The wealth of information delivered about North Atlantic hurricanes at this time of year may be overwhelming. Experienced reinsurance buyers with hurricane exposure understand that nothing is certain. To compound the certainty challenge, as hurricane science and modelling advance, analytical outputs have become more nuanced, and therefore require greater judgement to interpret and apply to reinsurance purchasing decisions.
To help make sense of the nuanced outputs of hurricane-season analysis, Willis Re held a hurricane-focussed webinar just weeks into the 2026-7 Atlantic wind season. It brought together foremost experts in these sciences from Willis Re and the Willis Research network, which concentrates the knowledge of multiple specialists around the world for the benefit of Willis Re and WTW clients.
Leading hurricane scientists Dr James Done and Professor Ralph Toumi provided an overview of the latest hurricane modelling approaches. The pair explained how forecasting teams around the world used models to arrive at the flurry of forecasts that suggest a lower-than-average number of hurricanes will form during the 2026-7 season. They then considered some of the many anomalies surrounding the headline predictions.
Jessica Boyd, Head of Model Research at the Willis Research Network, and Dr Cameron Rye, Head of Natural Catastrophe Analytics at Willis Re, joined the discussion. They concluded the webinar by bridging the divide between science and commercial risk transfer. They eloquently put the scientists’ discussion into a commercial context. In part two of our reporting about the webinar, we present their advice.
No time for complacency
The current low-season forecast for Atlantic hurricanes follows several years of relatively low insured hurricane losses. Over the same period, the cost of Severe Convective Storms has been relatively high, exceeding $50 billion in each of the past three years. Because of that, the industry’s attention has understandably shifted to the latter.
“That doesn’t mean the underlying risk from hurricanes has gone away”, Dr Rye warned. With exposure growth, climate change, and inflation, the potential for a very large loss from a single North Atlantic storm remains high.
Dr Rye pointed out that the average global insured annual loss from across the perils has been between $120 billion and $140 billion for several years, but a loss of that magnitude from a single hurricane should not be unexpected. “The return period for that Atlantic hurricane loss is 15 or 20 years,” he said. “The reinsurance industry and beyond should be preparing actively for such a storm.”
Catastrophe models are designed to quantify the potential financial impact of catastrophic events on re/insurers. They do so by generating outputs including return periods for specific potential losses and estimates of annual average losses arising from specific events over a period of multiple years.
Some of the components overlap with climatologists’ hurricane models, but primarily they are complementary tools,” Jessica Boyd said. “Cat models may be used to price risk on an annual basis, whilst seasonal forecast models can be used to better understand how current climate conditions could affect baseline hurricane risk for the season ahead”, she explained. “They should be used in conjunction.”
Practical magic
She noted that the largest number of hurricane-strength landfalls experienced in a single season in the US is six, but that nothing physically prohibits seven or eight hitting the coastline. Cat models provide a set of physically plausible options that include these scenarios, allowing the impact of multiple landfalls to be quantified, and therefore included in tail-value pricing assumptions, reinsurance buying, capital allocation, and other insurance functions.
Climate change is another factor. “The historical record is not necessarily reflective of the climate regime we’re in at the moment”, Jessica said. “If we look only at history, we’re looking at different climate conditions.” To consider what’s likely to happen tomorrow or next year, climate models are essential to an accurate understanding of the changes and their impacts, and therefore the actual risk.
“Exposure is changing in terms of where people are building and the cost to rebuild properties. A lot of variation must be considered when modelling.”
Rebuild costs could come as a surprise in the wake of the next big hurricane, Jessica said. “We’ve seen huge increases in the rebuild cost of asphalt roofs, for example”, she pointed out. “If a big hurricane hits, many insurers will experience higher-than-expected claims. The cost may not be very well reflected in their calculations”, she said.
Multiple views
Dr Rye said having multiple perspectives, sometimes overlapping, sometimes different but equally plausible, may have great value. They inform multiple views of risk, which can grant insurers a broader understanding of their portfolio and its potential interaction with natural perils.
“When we make reinsurance decisions supported by multiple models which were built using different assumptions, we can very much better understand the uncertainties and what’s driving the tail risk”, he said. “They support sensitivity testing, perhaps of different reinsurance structures, to look at programme optimisation.”
Some models may, by design, perform better for certain lines of business or geographies than others. That’s another reason it is important for insurers to be presented with and consider a range of options to help them find the best solution to fit their own view of risk.
A key goal is to understand the different assumptions and how they impact results. Insurers will wish to match their own risk appetite with model assumptions, perhaps based on loss experience.
That’s just one area where the natural perils reinsurance experts at Willis Re add tremendous value.
