Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | home
METHODOLOGY
Development stages of this study are as follows: 1. Identify and inventory those spatial hazards which influence the fire risk to the wildland urban interface, 2. Delineate and map the wildland urban interface in five counties (Benton, Carroll, Crawford, Madison, and Washington) in Northwest Arkansas, 3. Use spatial analytical tools to measure and model varying degrees of wildfire risk potential to the wildland urban interface.
Hazards to rural populations living in the wildland urban interface are those features which directly relate to the magnitude of threat from wildfire. Some examples of hazard types found in rural NW Arkansas are: distance to a rural volunteer fire department (RFD) or the absence of fire protection, and access to the fire scene, i.e. narrow-winding roads, types of forest fuels in close proximity, and history of wildfire in the study area. Land ownership will often predetermine the amount of resources available for fire suppression, boundaries, ownership boundaries will be mapped but not incorporated into any models.
The Risk Potential Map was developed from the following categories or layers: Ownership Layer (Federal, State, Private), Rural Structures Layer (Residence), Historical Fires Layer, Fuel Type Layer, Rural Fire Departments Layer (geographic location and protection area), and Road Types Layer (blacktop, gravel, graded, and non-maintained) were generated. These Layers were placed into their appropriate Potential Component.
The Risk Potential Map was compiled from the Threat Matrix and the Suppression Matrix. The Risk Potential Map categories has three levels of risk: high, medium, and low. A decision matrix was used to reduce the potential number of categories to a manageable number.
1. Identify and Inventory Components
The Rural Fire Departments layer (Response Potential Component) was subdivided into two categories: 5 mile protection zone (1) and greater than 5 miles (2). Residences located more than 5 miles from the nearest rural volunteer fire department will be considered at risk to wildfire. Roads Types (Access Potential Component) were subdivided into four categories: All Weather roads which were the paved and gravel roads (1), Seasonal roads which were the graded roads (2), Limited Access roads which were those roads which were non-maintained (3), and those areas with no access or were greater than 300 feet from any road type (4). Note: The driveway to each residence is not part of the model component primarily because there is no data set available but also because of the variability of access and condition. The 6 Fuel Types used in the Fuel Potential Component were reduced to three volatility classes: low (1), medium (2), high (3). These volatility classes are based upon rate-of-spread (ROS) of each fuel class.
Statewide vegetation and a statewide forest fuel data sets are available. The statewide forest fuel dataset contains 6 fuel types (Table 2) as defined by Anderson (1982). Presence and proximity of each of these fuel types to the wildland urban interface will impact the overall risk. Historical fire data (1992 - 1996) was used to identify areas more susceptible to wildfire, due primarily to human behavior as stated in Kluender et al (1988) study.
2. Mapping the Wildland Urban Interface
Rural structures were extracted from one of the 63 layers of the AHTD dataset, the dataset is in Intergraph Microstation drawing format (dgn). The AHTD dataset was digitized from 1994 aerial photography and from 7.5 minute USGS quadrangles at a scale of 1:2400. The structures were identified as either residence or poultry house by their layer number and icon type. The residence type was extracted as a single thematic layer for GIS analysis. Only residences are being mapped in this project.
The AHTD data set contains other separate layers: each road type, incorporated boundaries, and political ownership e.g. Ozark National Forest. These data points were buffered based on preset criteria (see each Component listed below for amount of buffering). Arkansas' forest fuel types were developed from an existing statewide vegetation data set. The statewide vegetation data set was developed from 1992 Landsat Thematic Mapper satellite imagery. Rural Volunteer Fire Departments (RFD) were located geographically using global positioning system technology. The RFD data set will be used to delineate rural areas with residences at higher risk because of the distance a fire truck must travel to reach the residence.
3. Modeling Risk Potential in the Wildland Urban Interface:
The cartographic model (Figure 2) shows the relationship of each of the various data sets. The permanent Risk Potential Map was derived by joining two matrices, the Threat Matrix and the Suppression Matrix. The Threat Matrix was compiled from the Ignition Potential Component & the Fuel Potential Component. The Suppression Matrix was compiled from the Access Potential Component & the Response Potential Component. The different Potentials are comprised of permanent hazards which do not change substantially over time, such as road infrastructure, residential construction, forest fuels, etc. The Behavior Potential is made up of those components which cause daily or seasonal changes in wildfire behavior, such as weather, climate, and terrain.
Components of The Wildland Urban Fire Risk Model:
Wildland fires are influenced by five different components: Fuel Potential, Ignition Potential, Response Potential, Access Potential, and Behavior Potential. Four of these components (Fuel, Ignition, Access, and Response) are macroscale potentials with minimal temporal changes. Behavior Potential on the other hand is mesoscale and tends to be temporary in nature, changing on a daily basis depending on current weather conditions.
The first step in this project was to develop a fuel data set from a reclassification of landcover types using Anderson's (1982) 13 Fire Behavior Fuel Models (Table 2). Each landcover type was evaluated and identified as one of Anderson's Fuel Models. Note: Anderson's (1982) Fire Behavior Fuel Models are used nationwide but were developed from fuel types existing in western fire regimes. Wildfire experts with the Arkansas Forestry Commission (AFC) aided in the reclassification effort. This task was subjective, based on the experience of AFC personnel and it generalizes such fuels as ground litter, understory, and grasses. The development of the Arkansas fuel map (Figure 9) was a direct result of the reclassified vegetation classes (Table 3). Table 4 shows the actual associations between Anderson's (1982) fuel models and the vegetation develop by the Arkansas GAP Analysis Project (Dzur, et.al, 1995). Table 17 breaks down the fuel models found in District 6 by area and Table 15 breaks down the fuel models statewide. Many variables such as rainfall, drought, and grazing actually determine the volatility of grass fuels. Determination of these variables is beyond the scope of this project and therefore was not modeled. To reduce the number of categories each of Anderson's fuel models was reduced to one of three volatility classes. The volatility class was based upon the fuel model's behavior or rate-of-spread. Three volatility classes were developed: low (1), moderate (2), and high (3).
Table 2. Anderson's 13 Fuel Models Description (plus water):
FM 1. Short Grasses (agriculture and improved pasture) Fire spread is governed by the fine, very porous, and continuous herbaceous fuels that have cured or are nearly cured. Fires are surface fires that move rapidly through the cured grass and associated material. Very little shrub or timber is present, generally less than one-third of the area. Grasslands and savanna are represented along with stubble, grass-tundra, and grass shrub combinations that met the above constraint. Annual and perennial grasses are included in this fuel model. Fire Behavior: Surface fires that can burn very rapidly. |
FM 3. Tall Grasses - (sage grasses, unimproved pasture,
pine plantation) Fires in this fuel are the most intense of the grass group and display high rates of spread under the influence of wind. Wind may drive fire into the upper heights of the grass and across standing water. Stands are tall, averaging about 3 feet (1m), but considerable variation may occur. Approximately one-third or more of the stand is considered dead or cured and maintains the fire. Wild or cultivated grains that have not been harvested can be considered similar to tall prairie and marshland grasses. Fire Behavior: Surface fires that can spread easily. Clumps of fuels that generate higher intensities may produce firebrands. |
FM 4. Brush/Cedar Glades (dense brush intermixed with cedar) Fire intensity and fast-spreading fires involve the foliage and live and dead fine woody material in the crowns of a nearly continuous secondary overstory. Dead woody material in the stand significantly contributes to fire intensity. A deep litter layer may also hamper suppression efforts. Fire Behavior: Very high to extreme rates of spread can be experienced in this model. Very high intensities make control efforts difficult. |
FM 6. Intermediate Brush/Hardwood (clearcuts > 1 yr.) Brush height between 2 to 6 feet. Foliage is generally flammable although moderate to strong winds may be required to carry fire in the crowns. Fire Behavior: Fire carries through the shrub layer, but drops to the ground at low wind speeds or openings in the stand. |
FM 7. Pine Timber (needle litter) Communities of palmetto-gallberry understory with pine overstory usually found in the coastal plains. Fire burns through the surface and shrub strata with equal ease, and can occur at higher dead fuel moisture contents because of flammable nature of live foliage and other live materials. Stands are generally between 2 to 6 feet high. Fire Behavior: Rate of spread and fire intensity are both moderately high. |
FM 8. Closed Timber Litter (pine with hardwood understory) Closed timber overstory, short-needled conifers or hardwoods that have leafed-out support fire in the litter layer. Litter mainly composed of pine needles, leaves, and twigs. Fire Behavior: Slow burning surface fires with low flame heights are typical. |
FM 9. Hardwood Litter (seasonal fuels - leaves) Oak/hickory types are best represented in this fuel model. Also other hardwoods and loosely compacted litter under long-needled conifers found in plantations. Fire spread is primarily in surface litter such as concentrations of dead, dry leaves in fall or spring. Stands can be hardwoods, mixed hardwood/conifers, or long needle conifers. Fire Behavior: Fires run through the surface litter and possibly torch out trees, spot, and crown where concentrations of dead-down woody materials are encountered. |
1. AFC's District 6 Vegetation Layer
The statewide vegetation data layer used for this study was developed
by Dzur et al, (1995) at the Center for Advanced Spatial Technologies at
the University of Arkansas. The statewide vegetation map was developed using
Landsat Thematic Mapper satellite imagery and groundtruthed using ancillary
datasets provided by members of the Arkansas GAP Consortium. The dataset
was classified using the vegetation cover types developed by the Arkansas
GAP Vegetation Committee (Table 3). The 5-county area comprising AFC Fire
District 6 was re-sampled from the statewide forest fuel dataset into a
smaller regional data set.
Table 3: GAP Vegetation Classes
| 1. Pinus echinata | 20. Quercus falcata var. pagodifolia |
| 2. Pinus taeda | 21. Celtis laevigata |
| 3. Juniperus virginiana (pure stands) | 22. Quercus nuttallii |
| 4. Quercus spp. - Pinus echinata - Carya spp. | 23. Quercus phellos |
| 5. Pinus taeda/echinata Quercus spp. | 24. Liquidambar styraciflua |
| 6. Juniperus virginiana (mixed w/hardwood) | 25. Taxodium distichum |
| 7. Fagus grandifolia | 26. Taxodium distichum (dominate spp.) |
| 8. Quercus alba | 27. Nyssa |
| 9. Quercus rubra - Quercus spp. | 28. Tall Grasses |
| 10. Quercus falcata - Quercus spp. | 29. Arundinaria gigantea |
| 11. Quercus stellata | 30. Salix - Populus |
| 12. Juniperus virginiana - Quercus spp. | 31. Betula - Platanus - Acer |
| 13. Pinus echinata - Quercus spp. | 32. Bare |
| 14. Juniperus ashei - Quercus spp. | 33. Water |
| 15. Quercus spp. - Carya texana | 34. Agriculture (Wet crops) |
| 16. Mixed shrub species (uplands) | 35. Agriculture (Dry crops) |
| 17. Mesic Prairie | 36. Agriculture (pasture) |
| 18. Quercus lyrata | 37. Urban Commercial/Industrial |
| 19. Carya aquatica | 38. Urban Residential |
2. AFC's District 6 Forest Fuel Layer
The AFC Fire District 6 fuel class (Table 4) was a reclassification, or map recoding, of each vegetation community in the vegetation dataset into one of the 13 fire behavior fuel models developed by Anderson (1982). The 43 vegetation classes (Table 3) were reclassified into the 13 fuel models (Table 2). Each fuel type was evaluated based on its potential volatility as an energy source. Each fuel model will consists of one of the three volatility classes, low (1), medium, (2) and high (3). Each fuel model was assigned a volatility class described in Table 5, based upon interpretation of the descriptive nature of each of Anderson's 13 fire behavior fuel models (Table 2), rate of spread was the primary indices used. Of note is the lack of understory information in Dzur et al (1995) vegetation data set, the results of using satellite imagery's spectral signature of the canopy. GAP vegetation classes were evaluated using the assumed presence of the understory, if that was part of Anderson's fuel model description. For example fuel models 7 and 8 are distinguished primarily by presence or lack of an understory.
Table 4: Fire District 6 Fuel
Categories
| Fuel | Vegetation Classes | Fuel | Vegetation Classes |
| 8 | 1. Shortleaf Pine | 9 | 20. Cherrybark Oak - Mixed Oak |
| 8 | 2. Loblolly Pine | 9 | 21. Sugarberry |
| 4 | 3. Eastern Red Cedar (Upland) | 9 | 22. Nutall Oak - Mixed Ash/Oak |
| 9 | 4. Oak - Hickory - Pine (Upland) | 9 | 23. Willow Oak - Mixed Oak/Hickory |
| 8 | 5. Lob/Shortleaf - Oak (Lowland) | 9 | 24. Sweetgum |
| 8 | 6. Eastern Red Cedar (Lowland) | 9 | 25. Bald Cypress - Mixed Hardwood |
| 9 | 7. American Beech - Holly | 9 | 26. Bald Cypress |
| 9 | 8. White Oak - Mixed Oak | 9 | 27. Water Tupelo - Bald Cypress |
| 9 | 9. N. Red Oak - Mixed Oak | 3 | 28. Tall Grass (Moist) |
| 9 | 10. S. Red Oak - Mixed Oak | 1 | 29. Tall Prairie Grass |
| 9 | 11. Post Oak | 9 | 30. Mixed Willow/Cottonwood |
| 9 | 12. Eastern Red Cedar - Mixed Oak | 9 | 31. Sycamore - Sugar Maple |
| 8 | 13. Shortleaf Pine - Mixed Oak | 1 | 32. Bare |
| 6 | 14. White Cedar - Mixed Oak | 14 | 33. Water |
| 9 | 15. Oak - Hickory (Black) | 1 | 34. Agriculture - Wet |
| 1 | 16. Hawthorn - Sparkleberry | 1 | 35. Agriculture - Dry |
| 3 | 17. Tall Mesic Prairie | 3 | 36. Agriculture - Pasture |
| 9 | 18. Overcup Oak - Water Hickory | 1 | 37. Urban Commercial |
| 9 | 19. Oak - Hickory - White Ash | 1 | 38. Urban Residential |
Table 5. Volatility Class of Anderson's Fuel Models
| Fuel Model | Volitity Class (Code) |
| FM 1 Short Grasses | Medium (2) |
| FM 3 Tall Grasses | Medium (2) |
| FM 4 Brush/Cedar Glades | High (3) |
| FM 6 Intermediate Brush/Hardwood | Medium (2) |
| FM 7 Pine Timber (needle litter) | High (3) |
| FM 8 Closed Timber Litter (pine with hardwood understory) | Low (1) |
| FM 9 Hardwood Litter (seasonal fuels - leaves and twigs) | Medium (2) |
B. Ignition Potential Component:
The Ignition Potential Component is comprised of known areas of ignition. Historical fire data from 1992 to 1996 was used as the primary data source for the development of the Ignition Potential model. In addition to the limited temporal nature of the data set an average wildfire size was applied to each wildfire location. Each pixel will be evaluated and assigned a code of 1 for high ignition probability or a 2 for low ignition probability.
Fire Incident Reports from 1992 to 1996, provided by the Arkansas Forestry Commission, were in Dbase2 format. Each fire record contained 90 fields of information, such as location, source, fuel type, response, date and time, etc. The location of each fire was referenced to a 40 acre land parcel, this parcel was described using the Public Land Survey System (PLSS), i.e. NW corner of the NE corner of Section 12, Township 13 North, and Range 33 West. Each fire located in the study area was georeferenced, mapped, and analyzed for average fire size. Twenty acres was the average fire size in district 6 from 1992 to 1996 as determined by analysis of 847 wildfire incident records. Each fire location was buffered 20 acres.
C. Response Potential Component
This component was used to identify those rural areas which were within
5 miles of a rural fire department. Those residences not within the 5 miles of a rural volunteer fire department were identified as residences at risk due to lack of fire protection. Areas within 5 miles were assigned a code of 1 while areas outside of 5 mile response areas were assigned a code of 2.
Volunteer fire departments in the study area were plotted with 5 mile buffered rings. RFD's in Washington and Carroll were located by the AFC personnel using Navstar Global Positioning System, the other RFDs were located on maps by County Fire Coordinators. The buffered RFD areas identified the level of fire protection for each counties' wildland urban interface. Each RFD locations was verified by District 6 personnel.
D. Access Potential Component:
Four types of roads were identified in the AHTD dataset: 1) paved, 2) all weather gravel, 3) graded, and 4) non-maintained. Four classes of 'Access' were developed from these four road types: 1. The All Weather Access class is made up of both paved and all weather gravel road types. 2. The Seasonal Access class is made up of graded road type. 3. The Limited Access class was the non-maintained road type. 4. The No Access class was those areas that were more than 300 feet from any Access class, driveways were not a factor in this study. Each road class was buffered out 300 feet, this is the maximum distance from which RFD can operate. All Weather Access was assigned a code of 1, Seasonal Access was assigned a code of 2, Limited Access was assigned a code of 3, and No Access was assigned a code of 4.
Three levels of ownership exist in the study area: 1. Federal, 2. State, and 3. Private. Each will be mapped and labeled with a code from 1 to 3, Federal (1), State (2), and Private (3). These assignments are for methodological modeling and may not represent actual levels of response. There also exist agreements between the AFC, the USFS, and individual RFDs which provide varying degrees of support in combating wildfires. Those interface areas which are within either state or federal ownership were masked out and not included.
2. Political and Administrative Boundaries
Political and administrative boundaries will be included for reference. State, county, and incorporated boundaries were mapped. AFC District 6 boundary was mapped. The RFD district boundary in each county were mapped. The location of all fire departments in the study area was also mapped.
Using a decision matrix the Risk Potential Map was developed from two separate matrices, the Threat Matrix and the Suppression Matrix. The categories for the Risk Potential matrix are: high, moderate, and low.
The Threat Matrix (Table 6) is a 2 by 3 matrix of the interaction between the Ignition Potential Component categories and of the Fuel Potential Component volatility categories. Table 6 shows the potential values used to qualify the risk of wildfire ignition and wildfire intensity based on occurrence of wildfires between 1992 and 1996 and the volatility of different forest fuels. For example if there was a high risk from Ignition Potential and a high risk from the Fuel Potential Component volatility then the threat risk would be high.
IGNITION POTENTIAL |
HIGH VOLATILITY |
MODERATE VOLALITILITY |
LOW VOLATILITY |
HIGH |
HIGH RISK |
HIGH RISK |
LOW RISK |
LOW |
MODERATE RISK |
LOW RISK |
LOW RISK |
The Suppression Matrix (Table 7) is a 2 by 4 decision matrix of the interaction between the Response Potential Component categories and of the Access Potential Component categories. Table 7 shows the potential risk categories assigned by the matrix which qualifies the risk to rural population, once a fire has been reported. For example if a rural residence is greater than 5 miles from a rural fire station it is considered to be at a high risk and if the roads leading to the rural residence are limited (no-maintenance), the risk due to access is considered to high, then the overall suppression risk is also categorized as high. Conversely, if a rural residence is within 5 miles of a rural fire station and if the rural residence is accessible by all weather road then the suppression risk is considered low. Any rural residence which is located greater than 300 feet from any main road is considered as high no matter the distance to the nearest rural fire station.
RESPONSE POTENTIAL |
ALL WEATHER ACCESS |
SEASOAL ACCESS |
LIMITED ACCESS |
NO ACCESS |
WITHIN 5 MILES |
LOW RISK |
LOW RISK |
MODERATE RISK |
HIGH RISK |
OUTSIDE 5 MILES |
MODERATE RISK |
HIGH RISK |
HIGH RISK |
HIGH RISK |
The Risk Matrix (Table 8) collapses the Threat Matrix and Suppression Matrix into a final matrix for the overall Risk Potential. Table 8 shows the potential risk categories assigned by the this matrix which qualifies the degree of risk posed. First by the threat from wildfire ignition and the volatile nature of the fuels and secondly by the prospect of adequate suppression forces gaining access to the wildfire scene in a timely manner. For example if the Suppression Matrix had a high geospatial risk potential and the Threat Matrix had a corresponding high geospatial risk potential then the Risk Potential for that location was considered high.
THREAT MATRIX |
SUPPRESSION MATRIX HIGH |
SUPPRESSION MATRIX MODERATE |
SUPPRESSION MATRIX LOW |
HIGH |
HIGH RISK |
HIGH RISK |
MODERATE RISK |
MODERATE |
MODERATE RISK |
MODERATE RISK | MODERATE RISK |
LOW |
MODERATE RISK |
LOW RISK |
LOW RISK |
Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | home
