Kaan Ozbay, Ph.D.

1
Modeling of Transportation Evacuation Problems
for Better Planning of Disaster Response
Operations
M. Anil Yazici Graduate Assistant, Rutgers
University, Civil Environmental Engineering
Dept. 623 Bowser Road, Piscataway,
NJ yazici_at_eden.rutgers.edu

• Kaan Ozbay, Ph.D.
• Associate Professor,
• Rutgers University,
• Civil Environmental Engineering Dept.
• 623 Bowser Road, Piscataway, NJ
• kaan_at_rci.rutgers.edu

2
Evacuation?

• mass physical movements of people, of a
  contemporary nature, that collectively emerge in
  coping with community threats, damages, or
  disruptions
• by E. L. Quarantelli, founder of Disaster
  Research Center.

3
Strategies Against a Disaster

• Control of the threatening event itself
• Control of human settlement patterns
• Development of forecasting techniques and warning
  systems that generate a protective response by
  those whose threatened
• ?Subjects of disaster preparedness
• Reference
• Perry, R., Lindell, M., and Greene, M. (1981).
  Evacuation planning in emergency management.
  Lexington Books, Lexington, Mass.

4
Types of Evacuations

• Voluntary
• Recommended
• Mandatory
• ? The issue of such evacuation orders involve
  legal aspects heavily
• Reference
• Wolshon B., Urbina E., Levitan M., National
  Review of Hurricane Evacuation Plans and
  Policies, LSU Hurricane Center Report, 2001.

5
Evacuation Modeling

• 1970s first attempts mostly for hurricane
  evacuation
• 1979, a milestone Nuclear accident in Three
  Miles island Evacuation studies focus on nuclear
  plant threats
• 1990s, emphasis is directed towards hurricanes
  again
• Recent Tsunamis and earthquakes in Asia brought
  the network connectivity issue into consideration
• What will happen to all those evacuated people? ?
  Shelter/supply location-allocation.
• Selected References
• Chester G. Wilmot and Bing Mei, Comparison of
  Alternative Trip Generation Models for Hurricane
  Evacuation, Natural Hazards Review, November
  2004, pp 170-178.
• Sherali, H. D., Carter, T. B. and Hobeika, A. G.,
  A Location-Allocation Model and Algorithm for
  Evacuation Planning under Hurricane/Flood
  Conditions, Transportation Research Part B, Vol.
  25(6), 1991, pp.439-452.
• Chang S.E. and Nobuoto N., Measuring Post
  Disaster Transportation System Performance the
  1995 Kobe Earthquake Comparative Perspective,
  Transportation Research PartA, Vol35, 2001,
  pp.475-494.

6
3 Critical Questions

• What is the clearance time required to get the
  hurricane-vulnerable population to safe shelter?
• Which roads should be selected?
• What measures can be used to improve the
  efficiency of the critical roadway segments?
• Reference Donald C. Lewis, Transportation
  Planning for Hurricane Evacuations, ITE Journal,
  August 1985, pp31-35

7
Evacuation Modeling, A Simple Scheme
Operational and Structural Aspects
Demand Generation
Contra-flow
Evacuation Demand
Shelters
Destination and Route Assignment
Supply Logistics
Sensitivity of Behavioral Models
Assignment Under Link Capacity Uncertainties
8
Major Parameters Affecting Evacuation Demand
under Hurricane Conditions

• Baker (1991) studies 12 hurricanes from 1961 to
  1989 in almost every state from Texas through
  Massachusetts.
• Risk Level (Hazardousness) of the area
• Actions by public authorities
• Housing
• Prior perception of personal risk
• Storm specific threat factor
• Reference
• Earl J. Baker, Hurricane evacuation behavior,
  International Journal of Mass Emergencies and
  Disasters, Vol.9, No.2, 1991, pp 287-310

9
Evacuee Behavior

• Individual decision process consists
• Whether to evacuate
• When to evacuate
• What to take
• How to travel
• Route of travel
• Where to go and
• When to return
• References
• Alsnih R., Stopher P.R., A Review of the
  Procedures Associated With Devising Emergency
  Evacuation Plans, TRB Annual Meeting, 2004.
• Sorensen, J.H., Vogt, B.M., and Mileti, D.S.
  (1987), Evacuation An Assessment of Planning
  and Research, Oak Ridge National Laboratory,
  report prepared for the Federal Emergency
  Management Agency Washington D.C.

Relates to Evacuation Demand
Relates to Traffic Assignment
10
Approaches for Determining Evacuation Demand

• Empirical, expertise based approaches
• Sigmoid response curves (S-Curves)
• Artificial Neural Network Models
• Hazard / Survival Models
• Logit Models

• References
• Haoqiang Fu, Development of Dynamic Travel
  Demand Models For Hurricane Evacuation. PhD
  Thesis, Louisiana State University, 2004.
• Mei B., Development of Trip Generation Models of
  Hurricane Evacuation. MS Thesis, Louisiana State
  University, 2002.

11
Related Studies Carried Out by the Rutgers CEE
Research Team

• Evacuation Demand Analysis
• Ozbay K., Yazici M.A. and Chien S. I-Jy. Study
  Of The Network-Wide Impact Of Various Demand
  Generation Methods Under Hurricane Evacuation
  Conditions. Proceedings of the 85th Annual
  Meeting of the Transportation Research Board,
  Washington, D.C., 2006.
• Ozbay K. and Yazici M.A., Analysis of
  Network-wide Impacts of Behavioral Response
  Curves for Evacuation Conditions, Proceedings of
  the IEEE ITSC 2006 Conference, 2006.
• DTA with Stochastic Network Link Capacities
• Yazici M.A. and Ozbay K., Determination of
  Hurricane Evacuation Shelter Capacities and
  Locations with Probabilistic Road Capacity
  Constraints, Accepted for Presentation at the
  86th Annual Meeting of the Transportation
  Research Board, Washington, D.C., 2007.
• Shelter Supply Logistics
• Ozbay K. and Ozguven E.E., A Stochastic
  Humanitarian Inventory Control Model for Disaster
  Planning, Accepted for Presentation at the 86th
  Annual Meeting of the Transportation Research
  Board, Washington, D.C., 2007.

12
Simple Evacuation Network for Multiple Origin
Single Destination
Destination
Evacuation Routes
Demand Origin
Source NJ Office of Emergency Management
13
Multiple-Origin Multiple-Destination Cell
Transmission Model
Source Yazici M.A. and Ozbay K., Determination
of Hurricane Evacuation Shelter Capacities and
Locations with Probabilistic Road Capacity
Constraints, Accepted for Presentation at the
86th Annual Meeting of the Transportation
Research Board, Washington, D.C., 2007.
14
Simple SO DTA Formulation
SO DTA in Compact Format
SO DTA with Probabilistic Capacity Constraints
15
Demand Sensitivity Analysis

• Cell Transmission Based (CTM) System Optimal
  Dynamic Traffic Assignment (SO DTA) is used.
• Choice of demand model changes the evacuation
  performance measures significantly (e.g.
  Clearance Times, Average travel times).
• Even using simplistic S-Curve only , under
  Rapid-Medium-Slow response, the results change
  significantly.
• Demand loading scheme plays a very important role.

16
Stochastic Link Capacity Analysis

• Singlr demand profile ? S-Curve is used within
  CTM based SO DTA framework.
• Probabilistic link capacities are assigned to
  represent flooding, incidents etc. on the network
  during evacuation
• SO DTA formulation is extended with probabilistic
  capacity constraints and pLEP method proposed by
  Prekopa is used for solution.
• The network flows change considerably when
  probabilistic analysis is performed.
• The required capacity of the shelters also change
  with probabilistic assignment.

17
Summary of Important Findings

• The demand sensitivity analysis show that the
  choice of demand curves impact clearance and
  average travel times, especially in case of a
  link capacity reduction.
• The probabilistic SO DTA shows that overall
  network flows and the number of people arriving
  each shelter are mainly affected by the
  probability of link failures.
• The number of people in each shelter is the main
  component required for the determination of
  required supply (logistics) as well as the
  structural and operational aspects of these
  shelters.

18
Future Work

• Modify existing demand models based on available
  data to fit NJ facts.
• Run evacuation scenario using a micro-simulation
  model for comparison with the analytical results
  obtained from the SO-CTM model
• Extend the probabilistic link capacity analysis
  to include other stochasticities such as demand
  uncertainty.
• Test robustness of the results for a more
  accurate and real size network

19
Thank you