Learning Center
Plans & pricing Sign in
Sign Out



									Analysing Evacuation Decisions using
  Multi-Attribute Utility Analysis
              Paul Kailiponi
              CRISIS Centre
          Aston Business School
             Aston University
• ERGO Project
• The Evacuation Problem
• Decision Components
   – Objective Function
   – Probability Function
• Illustrative Example (identifying risk thresholds)
• Substantive Uses of Decision Model
• Future Improvements to Model
    Evacuation Responsiveness by
  Government Organizations (ERGO)
• European Commission Project
• Project Goals
  – Models for public preparation
  – Analytical Models
  – Substantive (real) aids for Evacuation
• Explicit Practitioner Participation
                  ERGO (cont.)

80 interviews, approximately 150 documents, other media data
            The Evacuation Problem
• Evacuation Decision - When do we start evacuating an
1. How long does it take to evacuate?
  –       Oak Ridge Evacuation Modeling System (OREMS)
  –       Configurable Emergency Management & Planning System
          (CEMPS) (Pidd et al., 1996)
  –       Examples from ERGO Countries
      •      Spain, Japan, Iceland
2. When is the risk of a hazard high enough to call for an
  –       Hurricane Evacuation Decisions (Regnier, 2008)
  –       Decision Analysis
             Decision Analysis
• Multi-Attribute Utility Theory (MAUT)
• Evacuation Decision-making Characteristics
  – Multiple, Conflicting Objectives
  – Uncertain Outcomes
• Decision Model Creation
  – Objective Function
  – Probability Function
     Objective Function Assessment
•   What do emergency managers
    care about when faced with
    potentially catastrophic disasters?
•   Elicitation Process
     – Broad range of stakeholder
     – Maximize confidence that all values
       are identified (Bond, 2007)
•   Utility assessment for each
•   Weights created for the
    importance of each objective
•   Identification of Objective Trade-
•   Multi-Attribute Utility Function
    created from preliminary utility
           Probability Assessment
• Hazard Profile
   – Region and hazard specific
   – Casualty rates due to hazard
• Evacuation Behaviour
   – Official orders/information (Burnside, 2007)
   – Visual Clues (Perry, 1983)
• Probability Function for Example Model
   – Storm surge probability taken from Hamburg during ERGO data-
     gathering visits
   – Forecasts at 12 & 9 hours normally distributed with S.D. Of
     50cm and 30cm respectively
   – Public Reaction to Evacuation Orders drawn from limited
Illustrative Evacuation Decision Model
• Identify Risk Thresholds
• Four Evacuation Strategies
   – No Action, Advisory, Mild Evacuation Order, Urgent
     Evacuation Order
   – Strategy chosen affects the percentage of the public that
   – Strategy chosen affects the economic/organizational costs
   – Casualty rates affect the percentage of public that DO NOT
     evacuate & lead to life costs
• Optimal Decisions at 12 & 9 hour forecasts
• Flood defences
   – Dykes at 8 metres
Example Influence Diagram
Results – 12 hour forecast
Results – 9 hour forecast
            Sensitivity Analysis
• Parameters where slight variation in values
  leads to changes in the optimal decision
• Key parameters in Example Evacuation
  Decision Model
  – Objective weight (life costs)
  – Non-evacuee casualty rates
• Represent areas in which the respective
  assessments must be verified
Substantive Benefits of MAUT Process
• Explicit identification of objectives
• Value-focused creation of strategies
• Scenario building
• Quantitative assessment of trade-offs between
• Identification of risk thresholds
• Evaluation of evacuation mitigation policies
• A model based on expert participation
  throughout the process
• MAUT process is appropriate for any decision
  with multiple conflicting objectives and
  – Nuclear Disaster (French, 1996)
  – Anti-Terrorist Analysis (Keeney, 2007)
  – Fire Service Analysis (Swersey, 1982)
• Dependent on participation by decision-
• Application to Evacuation Decisions

To top