Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Development of the Pennsylvania Statewide Commodity-Based
Freight Model
Wade White, AICP, CitilabsBrian Wall, Pennsylvania Dept. of Transportation
Patrick Anater, Gannett Fleming
11th TRB National Transportation Planning Applications Conference
May 7, 2007
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
PennDOT and Analytical Tools—A Historic Perspective
PennDOT’s Credibility & Funding Crises—1970s PennDOT & State Transportation Advisory Committee
– New Directions for PennDOT– Program and Budgetary Control– Strategic Planning & Management
This “Model” was successful for a generation Resource Constraint Renews True Planning Focus PA Mobility Plan as the catalyst for improving
analytical capabilities
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Public Input to the Planning ProcessPublic Input to the Planning ProcessImportance and Satisfaction with Travel OptionsImportance and Satisfaction with Travel Options
2.8 3.8
General Public, n=751
C+
C
1
3
4
5
6
9
Low Grade/High Importance
10
Somewhat IMPORTANCE Very Important (Scale of 1 - 4) Important
B-
High Grade/High Importance
Low Grade/Low Importance
High Grade/Low Importance
8
7
11
2
1. Roads for cars2. Roads for trucks3. Public transit (buses & rail)4. Bicycling facilities5. Pedestrian facilities6. Airline services7. Deep water port facilities8. Bus service between cities9. Passenger train service
between cities10. Rail freight service11. Ride sharing (car and van
pool)
Travel Options (1-11)
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Overview of the Model Development
Identification of data needs Analysis of the freight and GIS databases (Global
Insight, CFS, Freight Analysis Framework) Identification of commodity groupings and
development of zone systems, base year matrices and networks
Development of model structure within Cube Cargo Linkage of data and model adjustment
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Basic Characteristics
3 line-haul modes:– Road– Rail– Inland Waterway– Intermodal is estimated transport logistic
chain models
Ten (10) commodity classes Estimates annual commodity flow
matrices by mode and commodity class
Estimates annual and daily truck matrices by large and small truck class
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Geography
Within Pennsylvania- – Consistent with Statewide Travel
Demand Model– 1001 Zones
Border States Defined same as Travel Model
Other States- Individual Zones
4 external zones; 3 for Canada and 1 for Mexico
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Logistics Nodes
Ports– Port of Wilmington,
DE– Port of Baltimore,
MD– Port of NY/NJ– Port of Philadelphia,
PA– Port of Pittsburgh,
PA– Port of Erie, PA– Port of Cleveland,
OH
Rail Yards– NS Rutherford Yard- Harrisburg, PA– NS Trafford PA Intermodal Yard– Newark/Elizabeth NJ Intermodal Yards– NS Bethlehem PA Yards– NS Taylor, PA NS Yard– Conrail, CSX, NS Philadelphia, PA Rail
Yards– NS Harrisburg Yard
Distribution Centers– Clearfield, PA Walmart Distribution
Center– Bedford, PA Walmart Distribution
Center
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Commodity Groups
1 Unprocessed Agricultural/Fishing Products 2 Unprocessed Ores & Petroleum 3 Coal 4 Processed Food & Tobacco 5 Textiles & Apparel 6 Lumber & Wood Products 7 Chemical, Petroleum or Coal Products 8 Clay, Glass, Concrete, Stone & Leather 9 Machinery & Metal Products 10 Miscellaneous
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
STC Codes, Model Groups and Descriptions
1 1 Farm Products 8 1 Forest Products 9 1 Fresh Fish Or Marine
Products 10 2 Metallic Ores 11 3 Coal 13 2 Crude Petroleum Or
Natural Gas 14 2 Nonmetallic Minerals 19 2 Ordnance Or Accessories 20 4 Food Or Kindred Products 21 4 Tobacco Products 22 5 Textile Mill Products 23 5 Apparel Or Related
Products 24 6 Lumber Or Wood Products 25 6 Furniture Or Fixtures 26 6 Pulp, Paper Or Allied
Products 27 6 Printed Matter 28 7 Chemicals Or Allied
Products
29 7 Petroleum Or Coal Products 30 8 Rubber Or Misc Plastics 31 8 Leather Or Leather Products 32 8 Clay, Concrete, Glass or
Stone 33 9 Primary Metal Products 34 9 Fabricated Metal Products 35 9 Machinery 36 9 Electrical Equipment 37 9 Transportation Equipment 38 9 Instrum, Photo Equip,
Optical Eq 39 9 Misc Manufacturing Products 40 10 Waste Or Scrap Materials 41 10 Misc Freight Shipments 43 10 Mail Or Contract Traffic 46 10 Misc Mixed Shipments 49 10 Hazardous Materials Or
Substances
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Basic Structure 3 groups:
– Network costs: used to estimate zone to zone matrices of door-to-door travel time and cost by each of the line-haul modes for each commodity class. Line-haul modes are Truck, Rail and waterway.
– Cube Cargo: estimation of commodity and truck flows
– Trip Table Preparation: Prepare Truck Trip Tables for PATDM Assignment
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Network Cost
Road: applies Cube Voyager programs to the roadway network to estimate paths
Rail: applies Cube Voyager to the rail network to estimate paths
Water: network of major waterway services and rivers
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Truck, Rail, and Water Cost Estimation
Estimation of travel times and costs via:– Network based travel time
with user assumptions on:• Pickup and drop-off time• Driver rules: Federal Hours of
Service (truck)• Average speed
– Network based distance used with cost parameters:
• Cost per ton-mile by commodity type
– Based on user assumptions and published cost data
– Network developed from FHWA roadway and rail network
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Network Costs
The network cost group provides:– Zone to zone door-to-door cost per ton by mode and commodity
group– Zone to zone door-to-door travel time by mode and commodity
group– Zone to zone travel distance by mode
Data are important elements in the estimation of the distribution, mode choice and transport-logistics nodes models
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Model Process Generation: estimates annual tons
of commodities produced and consumed by zone by commodity class
Distribution: distributes goods by commodity class
Mode Choice: estimate modal shares of long-haul flows
Logistics Nodes Model: partitions the long-haul matrices by mode and commodity class into direct flows and transport chain flows
Fine Distribution Model: for each of the matrices redistributes from coarse zones to the fine zones
Vehicle Model: converts the estimated annual commodity flows by truck into number of heavy trucks and light trucks
Service Model: estimates daily urban service truck trips
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Generation Regression models on
socioeconomic attributes (zonal data) and constants by commodity class and country
Use of special generators to represent external generated commodities: ports by location of facility and commodity class
Trend rates to represent production efficiencies and other factors not represented in the regression models by commodity class and country
User specified values for the amount of production exported to external zones and the amount imported to the internal zones by commodity class
Trend rates to represent trends in the level of import and export. Internal Area External Area External Zone TLNStudy Area
External zones controlled with ‘shift’/ “singpoint” variables to fix imports and exports by commodity class and trend rates
TLN have no production and consumption
Production and consumption by commodity class and country based on socioeconomic attributes of the zones and trend rates (efficiencies)
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Distribution User assumption on percentage of
goods that are to be considered short-haul and long-haul by commodity class
Trend rates to represent changes in short-haul and long-haul percentages by commodity class
Short-haul trips are considered to be transported by truck and are distributed using gravity models by commodity class Impedance is cost
Segments the remaining long-haul flows into those remaining ‘internal’ and those remaining ‘external’ by commodity class
Adjusts internal and external fractions by user assumption on trends by commodity class
Distributes internal, import and export long-haul flows using gravity models by commodity class
Impedance is a generalized cost using a linear combination of time, distance and cost by mode weighted by the mode choice coefficients
Internal Area External Area External Zone TLNStudy Area
Set assumption (%) and trend rates on what is short- and long-haul flow by commodity class
Short-haul flows will be truck only and distributed with gravity models
External fractions set by user plus trends
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Mode Choice
For Long-Haul Only Multinomial logit models
stratified by commodity and distance class
Choice models use by mode and commodity class– Time– Cost– Constant
Internal Area External Area External Zone TLNStudy Area
Estimate percentage truck, rail and waterway by commodity class based on door-to-door shipment time and shipment cost and constant
Only for long-haul flows
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Logistics Node Model
Partitions the long-haul matrices by mode and commodity class into direct and TLN flows
Definition of zone location of TLNs and the zones that they serve
Definition of directionality of TLN flows and selection of TLN
Product is matrices by commodity group segmented into:
– Long-haul direct flows by mode
– Long-haul to/from TLN flows by mode
– Short-haul to/from TLN flows by truck
Internal Area External Area External Zone TLNStudy Area
Define location of TLN
Define service area of TLN
Partitions into Long-Haul Direct Flows by mode
Partitions into Long-Haul TLN Flows and Short-Haul TLN Flows by mode
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Fine Distribution Model Distributes models from
coarse zone system to fine zone system
The fine zones are smaller and nested under a coarse zone. These flows are distributed to the fine zones encompassed by the coarse zone using:
– a weight to establish a small sub-matrix of the fine zone matrix based on parameters and fine zone level zonal data
– and a gravity model using, distance as the impedance, to infill the individual cell values
– It is possible to override these models to represent particular points in the system
Fine ZoneCoarse Zone
30
10
15
25
25
25
25
35
10
Allocate Origins with Weights based on socioeconomic data
Allocate Destinations with Weights based on socioeconomic data
Distribute from fine origin to fine destination using gravity models
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Review: bringing the flows to the vehicle model Generation gives P & C by zone and commodity class
Distribution distributes two sets of matrices:– short-haul flows by commodity class which are assumed to be truck flows;
and– long-haul flows by commodity class which go to mode choice
Mode Choice splits the long-haul flows into long-haul flow matrices by mode and commodity class
The long-haul modal matrices from Mode Choice are segmented into flows that:– travel directly from zone of production to zone of consumption (Direct long
haul flows by commodity class) and,– flows that will use a TLN. The flows that go via TLNs are segmented into:
• short haul segment by mode and commodity class• long-haul segment by mode and commodity class
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Generation
P & A by CC
Distribution
Direct Short-Haul Flows by CC by Truck
Long-Haul Flows by CC
Mode Choice
Long Haul Flows by Mode & CC
TLN
Direct Long-Haul Flows by Mode & CC
Short-Haul Flows to TLN by Truck & CC
Long-Haul Flows to TLN by Mode & CC
Fine Distribution
Direct Short-Haul Flows by CC by Truck
Direct Long-Haul Flows by Mode & CC
Short-Haul Flows to by truck & CC
Long-Haul Flows to TLN by Mode & CC
Fine zone level
Coarse zone level
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Vehicle Models The vehicle models estimate matrices of light and heavy truck
trips by using vehicle models and load factors.
Two vehicle models:– Touring Vehicle Model
• Used for flows where the origin is a TLN• And, where the user selects specific zones
– Standard Vehicle Model• Used in all other cases
Results are combined to provide a truck trip matrix for assignment– Heavy long-haul trucks
– Heavy short-haul trucks
– Light short-haul trucks
By default, these matrices are annual truck flows (resulting from the annual commodity flows). Matrix manipulation can be used to estimate daily and hourly flows by season if so desired.
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Touring Vehicle Model
Vehicles are assumed to have the same start and end zone but make intermediate stops to load and unload.
Heavy computations so limit use to TLN and selected zones.
Estimates number of vehicles based at the origin using the flows from that zone and average load factors.
Internal Area External Area External Zone TLNStudy Area
Generated tour from a TLN and back doing pickups and drop-offs
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Standard Vehicle Model
Model assumes that all vehicles make trips of the form A-B-A.
Return load is a function of the commodity flow in the opposite direction.
Can modify using ‘big zones’ enlarging the area considered for a return load.
Internal Area External Area External Zone TLNStudy Area
By default, the standard model creates direct round trips between the two zones. The probability of a return load depends on the flow of goods in the ‘back direction’
With the use of ‘Big Zones’ can include neighboring zones when calculating probability of a return load. This generates a simple tour.
A ‘Big Zone’
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Service Trips All modeling to this point concerns the movement of
goods.
The service model is used to estimate urban service truck trips such as:– Repair men (e.g. elevator repair)– Small shopkeeper taking goods from a wholesaler to a local restaurant, etc.
Used directly on the fine zone system
Estimates generation using regression models based on zone type and socioeconomic data
Trips are distributed using gravity models.
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Truck Trip Table Creation
Matrix manipulation to estimate average daily truck trips for assignment
Single Unit (Light Trucks)
Combination Unit (Heavy Trucks)
Through (E-E) Truck Trips
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Running the model – base year
A user menu allows the modification of various attributes for the scenario:– Input highway network– Truck pick-up and drop-off delay
time– Driver parameters (break and
sleep)– Costs per ton-mile by mode– Rail pick-up and drop-off delay– Waterway pick-up and drop-off
delay– Average rail travel speed
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Potential Use for the PA Statewide Freight Model
To Answer Policy Questions
– Effects of alternative growth scenarios on freight movement• What if regional development patterns change?• What if major freight facilities are developed?
– Effects of alternative policies on freight movement• What if tolls were increased?• What if the price of fuel continues to increase?
– Impacts of major projects on freight movement• What if a two-lane US highway was widened to four lanes?• What if major access improvements to a region were advanced?
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Outputs for PennDOT’s Use
Region to Region Commodity flows– 10 Regions
Tonnage and Value by Mode Key Freight Corridors Future expected growth System Impacts
0%
50%
100%
150%
200%
250%
% C
han
ge 2
002 to
2030
Unprocessed Agricultural/ Fishing Products
Unprocessed Ores & Petroleum
CoalProcessed Food & Tobacco
Textiles & Apparel
Lumber & W
ood Products
Chemical, Petroleum
or Coal Products
Clay, Glass, Concrete, Stone & Leather
Machinery & Metal Products
Miscellaneous
Commodity
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Changes in Organizational Thinking
Current LRTP serves as a change agent
Major awareness raising challenges
Creating a culture valuing analytically based planning
TDM in an emerging Multi-Modal Agency
Leveraging strong PennDOT—MPO and RPO Partnership
Paradigm shifts for evaluation—Core System, State of the System, Performance Based
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Changes in PennDOT Capabilities
Leadership Support
Analytical Skill Building from the ground up
Applied Training—Learning through implementation
Capacity Building Bridging Agency & Consultant
Marketing capabilities to PennDOT staff and leadership
Continuing to Expand Horizons—Applying and expanding the tools as warranted
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Implications
TDM and freight modeling can’t be in a vacuum
Strategic Integration of T-O-P Thinking– Technical, Organizational, and People
Building bridges among a wide range of organizations (internal & external)
Communicating Progress & Raising Awareness
Fostering Experimentation, Innovation, and Risk Taking
Pennsylvania Statewide Freight ModelPennsylvania Statewide Freight Model
Questions?