1
2
3 SENSITIVITY ANALYSIS of FIRE BEHAVIOR MODELING with
4 LIDAR-DERIVED SURFACE FUEL MAPS
5
6 Muge Mutlu, Sorin C. Popescu, and Kaiguang Zhao
7
8 Muge Mutlu
9 Graduate Research Assistant
10 Spatial Sciences Laboratory
11 Department of Ecosystem Science and Management, Texas A&M University
12 1500 Research Parkway, Suite B 215
13 College Station, TX
14 Tel: (979) 458-0742
15 Fax: (979) 862-2607
16 mugemutlu@tamu.edu
17
18
19
20
21
22
23
24
25
26
27
1 SENSITIVITY ANALYSIS of FIRE BEHAVIOR MODELING with
2 LIDAR-DERIVED SURFACE FUEL MAPS
3
4 Muge Mutlu*, Sorin C. Popescu, and Kaiguang Zhao
5 Spatial Science Laboratory, Department of Ecosystem Science and Management, Texas A&M University,
6 1500 Research Parkway, Suite B221, College Station, TX 77843
7
8 Abstract
9 The overall aim of this project is to model fire behavior using FARSITE (Fire Area
10 Simulator) and investigate differences in modeling outputs using fuel model maps, which
11 differ in accuracy, in east Texas. Each year, wild-land fires burn millions of hectares of
12 forest worldwide. Fire managers need to provide effective methods for mapping fire fuels
13 accurately. Fuel distribution is very important for predicting fire behavior. This simulator
14 model requires as input spatial data themes such as elevation, slope, aspect, surface fuel
15 model, and canopy cover along with separate weather and wind data. Seven fuel models,
16 including grass, brush and timber models, are identified in the study area. To perform
17 modeling sensitivity analysis, two different fuel model maps were used, one obtained by
18 classifying a QuickBird image and the other obtained by classifying a LIDAR (LIght
19 Detection and Ranging) and QuickBird fused data set. Our previous investigations
20 showed that LIDAR improves the accuracy of fuel mapping by at least 13 %. According
21 to our new results, LIDAR derived variables also provides more detailed information
22 about characteristics of fire.
23 Keywords: LIDAR, Fuel model, FARSITE, QuickBird
24 *Corresponding author. Phone: +1 979 458-0742 Fax: +1 979 862-2607
25
26
2
1 1. Introduction
2 Forest fires destroy many houses and natural resources such as plant and animal
3 life each year. To decrease the loss of lives and property because of the wildfires, fire
4 managers need to actively evaluate fire risks. However, managing fire risks is very
5 difficult because fuel hazards are always changing. Fire behavior is very sensitive to
6 changes in land growth, fuel risks, weather and wind conditions, and topography (Keane
7 et al., 1998). Therefore, fire managers must provide more accurate fuel model predictions
8 (Mutlu et al., 2008). Humans are the primary modifiers of fuel sources and ignition
9 source vectors for the propagation of fire (Pyne, 1992). To reduce the threat of the
10 wildfires, Texas fire managers need a tool that helps them assess fire risks more
11 accurately. This study will show the significance of using accurate input data layers
12 derived from LIDAR remote sensing technique into FARSITE software for realistic
13 predictions of fire growth.
14 Surface fuels are the greatest concern since they are major contributors to the
15 intensity and spread of fires. According to Anderson (1982), grass, brush, slash and
16 timber are the four major groups of surface fuels and thirteen surface fuel models are
17 identified for the United Stated, each varying in amount, size, and arrangement of the fuel
18 model (Anderson, 1982). Table 1 illustrates the description of each fuel model. Seven
19 fuel models are available in our study area: grass fuel models 1 and 2, brush fuel models
20 4, 5, and 7, and timber litter fuel models 8 and 9.
21 Important assessments include the potential size, rate, and intensity of a wildland
22 fire (Keane et al., 2000). Recent advances in computer software technology have allowed
23 development of several spatially explicit fire behavior simulation models, which predict
3
1 the spread and intensity of fire (Andrews and Queen, 1999). Some of this software can be
2 used to predict future fire growth and compute possible parameters of wildland fires for
3 real-time simulations (Campbell et al., 1995; Richards, 1990). An example of such
4 software is FARSITE, a spatially explicit fire growth model developed by Finney (1994).
5 FARSITE is a two-dimensional deterministic fire growth simulation model (Finney,
6 1998). This software incorporates models for surface fire (Rothermel, 1972), spotting
7 (Albini, 1979), crown fire (Wagner, 1993), and fuel moisture (Nelson, 2000). FARSITE
8 produces maps of fire growth and behavior in vector and raster format by using Huygens‟
9 Principle (Stratton, 2004; Finney, 1998). In addition, to calculate surface fire spread,
10 FARSITE implements the Rothermel (1972) equations (Miller and Yool, 2002). Many
11 wildland fire managers use this powerful tool to simulate characteristics of prescribed
12 wildfires (Finney, 1998; Grupe, 1998). FARSITE is specially designed for forest fire
13 modeling.
14 Satellite technology can assist in providing data for the FARSITE software
15 (Cheuvieco, 1997). LIDAR remote sensing is an advance technology for forestry
16 applications over the large areas. The airborne LIDAR has big potential for the direct
17 measurement of vertical forest structure and it allows researches to measure the three-
18 dimensional distribution of forests. More accurate and efficient estimation of canopy fuel
19 characteristics can be obtained using LIDAR techniques over large areas of forests
20 (Andersen et al., 2005). LIDAR is an active remote sensing technique that transmits
21 lasers to an object and measures the distance between the sensor and the object sensed.
22 This technology is useful for high-resolution topographic mapping and accurate
23 measurements of surface elevations. Airborne LIDAR systems can be used for fire
4
1 detection, location, and mapping (Justice et al., 1993), for burned area assessment, and,
2 important to this study, for fuel mapping (Keane et al., 1998). Multispectral image
3 classification is the most important part of digital image analysis. Mutlu et al. (2008)
4 specifically mapped fuels for FARSITE software use and their results are used in this
5 paper. The authors applied supervised image classification to determine which classifier
6 is more efficient and useful for two different fuel model maps that they created. These
7 two fuel model maps include seven fuel models shown in Table 1. The first fuel model
8 map was obtained by classifying only a high-resolution QuickBird satellite image and the
9 second one was obtained by classifying a LIDAR and QuickBird fused data set. The
10 investigations of Mutlu et al. (2008) show that LIDAR improves the accuracy of fuel
11 mapping by at least 13%.
12 Our research is unique and differs from other studies by using LIDAR-derived
13 input data layers to set initial conditions for fire simulations in FARSITE. We developed
14 all the spatial data layers including the fuel model map, canopy cover, DEM, slope and
15 aspect, required by FARSITE using LIDAR remote sensing techniques and multispectral
16 data.
17 The main objectives of this paper are to model fire behavior using FARSITE and
18 investigate differences in modeling outputs using fuel model maps, which differ in
19 accuracy, in east Texas. This paper will proceed as follows: Previous studies, study area,
20 materials and methods, fire simulation, processing approach, results and discussions, and
21 conclusions.
22 2. Previous Studies
23 In order to run FARSITE, spatial data derived from GIS (Geographic Information
5
1 Systems) and/or remote sensing is required and should be imported into the program.
2 These data layers must be reliable for all lands and ecosystems (Keane et al., 2000). The
3 accuracy of the input data layers is very important for realistic predictions of fire growth
4 (Keane et al., 1998; Finney, 1998). The fuel model map is the key input for the FARSITE
5 simulation software and many fire managers do not have the fuel maps needed to run the
6 software for their area. Recently, FARSITE has been used by many fire managers all over
7 the world (Finney, 1998; Keane et al., 1998). Falkowski et al. (2005) evaluated the
8 accuracy and utility of imagery from the Advanced Spaceborne Thermal Emission and
9 Reflection radiometer (ASTER) satellite sensor, and gradient modeling, for mapping fuel
10 layers for fire behavior modeling with FARSITE and FlamMap (Flammability Mapping),
11 fire simulation software. They created the surface fuel models map using a classification
12 tree based on three gradient layers; cover type, potential vegetation type, and structural
13 stage. The final surface fuel model layer had an overall accuracy of 0.632. Stephens
14 (1997) used FARSITE to spatially simulate fire growth and behavior in mixed-conifer
15 forest and to investigate how silvicultural and fuel treatments affect potential fire
16 behavior in the North Crane Creek watershed of Yosemite National Park. Keane et al.
17 (2000) combined both gradient modeling and remote sensing to map fuels spatial data
18 layer required by FARSITE to spatially model fire behavior on the Gila National Forest,
19 New Mexico. They used sampled field data to guide the classification criteria for each
20 category and to assess the value of each category to the overall classification. They
21 created the three vegetation spatial data layers: potential vegetation type (PVT), structural
22 stage, and cover type (CT). All vegetation classifications were coded into Paradox
23 database queries which use canopy cover and plant type information along with other
6
1 relevant site descriptions. Then, they used these three vegetation layers to map fuels and
2 input layers required to run FARSITE. Stratton (2004) used FireFamily Plus to evaluate
3 historical weather and calculate seasonal severity and percentile reports. Then, they used
4 this information in FARSITE and FlamMap to model pre-treatment and post-treatment
5 effects on fire growth, spotting, fireline intensity, surface flame length, and the
6 occurrence of crown fire. Miller and Yool (2002) evaluated the sensitivity of FARSITE
7 to the level of detail in the fuels data, both spatially and quantitatively, which provided
8 land managers knowledge about the effectiveness of detailed fuels mapping in modeling
9 fire spread.
10
11 3. Study area
12
13 Fire behavior was modeled for an area in east Texas near Huntsville, centered
14 within the rectangle defined by 95° 24‟ 57” W- 30° 39‟ 36” N and 95° 21‟ 33” W- 30°
15 44‟ 12” N, covering about 47.15 km2. The study area contains part of the Sam Houston
16 National Forest, characterized by deciduous, coniferous, mixed stands, open ground with
17 fuels consisting of grasses and brushes. Fig. 1 represents the high resolution (2.5x2.5 m)
18 multispectral QuickBird image, owned and operated by DigitalGlobe, of the study area.
19 Yellow marks on the image illustrate the locations of field plots on entire study area.
20
21 4. Materials and Methods
22
23 Two different fuel model maps obtained from Mutlu et al. (2008) were used to see
24 the differences in fire growth with fuel model maps of different accuracies (see Fig. 2 (a)
25 and (b)).
7
1
2 4.1. Data
3 FARSITE version 4.1 was used in this study. This software requires eight data
4 layers which include Digital Elevation Model (DEM), slope, aspect, canopy cover, fuel
5 models map, weather, wind, and fuel moisture for surface fire simulations (Finney,
6 1995). Two different input data sets were created in this study to generate real-time fire
7 simulation outputs. In dataset I, all the spatial data layers required by FARSITE software
8 were developed using LIDAR remote sensing and GIS techniques. In dataset II, spatial
9 data layers were obtained from different sources.
10 For both datasets I and II, fuel moisture data was obtained using RAWS (Remote
11 Automated Weather Station) maintained by the Texas Forest Service. The weather and
12 wind information were gathered from the Texas Forest Service‟s weather station which is
13 part of the interagency RAWS network in Huntsville, Texas (http://www.fs.fed.us/raws).
14
15 4.1.1. Dataset I: LIDAR-derived dataset
16 LIDAR scanning data was provided by M7 Visual Intelligence Inc. in LAS
17 format. A total of 47 flight lines were obtained over 6474.9 ha area with 28 flight lines
18 obtained from East to West and 19 flight lines obtained from North to South in leaf off
19 condition. Based on the study of Mutlu et al. (2008), a LIDAR-QuickBird stack image
20 of ten bands was created by stacking the four bands of the QuickBird image with four
21 LIDAR height bins, one band from the canopy cover model, and one band from the
22 canopy cover variance. Popescu and Zhao (in press) generated the height bin approach
23 that was used to create LIDAR multiband data from scanning data. They created the
8
1 LIDAR bins by counting the occurrence of LIDAR points within each volume unit and
2 normalizing by the total number of points. Fig. 3 shows the LIDAR-QuickBird stack
3 image. This image was classified with Mahalonobis distance algorithm in ENVI (ITT
4 Visual Systems, Inc.) with an accuracy of 90%. A total of seven fuel models are
5 available in the study area: grass fuel models 1 and 2, brush fuel model 4, 5, and 7, and
6 timber litter fuel model 8 and 9, shown in Table 1.
7 Canopy cover, the horizontal percentage of area covered by tree crowns at the
8 stand level, was found using methods developed by Griffin et al. (in review). Their study
9 developed the use of airborne laser methods to evaluate various canopy parameters such
10 as canopy cover and Leaf Area Index (LAI). To summarize, canopy cover is estimated by
11 using LIDAR-derived height bins and calculating the percentage of laser canopy hits 2 m
12 above ground.
13 A digital elevation model was also derived from LIDAR. By using ENVI, slope
14 and aspect were derived from the DEM.
15
16 4.1.2. Dataset II: Multispectral data and other sources
17 The second map, shown in Fig. 2(b), was derived from QuickBird data at 2.5 m
18 resolution. Based on the report from Mutlu et al. (2008), a maximum likelihood image
19 classification technique was used to classify the multispectral image with an accuracy of
20 77%. This fuel model map includes seven fuel models as well.
21 Hemispherical photos were used to create canopy cover data expressed as a
22 percent density at point locations used for training the image classifier and for accuracy
23 assessment. Maximum likelihood supervised image classification technique was applied
24 to the QuickBird multispectral image to classify the canopy cover over the entire area.
9
1 Classification is achieved by determining a canopy cover percentage range. This
2 classification was implemented on a 1 – 4 scale including 1:0-20%, 2:21-50%, 3:51-80%,
3 and 4:81-100%.
4 The DEM was downloaded from the National Fire and Aviation Management
5 Web Applications (http://www.fs.fed.us/fire/planning/nist/wims_web_userguide.htm).
6 The resolution of DEM was at 30 m resolutions, and then converted to 2.5 m resolution to
7 match it with the resolution of multispectral imagery. The slope and aspect data were
8 derived from the DEM using ENVI 4.2.
9 5. Fire Simulation
10 FARSITE requires data input as an ASCII file format because ASCII text files
11 can be viewed or created with any text editor (Finney, 1995). Since all the data were in
12 the Band Sequential (BSQ) file format, the data were saved in Leica Geosystems‟
13 ERDAS Imagine (Leica Geosystems, Inc) image processing software image file format
14 using version 9.1, then converted to ARC GRID format for incorporation into the
15 development and implementation of the fire behavior model.
16 A total of 62 plots were measured in the study area and plot center locations were
17 used as ignition points with FARSITE simulations. FARSITE was run 124 times, 62
18 times on the dataset with the LIDAR-derived fuel model map and 62 times on the dataset
19 with the QuickBird-derived fuel map. The duration of each simulation was 72 h
20 beginning at 8:00 AM and ending at 8:00 AM three days later.
21 6. Processing Approach
22 The overall study steps are shown in Fig. 4.
23 7. Results and Discussions
10
1 Different accuracy maps provided different results depending on fuel model on
2 the study area, which were expected. However, the difference is much greater than we
3 expected. The average burn area time for fires in Texas is between 3-5 days (Mark
4 Stanford, personal communication, October, 2006). We have decided to run the
5 simulation for 72 h. The comparisons of histogram for burned area and perimeter results
6 are illustrated in Fig. 5 and Fig. 6, respectively, for 72 h. Based upon the fire simulation
7 results, fuel model map derived from LIDAR shows larger fire growth areas than the
8 other fuel model map derived from QuickBird imagery as shown in Fig. 5 and Fig. 6.
9 The estimated average fire growth areas from LIDAR-derived fuel model map
10 and QuickBird derived fuel model map were approximately 174.2 ha (430.6 ac) and
11 121.9 ha (301.3 ac), respectively. Apparently, there is a considerable difference between
12 the two outputs. There are some extreme situations in our results. For instance, while
13 368.4 ha (910.3 ac) were burned on the LIDAR-derived fuel model map on the ninth run,
14 112.1 ha (277.2 ac) were burned on the QuickBird-derived fuel model map. On the
15 thirty-eight run, the burned area is 337.1 ha (833.1 ac) on the LIDAR-derived fuel model
16 map and 75 ha (185.2 ac) on the QuickBird-derived fuel model map. The reason for the
17 difference is because both maps show different fuel models on the second ignition points,
18 Fuel Model #2, grass group, on the LIDAR-derived fuel model map and Fuel Model #8,
19 timber litter group, on the QuickBird-derived fuel model map. The same situation can be
20 seen in Fig. 6 for fire perimeters for both models. For example, on the ninth run, the fire
21 perimeter is 17.5 km on the LIDAR-derived fuel model map and 4 km on the QuickBird-
22 derived fuel model map. The fire perimeters are 8.8 km and 2 km on the LIDAR-derived
23 and the QuickBird-derived fuel model maps, respectively, on the thirty-eight run.
11
1 Cost of fire is one of the biggest issues in wildland fire management. Wildfires
2 can have significant local economic effects both long-term and short-term. Donovan and
3 Rideout (2003) used the Cost plus Net Value Change (C+NVC) model. The C+NVC
4 model has the potential to minimize wildfire cost by reducing the total amount of money
5 spent on presuppression, suppression, and NVC (net wild fire damages). They assume
6 that the cost of fire per a hectare is $100. Fire fighting resources, such as crews, dozer,
7 engine, tractor, cost approximately $3,800 based on National Wildlife Coordinating
8 Group fireline handbook (1998). The sum of all costs and damages is minimized. In our
9 case, if we take a look at the results, average burned area is 174.2 ha on the LIDAR-
10 derived fuel model map. This burned area will cost around $87,826 by summing
11 “$3800+$66600+$17426” based on Donovan and Rideout‟s calculations. In this case, the
12 NVC is $17,426 which was calculated by multiplying the total burned area times 100.
13 Table 2 shows the fire fighting resources needed for a 72 h burn time based on the
14 LIDAR-derived fuel model map. However, the average burned area is 121.9 ha on the
15 QuickBird-derived fuel model map and the cost of fire is $61,290. Since the average
16 burned area on the QuickBird fuel model map is smaller than that on the LIDAR-derived
17 fuel model map, we created another fire fighting resources table (Table 3) for a 72 h burn
18 based on the QuickBird-derived fuel model map, using fewer fire fighting resources.
19 Apparently, there is a significant difference between these two outputs.
20
21 8. CONCLUSIONS
22 This study indicates the influence of a more accurate fuel map on modeling fire
23 behavior and assessing fire risk. FARSITE was developed mainly to be used as a tool in
12
1 the fire management. FARSITE will assist fire managers with the mitigation of the
2 harmful effects of wildfire. Also, it gives the power of sound, accurate, and efficient fire
3 behavior modeling technology to forest fire fighters (Finney, 1998).
4 Airborne LIDAR systems can be used for fire detection, location, fuel mapping,
5 and burned area assessment (Justice et al., 1993). In this study, LIDAR data processing
6 by the height bin approach is used to generate accurate estimates of surface fuels
7 parameters and created a fuel model map. Many fire managers do not have the fuel maps
8 needed to run the fire simulation models for their area. We were able to develop all the
9 spatial data layers required by FARSITE using a unique remote sensing approach that
10 fuses innovative LIDAR-derived products and multispectral imagery. Results from this
11 study indicate the influence of a more accurate fuel map on modeling fire behavior and
12 assessing fire risk. According to results, LIDAR derived products were able to assess
13 fuel models with high accuracy and provide different fire perimeters and fire growth area.
14 For fire mitigation purposes, we need to know both fire perimeters and fire growth areas.
15 Fire growth area results are helpful to determine the cost of fire. Fire perimeter results are
16 important because they help in determining an optimal mix of fire fighting resources
17 needed to fight fires such as dozer, tractor, crews, helicopter, engines, hourly cost of
18 operating the resources, arrival time, etc. Using two different datasets, one derived from
19 LIDAR and the other one derived from QuickBird imagery and different data sources,
20 provided significantly different outputs. The differences could be attributed to different
21 fuel model map, canopy cover, DEM, slope, and aspect.
22 The cost of our LIDAR data were around $32,645 and the QuickBird data were
23 obtained for an approximate cost of $3,890 for the whole study area, 4741.83 ha. The
13
1 cost of first fuel model map which includes LIDAR and QuickBird data is $36,535. The
2 cost of second fuel model map which includes only QuickBird data is $3,890. Based on
3 this, LIDAR-derived fuel model map is more expensive than QuickBird-derived fuel
4 models map. However, if we look at the results LIDAR data will save us thousands of
5 dollars. The cost of LIDAR data for the entire study area is a lot less than the cost of
6 average burned areas.
7 Accurate estimation of fire growth area and the direction of fire growth is
8 extremely important information for the fire management process. With the knowledge of
9 this essential information will avoid any health risk for local people who live in that
10 vicinity. This information will be more useful if it is used by fire management
11 authorities. In case of fire, if the fire managers use the QuickBird-derived fuel model map
12 they may not be able to send enough sources to fight with the fire and this situation may
13 cause more serious problems. Small errors in fuel models may not be significant for a
14 small area; however, for large areas, small errors may change the result of fire behavior
15 and fire sizes significantly.
16
17 Acknowledgements
18 This project was founded by the Texas Forest Service (award # 02-DG-11083148-
19 050). We want to thank all Texas Forest Service personnel especially Curt Stripling for
20 his help on collecting and analyzing the field data and determining fuel models based on
21 these datasets. We thank Tom Spencer for determining fuel models for our study area and
22 thank Alicia M.R. Griffin for providing us a canopy cover data needed to run the
23 FARSITE software and Dr. Lewis Ntaimo for helping us with the cost analysis. We also
24 want to thank Texas A&M University Department of Ecosystem Science and
14
1 Management and the Spatial Science Laboratory for their support.
2
3 References
4 Albini, F.A., 1979. Spot fire distance from burning trees: A predictive model. General
5 Technical Report INT-56. Ogden, UT: USDA Forest Service, Intermountain
6 Forest and Range Experiment Station. 73.
7 Andersen, E. H., McGaughey, J, R., and Reutebuch E. S., 2005. Estimating forest
8 canopy fuel parameters using LIDAR data. Remote sensing of Environment, 94,
9 4, 441-449.
10 Anderson, H. E., 1982. Aids to Determining Fuel Models For Estimating Fire Behavior,
11 United States Department of Agriculture, Forest Science, General Tehcnical
12 Report INT-122.
13 Andrews and Queen, 1999. Fire Modeling and information system technology.
14 International Journal of Wildland Fire, 10, 343-352.
15 Campbell, J.; Weinstein, D.; Finney, M.,1995. Forest fire behavior modeling integrating
16 GIS and BEHAVE. In: Thompson, J.E. (Compiler): Analysis in support of
17 ecosystem management. USDA Forest Service Ecosystem Management Center
18 Report. Washington D.C. Pages 184-192.
19 Cheuvieco E., 1997. A review of remote sensing methods for the study of large wildland
20 fires. Megafires project ENV-CT96-0256. Spain. Pp.192
21 Donovan, G.H., and Rideout, D.B., 2003. An Integer Programming Model to Optimize
22 Resource Allocation for Wildfire Containment. Forest Science, 2, 49.
23 Fallowski, M. J., Gessler, P.E., Morgan, P., Hudak, A. T, and Smith A. M. S., 2005.
24 Characterizing and mapping forest fire fuels using Aster imagery and gradient
15
1 modeling. Forest Ecology and Management, 217, 2-3, 129-146.
2 Finney, M.A., 1994. Modeling the spread and behavior of prescribed natural fires. In
3 Proceedings of the twelfth conference on fire and forest meteorology, edited by J.
4 D. Cohen, J. M. Saveland, and D. D. Wade, 138–43. Bethesda, MD: Society of
5 American Foresters.
6 Finney, M.A., 1995. FARSITE Fire Area Simulator Version 1.0 User guide and technical
7 documentation. Systems for Environmental Management Report. Systems for
8 Environmental Management, Missoula, MT. 76.
9 Finney, M.A., 1998. FARSITE: Fire Area Simulator Model development and evaluation.
10 Research Paper RMRS-RP-4. Ogden, UT: USDA Forest Service, Rocky
11 Mountain Research Station. 47.
12 Griffin, A.M.R., Popescu, C.S., and Zhao, K. (in review). Using LIDAR and Normalized
13 Difference Vegetation Index to Remotely Determine LAI and Percent Canopy
14 Cover. Remote Sensing of Environment
15 Grupe, M. A. 1998. Assessing the applicability of the Terrestrial Ecosystem Survey for
16 FARSITE. Master‟s Thesis, University of New Mexico, Albuquerque, New
17 Mexico, 95.
18 Justice, C.O., Malingreau J.P, and Setzer A.W., 1993. Satellite remote sensing of fires:
19 Potentials and limitations in „Fire in the environment‟. 77-88
20 Keane, R.E., Garner, J. L., Schmidt, K.M., Long, D. G., Menakis, J.P., Finney, M.A.
21 1998. Development of the Input Data Layers for the FARSITE Fire Growth
22 Model for the Selway-Bitterroot Wilderness Complex, USA. USDA Forest
23 Service General Technical Report RMRS-GTR-3. 121.
16
1 Keane, R.E., Mincemoyer, S. A., Schmidt, K.M., Long, D. G., Garner, J. L. 2000.
2 Mapping vegetation and fuels for fire management on the Gila National Forest
3 Complex, New Mexico, [CD-ROM]. Gen. Tech. Rep. RMRS-GTR-46-CD.
4 Ogden, UT: U.S. Department of Agriculture, Forest Service, Rocky Mountain
5 Research Station, 126.
6 Miller, J.D., and Yool, S.R., 2002. Modeling fire in semi-desert grassland/oak woodland:
7 the spatial implication. Ecological Modeling, 153, 3, 229-245.
8 Mutlu, M., Popescu, C.S., Stripling, C., and Spencer, T. 2008. Mapping surface fuel
9 models using lidar and multispectral data fusion for fire behavior. Remote
10 Sensing of Environment. 112, 1, 274-285.
11 Nelson, R.M., Jr., 2000. Prediction of diurnal change in 10-h fuel stick moisture content.
12 Canadian Journal of Forest Research 30:1071–1087.
13 National Fire and Aviation Managament Web Applications
14 http://www.fs.fed.us/fire/planning/nist/wims_web_userguide.htm. Accessed on:
15 03/10/07.
16 Popescu, C.S., and Zhao, K., 2008. A voxel-based lidar method for estimating crown
17 base height for deciduous and pine trees. Remote Sensing of Environment, 112, 3,
18 767-781.
19 Pyne, S.J., 1992. Fire in America, a cultural history of wildland and rural fire. Princeton
20 University Press, 654 pp.
21 Richards, G.D., 1990. An elliptical growth model of forest fire fronts and its numerical
22 solution. Int. J. Numer. Meth. Eng. 30, 1163-1179.
23 Reeves, H.C. 1988. Photo Guide For Appraising Surface Fuels in East Texas. Center
17
1 For Applied Studies-School of Forestry. Stephen F. Austin State University.
2 Nacogdoches, Texas. 75962.
3 Remote Automated Weather Station. http://www.fs.fed.us/raws. Accessed on 5/12/07.
4 Rothermel, R.C., 1972. A mathematical model for predicting fire spread in wildland
5 fuels. Research Paper INT-115. USDA Forest Service, Intermountain Forest and
6 Range Experiment Station, Ogden, UT.
7 Stephens, S. L. 1997. Evaluation of the effects of sivilcultural and fuels treatments on
8 potential fire behavior in Sierra Nevada mixed-conifer forests. Forest Ecology and
9 Management, 105, 21-35.
10 Stratton, D.R., 2004. Assessing the Effectiveness of Landscape Fuel Treatments on Fire
11 Growth and Behavior. Forestry, October/November 2004. P.32-40
12 Wagner, W.., 1993. Prediction of crown fire behavior in two stands of jack pine. Can. J.
13 For. Res. 23, 442–449.
14 Wagner, W., Eberhofer, C., and Summer, G., 2004. Robust Filtering of Airborne Laser
15 Scanning Data for Vegetation Analysis. International Archivews and
16 Photogrammetry, Remote Sensing and Spatial Information Sciences. 36 (8), 56-
17 61.
18
19
20
21
22
23
18
1
2
3
4
5
6
7
8
9 Fig. 1. The location of our study area and false color composite of a QuickBird image
10 (Digital Globe, Inc)
11
12
13
14
15
16
17
18
19
20
21
22
19
1
2
3
4
5
6
7
8
9
¯ ¯
Legend Legend
FM # 9 FM # 9
FM # 8 FM # 8
FM # 7 FM # 7
FM # 5 FM # 5
FM # 4 FM # 4
FM # 2 FM # 2
10 FM # 1 FM # 1
11
12 (a) (b)
13
14 Fig.2. (a) The fuel map obtained by classifying a LIDAR and QuickBird fused
15 data set, (b) the fuel model map obtained by classifying a QuickBird image
3 3
16
17
18
19
20
21
22
23
24
20
1
2
3
Four QuickBird Bands
LIDAR-derived Canopy Cover
First Four LIDAR Height Bins
CHM (Canopy Height Model) Variance
4
5
6
7 Fig.3. LIDAR-QuickBird data fusion
8
9
10
11
12
13
14
15
16
17
18
19
20
21
1
22
Fig. 4. The overall study steps
23
1
2 Fig. 5. Comparison of burned areas for both fuel model maps.
3
4
5
6 Fig. 6. Comparison of fire perimeter results for both fuel model maps.
7
8