Making sense of hurricane analysis

Major forecasters expect a below-average North Atlantic hurricane season, but a quiet year by no means rules out a devastating, Category 5 landfall which could cost thousands of lives and billions of dollars. Meanwhile in the East and Central Pacific, the occurrence of a powerful El Niño supports an active Pacific tropical cyclone season. According to conventional wisdom, though, a powerful El Niño fosters a milder North Atlantic hurricane experience.

No matter what is forecast, 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 layered outputs of hurricane-season analysis, Willis Re held a 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.

The webinar moderated by Scott St. George, head of Weather and Climate Research at the Willis Research Network. He was joined by:

• Dr James Done, a Senior Fellow at the National Centre of Atmospheric Research in Boulder, Colorado, and an expert in the physical modelling of tropical cyclones.
• Professor Ralph Toumi, Director of Imperial College London’s Grantham Institute, one of the world’s leading tropical cyclone research centres.
• Jessica Boyd, Head of Model Research at the Willis Research Network.
• Dr Cameron Rye, Director of Natural Catastrophe Analytics at Willis Re.

The panellists set out to explain how the modelling and research communities make predictions about the behaviour of North Atlantic hurricanes. The aim of their conversation was and remains to helps insurers deploy science-based insights to make better decisions about the risk of hurricanes to insurance portfolios in North America, the Caribbean, and Central America.

Dr Done began by explaining that hurricanes respond to three environmental factors: ocean temperatures, winds aloft, and moisture in the atmosphere. The basis of hurricane forecasting lies in understanding how hurricanes in the historical record responded to these three factors.

One major class of modelling does this by applying statistical approaches. Forecasts are made based on current observations of the three environmental factors and historical responses to them, or to similar mixes of conditions. This approach is limited by the historical experience, however, and by the assumption that the future will resemble the past.

The second forecasting methodology uses process-based modelling. The approach captures the physical processes by which hurricanes respond the three environmental factors, then, based on the laws of thermodynamics, models the processes forward in time to forecast the hurricane season ahead.

This can help to overcome the challenge of the limited record of actual hurricane experiences. “Our detailed history of hurricanes is short, 150 years at maximum, with only about 50 years of reliable intensity data, since satellite measurements were made,” WRN’s Jessica Boyd told webinar viewers. “The power of process-based models is their ability to extrapolate beyond those experiential events to reveal credible alternative hurricanes.”

Dr Done argued that this approach provides a broader, more expansive view of what’s possible in the year ahead.

Regardless of the forecasting methodology deployed, everyone is calling for a quiet season, with few tropical cyclones and few hurricanes. “Nearly all forecasts call for below-normal numbers, somewhere in the low teens”, said Dr Done. “If this comes to pass, we will notice, because most of the past ten years have seen an above-average number of named storms.”

As has often been reported in recent months, expectations are for a powerful El Niño Southern Oscillation this year. Indeed, unusually warm ocean surface temperatures in the central and eastern tropical Pacific Ocean have already been noted, marking the commencement of the phenomenon. Such years tend to produce fewer North Atlantic cyclones.

That’s one reason the frequency of named storms is expected to be lower than average during the 2026-7 hurricane season, although forecast numbers remain high relative to the expected power of the El Niño.

The temperature of the North Atlantic is the other influencer. “Ocean heat is the local source of energy for hurricanes”, Dr Done told the webinar. “This year it looks to be about average, with some parts of the North Atlantic cooler than normal, and some parts warmer. It will probably provide its usual foundational support for hurricanes.”

Climate change may inject chaos into the El Niño-hurricane relationship. The norm was disrupted in 2023, when an El Niño coincided with the fourth-most-active hurricane season on record, because the Atlantic was very hot.

“Don’t expect a repeat”, Dr Done advised. “This year is not as hot, so we’re expecting the El Niño dampening effect to occur.”

That said, he admitted, modelers have little recorded experience of the forthcoming combination of opposing forces. That means their uncertainty about how the atmosphere will behave is higher. Here the process-based, extrapolating models come into their own, because they are able to forecast beyond the historical experience. This year they forecast that the hurricane season will be supressed by the expected strong El Niño. The lingering question is how much.

One strength of the Willis Research Network – which one of the world’s longest-running and most influential partnerships between academia and the insurance industry, and is available to support all Willis Re clients – is that it provides multiple perspectives on specific risks of significance to the insurance industry.

Professor Ralph Toumi is Director of Imperial College, London’s Grantham Institute, the largest European research group on tropical cyclones, and a founding director of the open-source Oasis Loss Modelling Framework. He zeroed in on the climate change point.

“Researchers draw completely different conclusions about the multiple factors involved in the interaction between climate change and hurricane season intensity, because they have made different assumptions”, he said during the webinar. “Unfortunately it’s not clear which assumptions we can eliminate. That creates challenge and confusion. It’s a complex, multifactorial system.”

For example, Professor Toumi said: “There’s a strong debate about whether climate change makes hurricane numbers go up or down, and about its influence over changes in tracks. However, we are quite sure that warming makes cyclones stronger.”

That is a useful conclusion, but on its own is not terrifically helpful. Meanwhile hurricane basin predictions are high profile, but the real interest is in landfall. Current pre-season forecast models simply cannot predict in-season weather patterns, or steering flows, months in advance. Predictability is at best a week or so, making it very difficult to forecast when, where, and how many hurricanes will come ashore.

Professor Toumi presented the hybrid ‘IRIS’ model to the webinar. The risk model is used by re/insurers to provide an alternative view of risk to those provided by major model vendors. It deploys a statistical data approach, in combination with the thermodynamic modelling characteristic of physics models. It incorporates a super-ensemble of all of the historical tracks since 1950, and includes slightly “perturbed” tracks at multiple different intensities, limited by the thermodynamics. That creates a very large counterfactual universe to help the Grantham Institute team focus on the “end game” that is landfall. It cannot predict landfalls, of course – no model can – but it helps analysts to understand them.

They use IRIS to analyse hurricane-strength cyclones at the point after their maximum intensity has passed, but before they make landfall. That, Professor Toumi said, “makes landfall a problem of hurricane decay, not of intensification”, which makes a big difference.

Modellers can change potential intensities within IRIS, then rerun the history to see the impact. Similarly, tracks or counts can be altered. “By isolating the components of the problem, we can explore potential impacts from new perspectives”, he explained.

The team re-simulates historical tracks, incorporating the well-established thermodynamic constraints which determine potential intensity, and allowing for hits and misses. They have tuned the model to focus separately on tracks during El Niño, neutral, and La Niña years.

In the real world, hurricane track characteristics are not evenly distributed between El Niño and La Niña, the Professor said. This can be simulated in the modelled world of IRIS. El Niño-like tracks can be weighted more, for example. When run against a basic damage model, the results of this exercise were surprising.

“If you experience fewer hurricanes because of El Niño, the output should be less damage overall, but we found more damage in multiple areas, including East Florida, and even the Northeast”, Professor Toumi revealed. Initially he thought it was an error, but investigation uncovered a climatic explanation.

El Niño drives high wind shear (sudden, drastic change in wind speed, direction, or both) across the world in the area known as the Main Development Region (or MDR), where many North Atlantic hurricanes form. That reduces storm genesis there, because the sudden, powerful wind cuts the tops off developing hurricanes, preventing them from developing further.

“They can’t really even get going,” he said. However, nearly half of the actual tracks in the model that made landfall in East Florida since 1950 had their genesis outside the MDR. The assumption could be that they were unaffected by the El Niño wind-shear effect, and threatened a highly developed coastline, leading to the higher-than-expected modelled losses.

Furthermore, the team found that the shear phenomenon reduces slightly north of the MDR. The shear effect is very strong in the Caribbean, but has the opposite impact as formations near Florida. In support of the shear hypothesis, Professor Toumi’s team discovered that hurricanes that made landfall in Texas and East Florida formed primarily within the MDR.

“We may see a major El Niño this season, but plenty of storms formed outside the MDR over the past 75 years, and their time from genesis to landfall in East Florida is remarkably short.” When the correct atmospheric conditions converge, that could occur any moment, and lead to an enormous hurricane hit on somewhere like Miami, Fort Lauderdale, or Palm Beach.

Professor Toumi emphasised the hypothetical nature of this conclusion, which could explain the disagreement between hurricane activity at the whole-basin level, and incidence on smaller geographic scales.

Another factor challenges the simplistic El Niño = fewer hurricanes correlation. The relationship looks strong for the past 50 years, but if you go back in time before 1980, it’s not such a good fit. “The shear mechanism still works in those earlier years”, Professor Toumi said, “so there’s clearly something else going on.”

That something was not revealed during the webinar, but serves to underline the fact that a specific temperature change as a result of El Niño does not ensure a direct reduction in hurricane numbers.

It also underlines the importance of realising that assumptions may not always play out as expected. As WRN moderator Scott St George concluded, “It’s certainly a good reminder to keep questioning our assumptions, and to be careful about the things that we think that we know.”

Watch out for the second article covering this webinar, where Willis Re experts Jessica Boyd and Dr Cameron Rye advise on the practical application of learnings from the scientific analysis of hurricanes.