Forest Fire Detection Economics

1
Forest Fire Detection Economics

• David L. Martell
• Faculty of Forestry University of Toronto

Robert S. McAlpine Ontario Ministry of Natural
Resources Fire Detection Workshop Hinton,
Alberta March 25, 2003
2
Overview

• Basic Concepts
• Detection Methods
• Detection Patrol Routing Problem
• Detection/Initial Attack System Model
• Conclusion

3
Life Cycle of a Forest Fire
4
Value of Detection System

• Need to assess detection system from an overall
  system perspective
• Detection system objective is to find fires such
  that they can be controlled at reasonable cost
  and impact
• Value of the detection system is the net
  reduction in total cost plus loss

5
Detection Considerations

• Value of the resource protected
• Visibility
• Probability of a fire occurring
• Expectations of fire behavior
• Potential for fire spread
• Coverage by unorganized detection

6
Detection Probability

• Partition the protected area into many small cells

7
Detection Methods

• Lookout Towers

8
Lookout Towers

• Strategic Decisions
• 1. How many towers?
• 2. What locations?

9
Fire Lookout Tower Location Models

• Partition protected
• area into a large
• number of small
• rectangular cells

• Identify potentially good tower sites

10
Tower Location Models

• 1. Minimize the number (or cost) of towers
  required
• to cover all cells
• - may require double coverage for triangulation
• 2. Maximize the number of cells seen by a
  specified number of towers
• - use potential damage estimates to weight cells

11
Aircraft

• Strategic Decisions
• 1. How many aircraft?
• 2. What hours?
• 3. What type?

12
Aircraft

• Tactical Decisions
• 1. When to dispatch
• 2. Where to fly

13
Detection Patrol Routing Problem

• Partition the protected area into a large number
  of small rectangular cells
• Predict the expected number of fires or
  probability of fires in each cell
• Use vegetation, fire weather and values at risk
  map to identify potentially critical cells that
  must be visited
• Develop the best patrol route(s) to visit all
  the cells that must be visited

14
Simple Detection Patrol Routing Problem

• 1. Should you dispatch a
• detection patrol?
• 2. If you dispatch
• detection patrol, at
• what time?

15
Simplifying Assumptions

• 1) Fire Started at 0800 hours
• 2) Forward Rate of Spread of the Fire 36 m/h
• 3) Fire Damage 200 per hectare burned up
  until the time of detection

16
Fire Loss Assuming Fire is Circular
17
Detection Probability Function
18
Detection Patrol Routing Problem

• Suppose you look at 1000
• Expected Cost (1,000 320 )0.2 (find at
  1000)
• Loss (1,000 11,720)(1-0.2) (pu
  blic at 2000)
• 10,440

19
Detection Patrol Routing Problem
20
Towers vs Aircraft

• Aircraft
• flexible
• inexpensive
• intermittent surveillance

• Towers
• fixed
• expensive
• constant surveillance

• Use in low value forest with small detection
  budget

• Use in high value forest if have a large
  detection budget

21
Measures of Detection System Effectiveness
Cost per unit area protected
(minimize with NO effort)
Cost per fire detected
(let the public find them all)
Hours flown per fire detected
(minimize with NO effort)
Percent of fires detected by airborne observers
(compete
with the public)
Average size at detection
(ignores travel time, spread
rate, etc.)
Find fires so you can put them out at
reasonable cost and damage (detection cost,
suppression cost, fire damage)
22
Detection/Initial Attack System Model

• Model that predicts the final sizes of historical
  fires given
• Actual fire report record
• Actual fuel and fire weather information
• Suppression by a perfect hypothetical initial
  attack crew
• Model provides an objective relative measure of
  how well the detection system worked on a single
  fire or collection of fires
• Does not indicate how well the system should
  perform

23
Fire Behaviour

• Fire Shape wind driven ellipse model
• Fire Growth FBP to predict area, perimeter
• Fire declared held when the fire line constructed
  equals 50 of the fire perimeter

24
Fire Suppression

• Rate of Line Construction
• RLC B0 B1 FI by fuel type

25
Simple Containment Model

• Hypothetical Final Size
• Predicted final size of a fire given the fire
  conditions and a hypothetical perfect initial
  attack crew that is dispatched as soon as the
  fire is reported
• Perfect Final Size
• Final size of a fire given detection as soon
  as the fire starts, and a hypothetical perfect
  initial attack crew that is dispatched as soon as
  the fire starts
• Detection Loss HF - PF (ha per fire)

26
Average Annual Results (1980 - 85)

• Year to year comparisons (e.g., before and after
  detection program changes) are valid
• Direct comparison between regions questionable
  (values at risk and fire loads differ)

27
How Well Should the Detection System Perform?

• Depends Upon
• Values at risk
• Number of fires per year
• Fire behaviour
• Public detection system
• Detection budget

28
Thank YouDiscussion