The Best Practice Model in New YorkThe Best Practice Model in New York

TMIP NCDOT PanelRaleigh, NC

Kuo-Ann ChiaoDirector of Technical Services

New York Metropolitan Transportation Council

Study AreaStudy Area

•• 28 counties 28 counties•• 3,586 Transportation Analysis 3,586 Transportation AnalysisZonesZones•• 4 time periods 4 time periods•• 8 trip purposes 8 trip purposes•• 10 modes 10 modes

Location DistributionLocation Distribution 1997 Household Travel Survey1997 Household Travel Survey

Highway NetworkHighway Network

•• Very large networkVery large network–– 52,794 links52,794 links–– 4,950 High-level facilities4,950 High-level facilities–– 26,385 Arterials26,385 Arterials–– 10,694 Centroid and external connectors10,694 Centroid and external connectors–– 10,765 Other10,765 Other

•• Unidirectional / Unidirectional / dualizeddualized coding coding•• Conflated the network geographyConflated the network geography•• GIS Network Developed in GIS Network Developed in TransCADTransCAD Software Software•• SOV, HOV2, HOV3+, taxi, truck, other commercialSOV, HOV2, HOV3+, taxi, truck, other commercial•• Attributes include capacities, initial speeds, lanes,Attributes include capacities, initial speeds, lanes,

parking availability, truck restriction, signal spacing,parking availability, truck restriction, signal spacing,Roadway Names, and functional classRoadway Names, and functional class

GIS Street Base – TIGER (or LION)GIS Street Base – TIGER (or LION)

Highway NetworkHighway Network

Uni-directional coding& / Ramps

Zones System–Census Tract BasedZones System–Census Tract Based

BPM zone boundaries

Link AttributesLink Attributes

Transit NetworkTransit Network

•• Extremely detailed transit coding based onExtremely detailed transit coding based oninformation from MTA and NJ Transitinformation from MTA and NJ Transit

•• Developed in Developed in TransCADTransCAD 4.0 4.0•• Each route variation coded as a distinctEach route variation coded as a distinct

route:route:–– 100 NYC subway routes100 NYC subway routes–– 900 Commuter rail routes900 Commuter rail routes–– 2,300 bus routes2,300 bus routes–– 50 ferry routes50 ferry routes–– Includes sidewalk network in ManhattanIncludes sidewalk network in Manhattan–– Walk access/egress linksWalk access/egress links–– Park - and - RidePark - and - Ride

Transit NetworkTransit Network

Transit NetworkTransit NetworkLower Manhattan Transit RoutesLower Manhattan Transit Routes

Highlights of NYBPMHighlights of NYBPM

•• Micro-Simulation choice modelsMicro-Simulation choice models•• Population synthesis and intra-householdPopulation synthesis and intra-household

travel interactionstravel interactions•• Journey-based travel units modeledJourney-based travel units modeled•• Non-motorized (pre-mode choice)Non-motorized (pre-mode choice)•• Mode-Destination Choice (nested logit)Mode-Destination Choice (nested logit)•• Stop frequency and location sub-modelStop frequency and location sub-model•• Full multi-modal analysis / assignmentFull multi-modal analysis / assignment

Journey Generation

Mode & Destination

Time of Day

Assignment

Mic

ro-S

imul

atio

n

Stop Freq & Location

General Modeling StructureGeneral Modeling Structure

Journey Generation

Mode & Destination

Time of Day

Assignment

SyntheticPopulation

AutoOwnership

JourneyFrequency

Socio-EconomicTargets

Accessibility

Seed PUMS

LUM

Stop Freq & Location

Journey GenerationJourney Generation

Journey Frequency

Pre-Mode Choice

Non-MotorizedDestination

MotorizedDestination

Motorized Mode

Attraction(Activity Size)

Stop Frequency

Stop Location Stop Density

MDC

Stops

PAP /TOD Journey TOD

Impedance (Skims)

1

2

3

Stop TOD

Modeling StructureModeling Structure

Workers Non-Workers Children

Man

dato

ryM

aint

enan

ceD

iscr

etio

nary

Work

School

At Work

M

M

M

SchoolSchool University University

Intra-Household Interaction

Indi

vidu

al T

ime-

Spac

e C

onst

rain

t

D

D

D

Journey Frequency ModelJourney Frequency Model

Mode Choice to WorkMode Choice to WorkNested StructureNested Structure

DriveAlone

Transit &Shared Ride

Taxi

DriveAlone

DriveAlone

Taxi2 Pass 3 Pass 4 Pass CommuterRail

Transit

2 Pass 3 Pass 4 Pass Walk toCommuter

Rail

Drive toCommuter

Rail

Walk toTransit

Drive toTransit

Taxi

Mode Choice to Work:Mode Choice to Work:Mode AvailabilityMode Availability

Drive aloneDrive aloneNo cars in householdNo cars in household

Walk & drive to transit & commuterWalk & drive to transit & commuterrailrail

Zero INV time in skimZero INV time in skim

Walk & drive to transit & commuterWalk & drive to transit & commuterrailrail

No walk access to transit atNo walk access to transit atdestinationdestination

Walk to transit & commuter railWalk to transit & commuter railNo walk access to transit atNo walk access to transit atoriginorigin

Modes UnavailableModes UnavailableCharacteristicCharacteristic

Destination-Choice Model:Destination-Choice Model:Utility ComponentsUtility Components•• Attraction-size variableAttraction-size variable•• Mode-choice log-sumMode-choice log-sum•• 3 River-crossing dummies3 River-crossing dummies•• Intra-county dummyIntra-county dummy•• Distance-based termDistance-based term•• 4 To-Manhattan dummies4 To-Manhattan dummies•• County-to-county k-factorsCounty-to-county k-factors

Dis

aggr

ega t

eD

isa g

gre g

a te

Ca l

ibr a

t ion

Ca l

ibr a

t ion

Ag g

reg a

t eA

g gre

g at e

Ad j

u stm

e nt

Ad j

u stm

e nt

Route-Deviation ConceptRoute-Deviation Concept

Origini

Destinjdij

Stopkdik dkj

Combined impedance: dik + dkj

Absolute route deviation: dik + dkj - dij

Relative route deviation: (dik + dkj – dij)/ dij

Stop Frequency by PurposeStop Frequency by Purpose

0%10%20%30%40%50%60%70%80%90%

Work-low

Work-med

Work-high

School Univ Atwork

Maint Discr

No stops Outbound Return Both

Stop Frequency by ModeStop Frequency by Mode

0%10%20%30%40%50%60%70%80%

Drivealone

Sharedride

Transit Commutrail

Taxi Schoolbus

Other

No stops Outbound Return Both

Stop Distribution by DurationStop Distribution by Duration

0%10%20%30%40%50%60%70%80%

< 1 h 1-2 h 2-3 h 3-4 h 4-5 h > 5 h

Activity duration, hours

Stop-Frequency Choice Model

Choice Alternatives Structural Dimensions Utility Components0

- No

stop

s

1 - O

utbo

und

2 - R

etur

n

3 - B

oth

WorkSchool

UniversityMaintenanc

eDiscretionary

. . . . . . . . .

Journey Purpose

Person Type

WorkSchoolUniversit

yMaintenance

DiscretionaryAt Work

Worker

Non-Worker

Child

Income

Car Sufficiency

Mode

SOV, Taxi HOV

Transit

Journey Distance

Stop-Location(Density) Log-Sum

HouseholdComposition

Other Journeys

At Work

Stop-Location Choice Model

Choice Alternatives Structural Dimensions Utility Components

5 miles

5 miles

20%

Journey Purpose

Person Type

WorkSchool

University

Maintenance

Discretionary

Worker

Non-Worker

Child

Mode

SOV, Taxi HOV

Transit

Journey Leg

Outbound Return

Stop Density (Size)

CombinedImpedance

Route Deviation

Stop Activity

WorkSchool

University

Maintenance

Discretionary

Stages of CalibrationStages of Calibrationand Validation Sourcesand Validation Sources

Disaggregate Calibration

by Purpose

Aggregate Calibration

Of Destination Choice

Aggregate Calibration

Of Mode Shares

Highway and Transit

Assignment

Household Survey

Household Survey;PUMS

Household Survey;PUMS

Traffic Counts;Screenline Database;MATRIX

MDC Calibration FrameworkMDC Calibration FrameworkMode ChoiceCalibration

DestinationChoice

Calibration

ApplicationDistance Term Adjustment to

Observed TLD

Intra County-Specific

Adjustment

To-ManhattanAdjustment

AdjustmentTo Aggregate

Targets

OPTION: Adjustment to Traffic Counts

BPM Structure –“GUI” for UserBPM Structure –“GUI” for UserDocumentationDocumentation

Procedures forProcedures forEstimating CongestionEstimating Congestion

HighwayHighwayNetworksNetworks

TransitTransitNetworksNetworks

Socio-Socio-EconomicEconomicDataData

BestPracticeModel

PPAQ /PEQUEST

UpdatedTraffic Data

County LevelCounty LevelDelaysDelays - Person Hours - Person Hours - Vehicle Hours - Vehicle Hours

Congestion MapsCongestion Maps - Link Level - Link Level - Level-of-Service - Level-of-Service Analysis Analysis

Reviews:

Agencies

Public

Comparative Skims: Times to MidtownComparative Skims: Times to MidtownSOV vs. Commuter RailSOV vs. Commuter Rail

How is BPM Better than Other ModelsHow is BPM Better than Other Models

Comparison of BPM versus Traditional ModelComparison of BPM versus Traditional Model

•• GIS BasedGIS Based•• Journey versus tripJourney versus trip•• MicrosimulationMicrosimulation - looks at each household and each - looks at each household and each

journeyjourney•• Walk Trips SeparatedWalk Trips Separated•• Travel InteractionTravel Interaction

–– Auto AvailabilityAuto Availability–– Family InteractionFamily Interaction–– Time constraintsTime constraints

Applications of BPM .. Applications of BPM .. NYMTC’sNYMTC’s Use Use

•• Conformity AnalysisConformity Analysis

•• Regional Transportation PlanRegional Transportation Plan

•• Congestion Management SystemsCongestion Management Systems

•• Testing Scenarios for emission reduction strategiesTesting Scenarios for emission reduction strategies

•• Request for Data Manipulation and Runs from otherRequest for Data Manipulation and Runs from otheragenciesagencies

Applications of BPM .. ProjectsApplications of BPM .. Projects

•• Tappan Zee BridgeTappan Zee Bridge•• GowanusGowanus Expressway Expressway•• Bronx Arterial NeedsBronx Arterial Needs•• BrucknerBruckner SheridenSheriden Expressway Expressway•• Long Island East Side StudyLong Island East Side Study•• Canal Area Transportation StudyCanal Area Transportation Study•• Lower Manhattan Development CorporationLower Manhattan Development Corporation•• Southern Brooklyn Transportation StudySouthern Brooklyn Transportation Study•• Regional Freight Plan StudyRegional Freight Plan Study•• Hackensack Meadowland Development Corp.Hackensack Meadowland Development Corp.

Model UpdateModel Update

•• Study of Post 9/11 Travel Pattern ChangesStudy of Post 9/11 Travel Pattern Changes•• New Set of Socioeconomic and DemographicNew Set of Socioeconomic and Demographic

ForecastsForecasts•• Collection of 2002 traffic and transit dataCollection of 2002 traffic and transit data•• Updated 2002 base year Model by January,Updated 2002 base year Model by January,

20042004

What’s NextWhat’s Next

•• Overcome the current problems Overcome the current problems

–– Very Complex Model – 9 million households, 25 million pairedVery Complex Model – 9 million households, 25 million pairedjourneys, 8 trip purposes, 4 time periods, 10 travel modesjourneys, 8 trip purposes, 4 time periods, 10 travel modes

–– Long Running Time – More than 100 hours for a single scenarioLong Running Time – More than 100 hours for a single scenariorun.run.

–– Hardware Needs - 2 GB RAM / Dual Processor / 1.5 Hardware Needs - 2 GB RAM / Dual Processor / 1.5 GhzGhz / 80+ / 80+GB GB HardriveHardrive

–– Software problems – Software problems – TransCADTransCAD version changed version changed

–– High turnover at consultant endHigh turnover at consultant end

Status of On-Going ImprovementsStatus of On-Going Improvements•• Speed up the running time Speed up the running time

–– Memory HandlingMemory Handling

•• allocated the memory only once, using a flag to determine if the memory had alreadyallocated the memory only once, using a flag to determine if the memory had alreadybeen allocatedbeen allocated

•• memory could be allocated in one blockmemory could be allocated in one block

–– Input/OutputInput/Output

•• Remove messages (one per 33 million lines in the HAJ trip file) to the screen, reducedRemove messages (one per 33 million lines in the HAJ trip file) to the screen, reducedprocessing time from 22 minutes to 20 secondsprocessing time from 22 minutes to 20 seconds

–– Parameter PassingParameter Passing

•• Passing information of a pointer to a structure rather than an entire structure (e.g., thePassing information of a pointer to a structure rather than an entire structure (e.g., thememory used to call about 260,000 times of one function with 92 bytes could bememory used to call about 260,000 times of one function with 92 bytes could bereduced significantly by passing a pointer to the structure that only requires 4 bytes)reduced significantly by passing a pointer to the structure that only requires 4 bytes)

–– In-lining Function CallsIn-lining Function Calls

•• Very short functions that are called frequently can cause bottlenecks (function consistsVery short functions that are called frequently can cause bottlenecks (function consistsof just a few lines (e.g., Calling a function, which was being called between 300,000 toof just a few lines (e.g., Calling a function, which was being called between 300,000 to600,000 times, was taking up 10% of the total program time. In-lining the function600,000 times, was taking up 10% of the total program time. In-lining the functionreduced it to 0.3% of the total program time)reduced it to 0.3% of the total program time)

–– Additional optimizationAdditional optimization

Model ImprovementsModel Improvements

•• Better Highway -Transit Connection – Bus Preload on highways Better Highway -Transit Connection – Bus Preload on highways•• Improve transit models Improve transit models•• Integrate BPM with the Land Use Model Integrate BPM with the Land Use Model•• Web Applications Web Applications

–– Model output analysis Model output analysis–– Model runs Model runs

•• Distributed Process Distributed Process•• Better GUI Better GUI•• More project applications More project applications