Fire screen: new tools give insurers better information for underwriting areas prone to brush and forest fires.

The property losses in recent years in California due to terrible brush and forest fires have insurers wondering if their underwriting practices are sufficient. The loss of homes and businesses in areas previously considered safe brings into question what was thought of as state-of-the-art tools, data and decision-making. The brush and forest fires of 2003 seemed to move in random patterns pushed by changing winds. Houses blocks away from brush and trees of any significance were destroyed. Valleys, slopes on all sides of hills, and high plateaus all suffered, as shown in "Victims in High and Low Places" on page 121.

If there is any good to come from the losses, it is that now experts have enough data to change the way property insurance is underwritten in brush and forest fire-prone areas. Geographic Information System tools can be used to analyze maps and data of burned areas, lost buildings, topography and wind patterns. The result is a new set of underwriting guidelines and new tools for insurers.

As society lays claim to undeveloped land, a conflict has erupted between the forces of nature and a desire to control the environment. One of those forces, fire, has been a part of the cycle of life as much as evaporation and rain. Early prairie fires burned for hundreds of miles. Native Americans used fire to clear land. Lightning strikes and other weather phenomena started fires that consumed thousands of acres each year.

This was not a problem until people moved into the natural path of the fires. Now society has a need to stop nature from taking its course. Insurers have a concomitant need to make sure they are appropriately compensated for assuming brush and forest fire-related exposure, and that they avoid writing policies for some risks altogether.

Losses have skyrocketed not only from people's encroachment into the woods, but also from the types of dwellings being built. Homeowners have made the transition from a city home with a smaller weekend residence nestled in the hills, to huge primary homes that rely on the forest to enhance the architecture. Nature has not changed in thousands of years. The thought that people can somehow modify the winds and keep fires from making their annual appearance is just that, a thought.

The Solution

Insurers are faced with these questions as they streamline the underwriting process. Decisions need to be made quickly with cold, hard data analyzed hundreds of miles from the risk location. The act of looking at maps and making subjective interpretations slows down the underwriting process and introduces the potential for error. Automated decision-making is a data-in/data-out process. If insurers have to look at a map, the automated systems have failed.

Insurers have all the tools they need to address this issue. From a single data entry point, every conceivable piece of information can be gathered, decisioning tools execute corporate underwriting guidelines, rates are determined and policies issued or declined. Underwriters have become accustomed to using geographically based information, such as the distance to the ocean, because there has not been data on brush and forest-fire potential that worked well in automated systems. Now there is.

Some progress toward the goals of living with nature and having insurers survive financially has been achieved with technologies such as satellite imagery. One of the first to attempt this approach was Insurance Services Office with the development of FireLine. This product used satellite imagery, slope and a digital road map. The result was a score that incorporates the fuel for brush fires (trees, brush and grass), the slope of the ground at the risk address and an indication of any problems in getting fire-fighting equipment to the risk. While FireLine was a good early attempt, it omits some factors and, some believe, overemphasizes others.

The next step in arriving at a complete solution was developed by the California Department of Forestry and Fire Protection and adopted by Marshall & Swift/Boeckh. That model starts with satellite imagery that sorts the vegetation into species by an analysis of the sunlight reflected from Earth to a satellite. The reflected light spectrum of each species is slightly different. In fact, spectrums can be divided into more than 500 species. They are then grouped by similarity of burn characteristics, and this is where the CDF model begins to deviate from ISO's.

Material Factors

Firefighters know there are two different types of wild land fires: brush or surface fires, and crown fires. To illustrate the difference, the vegetation species must be grouped by their ladder and crown index. The ladder index indicates how far up the tree or plant leafy vegetation (fuel) is found. For most pine trees, it is very close to the ground; for tulip poplars, it can be more than 100 feet before the first branch. The crown index tells how much leafy vegetation can be found in a certain species. Some trees will have a higher ratio of fuel than others. Brush fires can go through forests and never reach the crowns of the trees high above because they are ground level while forest fires consume the higher crowns.

The more complete system used by MS/B incorporates the species, and the ladder and crown index. But what about the concept of slope? The original thought was that since fire travels up hill much like heat going up a chimney, the risks located on hillsides were particularly susceptible to fires. Although it sounds like good science, an analysis of fire losses does not bear this out. While the cost to rebuild after a loss is higher on a sloped property, there does not seem to be a major increase in the likelihood that a home on a slope will be lost to brush and forest fires over those on flat ground. Similarly, no cases were found where firefighting equipment could not get to homes due to sharp curves in public roads used in previous models. Some private roads and long driveways have caused problems, but they are not part of any existing models used today.

It was also believed that the aspect or direction of the slope was important. Aspect is an important part in determining the origin and early spread patterns of fires. South-facing slopes in the Northern Hemisphere receive considerably more sunlight than northern ones. This makes them drier and makes them more likely to be the site of ignitions and early spread. Once a fire is well established, aspect is less important. The United States Department of Agriculture concluded that neither slope nor aspect had any major bearing on loss of vegetation due to fire.

Besides fuel, the dominant factor in fire spread is wind. But, wind is not a constant, in neither speed nor direction. In the California fires of October-November 2003, changes in winds hampered firefighters and sometimes even put them in hazardous situations.

The Santa Ana winds blow from the northeast to the southwest in southern California, but several other wind patterns that affect fires blow in many directions. Diurnal winds have more daily cycles and, while usually not as strong as the Santa Anas, play a major part in determining fire spread. At night, winds blow from the land to the ocean and from higher elevations down slopes to lowlands. During the day, as the land heats up, the air movement is reversed. The Coriolis effect creates a different wind pattern. The spinning of the Earth and the drag produced by it move wind in predictable patterns. In southern California, the Coriolis-effect winds can blow in the opposite direction of the Santa Ana winds. If Santa Anas slow down, the Coriolis winds take over and blow back in the opposite direction. A report by the Los Angeles Fire Department on the Bel Air-Brentwood fire of 1961 indicated, "As the strong northeasterly winds died during the evening, moderate breezes began to flow upslope from the southwest. This had the effect of carrying the fire northward up the canyons toward Mulholland Drive." In the development of predictive tools, dominant wind patterns should not be ignored. More advanced models like CDF's incorporate them plus localized seasonal weather information such as humidity and rainfall.

The effect of these wind patterns can be seen in burn patterns, ff the Santa Ana wind were the only one of concern, burn patterns on the ground would fan out in a single direction like a smoke plume from the ignition point in the same direction as the wind. "Where the Fire Went" on page 122 shows a time progression of a bum area. With a better understanding of the types of winds, a viewer can almost determine when one slowed down or stopped, and the other took over. Since insurers cover a property for an elongated period of time, the effect of all types of wind, not just the Santa Ana, must be considered.

The same technology that could help insurers better understand fire spread has created an interesting problem. Satellite imagery produces cells of data approximately the size of a baseball diamond. Imagine how many there would be for an entire state like California. The data files are huge. But the problem is that a small patch of very flat, nearly vegetation-free land can fall within a larger area of dangerous brush and trees. GIS technology converts the risk address into an exact place on the face of the Earth. If that point falls within the "good" cell and that information is returned to the underwriter or policy issuance system, the wrong decision may be made. In some studies done by MS/B, as many as 80% of the losses were in cells that were classified as Non-Fuel or Moderate hazard potential. Winds have been reported to blow firebrands (ignited matter) hundreds of feet. And what if a house a few blocks away from the edge of a brush/forest fire is ignited and the firebrands from it travel a block or two to set ablaze another home?

To overcome this, another piece of data, the distance to the nearest area of worse fire hazard, is needed to develop new underwriting guidelines.

The Next Step

Investigators have gained much knowledge of fire behavior and how it relates to property underwriting. The existing data is amazingly precise. The missing piece is an additional source of fuel for fire spread that until now has not been statistically included: structures themselves. Fire spread from home to home has been known to exist since people started to build cities. The same structure-to-structure fire that consumed such cities as Chicago, London and San Francisco transforms a brush/forest fire into a suburban conflagration. In the Bel Air-Brentwood fire, streets were reported to be "tunnels of heat and fire."

Insurers need a way to quantify the hazard level of a risk location that takes into account all of the vegetative, topographical and weather-related concerns as well as structures as fuel. The CDF style ranking of fire hazard potential needs to be modified by data on the structures between the risk and fire source (nearest worse area).

The missing index can grow in complexity based on an insurer's capacity to use the additional statistical data. The most basic index incorporates only the existence of deliverable addresses (homes usually) between the risk address and the hazardous area. Is there just a flat open area with no vegetation, or are there homes as possible fuel? But how many homes are there? The next level of sophistication requires that investigators understand the density of structures between the risk and the fire source. Homes with large spacing are less likely to spread fire than tightly packed developments.

The best answer is the one hardest to calculate. Besides knowing that homes are present, and if they are tightly packed, investigators need to know what they are made of. Masonry homes with the roofs are less likely to catch fire, and therefore spread fire, than frame homes with wooden siding and hand-split cedar shingles. Combined, these factors create an index that can be used in an automated underwriting system. MS/B is currently developing brush/forest fire indices for California and many other western states based on structure density and building characteristics.

Fires in the wild land are a fact of life. As society develops more rural areas, fire losses will continue to increase. Insurers must find a systematic approach to determine which risk is acceptable, and a defensible method of establishing a premium rate. Early attempts at using satellite imagery and GIS technology gave insurers great insight. More recent analysis and modeling have produced even better predictive tools. This trend of combining science with observations of the effect of brush and forest fires should lead to data that can be used by today's automated underwriting and policy issuance systems.

Insurers and others must never become complacent and believe that because they have analyzed past fire behavior and introduced cutting-edge technologies they are in control. The best they can hope for is to develop guidelines and processes that help keep everyone out of harm's way both physically and financially.

Victims in High and Low Places

The California brush and forest fires of 2003 seemed to move in random patterns. Valleys, slopes on all sides of hills and high plateaus all suffered. Houses in the illustration represent losses in Suncrest and Harbison Canyon, Calif.

[ILLUSTRATION OMITTED]

Where the Fire Went

The effect of wind patterns can be seen in burn patterns. In
this burn area progression in southern California, the fire
obviously moved in different directions at various times on
different days. Times are in military notation.

Riverside County

Pacific Ocean

Cedar Fire Progression

October 2003

October 25, 2400
October 26, 0200
October 26, 0300
October 26, 1000
October 26, 1900
October 26, 2100
October 26, 2200
October 26, 2400
October 27, 0600
October 27, 0900
October 27, 2200
October 28, 1600
October 28, 1900
October 30, 1800

Source: Marshall & Swift/Boeckh

Key Points

* Losses from brush and forest fires have skyrocketed, not only from people's encroachment into the woods, but also from the types of dwellings being built.

* Insurers now have sophisticated data-gathering and analysis tools to evaluate most factors that contribute to making particular areas prone to fire.

* An important piece of information missing from this analysis is a way to quantify structures as fuel.

Charles Sharp is a consultant with the Underwriting Decisions Group at Marshall & Swift/Boeckh, Princeton, N.J.