Modeling Fire Risk: A Hot Topic at RAA Cat Risk Management 2019

The 2017 California wildfire season was an unusually bad one, setting records with 9,133 fires burning more than 1.4 million acres of land and damaging or destroying more than 10,000 structures.

Wed Mar 13 15:27:37 EDT 2019
,

The 2017 California wildfire season was an unusually bad one, setting records with 9,133 fires burning more than 1.4 million acres of land and damaging or destroying more than 10,000 structures.

Then 2018 happened. It established a new record for burned acreage in a single California fire season: 8,527 fires burned 1.9 million acres, with 23,647 structures destroyed, of which 17,133 were residences. The Insurance Information Institute projects that insured losses from just two of the 2018 California fires — Camp and Woolsey — are likely to be among the costliest on record.

Given this one-two punch in catastrophic wildfire years for California alone, it was no surprise that wildfire modeling was a hot topic at the Reinsurance Association of America’s (RAA’s) 2019 Catastrophe Modeling Conference. There were six breakout sessions at the conference focusing on wildfire modeling, including one hosted by Pitney Bowes.

Wildfire modeling can help insurers and reinsurers manage risk by quantifying and mapping such variables as slope, aspect, vegetation type, burn frequency, distance to water, special wind event regions and distance to fire stations, among others. Depending on the modeling product and the algorithms it uses, the results are map-based polygons with associated risk scores.

Some modeling tools provide just a risk score. Others, for example Fire Risk Pro from Pitney Bowes, provide detail into the data behind the score. This allows cat modelers and underwriters to understand that a fire line exists because there is a lot of brush or fuel in an area. It also will show why a line ends — that the model predicts it won’t jump a major highway, for instance, or cross a ridge.

But knowing where wildfires are likely to occur, and the paths that they are likely to take, is only part of the story for underwriters. As urbanization and “suburban-ization” move closer to areas with high wildfire probabilities, traditional techniques like placing zip code-wide moratoriums on new business may exclude large numbers of potentially profitable policies. In areas like California, with high residential values, that can mean leaving five-figure premiums on the table.

The second part of the story is knowing with precision where the property under consideration is located relative to the fire risk area. It can come down to knowing which side of the highway an insurer or reinsurer is willing to cover, versus a zip code or an entire county. With the combination of precision location data and precision wildfire modeling, a company can find insurable “gems” that can be written profitably. This holds true no matter where the high-risk wildfire areas are located.

While the California fires garnered most of the headlines over the last two years, wildfires happen across the western US, along with other high-risk areas scattered across the southeast and the Atlantic coastline. In fact, according to data reported by the Insurance Information Institute, wildfires burned in every US state in 2018 except Delaware and the District of Columbia.

Carriers have been hurt when they took the position that fire risk didn’t apply to them. For example, although small relative to California burns, the 2016 Chimney Tops 2 Fire in western Tennessee cost 14 lives, burned 17,000 acres and damaged 2,500 structures, resulting in $992 million in insurance claims.

At Pitney Bowes, our fire risk hazard and risk assessment database covers the entire United States, including Alaska and Hawaii, and is refreshed annually. Insurers can assess risk exposure for new business throughout the geographic distribution of their insured network and identify potential damage to structures in areas where wild land fuels continue to accumulate. Reinsurers can use Fire Risk Pro to evaluate a carrier’s current book of business. The common factor across these use cases is using data and sophisticated modeling to make better business decisions, structure by structure as well as area by area.