You are on page 1of 45

CIVILENGINEERINGSTUDIES

 IllinoisCenterforTransportationSeriesNo.16Ͳ020
UILUͲENGͲ2016Ͳ2020
 ISSN:0197Ͳ9191


EVALUATIONOFSOFTWARE
SIMULATIONOFROADWEATHER
INFORMATIONSYSTEM

PreparedBy
WilliamR.Vavrik,Ph.D.,P.E.
CarmineE.Dwyer,P.E.
WouterC.Brink,Ph.D.
GreggLarson
AppliedResearchAssociates,Inc.




ResearchReportNo.FHWAͲICTͲ16Ͳ018

Areportofthefindingsof
ICTPROJECTR27Ͳ142
EvaluationofSoftwareSimulation
ofRoadWeatherInformationSystem



IllinoisCenterforTransportation
September2016
TECHNICALREPORTDOCUMENTATIONPAGE
1.ReportNo. 2.GovernmentAccessionNo. 3.Recipient’sCatalogNo.
FHWAͲICTͲ16Ͳ018  
4.TitleandSubtitle 5.ReportDate
EvaluationofSoftwareSimulationofRoadWeatherInformationSystem September2016
6.PerformingOrganizationCode

7.Author(s) 8.PerformingOrganizationReportNo.
WilliamR.Vavrik,CarmineE.Dwyer,WouterC.Brink,GreggLarson ICTͲ16Ͳ020
UILUͲENGͲ2016Ͳ2020
9.PerformingOrganizationNameandAddress 10.WorkUnitNo.
AppliedResearchAssociates,Inc. 
100TradeCentreDr.,Suite200 11.ContractorGrantNo.
Champaign,IL61820 R27Ͳ142
12.SponsoringAgencyNameandAddress 13.TypeofReportandPeriodCovered
IllinoisDepartmentofTransportation(SPR) Oct.1,2014–Sep.30,2016
BureauofMaterialsandPhysicalResearch 14.SponsoringAgencyCode
126EastAshStreet FHWA
Springfield,IL62704
15.SupplementaryNotes
ConductedincooperationwiththeU.S.DepartmentofTransportation,FederalHighwayAdministration.

16.Abstract
Aroadweatherinformationsystem(RWIS)isacombinationoftechnologiesthatcollects,transmits,models,anddisseminates
weatherandroadconditioninformation.SensorsmeasurearangeofweatherͲrelatedconditions,includingpavement
temperatureandstatus(wet,dry,snow),subsurfacepavementtemperature,windspeedanddirection,precipitation,waterlevel
conditions,humidity,andvisibility.Thesedataaretransmittedtoautomatedwarningsystems,trafficoperationscenters,
emergencyoperationscenters,androadmaintenancefacilitiesfordecisionsupport.TheEnhancedIntegratedClimaticModel
(EICM)isacomputerizedheatandmoistureflowmodelthatsimulateschangesinpavementandsubgradeproperties.Ithas
evolvedoverthepast40yearsandisakeymoduleintheAmericanAssociationofStateHighwayandTransportationOfficials
(AASHTO)PavementMEDesignsoftware.UsingtheEICMasasoftwareͲbasedRWIScan“virtualize”thedatathatwouldbe
gatheredbyconventionalRWIShardwareandsoftwaresystems.ThesoftwareͲbasedRWISstationswouldprovidecurrent
conditionsaswellaspavementtemperatureforecaststosupplementorreplacehardwareintheRWISnetwork.

TheobjectiveofthisstudywastoevaluatetheuseoftheEICMtodeterminepavementsurfacetemperatureforwinter
maintenanceoperations.DetailedpavementinformationatIllinoisDepartmentofTransportation,IllinoisTollway,andMcHenry
CountyRWISlocationswascollectedandusedtomodelpavementsurfacetemperatureswiththeEICM.Themodeledpavement
surfacetemperatureswerecomparedwiththemeasuredpavementsurfacetemperaturesfromtheRWISsensors.

Dataanalysisshowedthat,whenthepavementmaterialsareusedatthecorrectthicknessandrecommendeddefaultvaluesare
usedformaterialtypes,areasonablepavementsurfacetemperaturepredictioncanbeobtained.Usingtheserecommended
defaultvaluesforthermalconductivity,heatcapacity,andshortwaveabsorptivityresultsinareasonablemodeledpavement
temperaturethatcanbeusedforunderstandingfuturesnowandicepotentialfromforecastweatherdata.

17.KeyWords 18.DistributionStatement
Roadway,weather,RWIS,vRWIS,EICM,pavement,surface Norestrictions.Thisdocumentisavailablethroughthe
temperature,thermalconductivity,heatcapacity,thermal NationalTechnicalInformationService,Springfield,VA
absorptivity,icing,precipitation,freezing 22161.
19.SecurityClassif.(ofthisreport) 20.SecurityClassif.(ofthispage) 21.No.ofPages 22.Price
Unclassified Unclassified 34pp.

FormDOTF1700.7(8Ͳ72)     Reproductionofcompletedpageauthorized

ACKNOWLEDGMENT,DISCLAIMER,MANUFACTURERS’NAMES
ThispublicationisbasedontheresultsofICTͲR27Ͳ142,EvaluationofSoftwareSimulationofRoad
WeatherInformationSystem.ICTͲR27Ͳ142wasconductedincooperationwiththeIllinoisCenterfor
Transportation;theIllinoisDepartmentofTransportation(IDOT);andtheU.S.Departmentof
Transportation,FederalHighwayAdministration(FHWA).Additionalcooperationwasprovidedbythe
IllinoisTollwayandMcHenryCounty.

Membersofthetechnicalreviewpanel(TRP)werethefollowing:
x DerekParish,TRPChair
x RickWalker,TRPPastChair,IDOTD6retired
x LlyleBarker,NationalWeatherService
x RuebenBoehler,IDOT
x RyanCulton,IDOT
x BrentGronewold,IDOTD6
x ScottKapton,IllinoisTollway
x EdwardMarkison,McHenryCounty
x DeanMentjes,FHWA
x TimPeters,IDOTBLRS

Thecontentsofthisreportreflecttheviewoftheauthors,whoareresponsibleforthefactsandthe
accuracyofthedatapresentedherein.Thecontentsdonotnecessarilyreflecttheofficialviewsor
policiesoftheIllinoisCenterforTransportation,theIllinoisDepartmentofTransportation,orthe
FederalHighwayAdministration.Thisreportdoesnotconstituteastandard,specification,or
regulation.
Trademarkormanufacturers’namesappearinthisreportonlybecausetheyareconsideredessential
totheobjectofthisdocumentanddonotconstituteanendorsementofproductbytheFederal
HighwayAdministration,theIllinoisDepartmentofTransportation,ortheIllinoisCenterfor
Transportation. 

i
EXECUTIVESUMMARY
Overthepast40+years,acomputerizedclimaticmodelhasevolvedthatcanaccuratelypredictthe
temperatureinpavementsystemsbasedonatmosphericweatherdatainputsandpavementmaterials.This
modelistheEnhancedIntegratedClimaticModel(EICM).TheEICMstartedastheUniversityofIllinoisHeat
TransferModelin1969andhasundergonecontinuousimprovement,allowingittobeusedwithcurrent
pavementmaterialsandstructuredata,historicalweatherdata,andforecastatmosphericweatherdatato
forecastpavementtemperatureandmoistureconditions.WhenusedasarealͲtimemonitoringtool,theEICM
canevaluatetheprobabilityoficingconditionsatthepavementsurface.ThefiveweatherͲrelatedparameters
thatarerequiredtoruntheEICMareairtemperature,windspeed,percentageofsunshine,precipitation,and
relativehumidity.Theseinputsareusedtoestimatetheheattransferbetweentheroadandtheatmosphere.
UsingtheEICMallowspavementsurfaceconditionstobemonitoredonafrequentbasis,identifyingtimesof
highprobabilityofsurfaceiceformation.

TheEICMwasusedtomodelmeasuredpavementtemperatures.Theobjectiveofthisprojectwastoevaluate
theuseoftheEICMfordeterminingpavementsurfacetemperatureforwintermaintenanceoperations.
ModelingthepavementtemperaturescouldprovidevirtualRWISdataatacostthatisconsiderablylessthan
thecostofphysicalsensorsandsystems.Thestudyinvolvedcollectingpavementmaterialinformation,
modelingthepavementsoverthepast5yearsattheselocationswithactualatmosphericweatherdata,and
evaluatingthedifferencebetweenactualandEICMmodeledpavementtemperatures.

Comparisonsbetweenthemeasuredandpredictedpavementtemperatureweremadebycollectingmeasured
pavementtemperaturesandpavementstructureinformationforvarioussitesacrosstheStateofIllinois.A
totalof38siteswasselected,whichconsistedof25IllinoisDepartmentofTransportation(IDOT)sites,11
IllinoisTollwaysites,andtwoMcHenryCountysites.Thetemperaturedatawerecomparedusingthe
PavementMEdefaultparametersforthermalconductivity,heatcapacity,andabsorptivity.Additionally,the
modelwascalibratedforeachsitebyadjustingtheparametersdiscussedabove.

Twodifferentclimatedatasetswereusedtomodelthepredictedpavementtemperatures.Bothclimate
datasetsshowedverygoodresultswhencomparingthemeasuredandpredictedpavementtemperatures(R2>
0.8).Theaveragemeantemperaturedifferenceamongallsiteswas2.5°Fforalltemperaturesand1.1°Ffor
both±10°Fand±5°Ffromfreezing(32°F).Thecalibrationprocesswascompletedbyrunningadesignmatrixof
differentthermalconductivity,heatcapacity,andabsorptivityvalues.Thematrixconsistedof27total
combinationsforbothflexibleandrigidpavements.TheEICMwasexecutedforeachcombination,andthe
rootmeansquareerror(RMSE)andmeanerrorweredetermined.Thetoptencombinationsbasedonthe±5°F
RMSEwereusedtodeterminethenewrecommendedvaluesforIllinoispavements.Basedontheresults,the
recommendedvaluesforthermalconductivity,heatcapacity,andabsorptivityofpavementswithaPCC
surfacewere1.5,0.3,and0.85,respectively.Thevalueswere1.0,0.25,and0.85forflexibleandcomposite
pavements.

UsingtherecommendedthermalinputsandtheactualinͲplacepavementstructure(materialtypesandlayer
thicknesses)asinputstotheEICMyieldsreasonablemodelaccuracywithameanerrorthatisgenerallyless
than2°F.Understandingthatthemodelerrorisafunctionofthequalityoftheweatherdata,thequalityofthe
sensordata,andthevalidityofthemodel,anerroroflessthan2°Fisconsideredreasonableandisappropriate
foruseinavirtualRWIS. 

ii
CONTENTS
LISTOFFIGURES.................................................................................................................IV
LISTOFTABLES...................................................................................................................V

CHAPTER1:INTRODUCTION................................................................................................1
1.1WHATISTHEEICM?..............................................................................................................1
1.2APPLICABILITYOFEICMTOROADWEATHERFORECASTING.................................................2
1.3ROADWEATHERINFORMATIONSYSTEMS............................................................................3
1.4CURRENTUSE/COSTOFRWIS?.............................................................................................3
1.5PROJECTOPPORTUNITYSTATEMENT....................................................................................3
1.6PROBLEMSTATEMENT.........................................................................................................4
1.7PROJECTOBJECTIVE..............................................................................................................4

CHAPTER2:RWISDATACOLLECTIONANDSITESELECTION..................................................5
2.1GATHERINGDATAFROMILLINOISRWISSITES......................................................................5
2.2SELECTINGRWISSITESFORSTUDYANALYSIS.......................................................................6

CHAPTER3:CREATIONOFPREDICTEDDATA......................................................................10
3.1PAVEMENTDATACOLLECTION...........................................................................................10
3.2VIRTUALRWISSITECREATION............................................................................................11

CHAPTER4:ANALYSISOFPREDICTEDVERSUSMEASUREDDATA........................................13
4.1DATAPREPARATION...........................................................................................................13
4.2DATACOMPARISON...........................................................................................................15
4.3ANALYSISSUMMARY..........................................................................................................23

CHAPTER5:CALIBRATIONOFMODELINPUTS....................................................................25
5.1INTRODUCTION..................................................................................................................25
5.2CALIBRATIONPROCESS.......................................................................................................25
5.3CALIBRATIONDESIGNMATRIX...........................................................................................25
5.3CALIBRATIONRESULTS.......................................................................................................26
5.4SUMMARY..........................................................................................................................29

CHAPTER6:RECOMMENDATIONSANDCONCLUSIONS......................................................33

iii
LISTOFFIGURES
Figure1.Schematicofheattransferatthepavementsurface................................................................2
Figure2.Pavementsurfacesensor...........................................................................................................5
Figure3.Onewintermonth’shourlytemperaturesforoneRWISsitewithmissingdata......................6
Figure4.Onewintermonth’shourlytemperaturesfortwoRWISsiteswithcompletedata.................7
Figure5.IllinoisRWISsitesselectedformeasuredversuspredictedanalysis........................................9
Figure6.Pavementcoring......................................................................................................................10
Figure7.Removingbasecoursematerial..............................................................................................10
Figure8.KonicaMinoltaCMͲ2500cspectrophotometer.......................................................................11
Figure9.The38selectedsitesshowninARA’sVirtualRWISprogram..................................................12
Figure10.Sitelocations..........................................................................................................................14
Figure11.Site1timeͲseriespavementtemperaturedata....................................................................17
Figure12.Site1measuredversuspredictedpavementtemperaturedata..........................................18
Figure13.Measuredversuspredictedpavementtemperaturecomparisons......................................19
Figure14.Site16andSite18timeͲseriespavementtemperaturedata...............................................19
Figure15.Site1measuredversuspredictedpavementtemperatureusingNARRͲgenerated
climatedata............................................................................................................................................21
Figure16.Site1timeͲseriespavementtemperaturedatausingNARRclimatefiles............................22
Figure17.Temperaturedifferencehistogram.......................................................................................23
Figure18.SummaryofR2valuesforallprojectsites.............................................................................24
Figure19.RootmeansquarederrorcomparisonforthePavementMEDesigndefaults.....................27

 

iv
LISTOFTABLES
Table1.NumberofPotentialRWISStudySites.......................................................................................5
Table2.ProjectSites..............................................................................................................................13
Table3.AverageMeanTemperatureDifferences.................................................................................24
Table4.DesignMatrixforPCCandHMAPavements............................................................................26
Table5.PCCPavementCalibrationResults—TopTen...........................................................................28
Table6.ErrorResultsAfterCalibration..................................................................................................30
Table7.HMAPavementCalibrationResults—TopTen.........................................................................31
Table8.CompositePavementCalibrationResults—TopTen................................................................32
Table9.RecommendedThermalInputsfortheEICMinaVirtualRWIS...............................................33


v
CHAPTER1:INTRODUCTION
Overthepast40+years,acomputerizedclimaticmodelhasevolvedthatcanaccuratelypredictthe
temperatureinpavementsystemsbasedonatmosphericweatherdatainputsandpavement
materials.ThismodelistheEnhancedIntegratedClimaticModel(EICM).TheEICMstartedasthe
UniversityofIllinoisHeatTransferModelin1969andhasundergonecontinuousimprovement,
allowingittobeusedwithcurrentpavementmaterialsandstructuredata,historicalweatherdata,
andforecastatmosphericweatherdatatoforecastpavementtemperatureandmoistureconditions.
WhenusedasarealͲtimemonitoringtool,theEICMcanevaluatetheprobabilityoficingconditionsat
thepavementsurface.

Historically,theEICMhasbeenusedforpavementperformanceresearchandasakeycomponentin
pavementperformancemodelingandpavementdesign.TheEICMisoneofthekeymodulesinthe
AmericanAssociationofStateHighwayandTransportationOfficials(AASHTO)PavementMEDesign
software,whichisthecommercialversionoftheMechanisticͲEmpiricalPavementDesignGuide
(MEPDG).TheEICMhasbeendemonstratedintheliteraturetobeusedfor:
x Soilproperties—shrinkage,suction,moisture,andtemperature,freezeͲthawdamage
x Unboundmaterials—resilientmodulus,moisture,freezeͲthawdamage,frostpenetration,
seasonalvariationofmoduli
x Thermalstressesinconcretepavements—warping,curling,performancemodeling,cracking
x Thermalcrackingandthermalengineeringpropertiesofasphaltpavements

WHATISTHEEICM?
TheEICMisaoneͲdimensionalforwardfinitedifferenceheatandmoistureflowmodelthatsimulates
changesinpavementandsubgradeproperties.Itincorporatespatternsofrainfall,solarradiation,
cloudcover,windspeed,andairtemperatureatthepavementsurface.ThefiveweatherͲrelated
parametersthatarerequiredtoruntheEICMincludeairtemperature,windspeed,percentageof
sunshine,precipitation,andrelativehumidity.Theseinputsareusedtoestimatetheheattransfer
betweentheroadandtheatmosphere,asshowninFigure1.

TheEICMismadeupofthreemaincomponents:theclimateͲmaterialsstructuralmodel(CMSmodel)
developedattheUniversityofIllinois,thefrostͲheaveandsettlementmodel(CRRELmodel)
developedattheUnitedStatesArmyColdRegionsResearchandEngineeringLaboratory,andthe
infiltrationͲdrainagemodel(IDmodel)developedatTexasA&MUniversity’sTexasTransportation
Institute.TheEICMpredictstemperature,resilientmodulusadjustmentfactors,porewaterpressure,
watercontent,frostandthawdepths,andfrostheavethroughoutthecompletepavementand
subgradeprofilefortheentiredesignlifeofthepavementstructure.

OneofthelimitationsofthecurrentmodelsintheEICMisthetreatmentofprecipitationwhen
determiningthepavementsurfaceconditions.Currently,precipitationdataareusedtodetermine
infiltrationofmoistureintothepavement,nottoimpactthesurfacetemperatureofthepavement.

1
Thislimitationwillbeexploredandreviewedtodeterminehowbesttochangethemodelinputsor
themodelformulationsothatprecipitationistakenintoaccountintheheatbalanceatthepavement
surface.


Figure1.Schematicofheattransferatthepavementsurface.

1.2APPLICABILITYOFEICMTOROADWEATHERFORECASTING
TheEICMcaneasilyaccommodatehourlyormorefrequentweatherdatainputs,therebymakingit
adaptabletorealͲtimepredictionofpavementtemperaturesandprecipitationconditions.Usingthe
EICMallowspavementsurfaceconditionstobemonitoredonafrequentbasis,identifyingtimesof
highprobabilityofsurfaceiceformation.

AnadvantageofusingtheEICMͲbasedpavementforecastisthatahistoryofthepavementsurface
temperaturescanbemaintainedsothatconditionsthatleadtoicingcanbereadilyidentified.Oneof
theseconditionscouldoccur,forexample,insituationswhereextremecoldspellsarefollowedby
precipitationattemperaturesjustabovefreezing.Inthosesituations,thepavementsurfaceand
temperaturesasafunctionofdepthwilloftenremainbelowfreezing,andiceformswhenrainfall
occursjustabovethefreezingtemperature.

BecauseusingtheEICMͲbasedpavementforecastrequiresnophysicalhardwareinstallation,EICMͲ
basedpavementtemperatureandforecaststationsforroadweatherapplicationswouldbe
consideredsoftwareͲbasedorvirtualroadweatherinformationsystems.Itisanticipatedthat
softwareͲbasedsystemswouldhavesignificantadvantagesinanoverallwintermaintenance
program.

2
1.3ROADWEATHERINFORMATIONSYSTEMS
Aroadweatherinformationsystem(RWIS)isacombinationoftechnologiesthatcollects,transmits,
models,anddisseminatesweatherandroadconditioninformation.ThecomponentofanRWISthat
collectsweatherdataistheenvironmentalsensorstation(ESS).AnESSisasuiteofsensorsthat
collectsandtransmitspavementandmeteorologicaldata.SensorsmeasurearangeofweatherͲ
relatedconditions,includingpavementtemperatureandstatus(wet,dry,snow),subsurface
pavementtemperature,windspeedanddirection,precipitation(amount,occurrence,type),water
levelconditions,humidity,andvisibility.Thesedataaretransmittedtoautomatedwarningsystems,
trafficoperationscenters,emergencyoperationscenters,androadmaintenancefacilitiesfordecision
support.Weatherserviceprovidersalsousethedatatodeveloptailoredroadweatherproducts(for
example,pavementtemperatureforecasts).Alloftheseactivitiesmakeforsaferroadwayconditions
formotorists.

Inthepast,RWISwereusedalmostexclusivelybystateandlocaltransportationmaintenance
departmentstomakebetteroperationaldecisions.Thecollectedweatherdataallowedagenciesto
coordinateantiͲicingpractices;efficientlyplanwintermaintenanceroutes;reducetheamountof
chemicals,sand,andsaltusedinroadwayclearingoperations;andreducewearandtearon
maintenancevehicles.Now,stateandlocaltransportationagenciesaresharingweatherdatawitha
broaderaudienceofweatherdatausers,recognizingtheinherentvalueofabetterͲinformed
travelingpublic.

1.4CURRENTUSE/COSTOFRWIS?
Thecosttoprocure,implement,andmaintainanRWISisasignificantinvestment.AsingleRWIS
stationcanhaveaninitialcostof$20,000to$50,000perstation,dependingonthelocationandthe
numberofsensorsused.ThesoftwareandhardwaretoruntheRWIScanbeanadditional$20,000.
TelephoneandothercommunicationlinesarerequiredtotransmitthedatatotheRWIS.The
computersandhardwarehavealimitedusefullifeandneedtobeupdatedapproximatelyevery5
years.Annualsystemmaintenancecancost$1,000ormoreperstationperyear.Reducingthese
coststoauniformequivalentannualcostbringsthecostofownershipinarangefrom$1,000to
$10,000perstationperyear.

1.5PROJECTOPPORTUNITYSTATEMENT
UsingtheEICMasasoftwareͲbasedRWIScan“virtualize”thedatathatwouldbegatheredby
conventionalRWIShardwareandsoftwaresystems.SoftwareͲbasedRWISstationswouldprovide
currentconditionsaswellaspavementtemperatureforecaststosupplementorreplacehardwarein
theRWISnetwork.TheabilitytohaveasoftwareͲbasedRWISevery10milesoneveryinterstatein
Illinoisandhavestationsineverycountywouldprovideimprovedroadweathermaintenance
decisionsupporttotheIllinoisDepartmentofTransportation(IDOT)andlocalagencies.

3
1.6PROBLEMSTATEMENT
UsingtheEICMtomodelpavementconditionsandprovideapavementforecastrequiresthatthe
modeledpavementconditionsmimictheactualobservedpavementconditions.Theaccuracyofthe
modelrelatesdirectlytotheapplicabilityofusingitforwintermaintenanceopportunities.

TheaccuracyanEICMͲbasedweatherstationdependsonthequalityofthemodel,thequalityofthe
atmosphericweatherdataandforecast,andthepavementproperties.Becausethequalityofthe
modelishighandwelldocumentedintheliterature,andthequalityofatmosphericweatherdata
andforecastsiscoveredinotherdisciplines,thisstudyfocusedonunderstandinghowthepavement
materialpropertiesimpactthemodelaccuracy.

ThepavementpropertyinputsfortheEICMincludeheatcapacity,thermalconductivity,shortwave
absorptivity,porosity,unitweight,andpermeability.Veryfewofthesepavementpropertiesare
measuredinthepavementdesignorconstructionprocess.ThisprojectdeterminedthebestEICM
modelinputsusingdataavailablefromdesignorconstructionthatminimizemodelerror.Thestudy
involvedcollectingpavementmaterialinformation,modelingthepavementsoverthelast5yearsat
theselocationswithactualatmosphericweatherdata,andevaluatingthedifferencebetweenactual
andmodeledpavementtemperatures.

1.7PROJECTOBJECTIVE
TheresearchteamevaluatedtheuseoftheEICMtodeterminepavementsurfacetemperaturefor
wintermaintenanceoperations.DetailedpavementinformationwascollectedatIDOT,Illinois
Tollway,andMcHenryCountyRWISlocationsandusedtomodelpavementsurfacetemperatures
withtheEICM.Themodeledpavementsurfacetemperatureswerecomparedwiththemeasured
pavementsurfacetemperaturesfromtheRWISsensors.Theultimategoaloftheprojectwasto
determinewhetherappropriateEICMinputstominimizemodelerrorcanbedeterminedfromreadily
availabledata.

4
CHAPTER2:RWISDATACOLLECTIONANDSITESELECTION

2.1GATHERINGDATAFROMILLINOISRWISSITES
Tocomparemeasuredandpredictedpavementsurfacetemperatures,theresearchersgatheredall
historicalRWISdatafromthestudyparticipants(IDOT,IllinoisTollway,andMcHenryCounty).RWIS
stationscanhavemanydifferentsensorsprovidingweatherͲrelateddata.Thisstudywasprimarily
concernedwithtemperaturedatafrompavementsurfacesensors,suchastheoneshowninFigure2.


Figure2.Pavementsurfacesensor.

Aftertheresearchteamreceivedthedatasets,thedatafieldswerereviewedtodeterminewhich
stationshadpavementsurfacesensors.Somestationshadmultiplepavementsurfacesensors,some
hadnosurfacesensors,andothershadonlybridgedecksurfacesensors.Table1reportsthetotal
numberofRWISstationsforeachparticipatingagency,aswellasthenumberofstationswith
pavement(nonͲbridgedeck)surfacesensors.ThenumberofRWISsiteswithpavementsurface
sensorsrepresentsthepotentialsitestobeincludedinthemeasuredversuspredictedcomparisons.

Table1.NumberofPotentialRWISStudySites
NumberofRWISSiteswith
ParticipatingAgency TotalNumberofRWISSites PavementSurfaceSensors
IllinoisDepartmentofTransportation 58 49

IllinoisTollway 17 13

McHenryCounty 6 6


5
2.2SELECTINGRWISSITESFORSTUDYANALYSIS
Therewere68potentialstudysiteswithpavementsurfacesensors.Becausesensorscanmalfunction,
bedamaged,orgomissing,theresearchersanalyzedthehistoricalpavementsurfacetemperature
datafromeachsiteforcompletenessandreasonablevalueranges.Becausepredictedpavement
surfacetemperaturesforthisstudywerebasedoncurrentpavementstructures(currentmaterial
propertiesandlayerthicknesses),onlythefivemostrecent,fullͲyeardatasets(2009to2014)were
reviewed.Thegoalofthisreviewwastoeliminatesiteswithtoomuchmissingdata,particularlyin
themonthsinwhichfreezingandicingsurfaceconditionscanexist,andtoeliminatesiteswithoutlier
data.

Figure3isanexampleofonesitewithaconsiderableamountofmissingdataforagivenwinter
monthofoneyear.Figure4isanexampleoftwositeswithgooddata—acompletesetofdatawith
expected,reasonablevaluesforagivencoldermonth.

Missing Data


Figure3.Onewintermonth’shourlytemperaturesforoneRWISsitewithmissingdata.

6
Good

Good


Figure4.Onewintermonth’shourlytemperaturesfortwoRWISsiteswithcompletedata.

7
WhenRWISsiteswereselectedforpredictioncomparisons,arepresentativedistributionofother
factorswasconsidered.Thesefactorsincludedthefollowing:
x Locationsthroughoutthestate
x Pavementtype(asphaltconcrete,portlandcementconcrete,composite)
x Pavementthickness
x Trafficvolume

Attheconclusionofthemeasuredtemperaturedataanalysis,38RWISpavementsurfacesensorsites
wereselected.Thenumberofsitesselected,byagency,wasasfollows:
x IllinoisDepartmentofTransportation:25
x IllinoisTollway:11
x McHenryCounty:2

Figure5isamapofthe38sitesandshowsthedistributionofsitelocationsacrossIllinois.

8

Figure5.IllinoisRWISsitesselectedformeasuredversuspredictedanalysis.

9
CHAPTER3:CREATIONOFPREDICTEDDATA

3.1PAVEMENTDATACOLLECTION
PavementCoring
Toensurethepredictedpavementsurfacetemperaturesweredevelopedbasedonthecurrent,inͲ
placepavementstructure,theresearcherssentafieldcrewtoeachofthe38selectedRWISsitesto
collectpavementdata.Datacollectionincludedthefollowing:
x Coringthepavementstructuretodeterminepavementlayerthicknesses
x Augeringthebasecoursetodeterminebaselayerthicknessandmaterialtype
x Measuringpavementsurfacecolor

Figure5depictthecoringandauguringprocesses,respectively.


Figure6.Pavementcoring.


Figure7.Removingbasecoursematerial.

10
ThecorelogsfromtheRWISsitesarepresentedinAppendixAofthisreport.Theselogsnotethecore
locationandthicknessofthepavementlayers.Corephotographsarealsoprovided.

PavementColor
Fivepavementsurfacecolormeasurementswererecordedineachlocationforpotentialcorrelation
betweensurfacecolorandpavementthermalabsorptivity.Measurementsweretakenwitha
spectrophotometer,liketheoneshowninFigure8.Allmeasurementsweretakenpriortocoring.
Onewastakeninthecenterofthecorelocation,andtheremainingfourweretakenaroundthe
perimeterofthecorelocation.


Figure8.KonicaMinoltaCMͲ2500cspectrophotometer.

ThepavementcolorsfromtheRWISsitesarepresentedinAppendixBofthisreport.Theselogsnote
thelocationwherethecolormeasurementwastakenandthemeasuredpavementcolor.
Photographsarealsoprovidedwithastandardgrayscalecolorband.

3.2VIRTUALRWISSITECREATION
ThecollectedpavementdatawereusedtocreateinputfilesfortheEICMforeachofthe38sites.The
EICMinputfilewasthenusedtocreatetherespectivevirtualRWISsites.Figure9showsamapofall
38sitesinthevirtualRWISprogram.Duringthecreationofthesesites,thefivemostrecentyearsof
historicaldata(2009to2014)weregeneratedforcomparisontothemeasureddata.

11

Figure9.The38selectedsitesshowninARA’sVirtualRWISprogram.

12
CHAPTER4:ANALYSISOFPREDICTEDVERSUSMEASUREDDATA

4.1DATAPREPARATION
Table2providessummaryinformationonthe38sitesincludedinthestudy.Therawmeasured
pavementtemperaturedatawereextractedfromthevariousIDOT,IllinoisTollway,andMcHenry
Countydatabasesforprocessingpriortocomparingthemeasuredandpredicteddata.

Figure10showsthegeographicaldistributionoftheprojectsitesthroughoutIllinois.Thedifferent
colorsrepresentthedifferentdatasources.

Table2.ProjectSites
Site SiteID_ Pavement
Number Station SensID Lat Long Type
Site1 Tollway_IͲ355@DesPlainsRiverͲMP9.5 5_0 41.66389 –88.02479 PCC
Site2 Tollway_IͲ355@IͲ80ͲMP0.2 6_0 41.53845 –87.96227 PCC
Site3 Tollway_IͲ294@MileLongBridgeͲMP22.0 2_0 41.75408 –87.87294 PCC
Site4 Tollway_IͲ294@BensenvilleRRBridgeͲMP36.9 1_0 41.94062 –87.89380 PCC
Site5 Tollway_IͲ94@EdensSpurͲMP27.2 17_0 42.14887 –87.84001 Composite
Site6 Tollway_IͲ88@FoxRiverBridgeͲMP117.1 8_0 41.79595 –88.32345 PCC
Site7 Tollway_IͲ88@CountyLineRoadͲMP100.7 7_0 41.8745 –88.60807 Composite
Site8 Tollway_IͲ88@Route23ͲMP92.5 9_0 41.90289 –88.75318 Composite
Site9 McHenry_CH30@Westbound 665001_1 42.31308 –88.66394 AC
Site10 McHenry_CH30@Eastbound 665001_5 42.31303 –88.66390 AC
Site11 Tollway_IͲ90KishwaukeeRiverͲMP18.2 13_0 42.24713 –88.94598 AC
Site12 Tollway_IͲ355@ArmyTrailͲMP29.8 4_0 41.93167 –88.03780 PCC
Site13 Tollway_IͲ88@WinfieldRoadͲMP125.2 12_0 41.80652 –88.16683 PCC
Site14 IDOT_IͲ57@ILͲ9Paxton 593005_2 40.45452 –88.11241 Composite
Site15 IDOT_IͲ39@LincolnBridge 593000_0 41.32918 –89.07495 PCC
Site16 IDOT_IͲ39@LeeCountyͲMP81 559004_1 41.67969 –89.05265 Composite
Site17 IDOT_US20@nearEastDubuque 559000_0 42.46255 –90.57733 Composite
Site18 IDOT_US30@Clinton 559001_0 41.8379 –90.17255 PCC
Site19 IDOT_ILͲ9@McNaughtonBridge 584002_2 40.5725 –89.65180 PCC
Site20 IDOT_IͲ74@BrimfieldRd 584003_2 40.83507 –89.88917 Composite
Site21 IDOT_ILͲ9@SpoonRiver 584005_2 40.56775 –90.29451 AC
Site22 IDOT_USͲ136@LacyDitch 584007_2 40.29575 –90.07898 Composite
Site23 IDOT_USͲ136@MississippiRiver 587001_2 40.38905 –91.36910 AC
Site24 IDOT_IͲ172_ILͲ104@Quincy 587000_0 39.93503 –91.32418 Composite
Site25 IDOT_IͲ72@Barry 587002_0 39.71471 –91.06354 AC
Site26 IDOT_IͲ72_USͲ67@Jacksonville 587006_0 39.68697 –90.22854 Composite
Site27 IDOT_ILͲ100@JoePageBridge 588004_0 39.16033 –90.61611 Composite
Site28 IDOT_IͲ64@ILͲ160 588005_0 38.51710 –89.68644 AC
Site29 IDOT_SRͲ3@Gorham 596001_2 37.70056 –89.47248 Composite
Site30 IDOT_SRͲ146@EastCapeGirardeau 596000_0 37.29577 –89.50499 AC
Site31 IDOT_IͲ24@PulleysMill 596002_0 37.59015 –88.97875 Composite
Site32 IDOT_IͲ70@ILͲ140 595000_0 38.92359 –89.2463 Composite
Site33 IDOT_IͲ57@ILͲ16 594003_1 39.48346 –88.32161 Composite
Site34 IDOT_IͲ72@USͲ36USͲ51 594000_0 39.9083 –88.9553 Composite
Site35 IDOT_USͲ34@Monmouth 584001_2 40.90331 –90.6626 AC
Site36 IDOT_USͲ51@Heyworth 593002_0 40.31149 –88.98928 PCC
Site37 IDOT_ILͲ116@IͲ55 593003_0 40.87355 –88.67050 Composite
Site38 IDOT_IͲ57@USͲ45 594001_1 39.90551 –88.27932 Composite

13



Figure10.Sitelocations.

4.1.1MeasuredData
Theparticipatingagenciesprovidedthemeasureddatatotheresearchteam,andtheteaminternally
processedthedataforeachsitetomeethourlyformattingneeds.Thenumberofdatapointsforeach
sitewassignificantandvariedfromevery8to12minutes.Overall,morethan7.3milliondataentries
wereavailableforthe38sites.Thesevalueswerereducedtoeveryhourtobeconsistentwiththe
predictedvaluesinthePavementMEDesignsoftware.Afterthedatawereformattedtohourly
measurements,thenumberofobservationsdecreasedto1.12million.

Somesiteshadmoredatathanothers.Generallyspeaking,theIDOTsiteshadmoremeasureddata
thantheIllinoisTollwayandMcHenryCountysitesdid.

 

14
4.1.2PredictedData
ThepredicteddatawereobtainedbygeneratinganEICMinputfileusingtheactualpavement
thicknessandmaterialinformationandexecutingtheanalysissoftwaretoruntheEICM.The
predictedpavementtemperaturedatawereextractedfromtheEICMoutputfilesandstoredfor
furtheranalysis.

4.2DATACOMPARISON
Thedatacomparisonconsistedoftwodifferentclimatedatasets.Thefirstdatasetconsistedofthe
existingclimatestationsinthePavementMEDesignsoftware.Thesecondcomparisonusedthe
NorthAmericanRegionalReanalysis(NARR)datasettocreatenewclimatefiles.Thedetailsforeach
climatedatasetarediscussedbrieflybelow.

4.2.1PavementMEDesignClimateFiles
Theoriginalhourlyclimatedatabase(HCD)wascompiledinlate1990sfromtheNationalCentersfor
EnvironmentalInformation(NCEI),whichwasformerlytheNationalClimaticDataCenter(NCDC),to
supporttheFederalHighwayAdministration(FHWA)IntegratedClimaticModel(ICM).Originallyit
containedjust3years’worthofdata.Thelengthofthedatasetwaslimitedbywhatwasavailablefor
electronicdownloadfromtheNCEI.TheICM(and,therefore,thePavementMEDesignanalyses
dependentontheICM)requirescompleteinputdataoverthedesignperiodtoproperlyexecute.The
UneditedLocalClimatologicalData(ULCD)bydesignisarawdatasetwithminimalqualitycontrol;
datafieldswithmissingdataareleftblank.TheULCDalsocontainsasmallquantityoferroneous
data.EarlyuseoftheHCDrequireduserinputtofixmissingorobviouslyincorrectclimaticdata.

In2001,additionaldatawereaddedtotheclimaticdatabase,extendingtheperiodcoveredto6
years.BecausetheoriginalMEPDGsoftwaredidnothaveautilitytoedittheclimatefiles,itwas
determinedthatallmissingdatawouldbefilledpriortodeliverytotheuser.Asoftwareutilitywas
developedtoprogrammaticallyfillinmissingdataandcorrectbaddata.AmultiͲstepprocesswas
deployedthatidentifiedmissinganderroneousdataandcreatedalogfileofchangesmadetothe
ULCDtocreatetheHCD.Thisutilityhasbeenusedtofillandcorrectallcurrentlypublishedclimate
datafiles.

Thepreviousmethodforfixingthefilesconsistedofthefollowingsteps:
1. Findmissingdata.
2. Interpolatemissingdata,iflessthan12hours.
3. Ifmorethan12hours,repeatthedayfrompreviousgoodday.
4. Ifmorethanaweekismissing,markthemonthasincomplete.
5. Checkforvaluesoutofrange(relativehumidityofover100,temperaturesover130°F,etc.).
6. Createalogfileofanyinterpolationsorcorrections.

15
Stationswithincompletemonthsareavailabletobeusedinanalysesbutrequireinterpolation
betweennearbystationstocompletethemissingmonths.TheHCDhasbeenimprovedatdifferent
timestoincreasetheamountoftimeseriesdataandreduceinconsistenciesandanomaliesinthe
dataset.ThefollowingisashortdescriptionofthedatestheHCDwasmodified,andhow:
x In2006,theHCDwasupdatedusingthemethoddescribedabovefortheUnitedStatesand
CanadatoincludeclimaticdatauptoDecember31,2005.Atthattime,theHCDwas
recompiledwiththerawdatastartingfrom1995.
x In2013,theCanadianstationsintheHCDwereupdatedusingdatafromEnvironmentCanada,
whichaddeddatafromthe1940stoDecember2012.

TheHCDaccompanyingthePavementMEDesignsoftwareincludessomemissingand/orerroneous
data.Someoftheseanomalieshavebeenidentifiedandflagged.Oncetheyhavebeenflagged,the
missingand/orerroneousdatacanbepopulatedand/orreplacedtoprovideamorerepresentative
andaccuratedataset.TheclimatefilesincludedinthePavementMEDesignsoftwarewerepopulated
usingdatastartingin1995andendingin2005(U.S.sections).Someclimatefileshavemoredatathan
others.Thecurrentdatasetconsistsof1,083climatefilesfortheUnitedStatesandCanada.
Additionally,onlyclimatestationsthathavecompletemonthlydataareavailableasapresetstation
inthesoftware.Acompletedatasetforallmonthlyclimatemeasurementsisavailablefor870climate
stationsintheUnitedStatesandCanada.

OriginalClimateDataComparison
Figure11showsthetimeͲseriesdataforSite1.ThisIllinoisTollwayportlandcementconcrete(PCC)
siteislocatedonIͲ355atDesPlainesRivermilepost(MP)9.5.Asshowninthefigure,themeasured
temperaturedatarangefrom2012to2014.Forthisparticularlocation,dataaremissingforafew
monthsin2013.Theremayhavebeenerrorsinthesensors,orperhapsdatawerenotcollected
duringthattimeframe.Thefigurealsoshowsthepredictedpavementtemperatureforthedatesand
timeswhenmeasuredobservationswereavailable,aswellastheresidualerrorbetweenthe
measuredandpredictedpavementtemperatures.Basedontheresidualplot,theerrorislessfor
coldertemperaturesthanforwarmertemperatures.Thelowererrorsatcoldertemperaturesare
beneficialforthisprojectbecausedeterminingwhenfreezingwilloccurinthepavementisimportant.

Suchfigureswerecreatedforallsitesincludedinthestudy.Therestofthesefiguresusingthe
originalclimatedataarepresentedinAppendixC.

16

Figure11.Site1timeͲseriespavementtemperaturedata.

Additionally,foreachsite,aoneͲtoͲonecomparisonplotshowshowwellthepredictedpavement
temperaturescomparetothemeasuredpavementtemperatures.Figure12showsanexampleofthe
oneͲtoͲoneplotforSite1.Thefigureshowsaverygoodcorrelationbetweenthepredictedand
measuredtemperatureswithanR2of0.97for18,606datapoints.AppendixCcontainsthemeasured
versuspredictedpavementtemperaturefiguresfortheremainingsites.

17

Figure12.Site1measuredversuspredictedpavementtemperaturedata.

Overall,themajorityofthesitesshowedverygoodR2results.However,somedidnot.Particularly
notableareSite16(IDOTIͲ39atLeeCounty)andSite18(IDOTUSͲ30atClinton).Figure13showsthe
measuredversuspredictedtemperatureoneͲtoͲoneplotsforthesesites.TheR2valuesforthosetwo
siteswere0.28and0.096,respectively.Site16exhibitedwarmerpredictedtemperaturescompared
withthecorrespondingmeasuredtemperatures(overͲprediction).Alternatively,Site18showedan
underͲpredictionofthemeasuredtemperature—themeasuredtemperatureswerewarmerthanthe
predictedtemperatures.

TodeterminethecausefortheoverͲorunderͲprediction,thetimeͲseriesdatawereobserved.Figure
14illustratesthetimeͲseriescurvesforSite16andSite18.Thefiguresshowabnormaltemperature
measurementsforbothsites.Theredoesnotseemtobeaconsistenthourlytrendintheabnormal
data.Thecausesforthesemeasuredtemperatureanomaliesareunknown.Therefore,thesevalues
wereexcludedfromthefinalanalysiscomparisons.

Similaranalysiswasperformedontheothersitesforconsistency.Basedonthefindingsfromusing
theoriginalclimatedataandthemeasureddata,datacleaningwasperformedtoensurereasonable
andaccuratedatawereusedintheanalysis.Additionally,adifferentdatasetforhourlyclimatedata
wasusedtoimprovetheaccuracyofthepredictedpavementtemperatures.Theclimatedatadetails
andthecomparisonbetweenmeasuredandpredictedpavementtemperatureresultsarepresented
anddiscussedinthenextsection.

18
 
(a)Site16measuredversuspredicted (b)Site18measuredversuspredicted
pavementtemperature pavementtemperature

Figure13.Measuredversuspredictedpavementtemperaturecomparisons.

 
(a)Site16timeͲseriespavement (b)Site18timeͲseriespavement
temperaturedata temperaturedata

Figure14.Site16andSite18timeͲseriespavementtemperaturedata.

19
4.2.2NorthAmericanRegionalReanalysisDatabaseGeneratedClimateFiles
TheNARRdatabaseisusedprimarilyforatmosphericresearchrequiringhistoricalatmospheric
conditionsandtostudythevariabilityofclimateconditions.Thedatabasewasdevelopedbythe
NationalCentersforEnvironmentalPredictionstomodelorassimilateobservationaldatatoproduce
alongͲtermoverviewofweatheroverNorthAmerica.ThemodelisinitializedbyusingrealͲworld
temperature,wind,precipitation,andmoistureconditionsfromsurfaceobservations.

ManydifferentsourceswereusedtodeveloptheNARRdatabase.Someofthesesourceswerealso
usedinaglobalreanalysis,alongwithavarietyofadditionalsources.Thesesourcesincludethe
following:

x NationalCentersforEnvironmentalPrediction
x NationalCenterforAtmosphericResearch
x GlobalReanalysis
x ClimatePredictionCenter
x NationalEnvironmentalSatellite,Data,andInformationService
x EnvironmentalModelingCenter
x CenterforOceanͲLandͲAtmosphereStudies
x GreatLakesEnvironmentalResearchLaboratory
x LawrenceLivermoreNationalLaboratory

Additionaldetailsaboutwhichdatasetswereusedforeachsourcecanbefoundinthearticletitled
“NorthAmericanRegionalReanalysis”byMesingeretal.(http://journals.ametsoc.org/doi/pdf/
10.1175/BAMSͲ87Ͳ3Ͳ343).

TheNARRdataareavailablefora32×32km(20×20mile)gridacrossNorthAmerica.Thedataare
availablein3Ͳhour,daily,andmonthlyvaluesfrom1979topresent.Thelongertimeframeof
availableclimatedataisasignificantimprovementovertherangescurrentlyavailableinthe
PavementMEDesignsoftware.The37yearsofcontinuousdataaresignificantbecausetheclimateof
alocationisdefinedbasedontheweatherdatafromtheprevious30years.Therefore,theNARR
dataprovideamoreaccuraterepresentationofclimateforanylocationinNorthAmerica.
TheNARRdatasethasgonethroughseveralqualitycontrolchecksanddoesnotneedfurtherdata
smoothingorqualityassuranceandcontrol.Thisisalargeadvantage,giventheamountofclimate
dataneededforthePavementMEDesignclimatefiles.Additionally,aclimatefilecanbegenerated
foranylatitudeorlongitudeacrossNorthAmericabecausetheNARRdatasetisbasedonagrid
system,whicheliminatestheuseofaphysicalclimatestationthatmaynotbeclosetotheactual
pavementlocation.ItshouldbenotedthattheassimilationprocessintheNARRusestheavailable
observedvaluesinaparticular32×32kmgrid.Thenumberofavailableobservedvaluesandthe
topographycanaffecttheassimilationresultsandimpactthequalityofthemodelforsomelocations.

20
NARRͲGeneratedClimateDataComparison
TheNARRͲgeneratedclimatedatacompilationusesthenearestgridpointtothelocationofeach
projectsite.Thesesitesmaybedifferentfromthenearestclimatestation.Therefore,therearesome
differencesbetweentheoriginalPavementMEDesignclimatefileandtheNARRͲgeneratedclimate
file.Overall,theNARRdatabaseproducesabetterestimateofaregion’sclimatebecauseit
assimilatesclimatedatausingmanydifferentsourcesoveralongertimeperiod.

TheoneͲtoͲoneplotsandtimeͲseriesplotswerealsocreatedfortheNARRͲgeneratedclimatedata.
Theresultsweresimilartotheresultsfortheoriginalclimatedata.Figure15showsthemeasured
versuspredictedpavementtemperatureforSite1.Theresultsshowaverygoodcorrelation(R2=
0.95)betweenthemeasuredandpredictedpavementtemperature.Thescatteraroundthelineof
equalitywaslessatcoldertemperaturesthanatwarmertemperatures.Figure16showsthetimeͲ
seriestemperaturecurvesandtheresidualplotforSite1.Thefigureindicatesthatthereisasimilar
trendbetweenthemeasuredandpredictedtemperatures.Additionally,theresidualplotshowsthat
thereisarangebetweenͲ25and20°Fforthedifferencebetweenthemeasuredandpredicted
pavementtemperatures.Thelargerdifferencesoccurredduringthesummermonths.

AppendixBincludesthemeasuredversuspredictedandtimeͲseriesfiguresfortheothersites.


Figure15.Site1measuredversuspredictedpavement
temperatureusingNARRͲgeneratedclimatedata.

iv

21

Figure16.Site1timeͲseriespavementtemperaturedatausingNARRclimatefiles.

Tostudythedifferencesbetweenthemeasuredandpredictedpavementtemperatures,ahistogram
wascreatedtovisuallyseethespreadoftheresidualerror.Figure17showsthehistogramforSite1.
Theresultsshowthatthemeandifferencebetweenthemeasuredandpredictedpavement
temperaturewas1.3°F,withastandarddeviationof6°F.Thehistogrammakesiteasiertoseehow
manydatapointslieclosertozero.Ifadatapointisclosetozero,itmeansthereisalmostno
differencebetweenthemeasuredandpredictedpavementtemperature.

AppendixDcontainshistogramsfortheothersites.

22

Figure17.Temperaturedifferencehistogram.


4.3ANALYSISSUMMARY
BoththeoriginalandNARRͲgeneratedclimatefilesshowedadequatecomparisonsbetweenthe
measuredandpredictedpavementtemperatures.Manyofthepavementsitesweremissing
measureddata,especiallytheIllinoisTollwaysites.Themissingdatadonotaffecttheanalysis
becausethepredicteddatawerematchedonlywhenmeasuredtemperatureswereavailable.

Figure18showsthecomparisonofR2fortheoriginalandNARRdatabaseclimatefiles.Theresults
showthatthereisawiderspreadofR2forthepredictedpavementtemperatureswhenusingthe
originalclimatefiles.Overall,themajorityofthesitesshowedaverygoodR2.AlloftheR2values
weregreaterthan0.75whenusingtheNARRͲgeneratedclimatefiles.

Eventhoughthereseemstobeagoodcorrelationbetweenthepredictedandmeasuredpavement
temperatures,theresearchteamneededtodeterminewhethertheerrorcouldbereducedfurther.


Table3summarizestheaveragesofthemeantemperaturedifferencesforallsites,aswellasthe
averagesforthethreedifferentpavementtypes.ThehotͲmixasphalt(HMA)pavementsitesshowed
thelowestmeandifferenceforthetemperatures±5and±10°Ffromfreezing.Thedetailsonthe
calibrationofthepavementtemperaturepredictionsarediscussedinChapter5.OnlytheNARRͲ
generatedclimatedatawereusedinthemodelcalibrationreview.

23
 

 
(a)Originalclimatefiles (b)NARRdatabaseclimatefiles

Figure18.SummaryofR2valuesforallprojectsites.

Table3.AverageMeanTemperatureDifferences
Dataset All ±10°F ±5°F
AverageAmongAllSites 2.51 1.13 1.19
PCCOnly 2.18 1.35 1.48
HMAOnly 2.80 0.91 0.87
CompositeOnly 2.59 1.09 1.15

24
CHAPTER5:CALIBRATIONOFMODELINPUTS

5.1INTRODUCTION
Chapter4presentedthecomparisonbetweenthemeasuredandpredictedpavementtemperatures
forthe38sitesincludedinthisstudy.BasedontheresultspresentedinChapter4,theresearchers
establishedaprocesstoattempttoreducetheerrorbetweenthemeasuredandpredictedpavement
temperatures.

5.2CALIBRATIONPROCESS
Theprocessforcalibratingthepredictionmodelsconsistedofchangingtheinputvariablesinthe
PavementMEDesignsoftwarethataffectthepavementtemperaturepredictions—absorptivity,heat
capacity,andthermalconductivity.Theabsorptivityaffectshowmuchofthesolarradiationis
absorbedintothepavementmaterial.Theheatcapacityisequaltotheratiooftheheatadded
to/removedfromanobjecttotheresultingtemperaturechange.Thethermalconductivityisusedto
determinethepropertiesofamaterialtoconductheat.Thesethreevaluescanbechangedtostudy
theeffectsonpredictedpavementtemperature.

5.3CALIBRATIONDESIGNMATRIX
Eachofthethreeinputvariableswasadjustedtothreedifferentlevels.Thelevelswereselected
basedonthedefaultvaluesinthePavementMEDesignsoftwareandarealisticupperandlower
valueforeachvariable.Allcombinationswereanalyzed,andtheerrorwasdeterminedforeach
combination.

ThethermalconductivityandheatcapacityforPCCandHMApavementsdonotvarysignificantly
becausetheydependonthematerial.Therefore,avalueof±0.5wasusedforthermalconductivity
andavalueof±0.05wasusedforheatcapacity.Theabsorptivityvalueswereselectedbasedonthe
defaultsinthePavementMEDesignsoftwarewithadifferenceof±0.1.

Table4summarizesthefulldesignmatrixforPCCandHMApavements.All27combinationswere
analyzedforeachofthe38selectedsites.Theanalysisresultsweresummarizedtodeterminearank
orderofthelowesterrorforeachsite.Additionally,theresultswereshownforthreedifferent
temperaturecriteria.Thethreedifferentcriteriaconsistedofthefollowing:
x Allpavementtemperatures
x Pavementtemperatures±10°Ffromfreezing(32°F)
x Pavementtemperatures±5°Ffromfreezing

Theresultsforthevariouscombinationsaresummarizedanddiscussedafterthetable.
 

25
Table4.DesignMatrixforPCCandHMAPavements
Pavement Thermal Heat Absorptivity
Surface Conductivity Capacity 0.75 0.85 0.95
0.23 1 2 3
0.75 0.28 4 5 6
0.33 7 8 9
0.23 10 11 12
PCC 1.25 0.28 13 14 15
0.33 16 17 18
0.23 19 20 21
1.75 0.28 22 23 24
0.33 25 26 27
0.18 1 2 3
0.17 0.23 4 5 6
0.28 7 8 9
0.23 10 11 12
HMA 0.67 0.28 13 14 15
0.33 16 17 18
0.23 19 20 21
1.17 0.28 22 23 24
0.33 25 26 27


5.3CALIBRATIONRESULTS
Theresultsofthecalibrationrunsweresummarizedbycalculatingtherootmeansquarederror
(RMSE)foreachcombination(thermalconductivity,heatcapacity,andabsorptivity)andsite.The
resultswerealsosummarizedbypavementtype.Intotal,1,026EICMrunswereanalyzed.Sites6,7,
and11wereexcludedfromtheanalysisbecausetheydidnothavemeasuredtemperaturesduring
coldmonths.

PavementMEDesignDefaultValues
TheRMSEwascalculatedforallthesitesanddatasubsets.TheresultsareshowninFigure19.The
resultsshowalargedifferenceinRMSEwhencomparingthethreedifferenttemperaturecategories.
Thesedifferencesindicatethatthemodelismoreaccurateatcoldertemperaturesthanatwarmer
temperatures,asdiscussedpreviously.Additionally,thereisnodistinguishabledifferenceinRMSE
whencomparingthesitesfromIDOT,McHenryCounty,andtheIllinoisTollway,orwhencomparing
PCCandHMApavements.

26

Figure19.RootmeansquarederrorcomparisonforthePavementMEDesigndefaults.

ConcretePavementSections
Table5summarizesthePCCcalibrationresults.ThetoptenresultsbasedontheRMSEforthe±5°F
categoryareshown.Basedontheseresults,theaveragethermalconductivityvaluesamongallsites
was1.5,2.9,and0.89.TheRMSEdidnotvarysignificantlyamongthetoptenforanyofthethree
temperatures.Therefore,itisrecommendedthatthefollowingvaluesbeusedforPCCpavements:
x Thermalconductivity:1.50
x Heatcapacity:0.30
x Absorptivity:0.85

ThePavementMEDesigndefaultabsorptivityvalueisrecommendedbecauseofitslowimpactat
coldertemperatures.

27
Table5.PCCPavementCalibrationResults—TopTen
Rank Site1 Site2 Site3 Site4 Site12
1 1.75 0.33 0.95 1.75 0.33 0.95 1.75 0.33 0.95 1.75 0.33 0.95 1.75 0.33 0.95
2 1.75 0.33 0.85 1.75 0.28 0.95 1.75 0.28 0.95 1.75 0.28 0.95 1.75 0.28 0.95
3 1.75 0.28 0.95 1.25 0.33 0.95 1.25 0.33 0.95 1.75 0.23 0.95 1.75 0.23 0.75
4 1.75 0.28 0.85 1.75 0.33 0.85 1.75 0.33 0.85 1.25 0.33 0.95 1.75 0.28 0.85
PCC 5 1.25 0.33 0.95 1.75 0.23 0.95 1.75 0.28 0.85 1.25 0.28 0.95 1.75 0.33 0.85
6 1.25 0.33 0.85 1.25 0.28 0.95 1.25 0.28 0.95 1.75 0.33 0.85 1.75 0.23 0.85
7 1.25 0.28 0.85 1.25 0.33 0.85 1.25 0.33 0.85 1.75 0.23 0.85 1.25 0.33 0.95
8 1.25 0.28 0.95 1.75 0.28 0.85 1.25 0.23 0.95 1.75 0.28 0.85 1.25 0.28 0.95
9 1.75 0.33 0.75 1.75 0.23 0.85 0.75 0.33 0.95 1.25 0.23 0.95 1.25 0.28 0.85
10 1.25 0.33 0.75 1.25 0.28 0.85 1.25 0.28 0.85 0.75 0.33 0.95 1.75 0.28 0.75
Rank Site13 Site15 Site18 Site19 Standard
1 1.75 0.33 0.95 1.75 0.33 0.95 1.75 0.33 0.95 1.25 0.23 0.75  Average Deviation
2 1.75 0.33 0.85 1.75 0.28 0.95 1.25 0.33 0.95 0.75 0.23 0.75
3 1.75 0.28 0.85 1.25 0.33 0.95 1.75 0.28 0.95 0.75 0.28 0.75 Thermal
4 1.25 0.33 0.85 1.25 0.28 0.95 1.25 0.33 0.85 1.25 0.28 0.75 Conductivity 1.50 0.19
PCC 5 1.25 0.33 0.95 1.25 0.23 0.95 1.25 0.28 0.85 1.75 0.28 0.75
6 1.75 0.28 0.95 1.75 0.28 0.85 1.25 0.28 0.95 1.25 0.23 0.85 Heat
7 1.25 0.28 0.85 1.75 0.33 0.85 1.75 0.28 0.85 0.75 0.33 0.75 Capacity 0.30 0.02
8 1.75 0.23 0.75 0.75 0.33 0.95 1.75 0.33 0.85 0.75 0.23 0.85
9 1.25 0.28 0.95 0.75 0.28 0.95 1.25 0.23 0.85 0.75 0.28 0.85
10 1.75 0.28 75 0.75 0.23 0.95 0.75 0.33 0.95 1.25 0.28 0.85 Absorptivity 0.90 0.05

28
FullͲDepthHMAPavementSections
Table7summarizesthefullͲdepthHMApavementcalibrationresults.Thetoptencombinationsof
thermalconductivity,heatcapacity,andabsorptivityareshown.Therankingisalsobasedonthe
RMSEforthe±5°Fdatasubset.Theresultsshowthattheaveragesforthermalconductivity,heat
capacity,andabsorptivitywere1.08,0.25,and0.83,respectively.Basedonengineeringjudgement
andexperience,therecommendedvaluesforthesethreeinputsareasfollows:
x Thermalconductivity:1.00
x Heatcapacity:0.25
x Absorptivity:0.85
ThePavementMEDesigndefaultabsorptivityvalueisrecommendedbecauseofitslowimpactat
coldertemperatures,similartoPCCpavements.

CompositePavementSections
Table8summarizesthecompositepavementcalibrationresults.Thetop10combinationsare
presentedsimilartotheACandPCCdatasets.Theaveragethermalconductivity,heatcapacity,and
absorptivityforthecompositepavementsectionswere0.96,0.24,and0.86,respectively.The
recommendedvaluesforcompositepavementsarethesameasforACpavementsbecausethe
surfacelayerisalsoAC.

5.4SUMMARY
Thecalibrationprocessinvolveschangingthethermalconductivity,heatcapacity,andabsorptivityin
theEICM.Acalibrationmatrixwasdevelopedtodeterminetheeffectsofthesethreeinputvariables
onthepavementtemperaturepredictions.TheEICMwasexecutedforalltheselectedsites.Basedon
theresults,arecommendedvalueforeachparameterwasestablishedbasedonthepavementtype:
x PCCpavements
o Thermalconductivity:1.50
o Heatcapacity:0.30
o Absorptivity:0.85
x ACpavements(includingcomposite)
o Thermalconductivity:1.00
o Heatcapacity:0.25
o Absorptivity:0.85

ThePavementMEDesigndefaultabsorptivityvaluewasselectedforbothsurfacetypesbecauseofits
minimalimpactatcoldtemperatures.Overall,thePavementMEDesigndefaultvaluesprovideda
verygoodinitialpredictionofpavementtemperaturesfortheselectedpavementsites.Additionally,
allsiteswereanalyzedusingtheupdatedvalues.TheresultsaresummarizedinTable6andshowson

29
averageminimalerrorbetweenthemeasuredandpredictedpavementtemperatures.Theaverage
errorsarealsolowercomparedwiththedefaultvaluespresentedinTable3.

Table6.ErrorResultsAfterCalibration
Error type Agency AC Composite PCC
IDOT 3.9 3.9 4.8
Root Mean Square Error McHenry Co. 3.8 — —
(°F) Illinois Tollway — 4.8 4.1
Average 3.9 4.0 4.4
IDOT 0.7 0.9 1.0
Mean Error McHenry Co. 0.8
(°F) Illinois Tollway 2.4 1.7
Average 0.8 1.1 1.4

30
Table7.HMAPavementCalibrationResults—TopTen
 Site9 Site10 Site21 Site23 Site25 Site30 Site35
1 1.17 0.28 0.85 1.17 0.28 0.75 1.17 0.28 0.75 1.17 0.28 0.75 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.28 0.85
2 1.17 0.28 0.95 1.17 0.23 0.75 1.17 0.23 0.75 1.17 0.23 0.75 1.17 0.23 0.95 1.17 0.28 0.85 1.17 0.18 0.85
3 1.17 0.28 0.75 1.17 0.28 0.85 1.17 0.28 0.85 0.67 0.28 0.75 1.17 0.28 0.85 1.17 0.23 0.95 1.17 0.23 0.85
4 1.17 0.23 0.85 0.67 0.28 0.75 1.17 0.23 0.85 1.17 0.18 0.75 1.17 0.23 0.85 1.17 0.23 0.85 1.17 0.28 0.95
AC 5 1.17 0.23 0.75 1.17 0.18 0.75 1.17 0.18 0.75 0.67 0.23 0.75 1.17 0.18 0.95 0.67 0.28 0.95 1.17 0.18 0.75
6 1.17 0.23 0.95 1.17 0.23 0.85 1.17 0.28 0.95 0.67 0.18 0.75 0.67 0.28 0.95 0.67 0.28 0.85 1.17 0.23 0.95
7 1.17 0.18 0.75 0.67 0.23 0.75 0.67 0.28 0.75 1.17 0.28 0.85 1.17 0.18 0.85 1.17 0.28 0.75 1.17 0.23 0.75
8 0.67 0.28 0.85 1.17 0.28 0.95 0.67 0.23 0.75 1.17 0.23 0.85 0.67 0.23 0.95 1.17 0.18 0.85 0.67 0.28 0.85
9 0.67 0.28 0.75 0.67 0.28 0.85 0.67 0.28 0.85 0.67 0.28 0.85 1.17 0.28 0.75 1.17 0.23 0.75 1.17 0.28 0.75
10 1.17 0.18 0.85 0.67 0.18 0.75 1.17 0.18 0.85 1.17 0.18 0.85 1.17 0.23 0.75 1.17 0.18 0.95 0.67 0.23 0.85
 Average StandardDeviation
ThermalConductivity 1.08 0.04
HeatCapacity 0.25 0.01
Absorptivity 0.83 0.05 

31
Table8.CompositePavementCalibrationResults—TopTen
 Site5 Site8 Site14 Site16 Site17 Site20 Site22 Site24 Site26
1 1.17 0.28 0.95 1.17 0.18 0.85 0.67 0.18 0.75 0.17 0.23 0.75 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.28 0.85 1.17 0.23 0.95
2 1.17 0.23 0.95 1.17 0.23 0.85 0.67 0.23 0.75 0.17 0.28 0.75 1.17 0.23 0.95 1.17 0.23 0.95 1.17 0.23 0.95 1.17 0.23 0.85 1.17 0.18 0.95
3 0.67 0.28 0.95 0.67 0.28 0.85 1.17 0.18 0.75 0.17 0.18 0.75 0.67 0.28 0.95 1.17 0.18 0.95 1.17 0.28 0.85 1.17 0.18 0.75 1.17 0.28 0.95
4 1.17 0.18 0.95 1.17 0.28 0.85 0.67 0.28 0.75 0.17 0.28 0.85 1.17 0.18 0.95 0.67 0.28 0.95 1.17 0.23 0.85 1.17 0.23 0.75 0.67 0.28 0.95
5 1.17 0.28 0.85 0.67 0.23 0.85 1.17 0.23 0.75 0.17 0.23 0.85 0.67 0.23 0.95 0.67 0.23 0.95 0.67 0.28 0.95 0.67 0.28 0.75 0.67 0.23 0.95
6 0.67 0.23 0.95 1.17 0.18 0.75 1.17 0.28 0.75 0.67 0.18 0.75 1.17 0.28 0.85 1.17 0.18 0.85 0.67 0.28 0.85 1.17 0.28 0.75 0.67 0.18 0.95
7 1.17 0.23 0.85 0.67 0.23 0.75 0.17 0.28 0.75 0.67 0.18 0.85 1.17 0.23 0.85 1.17 0.23 0.85 1.17 0.18 0.95 0.67 0.23 0.75 1.17 0.18 0.85
8 0.67 0.28 0.85 0.67 0.18 0.75 0.17 0.23 0.75 0.17 0.18 0.85 0.67 0.28 0.85 1.17 0.28 0.85 1.17 0.18 0.85 0.67 0.28 0.85 1.17 0.23 0.85
9 0.67 0.18 0.95 0.67 0.28 0.75 1.17 0.18 0.85 0.67 0.23 0.85 1.17 0.18 0.85 0.67 0.28 0.85 1.17 0.28 0.75 1.17 0.18 0.85 0.67 0.23 0.85
10 1.17 0.18 0.85 0.67 0.18 0.85 1.17 0.23 0.85 0.67 0.23 0.75 0.67 0.23 0.85 0.67 0.23 0.85 0.67 0.23 0.85 0.67 0.18 0.75 0.67 0.18 0.85
 Site27 Site29 Site31 Site32 Site33 Site34 Site37 Site38

Composite
1 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.28 0.95 1.17 0.18 0.75 1.17 0.28 0.95 0.67 0.18 0.75 1.17 0.28 0.75
2 1.17 0.28 0.85 1.17 0.23 0.95 1.17 0.28 0.85 1.17 0.28 0.85 1.17 0.23 0.75 1.17 0.23 0.95 0.17 0.28 0.75 1.17 0.23 0.75
3 1.17 0.23 0.95 1.17 0.28 0.85 1.17 0.23 0.95 1.17 0.23 0.95 0.67 0.28 0.75 1.17 0.23 0.85 0.67 0.18 0.85 1.17 0.18 0.75
4 1.17 0.23 0.85 0.67 0.28 0.95 1.17 0.23 0.85 1.17 0.23 0.85 1.17 0.28 0.85 1.17 0.18 0.85 0.17 0.23 0.75 0.67 0.28 0.75
5 1.17 0.28 0.75 1.17 0.18 0.95 0.67 0.28 0.95 0.67 0.28 0.95 0.67 0.23 0.75 0.67 0.28 0.85 0.67 0.23 0.75 1.17 0.28 0.85
6 1.17 0.23 0.75 1.17 0.23 0.85 1.17 0.18 0.95 0.67 0.28 0.85 1.17 0.23 0.85 1.17 0.28 0.85 0.67 0.23 0.85 0.67 0.23 0.75
7 0.67 0.28 0.85 0.67 0.23 0.95 0.67 0.28 0.85 1.17 0.18 0.95 1.17 0.28 0.75 0.67 0.23 0.85 1.17 0.18 0.85 1.17 0.23 0.85
8 1.17 0.18 0.85 0.67 0.28 0.85 1.17 0.18 0.85 1.17 0.28 0.75 0.67 0.18 0.75 1.17 0.18 0.95 1.17 0.18 0.75 0.67 0.18 0.75
9 0.67 0.28 0.95 1.17 0.18 0.85 1.17 0.28 0.75 1.17 0.18 0.85 0.67 0.28 0.85 0.67 0.28 0.95 0.67 0.28 0.85 0.67 0.28 0.85
10 1.17 0.18 0.95 1.17 0.28 0.75 0.67 0.23 0.95 0.67 0.23 0.95 1.17 0.18 0.85 0.67 0.18 0.85 0.17 0.18 0.75 1.17 0.18 0.85 
 Average StandardDeviation
ThermalConductivity 0.96 0.22
HeatCapacity 0.24 0.01
Absorptivity 0.86 0.07 


32
CHAPTER6:RECOMMENDATIONSANDCONCLUSIONS
TheobjectiveofthisprojectwastoevaluatetheuseoftheEICMfordeterminingpavementsurface
temperatureforwintermaintenanceoperations.Modelingthepavementtemperaturescould
providevirtualRWISdataatacostthatisconsiderablylessthanthecostofphysicalsensorsand
systems.Byhavingawinteroperationssystemthatincludesbothphysicalsensorsandvirtual
stations,anagencycouldreducecostswhilemaintainingahighlevelofdatadensityandquality.

Throughadatacomparisonandcalibrationprocess,theresearchteamattemptedtodeterminethe
pavementmaterialsandthermalinputsthatshouldbeusedinestablishingavirtualRWISstationso
thatthetotalerrorisminimized.Table9liststherecommendedthermalpropertyinputs.

Table9.RecommendedThermalInputsfortheEICMinaVirtualRWIS
 ConcreteMaterials AsphaltMaterials
ThermalConductivity 1.50 1.00
HeatCapacity 0.30 0.25
ShortWaveAbsorptivity 0.85 0.85


UsingtherecommendedthermalinputsandtheactualinͲplacepavementstructure(materialtypes
andlayerthicknesses)asinputstotheEICMyieldsreasonablemodelaccuracywithameanerrorthat
isgenerallylessthan2°F.Understandingthatthemodelerrorisafunctionofthequalityofthe
weatherdata,thequalityofthesensordata,andthevalidityofthemodel,anerroroflessthan2°Fis
consideredreasonableandisappropriateforuseinavirtualRWIS.BecausethevirtualRWISwilluse
forecastatmosphericweatherdata,itisanticipatedthattheerrorinthemodelwillbemoreofa
functionofanerrorintheforecastdatathananerrorintheEICMmodelfordeterminingpavement
surfacetemperatures.

33

You might also like