Insights on the statistical variability of experimental fire behavior data using airborne-infrared. infrared

Transcription

1 Insights on the statistical variability of experimental fire behavior data using airborne-infrared infrared Douglas J. McRae, Susan G. Conard, Ji-zhong Jin, Anatoly I. Sukhinin, Tom W. Blake, and Galina A. Ivanova Natural Resources Canada Canadian Forest Service Wildland Fire Canada 2010, 6 October 2010, Kitchener, ON.

2 The problem TABLE 4. Fire impact on forest fuels and fire behavior characteristics of the Sharpsand Creek experimental fires in immature jack pine. Fire No b Total surface Fuel consumed (kg/m 2 ) Ground fuels Crown fuels Total fuel Depth of burn (cm) Rate of spread (m/min) Adapted from: Stocks, B.J Fire behavior in immature jack pine. Can. J. For. Res. 17: Frontal fire intensity (kw/m)

3 The 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. Obvious questions: 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?

4 The 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 versus m/min

5 The problem Fire behavior is difficult to quantify in the past because of: 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.).

6 FIRE BEAR Project (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.

7 FIRE BEAR Project (Fire Effects in the Boreal Eurasia Region) Rate of spread: 2.5 m/min Fireline intensity: 620 kw/m Rate of spread: 5.6 m/min Fireline intensity: 5220 kw/m Rate of spread: 4.9 m/min Fireline intensity: 2259 kw/m

8 Boguchany, Russian, June 18, 2002 Plot 1 3:29:11 PM 4 Geo-referencing fires

9 Boguchany, Russian, June 18, 2002 Plot 1 3:29:11 PM 4 Geo-referencing fires

10 Boguchany, Russian, June 18, 2002 Plot 1 3:31:43 PM Ignition line Wind

11 Boguchany, Russian, June 18, 2002 Plot 1 3:34:42 PM

12 Boguchany, Russian, June 18, 2002 Plot 1 3:37:15 PM

13 Boguchany, Russian, June 18, 2002 Plot 1 3:39:28 PM

14 Boguchany, Russian, June 18, 2002 Plot 1 3:39:32 PM

15 Boguchany, Russian, June 18, 2002 Plot 1 3:39:35 PM

16 Boguchany, Russian, June 18, 2002 Plot 1 3:39:35 PM

17 Boguchany, Russian, June 18, 2002 Plot 1 3:41:56 PM

18 Boguchany, Russian, June 18, 2002 Plot 1 3:44:07 PM

19 Boguchany, Russian, June 18, 2002 Plot 1 3:46:39 PM

20 Boguchany, Russian, June 18, 2002 Plot 1 3:49:14 PM

21 Boguchany, Russian, June 18, 2002 Plot 1 3:51:27 PM

22 Boguchany, Russian, June 18, 2002 Plot 1 3:53:45 PM

23 Boguchany, Russian, June 18, 2002 Plot 1 3:56:12 PM

24 Boguchany, Russian, June 18, 2002 Plot 1 3:58:34 PM

25 Boguchany, Russian, June 18, 2002 Plot 1 4:00:52 PM

26 Boguchany, Russian, June 18, 2002 Plot 1 4:03:27 PM

27 Boguchany, Russian, June 18, 2002 Plot 1 4:05:59 PM

28 Boguchany, Russian, June 18, 2002 Plot 1 4:06:05 PM

29 Boguchany, Russian, June 18, 2002 Plot 1 4:08:39 PM

30 Infrared data analysis 10,000 sample points/ha (1-m resolution) Isolines showing firefront locations (seconds)

31 Infrared data analysis Rate of spread (m/min)

32 Rate of spread 3000 All data All data (n = 21434) Number of observations Rate-of-spread classes (m/min)

33 Rate of spread 3000 All data All data (n = 21434) m/min (SE=0.08) Number of observations 1 Sample size = 507 Maximum value = 912 m/min Rate-of-spread classes (m/min)

34 Rate 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.

35 Rate of spread 3000 Data minus outer 10-m perimeter edge effect and erroneous high rates Data of minus spread outer (± 2 standard 10-m perimeter deviations edge of mean) effect m/min (SE=0.11) Number of observations >31 Rate-of-spread classes (m/min) `

36 Rate of spread Data minus outer perimeter edge effect and erroneous high Data rates minus of spread outer (± 210-m standard perimeter deviations edge of effect mean) and rates of spread greater than ± 2 standard deviations 7.9 m/min (SE=0.04) Number of observations >31 Rate-of-spread classes (m/min) `

37 Rate of spread % 27% m/min (SE=0.04) Number of observations Rate-of-spread classes (m/min)

38 Application to Models Rate of spread (m/min) RoS = (0.741 * ISI) R 2 =0.87; SE= % Error bars indicate one standard error from the mean ISI

39 Application to Models Data range for each fire X X Rate of spread (m/min) X X X X X X XX X X XX XX X X ISI

40 Depth of burn cm (SE=0.09) Number of observations DMC= 19 DC = Depth-of-burn classes (cm)

41 Depth of burn 60 Number of observations DMC= 19 DC = cm (SE=0.09) 6.3 cm (SE=0.15) DMC= 35 DC = Depth-of-burn classes (cm)

42 Total fuel consumption N u m b e r o f o b s e rv a t io n s kg/m 2 (SE=0.04) 2.0 kg/m 2 (SE=0.22) Fuel consumption class (kg/m 2 )

43 Total fuel consumption kg/m 2 (SE=0.04) 3.2 kg/m 2 (SE=0.23) Num ber of observations kg/m 2 (SE=0.22) DMC= 51 DC = Fuel consumption class (kg/m 2 )

44 Conclusions 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).

45 Thank you! Wildland Fire Canada 2010, 6 October 2010, Kitchener, ON.

46

47 Depth of burn 60 Number of observations DMC= 19 DC = cm (SE=0.15) DMC= 51 DC = Depth-of-burn classes (cm)

48 Depth of burn 60 Number of observations DMC= 19 DC = cm (SE=0.15) DMC= 51 DC = Depth-of-burn classes (cm)

49 Depth of burn Number of observations Low Moderate Severe 3% 76% 21% 6.3 cm (SE=0.15) DMC= 51 DC = Depth-of-burn classes (cm)

50 Total Fuel consumption Series1 Series2 Series kg/m 2 (SE=0.22)

51 Fuel consumption kg/m 2 (SE=0.04) Series1 Series2 Series

52 Fuel consumption kg/m2 (SE=0.23) Number of observations Fuel consumption class (kg/m2)

53 Results Table 2. Fuel consumption values (plot averages with standard error) observed for each experimental Siberian Scots pine fire Consumption (dry weight) by category (kg/m2) Fire No. Vegetation Dead & Down Litter Forest Floor Total ±0.01 (0-0.09) 0.28±0.10 ( ) 0.16±0.03 ( ) 1.45±0.08 ( ) 1.80±0.16 ( ) ±0.01 (0-0.12) 0.18±0.06 ( ) 0.11±0.02 ( ) 1.36±0.06 ( ) 1.68±0.12 ( ) ±0.04 ( ) 0.04±0.01 ( ) 0.18±0.04 ( ) 0.74±0.04 ( ) 0.93±0.04 ( ) ±0.01 ( ) 0.41±0.17 ( ) 0.30±0.06 ( ) 1.50±0.08 ( ) 2.03±0.22 ( ) ±0.01 ( ) 0.44±0.18 ( ) 0.25±0.25 ( ) 2.45±0.12 ( ) 3.03±0.23 ( ) ±0.02 ( ) 0.30±0.15 ( ) 0.11±0.01 ( ) 1.16±0.05 ( ) 1.50±0.15 ( ) Values in parentheses show the range in consumption values

54 Results Equilibrium (steady-state) fire behavior characteristics (plot averages with standard error) observed for each Siberian experimental Scots pine fire Fire No Depth of burn (cm) 5.6 ± 0.20 ( ) 4.4 ± 0.13 ( ) 3.3 ± 0.09 ( ) 4.6 ± 0.15 ( ) 6.3 ± 0.15 ( ) 4.2±0.10 ( ) Rate of spread (m/min) 7.9 ± 0.04 ( ) 4.9 ± 0.35 ( ) 2.5 ± 0.25 ( ) 2.0 ± 0.34 ( ) 5.6 ± 0.07 ( ) 6.5±0.05 ( ) Values in parentheses show the range in values Fireline intensity (kw/m) 2259 ± 189 ( ) 2259 ± 189 ( ) 620 ± 34 ( ) 1214 ± 149 ( ) 5220 ± 434 ( ) 2790 ± 341 ( ) Total fire intensity (kj/m2) ± 2316 ( ) ± 2316 ( ) ± 828 ( ) ± 4474 ( ) ± 4646 ( ) ± 3145 ( )