Examining fire spread patterns across a managed forest-land mosaic in northern Wisconsin, USA using a modeling approach

1
Examining fire spread patterns across a managed
forest-land mosaic in northern Wisconsin, USA
using a modeling approach
Jacob LaCroix, Daolan Zheng, Soung-Ryoul Ryu and
Jiquan Chen
2
Introduction

• Chequamegon National Forest, WI
• With FARSITE model (Finney 1994)
• Look at fire spread across landscape using
  standard fuel modules (Anderson 1982)
• Support research to other efforts of fire team,
  basic questions of fire spread

3
Objectives

• Examine 4 factors that influence the rate of fire
  spread on a landscape
• Interrelated at the landscape level

4
4 Factors to Investigate

• Influence/strength of fire behavior
• Can we make simplifying predictions for FML, FLI,
  HPA on ROS
• Demonstrate patch differences
• Is the fire behaving differently in each patch?
• Landscape Structure
• Are burned areas different for fires started in
  12 different locations?
• Influence/strength of 3 weather factors
• Wind, Rain, Temperature, influence on area of
  fire spread

5
Methods

• Gather data from FARSITE simulation at CNF, using
  2001 landscape and 2002, weather from the MHW
  met-station
• Place fires on the landscape in up to 12
  different locations, for 15 and 27 days long, 24
  hour burning period, same weather and starting
  moistures in the fuels in all cases
• Except last weather investigation
• For a modeler using Andersons (1982) fuels which
  FARSITE was designed to use without adjustment

6
Methods 1

• To determine the relative influences of 3 fire
  behavioral characteristics on the rate of spread
  to see if we can make simpler models
• Use Path Coefficient Analysis, to create a model
  and a table of direct and indirect affects for
  each variable, FLI, FML, HPA on ROS.
• These are the same 4 outputs as in a Fire
  Behavior Chart

7
(No Transcript)
8
What is Path Analysis

• Combines 3 statistical tools
• Linear Correlation
• Linear Regression
• Model Building
• Uses standardized Betas so variables in model are
  comparable
• Good tool for variables that are closely related,
  complex systems

9
Preliminary Results
10
Preliminary Results
Table 1. Path coefficient analysis of the
relationships between fire behavior
characteristics and rate of fire spread using
four standard fuel types in a FARSITE fire
simulation.
11
Methods 2

• To determine if fire is behaving differently in
  each patch/fuel type
• Use 4 separate, 1-way ANOVA tests, 1 for each
  variable of fire behavior, HPA, FLI, FML, ROS
• Verification of the FARSITE model for my own
  proof

12
Preliminary Results Table 2. Data collected from
fire spread simulation 27 day fire, samples
taken through- out the course of the fire
for each fuel category, 6 times.
13
Preliminary Results

• 1-way ANOVA analysis of the 4 patch types
• Hypothesis is that there is no significant
  difference between the HPA, FLI, FML, ROS among
  the 4 types of fuel/patches
• P values are HPA0.0001, FLI0.0001, FLM0.0001,
  ROS0.0007
• Every variable is significantly different between
  the 4 types of fuel/patches

14
Methods 3

• To determine if landscape structure affects how
  far fires spread, (area after the burn)
• Place 12 fires, started in different location on
  the landscape
• Compare the areas in Hectares of fires for
  significance
• Brainstorm ideas for statistical tests are to use
  t-test, comparison of slopes of a line, ANOVA,
  chi-squared, others?

15
Preliminary Results

• Earlier runs suggest that there will be
  difference in area in different starting points

16
Methods 4

• To determine the relative influences of 3 weather
  characteristics on the rate of spread of fire to
  see if we can make simpler models
• Separate out the affects of each component on
  area of fire in FARSITE first
• 3x3x327 landscape conditions
• Use Path Coefficient Analysis, to create a model
  and a table of direct and indirect affects for
  each variable, Wind, Rain and Temperature, on
  area of spread.
• Brainstorm, ideas about other statistical test to
  accomplish this, e.g. General Linear Model (GLM),
  ANOVA, others?

17
Weather Variables

• Use historic data from Ashland, WI to find ranges
  that are logical for Hi and Lo Wind, Rain, and
  Temperatures from the past 100 years
• http//mcc.sws.uiuc.edu/Precip/WI/470349_psum.html
  
• Change the ASCII file inputs in FARSITE
• run simulations 27x12324

18
Preliminary Results

• Running the model suggests differences in area in
  with different weather factors

19
FARSITE Outputs

• Show the 2001 landscape
• With Fire in HPA
• With Fire in ROS
• With Fire in FLI
• With Fire in FML

20
Figure 1. Shows the Chequamegon National Forest
Landscape with the 4 fuel types in different
colors. Brown brush. Green closed timber
litter no under-story, closed canopy, pine
needle and leaf duff. Red timber litter and
Under story hardwood leaves, young trees.
Yellow light logging slash.
21
Figure 2. Shows the Chequamegon National Forest
Landscape with the 4 fuel types in different
colors. Brown brush. Green closed timber
litter pine needle and leaf duff, closed
canopy. Red timber litter and under story,
hardwood leaves, young trees. Yellow light
logging slash. Also, the fire is placed on the
landscape with Heat per Area represented. Blue
Cool heat. Red Medium heat. Black High
heat.
22
Figure 3. Shows the Chequamegon National Forest
Landscape with the 4 fuel types in different
colors. Brown brush. Green closed timber
litter pine needle and leaf duff, closed
canopy. Red timber litter and under story
hardwood leaves and young trees. Yellow light
logging slash. Also, the fire is placed on the
landscape with Rate of Spread represented. Blue
Slow rate. Red Medium rate.
23
Figure 4. Shows the Chequamegon National Forest
Landscape with the 4 fuel types in different
colors. Brown brush. Green closed timber
litter pine needle and leaf duff, closed
canopy. Red timber litter and under story
hardwood leaves and young trees. Yellow light
logging slash. Also, the fire is placed on the
landscape with Fire Line Intensity represented.
Blue Slow rate. Red Medium rate.
24
Figure 5. Shows the Chequamegon National Forest
Landscape with the 4 fuel types in different
colors. Brown brush. Green closed timber
litter pine needle and leaf duff, closed
canopy. Red timber litter and under story
hardwood leaves and young trees. Yellow light
logging slash. Also, the fire is placed on the
landscape with Flame Length represented. Blue
Slow rate. Red Medium rate.
25
Acknowledgements

• This study is supported by a grant from the JFSP
• Co-authors and the LEES lab

26
Questions

• Help determining if these analysis are
  appropriate
• The best statistical tests to employ
• In landscape area objective and weather analysis
  objective