An Evaluation of National Fire Danger Rating System Components for Use in Prescribed Fire Decisions On the National Forests of Texas Terry G. Harris Fuels Specialist USDA Forest Service National Forests of Texas 415 South 1 st Street Suite 110 Lufkin, Texas 75901 Technical Fire Management - 12 Washington Institute HARRIS POLK RUSK TYLER BRAZORIA LIBERTY JASPER HOUSTON HARDIN SHELBY NEWTON PANOLA ANDERSON CHEROKEE WALKER TRINITY ANGELINA JEFFERSON SABINE CHAMBERS MONTGOMERY NACOGDOCHES GALVESTON SAN JACINTO ORANGE SAN AUGUSTINE . - , 45 ( / 69 ( / 69 ( / 59 ( / 59 " ! 19 " ! 103 " ! 7 " ! 21 " ! 87 " ! 87 " ! 147 " ! 7 " ! 21
28
Embed
An Evaluation of National Fire Danger Rating System ...€¦ · implementation of any prescribed fires. The NFGT prescribed fire parameters are summarized in Table 1. The National
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
An Evaluation of National Fire Danger Rating System Components for Use in Prescribed Fire Decisions
On the National Forests of Texas
Terry G. Harris Fuels Specialist
USDA Forest Service National Forests of Texas
415 South 1st Street Suite 110
Lufkin, Texas 75901
T e c h nical Fire Management - 12Washington Institute
NWS forecast of 15-20 MPH is accepted, includes gusts. ----------------------------- Texas State Air Quality Regulations.
TRANSPORT WIND SPEED (meters per second)
RO Sliding scale or State Regulations. ------------------------------------- 4 mps minimum
R8-5144 Exhibit 03. ------------------------------- Texas State Air Quality Regulations
MIXING HEIGHT (meters above ground level)
RO Sliding scale or State Regulations. ------------------------------------- 500 meters/agl minimum
R8-5144 Exhibit 03. ------------------------------- Texas State Air Quality Regulations.
SMOKE DISPERSION INDEX
RO >= 21 dispersion index or more restrictive State requirements
NWS does not provide Dispersion Index in Texas, State does not use. State regulations are more restrictive.
NFDRS: BURNING INDEX (BI)
RO ----------- Forest
90th percentile of Forest selected index, or indices. --------------------------------------- Forests: 65 BI Grasslands: 40 BI
---------------------------------------------- Exceptions must be approved by Regional Office.
PROBABILITY OF IGNITION. NFDRS (IC)
RO -------------- Forest
Forest to develop. ------------------------------------------ 50% maximum.
KBDI RO -------------- Forest
Forest to develop. ------------------------------------------ 550 maximum unless burn unit has received at least ¼ inch of rain within the previous 4 days.
---------------------------------------------- Exceptions must be approved by Fire Staff Office or Forest FMO.
DAYS SINCE RAIN
RO/Forest See KBDI above.
AMOUNT (inches) RO/Forest See KBDI above.
4
TABLE 2. National Fire Danger Rating System inputs.
NFDRS Inputs Definition
1-hour fuel moisture The moisture content of fuels consisting of dead herbaceous plants and woody vegetation.
10-hour fuel moisture The moisture content of dead woody fuels consisting of one-fourth to one-inch in diameter.
100-hour fuel moisture The moisture content of dead roundwood one to three inches in diameter.
1000-hour fuel moisture The moisture content of dead roundwood three to eight inches in diameter
Herbaceous fuel moisture The content of water of a live herbaceous plant expressed in percent.
Keetch-Byram Drought Index A number that represents the effect of evaporation and precipitation in cumulative moisture to approximately eight inches into the duff layer and upper soil layers.
Relative humidity The ratio of the amount of water vapor in the air necessary to saturate expressed as a percentage.
Temperature Temperature of the air
Wind For NFDRS calculations wind is measured at 20 feet above the ground or the average height of the vegetative cover.
Woody fuels moisture The content of water of live woody vegetation
5
TABLE 3. National Fire Danger Rating System outputs
NFDRS Outputs Definition
Burning Index A number related to the contribution of fire behavior to the effort of containing a fire. Scale is open-ended; thus it has no upper limit.
Ignition Component Rating of the probability that a firebrand will cause fire requiring suppression action. Scale of 0 to 100.
Energy Release Component A number related to the available energy (BTU) per unit area (square foot) within the flaming front at the head of a fire. Scale is open-ended; thus it has no upper limit
Spread Component A prediction of the rate of spread of a head fire. Scale is open-ended; thus it has no upper limit.
SCOPE
The scope of this analysis is limited to the relationships of NFDRS indices to
prescribed fire results on the National Forests in Texas. The findings may be
applicable to prescribed fires in similar vegetation types across the southeastern
coastal plains.
PROBLEM STATEMENT The National Fire Danger Rating System (NFDRS) is not designed to predict
behavior of an individual fire. However, inputs used to calculate its outputs are also
fire behavior factors, which suggest that NFDRS indices could be used as prescribed
fire parameters in the GO-NO/GO decision-making process for the NFGT. These
same inputs and outputs could be used to predict prescribed fire results. No other
studies of this have looked at using NFDRS indices as prescribed fire parameters on
6
the NFGT. The findings of this project could provide the NFGT with better
information available to conduct our program of work in prescribed fire.
GOAL STATEMENT
The goal of this analysis is to provide management the most effective and efficient
information available for use in the prescribed fire GO/NO-GO decision-making
process.
OBJECTIVES This project tests the null hypothesis of no difference between prescribed fire results
and NFDRS indices to determine if a significant relationship exists between the two.
Prescribed fire results for our statistical analysis were classified into three classes:
burns deemed successful, which met management objectives, burns where the fire
intensity was too cool, and burns where the fire intensity was too hot. The fires
classified as too cool or too hot did not meet management objectives. The statistical
method we used to test the null hypothesis was a one-way analysis of variance. (A
one-way analysis of variance is considered the appropriate statistical method for this
data, Bob Loveless, personal communication, January, 2005).
METHODS
Initially 130 prescribed fires over a 16 year period were separated into three classes:
35 fires in class 1 in which fire intensity was too cool to meet management
objectives, 30 fires in class 2 in which fire intensity was to hot to meet management
objectives and 65 fires in class 3 (objectives met) where management objectives
were met. The classification of each fire was based on post-burn evaluations
conducted by the responsible Burn Boss for each prescribed fire unit. Actual records
7
varied greatly. Most of the fires were several hundred acres or greater in size where
fuels, weather, time of day and other factors varied greatly over the burn area. The
13:00 hour fire weather observations for each day a prescribed fire occurred were
retrieved through the Weather Information Management System (WIMS) from the
following remote automated weather stations (RAWS) on the Forests. WIMS is a
web-based application used to collect, store, and manage current weather
information, as well as providing access to historical weather information. RAWS
are weather stations that tracks and stores weather observations. Our weather
observations were retrieved from the following RAWS: Conroe (415109), Sabine
South (413701), Lufkin (413509), Coldspring (414201), Sabine North (412901) and
Ratcliff (413302). The weather data from the nearest RAWS that existed at the time
was assigned to each prescribed fire. The following weather observations were
retrieved by WIMS for our analysis: temperature, relative humidity, wind, 1 – hour
Keetch-Byram Drought Index. These indices (variables) were analyzed to determine
if a statistically significant difference exists in regards to prescribed fire results
(classes). The key indicator to determine if a significant difference exists in step one
is the P-value. P-value must be less than or equal to alpha, 0.05 to show a significant
relationship between our variables and prescribed fire results (classes). Any variable
with a P-value greater than alpha was analyzed no further. A pair-wise comparison
was used in step two of statistical analysis to identify where the differences occurred
between prescribed fire classes (Rx classes). The results of this analysis are
displayed in Tables 4-12.
10
Assumptions The following assumptions have to be considered in this project: data used for the
analysis of variance analysis is normally distributed, National Fire Danger Rating
System process has not changed, and the weather conditions on the burn site did not
change from the weather observations collected at 1300 and the subjective method
used to classify prescribed fire results into different classes.
11
Analysis Results
TABLE 4. ANOVA and pair-wise comparisons for energy release components (ERC). ANOVA Source P-value
Class of Fires 0.0184
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between energy release component and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
3 39.954 A
2 37.700 AB
1 32.571 B
ERC can be used to distinguish class 1 (fire intensity to cool) from class 2 (fire intensity to hot) and class 3 (objectives met).
.
12
TABLE 5. ANOVA and pair-wise comparisons for ignition component (IC). ANOVA Source P-value
Class of Fires
0.0078
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between ignition component and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
2 21.833 A
3 21.077 A
1 14.000 B
IC can be used to distinguish class 1 (fire intensity to cool) from class 2 (fire intensity to hot) and class 3 (objectives met).
13
TABLE 6. ANOVA and pair-wise comparisons for Keetch-Byram Drought Index (KBDI). ANOVA Source P-value
Class of Fires
0.0032
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between Keetch-Byram Drought Index and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
2 215.80 B
3 150.80 A
1 140.51 A
KBDI can be used to distinguish class 2 (fire intensity to hot) from class 1 (fire intensity to cool) from) and class 3 (objectives met).
14
TABLE 7. ANOVA and pair-wise comparisons for 1-hour fuel moisture (1-FM). ANOVA Source P-value
Class of Fires
0.0043
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between 1-hour fuel moisture and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
1 9.4857 A
3 7.6462 B
2 7.3000 B
1-FM can be used to distinguish class 1 (fire intensity to cool) from class 2 (fire intensity to hot) and class 3 (objectives met).
15
TABLE 8. ANOVA and pair-wise comparisons for 10-hour fuel moisture (10-FM). ANOVA
Source P-value
Class of Fires
0.0057
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between 10-hour fuel moisture and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
1 11.171 A
3 9.6000 B
2 7.3000 C
10-FM can be used to distinguish each class from one another.
16
TABLE 9. ANOVA and pair-wise comparisons for 100-hour fuel moisture (100-FM). ANOVA
Source P-value
Class of Fires
0.0026
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between 100-hour fuel moisture and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
1 18.457 A
3 17.431 B
2 17.033 B
100-FM can be used to distinguish class 1 (fire intensity to cool) from class 2 (fire intensity to hot) and class 3 (successful burns).
17
TABLE 10. ANOVA and pair-wise comparisons for wind. ANOVA
Source P-value
Class of Fires
0.0261
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between wind and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
2 8.1667 A
1 6.6857 AB
3 6.4462 B
All-pairs comparison test reveals a significant difference between class 2 and class 3 prescribed fires. Wind can be used to distinguish class 2 (fire intensity to hot) from class 3 (objectives met).
18
TABLE 11. ANOVA and pair-wise comparisons for relative humidity (RH). ANOVA
Source P-value
Class of Fires
0.0027
The results of our ANOVA show a p-value which less than alpha (0.05) therefore a significant relationship does exist between relative humidity and prescribed fire classes. Pair-wise comparison. Class Mean Homogenous
Groups
1 56.200 A
2 47.554 B
3 43.633 B
RH can be used to distinguish class 1 (fire intensity to cool) from class 2 (fire intensity to hot) and class 3 (successful burns).
19
The results of our statistical analysis did not show a significant relationship for the variables summarized in Table 12. The P-value for these is equal to greater than alpha (0.05). TABLE 12. Variables with no significant relationship to prescribed fire results. Variable Source P-value
Spread Component
Class of Fires 0.3775
Burning Index Class of Fires 0.1428
Temperature Class of Fires 0.1302
1000-Fuel Moisture. Class of Fires 0.1907
Herbaceous Fuel Moisture Class of Fires 0.2037
Woody Fuel Moisture Class of Fires 0.2575
20
Discussion
Based on this analysis the following indices could be useful in distinguishing
between prescribed fires where the fire intensity was too cool (class 1) too hot (class
2), or met management objectives(class 3): energy release component, ignition
moisture, 100-fuel moisture, wind speed, and relative humidity. Several indices were
found not to be useful in distinguishing between any classes of our prescribed fires:
burning index, temperature, 1000-fuel moisture, herbaceous fuel moisture, and
woody fuel moisture. The only index that could be useful in distinguishing between
all three classes is 10-hour fuel moisture. Therefore, 10-hour fuel moisture is the
only indices that should be considered in the GO/NO-GO decision process to
distinguish between classes of prescribed fires. There are several limitations that
could have resulted in inaccuracies for our statistical analysis. The subjective method
used to classify our prescribed burns into different classes, changes in on-site
weather from the 1300 weather observations, data used for our analysis is normally
distributed and NFDRS process has not changed are all limitations for this project.
Based on this analysis I would recommend using 10-fuel moisture to distinguish
between the three classes of prescribed fires. Additional analysis would be necessary
before implementing this.
21
References
NWCG. 2003. Gaining Intermediate National Fire Danger Rating System S-491 Student Workbook. Schlobohm, Paul and others. NWCG 1982. Aids to Determing Fuels Models For Estimating Fire Behavior. Anderson Hal E. USDA Forest Service. 2002. Fire Family Plus User Guide Version 3.0. Rocky Mountain Research Station, Fire Science Labs For Environmental Management. USDA Forest Service 1996. National Forests & Grasslands in Texas Land and Resource Management Plan. USDA Forest Service. 2005. Behave Plus Fire Modeling System Version 3.0 User Guide. Rocky Mountain Research Station. Andrews Patricia, Bevins Collin, Seli Robert. P 142 Statistix 8 Analytical Software User’s Manual. 2003. Berenson Mark and others. p 396 A Cartoon Guide To Statistics. 1993. Gonick Larry, Smith Woolcott. p 230. Modern Elementary Statistics. Ninth Edition. 1997 Freund John E. and Simon Gary A. P 588.