Wildfire modelling plays an important role in managing wildfire risk, but it is just one piece of a much larger puzzle. Effective risk management also relies on mitigation efforts and consumer education, insurance experts explained during an InsuranceERM webinar held earlier this week.
Emma Watkins, head of catastrophe risk at Lloyd’s of London, said wildfire risk modelling in the insurance market is “relatively nascent, but rapidly developing”.
In 2018, she explained only a small number of syndicates at Lloyd’s were using fully stochastic probabilistic models for wildfire risk.
However, by 2024, Watkins said: “Over 60% of the Lloyd’s market was using a stochastic probability model [for wildfire]. Wildfire, especially in the US, is increasingly a peak peril for the Lloyd’s market, so now we are incorporating it into our explicit, detailed modelling for both overseeing the market and in our own internal capital model.”
In Watkins’ view, the insurance market needs to do much more to understand, validate and have confidence in wildfire risk models that are released.
“What we do need to do, though, is make sure that we don't become overly reliant on probabilistic models. We will always need to consider pure aggregation of exposure, or realistic disaster scenarios so that we don't end up feeling surprised when things like the LA fires [ happen],” she added.
Alice Kane, superintendent of insurance for the US state of New Mexico, emphasised the need for a total system in responding to wildfire risk. She said this involves community rules that support wildfire mitigation, as well as building codes that meet [the US] Insurance Institute for Business and Home Safety standards.
Kane added: “We’re trying to educate consumers that the reinsurance market is critical to the direct insurers, and what risks they are able to write.”
Commenting on technology advances in wildfire risk management, Sarah Russell, managing director at Bellwether, a mapping and spatial analysis project at X, Alphabet’s moonshot factory, said: “The new technology landscape is not just in the modelling space. For example, there are environmentally friendly fire suppression chemicals and some people are using drones to follow and look for [wildfire] ignitions.”
Bellwether offers scores to clients that deal with the probability of wildfire at an address in the future.
During the webinar, Russell explained: “Our approach has been a machine learning approach. It's a subset of AI which essentially says, let's use algorithms and feed a ton of data into a machine and have it understand, based on all the data, the probability of an event in the future. This is not a prediction that something will happen. It's a probability between zero and one.”