1. Estimating Freight Flows in WA State: Case studies in data-poor and data-rich environments
Erica Wygonik, University of Washington
Presented on behalf of Derik Andreoli, Anne Goodchild, Eric Jessup, and Sunny Rose
Good morning.
I am Erica Wygonik, and I am here presenting work on behalf of my colleagues who could not attend.
I am presenting a summary of their work, and, while I am familiar with their research, I did not participate in conducting it. I will do my best to represent what they‘ve done and answer any questions you might have.
The 3rd Conference on Innovations in Travel Modeling 12 May 2010 1

2. Freight supports regional economies
Research Problem
Freight supports regional economies
Desire to justify investments targeting freight
Evaluate the impacts of network changes
Vulnerability to disruptions
Improvements and infrastructure needs
Limited by available data
overlooked in infrastructure evaluations
quantify the value of improvements
These corridor-level planning exercises need the equivalent of activity-based modeling for freight.
Where are vehicles going, why, when, what trips link with what others, etc.

3. State of Freight Modeling
Currently two primary modeling sources:
Commodity flow data
Gross vehicle volumes
Assume industries use infrastructure in the same way
Existing methods are too coarse for needed analysis
Commodity flow data spatially aggregate
Vehicle estimates are categorically aggregate
Basic problem in freight modeling:
Lack of quality disaggregate commodity flow data

4. Project Scope
Improve the representation of freight movement in statewide modeling
Work within existing data constraints
Study Washington State due to the frequent disruptions to key freight corridors
I-5 (flooding)
I-90 (avalanche)
Minimal travel diary info for freight movements, unlike the thin but reasonably regular collection for personal travel
Develop and implement a methodology to estimate industry-specific, corridor-level truck trips
winter season: Snoqualmie Pass closed ~370 hours (WSDOT GrayNotebook 2008a).
Affects I-90 (and 2 & 12)
Highway 410 and Highway 20 are closed seasonally every winter
longest closures are 3-5 days long
no way to turn around

5. Washington State Topography
SEATTLE
SPOKANE
CASCADES RANGE
YAKIMA
Cascades divides Washington in two:
western portion, what most of you think of with washington: rain, starbucks, grunge, twilight.
4 county puget sound nearly 50% of population in state
eastern portion, very arid, much of the agriculture
VANCOUVER
Map courtesy of geology.com

6. Washington State Infrastructure
Only 3 ways across the Cascades
SEATTLE
SPOKANE
YAKIMA
Cascade passes significantly limit east-west travel
in the winter, only 3 routes cross the cascades: I-90, Highway 2, and State Route 12
ALL are very prone to avalanche danger
frequent closings of varying lengths
Which helps more: information on closures or solutions to improve reliability
VANCOUVER
Map courtesy of Google maps

7. Focus on Two Sample Data Sources
Estimate statewide truck trips required for the operation of industries within Washington State
Data-rich industry: potato distribution
Production
Processing
Demand
Distribution
Capacity Ratios
Data-poor industry: diesel distribution
Use estimated origins & destinations
How to model flows?
Photo courtesy of WSDOT
Look at two situations with differing amounts of information
Provide variants on the methodology for different levels of data availability
Identify the level of effort required based on available data
Washington is 2nd largest producer of potatoes in the nation
Potato distribution (idealized case)
Fresh, frozen, dehydrated, and potato chips
Include use of intermediate facilities
Exporting from the state
Diesel distribution (realistic case)
From terminal rack to cardlock facilities
Truck serves last segment
Uses barge and pipeline as much as possible to get to terminals

8. Potato Industry Flow Estimation
Courtesy of the WA State Potato Commission 8

10. Started at the first step. We know where potatoes are produced. USDA
Have detailed location, detailed quantities
Hard to see  3 orders of magnitude difference
leads to very different truck volumes.
Different amounts are grown bc of climate, farming techniques, potato type
Justification for why we want to know this level of detail.
NW grows thin-skinned organic vs industrial used for processing.

11. We captured processing  more detail than commodity flow
Know where processing is and how it influences flows.
Since different potatoes are used for different purposes,
all frozen and dehydrated is on the east side of the state
west side is just potato chips
good bit of moving around
(higher transportation cost for those as finished goods)

12. Here is illustration of how one region’s production facilities serve demand
you can see the potatoes are going all over
We have large amounts of data. from Washington State Potato Commission
The industry is organized and informed and regularly collects data from its members.
The process is relatively straightforward once the data is collected: match origins to destinations based on shortest path.

13. With this info we can model the routes shipments take, and how they would divert in the case of a closure.
The line weight illustrates the number of trucks per day on each route. The red lines show the diversion routes – many trucks are diverted in the case of a cross-Cascades disruption, resulting in a near doubling of trip lengths.
PS this only shows the links/trucks that would be affected by a disruption. Doesn’t include west side only trucks. Is ~20% of potato-based truck trips in state.

14. Potato Industry Flows: Summary
Significant cross-Cascades travel
Low profit margins on potato shipments
Cannot afford to take detours
Waiting or failure to stock products are expensive
Very vulnerable to long closures

15. Diesel Industry Flow Estimation15

17. Mapping diesel flows
This map is the overall flow of diesel cargo in the state.
Long distance movements occur almost exclusively on barges and in pipelines to the distribution points – terminal locations. These modes are used because of the large volumes and high cost of movement. The terminals are all on bodies of water or at the end of pipelines and then the diesel is moved by truck to local points of sale.
We are modeling the truck movements of diesel as cargo – the green arrows moving away from the red dot terminal locations.
Why barge/pipeline? Volume too high to move by truck – not enough trucks. And barge/pipeline far more economical.
terminal racks
truck
pipeline
barge

18. CASCADES RANGE
The red dots representing the terminals have become blue triangles. These are the origins in our model. We know where all the terminals are.
The green dots are racks – points of sale of diesel to consumers – and serve as the destinations. These locations represent a subset of diesel points of sale in the state but serve trucking companies and sell the majority of the diesel.
We are approximating the origins and destinations for diesel using physical infrastructure (buildings).
Terminals and racks are known.
We do not KNOW which terminals serve which racks but we can make assumptions.
These are not all the diesel points of sale but sell the majority of diesel. They are cardlock facilities used by trucking companies. They offer discounts and simpler billing but are not truck stops.

19. SEATTLE SPOKANE CASCADES RANGE YAKIMA VANCOUVER
We can assume the racks are served by the closest terminal.
With those OD pairs we can identify the shortest path.
Reasonable since the product quality is relatively consistent and costs are directly tied to traveled distance. Obviously the assumption fails in some cases, but generally holds up.
The link widths represent the number of OD pairs that would use a link
 NOT WEIGHTED BY VOLUME
(we don’t know rates of consumption)
 still useful on a strategic level.
some routes are used heavily, while other parts of the infrastructure don’t matter.
Diesel distribution does not rely on the Cascades corridor. It distributes locally on trucks or uses other modes.
VANCOUVER

20. Adding a cross-cascades disruption has less impact here:
Only 1 rack in the base case is served by a terminal that would require cross-cascades travel. But that rack could easily switch terminals.
Increased by 13.3 minutes per trip

21. Diesel Industry Flows: Summary
Minimal cross-Cascades travel
Multimodal network avoids mountain passes
Distributed terminals provide buffers
Can estimate network segment importance using known information…
BUT cannot assess flows because of lack of information
Very large volumes are moved
Diesel is a higher-value industry, but potatoes are more sensitive to road network disruptions
(diesel distribution is HIGHLY vulnerable to pipeline and/or barge disruption)

22. Methodological Summary
Proposed methods evaluate infrastructure use with and without primary flow data
Locations of fixed infrastructure are generally available
Flow data is much harder to obtain
Allows evaluation of impact of disruptions
Requires two different metrics
Effectively supplements travel
data in a data-poor environment
We know where farms, terminals, processing facilities are.
We can use that information to supplement travel information in a data-poor environment.
Ongoing work
Improving routing logic for statewide multi-modal GIS freight model
Survey of shippers and carriers
Correlating supply chain factors to economic factors
Evaluating the use of truck GPS data to support the above activities
Developing a methodology to collect sub-national commodity flow data
Photo courtesy of Shell

23. Thank you Questions: Anne Goodchild annegood@uw.edu
I am happy to take your questions.

24. Data Industry Data GIS Model Potatoes:
Washington State Potato Commission data and expertise
Previous work by Dr. Jessup and WSDOT
Diesel:
Washington State Department of Ecology, Environmental Protection Agency, Department of Revenue
CFN and Pacific Pride networks
Interviews with Marketers and industry experts
GIS Model
Multimodal representation of the state freight infrastructure
Includes impedance factors to travel along links in the transportation system
Impedance factors – distance and speed.