Wildfire Risk Assessment

Record temperatures and minimal rainfall in the Austin area over the past summer have resulted in conditions ripe for an increase in the number of brush fires. The purpose of this project will be to quantify the risk of brush fires in Travis County, Texas. There are many factors that are indicative of an area prone to fire, and GIS allows these to be incorporated spatially to provide the best assessment of risk. By incorporating GIS data into a fire simulator, fire behavior and severity can be modeled. The completed project will model a fire that occurred on September 4, 2000.

There are many fire simulators available on the web. I had two requirements in my search for a program. One, that it have the ability to integrate GIS data, and two, that it be relatively inexpensive. The simulator I found, FARSITE, utilizes data that is pre-processed in a GIS environment. It is also free and can be downloaded from http://www.farsite.org/.

FARSITE was developed by Mark A. Finney at Systems for Environmental Management in cooperation with the National Interagency Fire Center (NPS), USFS Intermountain Fire Sciences Laboratory, and under contract from USFS Region 5. Because FARSITE is based on spatial data, it is a powerful tool for the fire manager. It has the ability to simulate spatial and temporal changes into the fire model. The user is free to dictate different weather conditions, fuels, and terrain. FARSITE also allows fire managers the ability to simulate attack scenarios, utilizing both land and air resources. Fire fronts are modeled using Huygens' principle of wave propagation. In order to model fire behavior in FARSITE, GIS data is required to be imported into the program. Required data fields include Elevation, Slope, Aspect, Fuel Types and Canopy Cover.

I chose an area in western Travis County in which to test the program. This area, which was subjected to a 10-12 acre brush fire on September 4, 2000, can be seen in the following picture.

The fire was located in an urban interface area on Hudson Bend, near Lake Travis. The picture shown is taken the day after the fire occurred. In the foreground, the burned area can be distinguished; in the background, the unburned area is evident. The fire started at approximately 2:00 pm and was extinguished by the Hudson Bend Fire Department around 8:00pm. Weather reports of wind conditions indicate that the initial direction of burn was to the northeast. Record high temperatures of 110^o F and changing wind directions resulted in a constantly changing fire line with multiple flame fronts. Since no property was damaged in this fire, the initial cause of the fire was not investigated.

The data fields I needed to model this fire in FARSITE were Elevation, Slope, Aspect, Fuel Types and Canopy Cover. I chose the USGS Mansfield Dam quadrangle at 30 meter resolution (1:24,000) as a template for downloading these files.

Finding a coverage for Fuel Models proved slightly more difficult. FARSITE requires that fuels be inputted using Anderson's Fire Behavior Fuel Model. This classification of different fuel types has 13 parameters based on a scientific measure of fuel loading. Fuel depth, fuel particle density and heat content of fuel are a few of the criteria used to classify fuels under this system (Weinstein, et al., no date). I was not able to find fuel data for this area which utilized this fuel classification system. The TNRIS-Digital Data Site has available a Land Use/ Land Cover file for the Austin area which uses Standard Anderson Criteria. Luckily, FARSITE allows for conversion between these two systems easily.

In order to obtain Canopy Cover data, which is expressed as a percent density, I had to interpolate this from the LU/LC file previous mentioned. This classification was implemented on a 1 - 4 scale. 1: 0-20%, 2: 21-50%, 3: 51-80%, 4: 81-100%.

The DEM for Mansfield Dam was imported as an USGS DEM into ArcView. It was already projected in UTM coordinates, so I used that as my standard and projected all other data files into the UTM Zone 14 coordinate system.

Using ArcView Spatial Analyst, I was able to derive Slope and Aspect from the DEM of Mansfield Dam. This ensured that these files were of the same resolution and extent as required by FARSITE.

The Land Use/Land Cover file downloaded from TNRIS for the Austin Area was very large. In order to make it useable, I projected it in UTM coordinates and clipped it using the geoprocessing extension in ArcView. The extent of Mansfield Dam quadrangle is outlined below in red.

Using the Convert to Grid command in ArcView, I was able to obtain a LULC file with the same resolution and extent as the Digital Elevation Model.

Since no data exists for Canopy Cover for this region, I interpolated this information from the LULC file previously mentioned. Based on the fuel type specified, a percent density was assigned. To do this, I resaved the LULC file under a new name. I then opened the legend editor in ArcView and changed the value labels for each fuel type to a corresponding canopy cover density (shown below).

Once I had all the data I needed, I had to export all of the shapefiles into ASCII grid files in order to import them into FARSITE. The resulting landscape (.LCP) file that was created in FARSITE is shown to the right below. This image incorporates all of the data from the five files on the left.

FARSITE also requires weather, wind and fuel moisture files to be created. I estimated weather conditions for September 4, 2000 based on personal observations from the day of the fire, KXAN Live WeatherNet at Lake Travis Middle School, and the Hudson Bend Fire Department's Internal Report of the fire. Since I had no data on fuel moisture for this time period, I used an example file provided in FARSITE.

In order to have an accurate area for the fire ground, the perimeter of the burned area was traversed with a Garmin III handheld GPS unit. Measurements were taken every ten meters using decimal degree coordinates. These points were entered into a database (.dbf) file in Excel, added into ArcView using Add Event Theme and converted to a shapefile. A closer view of the extent of the fire can be seen below. This shapefile was added into the .LCP file in FARSITE, which allows up to 5 vector files to be overlain on a landscape file.

I attempted to convert these points into a polyline theme and export that into FARSITE. I was able geocode a polyline in ArcView; however, FARSITE would not let me import it.

Click here to see the location of the fire in relation to a map of Travis County.

Using an approximate ignition point based on information from firefighters on scene, I ran the FARSITE simulation. The duration of the fire was set from 2:00pm - 7:00pm. The internal time step for the model was 15 minutes; however, the time step visible in the .gif below is 1 hour. The contour lines indicate the extent of the fire. FARSITE also allows exportation of files back into ArcView for further post-processing.

In conclusion, GIS coupled with fire simulator can be a powerful tool for Fire Departments around the country. Provided the appropriate data is available, FARSITE is an extremely user friendly application that has the ability to model all aspects of fires. While this project did not consider different attack scenarios, FARSITE is capable of simulating both land and air attacks.

The FARSITE simulation of the Labor Day Fire at Hudson Bend was very accurate, considering the small extent of the fire compared with the low resolution data provided. I believe FARSITE overestimated the extent of the fire because no attack scenarios were inputted into the model when in reality 14 firefighters, 4 engines and numerous brush trucks fought the fire continuously from 3:45 until it was extinguished.

In order to assess risk, I had hoped to incorporate fire hydrant and road shapefiles to make some inferences about possible methods of attack. While FARSITE does allow importation of vector files, the shapefiles I had were not compatible with the format used. Also, since FARSITE is intended for use in large wildland fires, and is geared toward large scale attack scenarios, it does not utilize these hydrant and road files in planning attacks.

Brian Quinlan and Kenny Pailes with the Hudson Bend Fire Department for providing the incident report for the fire and for assistance with GIS implementation.

Karen Roberts with the Texas Department of Transportation for help in data acquisition.