2 TRB Daytona Stryker

Ohio Statewide Travel Model:Framework, Freight, and Initial Calibration: 

Ohio Statewide Travel Model: Framework, Freight, and Initial Calibration 11th National Transportation Planning Applications Conference May 6-10, 2007, Daytona Beach, Florida Session 6:

Acknowledgements: 

Acknowledgements This presentation was primarily developed by Pat Costinett.

Topics: 

Topics Ohio Statewide Modeling Framework Micro-simulation Integrates: Economic Land use Transport Models Aggregate Commercial Model (ACOM) Preliminary calibration results

General Model Structure: 

General Model Structure Integrated micro-simulation based Model economic activity & land use Build synthetic population Tour-based Home tours Establishment/Work tours Aggregate commodity movements

Model Components & Flows: 

Model Components & Flows

Model Components & Flows: 

Model Components & Flows Economic Activity by Geography

ISAM: 

ISAM Input-output economic model Represents trading commodities Exogenous to the model system

Slide8: 

1 = Model Area

ISAM: 

ISAM Input-output economic model Represents trading commodities Exogenous to the model system Region to region commodity flows Shares of commodity flows from the model area to regions

Economic Activity & Land Development: 

Economic Activity & Land Development Approximately 700 districts and 4000 zones Distribution of economic activities & flows by sector to analysis districts Production of goods & services by zone Consumption demand for goods & services by zone Flows of commodities (goods, services & labor) among zones In response to exchange prices Interacting with a grid-based representation of land supply, develop types, zoning, water & sewer service, flood plains, steep slopes, other protected land uses and land prices

Economic Activity & Land Development: 

Economic Activity & Land Development Results: Flows of commodities between districts Floor space allocated to activities by zone

Model Components & Flows: 

Model Components & Flows Synthetic Population

Model Components & Flows: 

Model Components & Flows Transport Models

Types of Trip Making Modeled: 

Types of Trip Making Modeled Personal Travel /Household Travel (PT): person movements arising from household (or population) production and consumption, separated into short distance (50 mi or less) and long distance Visitor Travel (VM): person movements made by non-residents staying at locations in the internal model area Business/Services Travel (DCOM): movements arising as part of the rest of the ‘business cycle’ apart from the physical delivery of commodities Goods Transport (ACOM): shipments of commodities arising from economic activity production and consumption

Short Distance Tours (SDT): 

Short Distance Tours (SDT) Weekday travel behavior of persons for all purposes: work & school, shopping, recreation, other Based on large sample of one-day travel diary data of urban and rural households in the model area (~15000 households) Higher decision models are informed by lower level accessibility measures Model components include: Work place location Household auto ownership Full day activity pattern choice Primary activity location choice Primary activity schedule choice Tour mode choice Intermediate stop location and arrival/departure time choices Trip mode choice Work-based sub-tours activity duration and location choice

Long-Distance Tours (LDT): 

Long-Distance Tours (LDT) Infrequent occurrence but important element of statewide/intercity travel demand Omitted from conventional models Household survey of long-distance travel Two-week retrospective survey of 6000 households Four week prospective survey of 2000 households Survey-based model derivation incorporating full micro-simulation of households & persons Model components include: Choice to make a LDT or not, whole household tour, individual business tour or individual other tour Tour pattern choice, depart, arrive, round trip, away on travel day Time-of-day tour scheduling Primary destination choice Intermediate stop frequency and destination choice Mode choice

Commercial Travel: 

Commercial Travel Incorporates long-haul commodity shipment, localized goods delivery, service provision & work-related tours Long-haul shipment related directly to commodity flows Establishment survey of goods delivery, service provision & work-related tours Micro-simulation of commercial tours for each employee (a first at this scale)

Why a freight model?: 

Why a freight model? Need to be consistent with economic models Freight movements are important to Ohio: Interest in impact of Turnpike tolls on trucks. Interest in road-rail diversion. Relatively large impact on traffic LOS

Underlying “Theory”: 

Underlying “Theory” Commodities are carried by trucks, rail, and other modes Commodity flow patterns determine truck flow patterns Truck characteristics vary substantially by commodity type and shipment distance Mode share Average value per ton Size mix Average payload weight Unlike personal travel, commodity shipment choices are influenced very little by network LOS measures

What does it do?: 

What does it do? ACOM translates dollar flows of commodities from ISAM and AA into truck trips by four size categories ISAM for E-E AA for I-I Both for E-I and I-E

ACOM and Economic Models Relationships: 

ACOM and Economic Models Relationships Internal to Internal External to Internal External to External Internal to External ISAM AA AA ISAM

What does it do?: 

What does it do? ACOM translates dollar flows of commodities from ISAM and AA into truck trips by four size categories ISAM for E-E AA for I-I Both for E-I and I-E These trips are different than service and sales calls made by employees covered in DCOM. Minimal overlap between ACOM and DCOM

General Model Flow: 

General Model Flow ISAM AA Distance

External to External Flow: 

External to External Flow ISAM AA Distance

Internal to Internal Flow: 

Internal to Internal Flow ISAM AA Distance

Internal to External Flow: 

Internal to External Flow

ETAZ and TAZ Weights: 

ETAZ and TAZ Weights ETAZ – based on Average production and consumption per employee by category from AA and Employment by category by ETAZ TAZ – based on production and consumption summary by TAZ from AA

$ Flows to Truck Trips by Size: 

$ Flows to Truck Trips by Size Truck $’s to Truck tons Split Truck tons by Truck Size Convert to Truck trips Distance Factors Total $’s to Truck $’s Convert to time periods by STCC

Calibration: 

Calibration Each of the models uses a gamma function to calculate deterrence as a function of distance and three parameters The parameters can be adjusted up or down to match trip lengths and distribution shapes Calibration Targets Ohio county to external state for Statewide Cordon Roadside Survey Selected MPO County to other Ohio counties truck trips from MPO Roadside Surveys Average trip lengths by area from CFS97 and Transearch

Average Truck Trip Lengths: 

Average Truck Trip Lengths

Average Truck Trip Lengths: 

Average Truck Trip Lengths

System Calibration Process and Targets: 

System Calibration Process and Targets The Statewide Modeling System

Four Stages of Parameter Development: 

Four Stages of Parameter Development S1: Parameter estimation - parameters are developed for each module separately and individually. Statistical methods are used to estimate appropriate values where suitable data are available. S2: Initial calibration - also involves the fit of each module in isolation but inputs include those provided by other modules. Parameters are adjusted to match module-specific targets. S3: Base year calibration – consolidating results of full model chain run and comparing to observed system flows. Selected S2 parameters revisited considering relative LOC. S4: Temporal(+) calibration – evaluation of model forecasts in comparison to independent forecast results. Selected S3 parameters revisited considering relative LOC.

S3 Calibration OD Checks: 

S3 Calibration OD Checks Total auto and total truck trips crossing model area and Ohio cordons versus counts Ohio county to external state auto and truck trips versus roadside survey for Ohio cordon For counties entirely within MPO roadside survey cordon, OD flows to counties entirely outside MPO cordon versus MPO roadside survey

MPO Roadside Survey Cordons: 

MPO Roadside Survey Cordons

Slide36: 

OD Analysis Districts

Initial results for auto vehicle trip OD (1): 

Initial results for auto vehicle trip OD (1)

Initial results for auto vehicle trip OD (2): 

Initial results for auto vehicle trip OD (2)

Acceptable Error by Volume: 

Acceptable Error by Volume Source: ODOT Assign2000.doc, by Greg Giaimo

S3 Calibration Global Assignment Checks: 

S3 Calibration Global Assignment Checks VMT by FUNCLASS Model Area Ohio MPO county groups Major Screenline Volumes by FUNCLASS Model Area cordon Ohio cordon MPO cordons Source of independent VMT estimates? Counts versus “counts”

Initial Unconstrained Auto Assignment ResultsSum of Link Flows for Links with Actual Year 2000 Counts (20,751 links): 

Initial Unconstrained Auto Assignment Results Sum of Link Flows for Links with Actual Year 2000 Counts (20,751 links)

Initial Unconstrained Auto Assignment ResultsSum of Link Flows for Links with Actual Year 2000 Counts (20,751 links): 

Initial Unconstrained Auto Assignment Results Sum of Link Flows for Links with Actual Year 2000 Counts (20,751 links)

Initial Unconstrained Auto Assignment ResultsAll Links: 

Initial Unconstrained Auto Assignment Results All Links

Conclusions: 

Conclusions This framework allows us to be consistent. Calibration results look good so far. More work to be done.

Questions for Pat?: 

Questions for Pat?

S3 Calibration Detailed Assignment Checks: 

S3 Calibration Detailed Assignment Checks VMT by FUNCLASS by county Major Screenline Volumes by FUNCLASS by segment Model Area cordon Ohio cordon MPO cordons Percent RMS by volume range – see acceptable RMS error table Plots of assigned flows versus counts Scatterplots by FUNCLASS Network links plots of differences Source of independent VMT estimates? Counts versus “counts”

Acceptable RMS Error: 

Acceptable RMS Error Source: ODOT Assign2000.doc, by Greg Giaimo

S4 Temporal Calibration Components: 

S4 Temporal Calibration Components 1990 to 2000 AA-LD application 2000 to 2030 Base Forecast Future Scenario Forecasts Turnpike Toll Scenario Five-County Corridor Scenario High Speed Rail Scenario

1990 to 2000 AA-LD Application: 

1990 to 2000 AA-LD Application Concept - Begin with 1990 land use, socioeconomic characteristics, roadway network Compare to 2000 socio-economic characteristics by AMZ Calibrate AA-LD parameters to get acceptable results – e.g. AA inertia terms, LD transition factors Problem – LD depends on floorspace prices from AA to influence development choices We have too little floorspace price data for AA price calibration and additional Assessor data collected by ODOT for additional counties has not improved the situation We are still working on ways to overcome this problem

Forecast Scenarios: 

Forecast Scenarios

2000 to 2030 Base Forecast: 

2000 to 2030 Base Forecast Purpose: evaluate model performance versus independent population forecasts Inputs: 2010 and 2020 roadway networks, development overrides by TAZ Target: ODOD population forecasts by county Related activity: Reconcile ISAM/SPG1 forecasts for Model Area with ODOD population forecasts Potential parameter adjustment focused on AA-LD response

Turnpike Toll Scenario: 

Turnpike Toll Scenario Purpose: evaluate model performance versus independent toll change response Inputs: 2010 and 2020 roadway networks, development overrides by TAZ, change in Turnpike tolls and truck speeds in 2005 Target: Turnpike volumes by weight class Potential parameter adjustment focused on value-of-time for trucks by Turnpike weight class

Five-County Corridor Scenario: 

Five-County Corridor Scenario Purpose: evaluate model response to corridor roadway improvements Inputs: 2010 and 2020 roadway networks, development overrides by TAZ Target: Nothing specific Potential parameter adjustment focused on AA-LD response

High Speed Rail Scenario: 

High Speed Rail Scenario Purpose: evaluate model performance versus independent HSR forecasts Inputs: 2010 and 2020 roadway networks, development overrides by TAZ, HSR ALT 1 & 2 intercity rail networks Target: Rail ridership forecasts from Ohio Rail Hub Strategic Study Potential parameter adjustment focused on LDT mode choice parameters