Insights on the Insights on the statistical variability statistical variability of experimental fire of experimental fire behavior data using behavior data using airborneairborne--infraredinfrared
Douglas J. McRae, Susan G. Douglas J. McRae, Susan G. ConardConard, , JiJi--zhongzhong Jin, Anatoly I. Jin, Anatoly I. SukhininSukhinin, , Tom W. Blake, and Galina A. Tom W. Blake, and Galina A. IvanovaIvanova
Natural Resources Canada
Canadian Forest Service
Wildland Fire Canada 2010, 6 October 2010, Kitchener, ON.
Adapted from: Stocks, B.J. 1987. Fire behavior in immature jack pine.
Can. J. For. Res. 17: 80–86.
The problemThe problem
TABLE 4. Fire impact on forest fuels and fire behavior characteristics of the SharpsandCreek experimental fires in immature jack pine.
2910.663.011.470.000.730.7418
2599027.304.643.361.111.310.9414
1713620.164.633.001.041.300.6612
4090349.443.842.921.400.970.5511b
5992.101.910.950.000.420.537
1078514.643.742.601.270.940.395
471710.742.011.550.890.440.222
Total fuel
Crown fuels
Ground fuels
Total surface
Frontal fire
intensity (kW/m)
Rateof
spread (m/min)
Depth ofburn (cm)
Fuel consumed (kg/m2)
Fire No.
The problemThe problem
Assumption:
Values for fire behavior databases are given as though they are a ‘magical’constant for a particular fuel type and burning condition rather than an average with statistical ranges associated with it.
1. How accurate are these averages?
2. What is the standard deviation/standard error level of each average?
3. What is the absolute range in these values?
Obvious questions:
The problemThe problem
Reason for questions:
1. Suppression activities – Concerns regarding rate of spread/intensity values: is it safe to send fire crews to work on a firefront?
Rate of spread average of 6.5 m/min
5.0 - 8.0 versus 5.0 – 15.0 m/min
• The lack of adequate sampling due to field expenses and expense of
monitoring equipment.• Equipment failures.• Reliance on visual observations based on fixed-point measurements.• Research personnel safety concerns.• Mother nature never cooperates (e.g., wind lulls and gusts, wind direction
changes, etc.).
Fire behavior is difficult to quantify in the past because of:
The problemThe problem
FIRE BEAR ProjectFIRE BEAR Project(Fire Effects in the Boreal Eurasia Region)(Fire Effects in the Boreal Eurasia Region)
• To better understand fire in central Siberia, the FIRE BEAR Project was created as a forest fire research study to provide answers to basic questions on fire management.
• Replicated 200 x 200-m experimental burn plots on Scots pine (Pinus sylvestris) / lichen (Cladonia sp.) / feather moss (Pleurozeum schreberi) forest sites.
• Fuel and fire behavior on these fires was quantified.
FIRE BEAR ProjectFIRE BEAR Project(Fire Effects in the Boreal Eurasia Region)(Fire Effects in the Boreal Eurasia Region)
Rate of spread: 2.5 m/minFireline intensity: 620 kW/m Rate of spread: 4.9 m/min
Fireline intensity: 2259 kW/m
Rate of spread: 5.6 m/minFireline intensity: 5220 kW/m
Boguchany, Russian, June 18, 2002
Plot 1 3:29:11 PM
1000
900
800
700
600
500
400
300
200
100
4 Geo-referencing fires
Boguchany, Russian, June 18, 2002
Plot 1 3:29:11 PM
1000
900
800
700
600
500
400
300
200
100
4 Geo-referencing fires
Boguchany, Russian, June 18, 2002
Plot 1 3:31:43 PM
1000
900
800
700
600
500
400
300
200
100
Ignition line
Wind
Boguchany, Russian, June 18, 2002
Plot 1 3:34:42 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:37:15 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:39:28 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:39:32 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:39:35 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:39:35 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:41:56 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:44:07 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:46:39 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:49:14 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:51:27 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:53:45 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:56:12 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 3:58:34 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 4:00:52 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 4:03:27 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 4:05:59 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 4:06:05 PM
1000
900
800
700
600
500
400
300
200
100
Boguchany, Russian, June 18, 2002
Plot 1 4:08:39 PM
1000
900
800
700
600
500
400
300
200
100
Isolines showing firefront locations
(seconds)
Infrared data analysis
10,000 sample points/ha
(1-m resolution)
Rate of spread(m/min)
Infrared data analysis
All data
Rate of spreadRate of spread
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Rate-of-spread classes (m/min)
Nu
mb
er
of
ob
serv
ati
on
s
≥31
All data (n = 21434)
All data
Rate of spreadRate of spread
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Rate-of-spread classes (m/min)
Nu
mb
er
of
ob
serv
ati
on
s
8.9 m/min(SE=0.08)
All data (n = 21434)
Sample size = 507
Maximum value = 912 m/min
≥31
Rate of spreadRate of spread
Variability in rates of spread can be caused by:
• Differences in fuel structure.• Differences in soil (ground fuel) moisture.• Gusts and lulls in wind speed.• Changes in wind direction.• Channeling and acceleration effect on wind.• Junction zone effects.• Edge effect of experimental plot.• Impact of tree density on solar radiation and fuel dryness.• Spotting.• Analysis problems.
Data minus outer 10-m perimeter edge effect and erroneous high
rates of spread (± 2 standard deviations of mean)
>31`
Rate of spreadRate of spread
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Rate-of-spread classes (m/min)
Nu
mb
er
of
ob
serv
ati
on
s
Data minus outer 10-m perimeter edge effect
9.0 m/min(SE=0.11)
≥31
Data minus outer 10-m perimeter edge effect and erroneous high
rates of spread (± 2 standard deviations of mean)
>31`
Rate of spreadRate of spread
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Rate-of-spread classes (m/min)
Nu
mb
er
of
ob
serv
ati
on
s
Data minus outer 10-m perimeter edge effect and rates of
spread greater than ± 2 standard deviations
7.9 m/min(SE=0.04)
≥31
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Rate-of-spread classes (m/min)
Nu
mb
er
of
ob
serv
ati
on
s
Rate of spreadRate of spread
63% 27%
7.9 m/min(SE=0.04)
ISI
0 2 4 6 8 10
Rate
of
sp
rea
d (
m/m
in)
0
1
2
3
4
5
6
7
8
9
10
Application to ModelsApplication to Models
RoS = 1.008 + (0.741 * ISI)R2=0.87; SE=0.90
95%
Error bars indicate one standard error from the mean
ISI
0 2 4 6 8 10
Ra
te o
f s
pre
ad
(m
/min
)
0
5
10
15
20
25
Application to ModelsApplication to Models
Data range for each fire
X
X
X
X
X
X
X
X
X
X
X
XX XX
X
XXX
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth-of-burn classes (cm)
Nu
mb
er
of
ob
serv
ati
on
s
0 21 3 1554 9876 13 14121110
Depth of burnDepth of burn
DMC= 19DC = 189
3.3 cm(SE=0.09)
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth-of-burn classes (cm)
Nu
mb
er
of
ob
serv
ati
on
s
0 21 3 1554 9876 13 14121110
6.3 cm(SE=0.15)
Depth of burnDepth of burn
DMC= 19DC = 189
DMC= 35DC = 399
3.3 cm(SE=0.09)
0
5
10
15
20
25
30
1 2 3 4 5 6 7 8 9 10 11 12 13
Fuel consumption class (kg/m 2)
Nu
mb
er
of
ob
se
rva
tio
ns
1.0 kg/m2
(SE=0.04)
Total fuel consumptionTotal fuel consumption
2.0 kg/m2
(SE=0.22)
0 21 3 54 9876 121110
0
5
10
15
20
25
30
1 2 3 4 5 6 7 8 9 10 11 12 13
Fuel consumption class (kg/m 2)
Nu
mb
er
of
ob
se
rva
tio
ns
3.2 kg/m2
(SE=0.23)
DMC= 51DC = 393
Total fuel consumptionTotal fuel consumption
1.0 kg/m2
(SE=0.04)
2.0 kg/m2
(SE=0.22)
0 21 3 54 9876 121110
ConclusionsConclusions
Fire behavior
• Fire behavior is a highly variable phenomena at the microsite level (e.g., 1- m pixel).
• Due to the lack of statistical data in the past, most current models do not indicate the actual ranges of fire behavior.
• Remote sensing using infrared cameras allows for multiple sampling to take place, which allows for adequate sample numbers to allow for statistical analysis.
• For fire crew safety, realize that there is a range of possible values around any average!
• Other applications for infrared monitoring (e.g., fuel consumption, carbon emission).
Thank you!
Wildland Fire Canada 2010, 6 October 2010, Kitchener, ON.
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth-of-burn classes (cm)
Nu
mb
er
of
ob
serv
ati
on
s
0 21 3 1554 9876 13 14121110
6.3 cm
(SE=0.15)
Depth of burnDepth of burn
DMC= 19DC = 189
DMC= 51
DC = 393
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth-of-burn classes (cm)
Nu
mb
er
of
ob
serv
ati
on
s
0 21 3 1554 9876 13 14121110
6.3 cm
(SE=0.15)
Depth of burnDepth of burn
DMC= 19DC = 189
DMC= 51
DC = 393
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth-of-burn classes (cm)
Nu
mb
er
of
ob
serv
ati
on
s
0 21 3 1554 9876 13 14121110
Depth of burnDepth of burn
Low Moderate Severe
3% 76% 21%
DMC= 51
DC = 393
6.3 cm
(SE=0.15)
71 22.5 90 52 52
50 23.5 70 44 44
52 24.5 55 35 35
44 25.5 44 27 27
35 26.5 50 35 35
27 27.5 41 30 30
35 28.5 42 32 32
30 29.5 45 32 32
32 30.5 507 377 46
32 Total 21434 14015 13684
46
0
2
4
6
8
10
12
14
16
18
1 2 3 4 5 6 7 8 9 10 11 12 13
Series1
Series2
Series3
2.0 kg/m2
(SE=0.22)
Fuel consumptionFuel consumption
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Series1
Series2
Series3
1.0 kg/m2
(SE=0.04)
Fuel consumptionFuel consumption
0
5
10
15
20
25
30
1 2 3 4 5 6 7 8 9 10 11 12
Fuel consumption class (kg/m2)
Nu
mb
er
of
ob
se
rva
tio
ns
3.2 kg/m2
(SE=0.23)
Fuel consumptionFuel consumption
ResultsResults
Values in parentheses show the range in consumption values
Table 2. Fuel consumption values (plot averages with standard error) observed for each experimental Siberian Scots pine fire
1.50±0.15
(0.53-7.79)
1.16±0.05
(0.12-3.65)
0.11±0.01
(0.01-0.27)
0.30±0.15
(0.01-7.33)
0.02±0.02
(0.00-0.13)
20
3.03±0.23
(1.17-9.79)
2.45±0.12
(0.09-10.36)
0.25±0.25
(0.07-0.51)
0.44±0.18
(0.00-7.55)
0.07±0.01
(0.00-0.13)
14
2.03±0.22
(0.26-10.47)
1.50±0.08
(0.02-5.53)
0.30±0.06
(0.02-1.50)
0.41±0.17
(0.00-7.43)
0.07±0.01
(0.00-0.16)
13
0.93±0.04
(0.49-12.43)
0.74±0.04
(0.00-2.00)
0.18±0.04
(0.07-0.37)
0.04±0.01
(0.00-0.30)
0.00±0.04
(0.00-0.12)
3
1.68±0.12
(0.32-4.32)
1.36±0.06
(00.0-4.90)
0.11±0.02
(0.00-0.28)
0.18±0.06
(0.00-2.28)
0.03±0.01
(0-0.12)
2
1.80±0.16
(0.34-5.23)
1.45±0.08
(0.03-4.63)
0.16±0.03
(0.03-0.55)
0.28±0.10
(0.02-3.28)
0.03±0.01
(0-0.09)
1
Total
Forest
FloorLitterDead & DownVegetation
Consumption (dry weight) by category (kg/m2)
Fire
No.
ResultsResults
Equilibrium (steady-state) fire behavior characteristics (plot averages with standard error) observed for each Siberian experimental Scots pine fire
25 757 ± 3145
(6629-157376)
2790 ± 341
(718-17049)
6.5±0.05
(2.4-16.5)
4.2±0.10
(1.7-8.3)20
55 938 ± 4646
(19344-196851)
5220 ± 434
(805-18372)
5.6 ± 0.07
(3.6-14.8)
6.3 ± 0.15
(1.2-15.0)14
36 450 ± 4474
(1564-209443)
1214 ± 149
(52-6981)
2.0 ± 0.34
(0.8-5.7)
4.6 ± 0.15
(0.9-10.5)13
14 878 ± 828
(6230-24920)
620 ± 34
(260-1038)
2.5 ± 0.25
(1.1-9.4)
3.3 ± 0.09
(1.1-5.9)3
27 662 ± 2316
(25235-83261)
2259 ± 189
(207-6800)
4.9 ± 0.35
(1.2-9.9)
4.4 ± 0.13
(1.1-9.8)2
27 662 ± 2316
(25235-83261)
2259 ± 189
(207-6800)7.9 ± 0.04
(1.3-17.9)
5.6 ± 0.20
(0.5-9.4)1
Total fire intensity
(kJ/m2)
Firelineintensity
(kW/m)
Rate of spread
(m/min)
Depth of burn
(cm)
Fire
No.
Values in parentheses show the range in values