Exploring Bicycle Route Choice

Behavior with Space Syntax Analysis

Zhaocai LiuZiqi Song,Ph.D.

AnthonyChen,Ph.D.Seungkyu Ryu,Ph.D.

DepartmentofCivil&EnvironmentalEngineeringUtahStateUniversity 1

Motivation

• Cyclingcanimproveurbanmobility,livabilityandpublichealth,anditalsohelpswithreducingtrafficcongestionandemissions.Understandingtheroutechoicebehaviorofcyclistscanpromotebicycletransportation.

• Cyclists’routechoicebehaviorisinfluencedbymanyfactors.Travelers’cognitiveunderstandingofthenetworkconfigurationhasbeenoverlookedbypreviousstudies.

• Spacesyntaxtheorycananalyzetravelers’cognitiveunderstandingofthenetworkconfiguration.

• Thecombinationofspacesyntaxtheoryandotherbicycle-relatedattributescanprovidebetterexplanatorypowerinmodelingcyclists’routechoicebehavior.

2Asaresult,wewanttoexploretheapplicationofspacesyntaxinmodelingbicycleroutechoicebehavior.

SpaceSyntaxTheory

•IntroducedbyHillerandHanson1984.•Originallyusedinarchitecturetomodeltheinfluenceofthespacestructureofabuildingonthemovementofpeopleinit.•DevelopedattheSpaceSyntaxLaboratoryatUniversityCollegeLondon.•Hasbeenappliedinurbanplanning,transport,socialinteractionandspatialeconomics

3

ApplicationsinTransportation

Theprocedureoftraveldemandestimationwithspacesyntax:

1) Representthenetworkwithagraphbyso-calledaxialanalysis

2) Measureconfigurationthroughtopologicaldistanceinthegraph,withoutmetricweighting

3) Predicttrafficflowdistributionbasedontheconfigurationalmeasurements.

4

AxialAnalysis

UnitSpace:Axialline

Definition:Thestraightroadsegmentthroughwhichtripmakersfindtheirextentofvisibility.

AxialMap:Urbanspacesuchasroadsandstreetsaremodeledbyaxiallines.

DualGraph:Eachaxiallineisrepresentedasanode,andtheintersectionsbetweenaxiallinesarerepresentedaslinks.

(a)Roadnetwork(b)Axialmap(c)Graphrepresentation

5

AccessibilityofUnitSpace:Integration

6

1.MeanDepth(MD)

SpaceSyntaxtypicallydescribethetopologicalconnectionsofunitspacethroughthenotionofdepthanalysis.

Whenmovingfromonespacetoitsconnectedspace,thereisatransitionofspace.Inspacesyntax,thetransitionofspace,whichisalsocalledsteporturn,istheunitofmeasurementof“distance”.

Thedistancefromonespacetoanotherspaceiscalleddepth.Themeandepthfromonespacetoallotherspacecanrepresenttheconnectivityofthespaceinthesystem.

3

1

2

1Step

1Step

1

(1,2) 1(1,3) 2

1 2 1.52

dd

MD

==+

= =

2

(2,1) 1(2,3) 1

1 1 12

dd

MD

==+

= =

( , )

1

( , ) the steps between space and

i kk

d i kMD

k

d i k i k

¹=-

=

å1

1Step

2

3

AccessibilityofUnitSpace:Integration

7

2.RelativeAsymmetry(RA)

Whenaspaceisdirectlyconnectedtoallotherspaces,ithasthelowestmeandepth.Inspacesyntax,weconsiderthisspacehasthehighestsymmetricity.

Whenaspaceneedtotravelthelongesttopologicaldistanttoreachotherspaces,ithasthehighestmeandepth.Inspacesyntax,ithasthelowestsymmetricity.

1( 1)( ) 11

nMD lowestn-

= =-

1

2

3

n

32

1

1Step

n

n

1

2

Depth(k-1)

Depth11

2

3k

1(1) 2(1) ( 1)(1)( )1 2n nMD highest

n+ + + -

= =-L

( )( ) ( )1 2( 1)

/ 2 1 2

kk

k k

MD MD lowestRAMD highest MD lowest

MD MDn n

-=

-- -

= =- -

AccessibilityofUnitSpace:Integration

8

3.RealRelativeAsymmetry(RRA)

Therelativeasymmetries(ofunitspaces)oftwodifferentsystemcannotbecompared,becausethesizeofasystem(nvalue)alsoinfluencestheaccessibilityoftheunitspaces.Thus(Hillieretal.1984)proposedafactorDn torelativise theRA.

222 log 1 13

( 1)( 2)n

kk

n

nnD

n nRARRAD

æ öæ + öæ ö - +ç ÷ç ÷ç ÷è øè øè ø=- -

=

AccessibilityofUnitSpace:Integration

9

4.Integration

Itdescribeshowclosely(ordistantly)thespaceistopologicallyaccessiblefromallotherspaceswithinagivensystemaddressingitssymmetricity andsize.

1kIntegrationRRA

=

GlobalIntegrationandLocalIntegration

10

•GlobalIntegration:itmeasureshowcloselyordistantlyeachspaceisaccessiblefromallotherspacesofasystem.

•LocalIntegration:integrationanalysisisrestrictedatalowerdepthofconnectivitytodeterminetheaccessibilityofthespaceatalocalorneighboringlevel.Forinstance,inanintegrationradius-3analysis,onlythespacethatarethreedepthsawayareconsidered.

AngularSegmentAnalysis

11

•IntroducedbyTurner(2001)•Axiallinesarebrokenintosegments•Stepbetweentwoconnectedsegmentsisweightedbasedontheanglebetweenthem.

(a) 1.0 step between line 1

and 2

(b) 0.5 step between line 1

and 2

(c) 𝜃/90 step between line 1

and 2

TravelDemandEstimation

12

•Spacesyntaxhasbeenusedtomodeldifferentmodeoftransportationincludingvehicle,metro,bicycleandpedestrian.

•Globalintegrationßà Vehiculartraffic•Localintegrationßà Pedestriantraffic

•Regressionanalysisisusedtocalibratethecorrelationbetweenintegrationandactualtrafficvolume.

No. Source Studyarea R-square Remarks1 Hillier1998 BalticHousearea 0.773 Pedestrian2 Hillieretal.1987 Bransbury 0.6413 Hillier1998 Santiago 0.54 Pedestrian4 Hillieretal.1987 Islington 0.536 Pedestrian

5 Eisenberg2005 Waterfront,Hamburg 0.523 Pedestrian

6 Peponisetal.1997 SixGreektowns 0.49 Pedestrian

7 Karimietal.2003 CityIsfahan 0.607 Vehicular

8 Peponisetal.1997 Buckhead,Atlanta 0.292 Vehicular

9 Paul2009 CityofLubbock,Texas 0.18 Vehicular

SpaceSyntaxinModelingBicycle

•Limitedworkshavebeendone(i.e.,Raford etal.2007;McCahill andGarrick2008;Manum andNordstrom2013)•Theresultsarenotasidealasexpected.•Bicycletrafficisatransportationmodethatfallssomewherebetweenvehicularandpedestriantraffic.Aspecificprocedureandproperspacesyntaxmeasurementneedstobedetermined.•Otherbicycle-relatedattributesalsoneedtobeconsidered.

13

Methodology

Bicycle-relatedAttributes

(1) Linkcognition:representedbyspacesyntaxmeasurements(2)Segmentbicyclelevelofservice(BLOS):evaluatedbasedon

HCM(2010)(3)Motorvehiclevolume(4)Linkpollution:estimatedbasedonanonlinearmacroscopic

modelofWallaceetal.(1998)(5)Presenceofbicyclefacilityonalink(6)Averageslopeofterrainonasegment

14

Methodology

15

StatisticalModeling

Linearregressionwasusedtoanalyzetherelationshipbetweenbicyclevolumeandvarioussegmentattributes.

𝑌" = 𝛽% + 𝛽'𝑋'" + 𝛽)𝑋)" +⋯+ 𝛽+𝑋+"where𝑌"=thebicyclevolumeonlink𝑎𝑋+" =thevalueofexplanatoryvariable𝑚 onlink𝑎.𝛽+ =modelcoefficientforvariable𝑚.

CaseStudy

16

BicycleCountsinSaltLakeCity

Date:Sep15th(Tue),16th(Wed),17th(Thu),19th(Sat),and20th(Sun).Location:19intersectionsDuration:2hourseachday,5-7pmonweekdays,12-2pmonweekends

Statistic All Counts Weekday Weekend

Number of counts 95 57 38

Minimum 2 7 2

Maximum 161 129 161

Median 47.0 47 42

Mean 54.8 54.1 55.9

Standard deviation 35.7 31.3 41.8

SummaryStatisticsfor2-hourBicycleCounts

CaseStudy

17

FIGURE 4.1 Map of Locations of Bicycle Counts

CaseStudy

18

SpaceSyntaxAnalysis

Globalintegrationandlocalintegrationwithametricradiusof3kilometers(1.86miles)werecalculatedusingsegmentanalysis.

(a) Global Integration

(b) Local Integration

5.1 to 1069.27 5.1 to 2962.79

CaseStudy

19

SupplementaryData

• Motorvehiclevolume:annualaveragedailytraffic(AADT)datafromUDOT.

• Dataaboutspeedlimitandnumberoflanes:UDOT• Bicyclelanedata:SaltLakeTransportationDivision• Terrainslopedata:digitalelevationmodel(DEM)from

UtahAutomatedGeographicReferenceCenter(AGRC)• Otherdata:estimatedbasedonHCM(2010)

CaseStudy

20

RegressionAnalysis

Localintegrationismoreappropriateinmodelingbicycletraffic.

ModelVariable

Coefficients

GlobalIntegrationModel LocalIntegrationModel

Constant

18.877

(0.325)

7.056

(0.332)

IntGa

0.001

(0.617) -

IntLa -

0.010

(0.005)

R-square 0.016 0.396

F-statistic

0.261

(0.617)

10.502

(0.005)

CaseStudy

21

RegressionAnalysis

Thecombinationoflocalintegrationandmotorvehiclevolumesprovidesastatisticallysignificantmodelwhichhasmoreexplanatorypower.

Othermodelsareeitherstatisticallynon-significantorunreasonable.

Model1 Model2 Model3

Constant

264.884

(0.075)

137.917

(0.333)

18.200

(0.039)

IntLa

0.014

(0.002)

0.012

(0.002)

0.013

(0.001)

BSega

-20.242

(0.242)

-8.295

(0.648)

Motva

-0.001

(0.497)

-0.001

(0.191)

-0.001

(0.041)

PSega

-4526.53

(0.068)

-2107.57

(0.352)

BikeLa

0.298

(0.970) -

Slopea

4.978

(0.034) -

R-square 0.731 0.588 0.547

F-statistic

4.986

(0.011)

4.635

(0.015)

9.055

(0.003)

Summary

• Localintegration,whichdescribestheaccessibilityataneighboringlevel,ismoreusefulinmodelingbicycleroutechoice.

• Spacesyntaxtheoryispromisinginanalyzingcyclists’cognitiveunderstandingofthenetworkconfiguration.

• Thecombinationofspacesyntaxmeasurementandotherbicycle-relatedattributescanimprovetheexplanatorypoweroftheregressionmodel.

22

23