareasofresearch)andtheimpactoftheresearch
that has been published in the leading journal Accident Analysis and
Prevention (Zou and Vu, 2019; Zou et al., 2020). These authors identified
“crash-frequencymodelinganalysis”tobethecoreresearchtopicinroad
safetystudies,henceshowingtherelevanceofthematerialcoveredinthis
textbook.
Althoughdesignandapplicationmanuals,suchastheHighwaySafety
Manual (HSM) (AASHTO, 2010) or the Road Safety Manual (RSM)
(PIARC,2019),specializedtextbooks,suchastheonebyHauer(1997)on
before-after studies or Tarko (2020) on surrogate measuresof safety, and
review papers (see Lord and Mannering, 2010; Savolainen et al., 2011;
Mannering and Bhat, 2014), already exist, there is not a single source
availablethatcoversthefundamental(andup-to-date)principlesrelated
totheanalysisofsafetydata.AsdiscussedbyZouandVu(2019),thefield
2https://visionzeronetwork.org/.