Wednesday, August 08, 2007

The West is Burning

The West is Burning: Exploring AIR’s U.S. Wildfire Model



This summer, much of the western United States is in flames. Wildfires spurred by above-normal spring and summer temperatures and one of the worst droughts in decades have charred more than 4.5 million acres across the 11 western states. Between July 16th and 19th alone, 1000 new blazes ignited, prompting fire officials to raise the nation’s fire preparedness level to 5, its absolute highest.

The rise in singed acreage is largely due to bone-dry conditions, exacerbated by the Forest Service's policy of quickly halting fires that threaten residential areas. This policy permits accumulation of dry brush—perfect fire fuel. The most common cause of wildfires today, however, is human activity, including illegal campfires in heavily vegetated areas. The latter was the source of one of the summer’s most costly blazes, which occurred in late June near South Lake Tahoe, California. Dubbed the “Angora Fire,” the event destroyed 329 buildings and ravaged 3,100 acres. The U.S. Forest Service had initially hoped to contain it within a week, but what began as a small brushfire soon intensified into a full blown inferno. It raged for more than ten days, destroying nearly every structure in its path, before crews finally contained it. At the time, AIR estimated that total insured losses from the fire would likely exceed $150 million.





The Angora Fire highlights a worrisome trend: while the frequency of U.S. wildfires has remained relatively constant over the past several decades, wildfire-driven losses have significantly increased. Since 1980, wildfires in the U.S. have been responsible for insured property losses exceeding $10 billion.


This increase in losses is driven by the increase in the number and value of exposed properties in a high-risk construction area known as the wildland/urban interface (WUI)—a buffer zone where human development intersects dense woodland vegetation. Fire here can move readily between structural fuel and vegetation fuel, facilitating unusually rapid spread. Well over half of California wildfires occur in WUI zones.

Meanwhile, Americans continue to move in, building well-equipped first and second homes, particularly in the fire-prone West. Contractors in the WUI aren’t shirking from building opportunities either; between 1990 and 2000, the rate of construction on WUI-designated land exceeded that of non-WUI areas by a factor of three. The rush to build results in a convergence of risk factors: an unprecedented accumulation of fuels in areas of increasing population and property. This trend is similar to the one seen on coasts where Americans have an undaunted inclination to build vacation houses on hurricane-prone beaches.




Complicating risks for residents in fire-prone zones is the unpredictable nature of the wildfire peril; a sudden ignition can result in either a moderate blaze or a catastrophic event depending on atmospheric conditions, topography, and moisture content of local vegetation. Furthermore, fires can drive losses disproportionate to their size. For example, the 2003 Cedar fire was the largest in California history, destroying more than 270,000 acres. But it caused just $1.2 billion in insured losses, making it far less devastating than the $3.1 billion Oakland Hills fire in 1991, which scorched less than one percent of the acreage but ranks as the most expensive fire event in U.S. history.

According to the National Interagency Coordination Center (NICC), many tens of thousands of fires occur each year. And the risk of catastrophe property losses is increasing as population growth in the WUI continues apace. To help insurers more accurately quantify potential losses and support underwriting guidelines, AIR released its US wildfire model in 2006.


The AIR Wildfire Model for California




The AIR wildfire model utilizes a fire spread algorithm to simulate how a blaze will spread once it is ignited, as well as extensive maps outlining wildfire history to aid in predicting where fires will start. Historical data was pooled from, among others, two complementary sources: the California Department of Forestry and Fire Prevention (DCFFP) and the U.S. Forest Service (USFS).

In addition to historical information, the model is shaped by five key factors—the features governing wildfire frequency and severity. These are 1) ignition number and location, 2) vegetation, 3) weather conditions, 4) topography, and 5) fire suppression activities.

Ignition

Ignition can result from natural causes, like dry lightning—the source of most of this season’s wildfires—or human activity, such as the illegal campfire that spurred July’s Angora blaze. As previously mentioned, the vast majority of loss-causing wildfires occur in areas designated as WUI. WUI zones were originally designated as such by a 2001 U.S. Department of Agriculture and U.S. Department of the Interior report on communities at risk from fire. WUI boundaries are not well defined, however, and continue to change as the population disperses.

The maps below show the locations of historical California wildfires. The leftmost map shows California Department of Forestry and Fire Prevention (CDFFP) data overlaid on California WUI areas, shown in green. The CDFFP data consists of locations of fires larger than 300 acres from 1900 to the present, and fires larger than 20 acres from 1905. The center map adds the U.S. Forest Service (USFS) data for fires exceeding one acre, and the rightmost map shows all USFS-reported fires, regardless of size. The USFS data only covers fires reported from 1986-1995.


Vegetation (Fuel)

Fuel, another model parameter, is classified into different types of wildland vegetation, including coniferous trees, grass, and chaparral. Each burns at a unique rate and generates flames of different intensities. Grass, for example, does not hold moisture particularly well and can dry out even in a short drought, quickly transforming into ideal tinder through which flames can zoom.






In forests, fire spread rates are more complex. Forest fires may be fairly slow to spread depending on the undergrowth, but they can also be exceedingly difficult for fire-fighters to access and put out. Additionally, if forest fires manage to spread vertically into the canopy, they may become full-fledged crown fires, which move at incredible rates through treetops and can be virtually uncontrollable.

Weather Conditions

Wildfires in California are highly seasonal due to variation in temperature and precipitation within the year. During the six months from May through October, the average maximum temperature rises considerably. During the same time, little or no rain falls. This lack of precipitation removes moisture from vegetation until it reaches a very dry state, which in turn increases both the probability that wildfires will occur, and the rate at which they spread.




Seasonal winds represent another influential weather component. They can quickly revitalize a blaze, depending on their speed and direction. The Cedar fire of 2003 provides an excellent demonstration; its early growth was driven primarily by the presence of strong Santa Ana gusts—hot, dry easterly winds unique to California. Fortunately, Santa Anas are often short-lived. In the case of the Cedar fire, they were soon replaced with westerly winds from the Pacific Ocean, and this reversed the fire’s direction. The AIR U.S. wildfire model allows for such shifts. It incorporates historical data on average hourly wind speed and direction from NOAA Cooperative Observation Program (COOP) weather stations.

Scientists understand how fuel and weather affect a fire once it is ignited, but understanding just how it will spread so as to predict its final footprint is still a challenge. Fortunately, scientists have a considerable amount of data on fire size, location, and shape, as well less extensive data on wildfire duration and associated wind speeds. The model inputs these variables in the fire spread algorithm, which successively refines the initial approximation that on flat ground, and in uniform fuels, a fire will spread in an elliptical pattern. Wind speed and direction, along with local slope, are the major determinants of the ellipse’s shape. Vegetation factors in, too; as was noted above, fire spreads faster in some fuels than others, and if a sufficiently large region of non-fuels is encountered by a moving blaze, spread is halted accordingly.

Topography

Topography also affects how fire spreads. Fire travels upslope more readily than it does on level ground because flame advancing at an upward angle encounters a larger cross-sectional area of fuel than flame on level or downward sloping ground, feeding its growth. Additionally, trees and grasses immediately upslope from a smoldering fire are preheated by sweltering winds. These winds rob fuels of their moisture, turning them into perfect tinderboxes. The AIR model incorporates topographic data from the U.S. Geological Survey (USGS). These data are used to magnify fire spread when flames are traveling up-slope and to reduce it if a fire site is on level or downward sloping ground.


Fire suppression

The fifth and final parameter contributing to the overall wildfire picture is fire suppression—the suite of tactical techniques used by firefighters to halt expanding flame. These include clearing underbrush, airdropping fire retardants, and building fire breaks, or gaps in vegetation.



Decisions regarding where and when to deploy firefighting resources are influenced by fuel conditions and weather forecasts. Officials review information on how a fire is developing, too, and where it is headed in order to identify potential locations at which to fight a particular blaze. In unpopulated areas, wildfires are often allowed to burn out. This eliminates fuel buildup. In areas closer to human habitation, however, fire policies are aimed at extinguishing fires as quickly as possible. AIR’s fire spread algorithm approximates human decision-making with respect to fire suppression based on such trends.

Damage Estimation



Using a synthesis of the factors outlined above, AIR scientists are able to simulate a wildfire’s perimeter—the area a fire could ultimately affect. A complementary analysis predicts the intensity with which the fire will impact points in the perimeter’s interior. Predictor variables (flame length, fire spread rate, heat per unit area, etc.) are computed using the USFS wildfire simulation program, FlamMap, as well as topographical and road access data provided by the ISO FireLine tool. Combinations of these variables are analyzed in order to determine how they compare to historical fire pictures described in damage reports. In this way, a point-intensity index with a value between zero and one can be derived.

This value is further refined using AIR damage functions, which produce estimates of fire-driven loss by incorporating the structural characteristics of specific properties, including vulnerability of different roofing and siding materials. Since people continue to populate fire-prone woodland spaces, engineers have identified construction materials that are fire-resistant. Unfortunately, many homes today still have wood siding and either wood or asphalt shingle roofs, both of which are highly susceptible to ignition.

In addition to building materials, AIR damage functions also account for set-back distances—cleared space separating a home from the surrounding vegetation. A post-disaster survey conducted by AIR investigators revealed that most of the fire-ravaged homes in the Angora blaze did not have these cleared spaces. Of the homes that did survive with minimal damage, complete or partial setbacks were in place.

Conclusion

The 2007 wildfire season is shaping up to be one of the most destructive in recent memory, threatening homes, power lines, and communication towers across the western U.S. The AIR wildfire model helps to quantify the risk driven by such catastrophic events.



.MGW.