The growing impact of secondary perils such as flood and wildfire has become one of the most pressing challenges facing the insurance industry. Over recent years, both the economic and insured losses associated with these events have increased markedly. With climate dynamics continuing to evolve, the frequency and severity of floods and wildfires look set to rise further.
While the industry increasingly recognises this shift, awareness alone is not enough. Insurers must strengthen their approach to exposure management, modelling and portfolio stress testing, if long-term resilience is to be maintained.
Wildfire: Recognising Exposure and Modelling Gaps
The 2025 Los Angeles wildfires, with economic losses estimated at around $30 billion, served as a global wake-up call. Yet, as Nalan Senol Cabi, VP, Catastrophe Research Engineer at Arch notes, this was not an unforeseen anomaly. From her perspective, the event confirmed two well-established trends.
On the exposure side, urban development patterns are fundamentally reshaping wildfire risk. Between 1990 and 2020, the total area of the US wildland–urban interface (WUI) grew by 31%, while the number of homes in these zones increased by 46%, reaching 44 million. [1]
As Nalan explains, “We aren’t just seeing ‘more fire’; we are seeing human settlement aggressively encroaching into the fuel source.” The scale of destruction in Los Angeles, with more than 16,000 structures lost and over 42,000 insurance claims made[2], illustrates how concentration of exposure, amount of combustible fuel and severe winds can rapidly translate into systemic loss.
The second challenge lies in catastrophe modelling. According to Nalan, “The biggest gap today is urban conflagration. Current cat risk models are quite sophisticated at predicting how fire spreads through vegetation in the wildland. But there are challenges with the transition phase, i.e. when the fire leaves the trees and enters the city.” Once fire penetrates dense urban areas, structure-to-structure spread follows very different physical dynamics, and this complexity means tail risk is often underweighted.
Despite this, there are reasons for optimism. Nalan believes technological advances will accelerate progress. “I believe wildfire modelling is where flood modelling was five years ago,” she says, “but because we can now use AI to ingest these dynamic fuel changes and atmospheric signals such as Vapor Pressure Deficit (VPD) which is on the rise and it is directly correlated with wildfire risk, the catch-up speed of the models is going to be exponential.”
She adds, “We will solve the urban conflagration problem much faster than we solved the flood plain problem. One can argue we haven’t solved that yet but I’m certainly optimistic.”
Flood: Recognising the Protection Gap
Flood risk presents a different but equally complex challenge. In 2025 alone, approximately 200 largescale flood events were documented globally. Yet insured losses remained relatively contained.
This, Nalan warns, can be misleading. “We cannot mistake a lack of insured loss for a lack of risk. The hydrologic activity was high in 2025.” Many events occurred outside major exposure centres or in regions with limited insurance penetration.
Several events highlighted critical vulnerabilities. In Texas, the July flash flooding at Camp Mystic saw the Guadalupe River rise approximately 26 feet in under 45 minutes. In the UK, Storm Claudia caused severe riverine flooding in Monmouth, where water levels exceeded anything recorded since monitoring began in the 1970s.
Flood risk linked to tropical cyclones further complicates the picture. Nalan notes, “The current atmospheric conditions are making storms linger around land and dropping tremendous amounts of rain which can cause major floods.”
Hurricane Melissa in Jamaica and Cyclone Alfred in Australia both demonstrated how slow-moving systems with high moisture content can produce catastrophic flooding, even when wind damage is limited.
Beyond modelling, protection gaps remain a key concern. Cyclone Ditwah in Sri Lanka, for example, was one of the year’s worst humanitarian disasters, yet insured losses were minimal. As Nalan stresses, “The lack of insured loss there is a major protection gap.” Even mature markets face uncertainty, with mechanisms such as the US National Flood Insurance Program relying on short-term reauthorisations.
Taking Decisive Action
Taken together, flood and wildfire underscore the need for a more proactive industry response with greater portfolio interrogation, Nalan believes.
For flood, this means identifying the stochastic scenarios that break individual portfolios, conducting counterfactual analyses – ‘What if that rural Texas flood had hit Houston?’ – and prioritising high-quality geocoding. “Similar to wildfire, flood is a location specific, high-resolution peril,” says Nalan. “If you are sitting on ‘unknown’ secondary modifiers or zip-code level geocoding in 2025, you are underwriting blind.”
For wildfire, the focus must be on the wildland urban interface (WUI). Until models fully capture urban conflagration risk, insurers should manually stress test scenarios. “Model the fire jumping the highway into the suburbs,” Nalan urges, “and look at your potential losses.”
Ultimately, to ensure future resilience of the insurance industry, it is imperative that the industry acts decisively now by leveraging existing cat models, counterfactual scenarios and data betterment to head-off these rapidly evolving risks in the future.
[1] Los Angeles County Economic Development Corporation
[2] Los Angeles County Economic Development Corporation