Fewer hurricanes, lower risk?

Willis Senior Fellow Dr. James Done shares his views on the 2026 North Atlantic hurricane season, which began on June 1 and will extend through the end of October.

The North Atlantic hurricane season is officially underway. What can we expect this year and what should we be paying attention to?  As of the time of writing, dry and stable air is keeping things quiet over the North Atlantic. Meanwhile the equatorial Pacific Ocean is in a state of transition towards El Niño.

Almost all major forecasting centers have issued forecasts for a below-average hurricane season for the North Atlantic. In their April 9 forecast, Colorado State University (CSU) called for 13 named storms, six hurricanes and two major hurricanes. You can compare that prediction with average numbers over the period 1991-2020 of 14 named storms, seven hurricanes and three major hurricanes. The forecast from the National Oceanic and Atmospheric Administration (NOAA), which was issued on May 21, also leans towards a quiet season. They expect a 35% chance of a near-normal season, a 10% chance of an above-normal season, and a 55% chance of a below-normal season.

It’s a very different story over the East and Central Pacific. Here, El Niño supports active seasons. Hawaii may well be the U.S. state that experiences the most hurricane activity this year. Indeed, El Niño years on average see three times as many tropical storms in the vicinity of Hawaii compared to La Niña years.

As a climate scientist who has studied hurricane seasons for twenty years, I constantly remind myself that the historical record of hurricanes and other tropical storms is actually quite short. We only have a small sampling of what the hurricane season can throw at us. Every hurricane season is unique because of day-to-day, month-to-month and year-to-year changes in the larger environmental factors that influence the beginning, middle and end of each storm. And even if we had exactly the same environments from one year to the next, hurricane seasons would still be different due to the randomness of the hurricane response to the environment. The ultimate goal of seasonal hurricane forecasting is to anticipate how all these environmental conditions will influence the upcoming season.

This season, I’m paying close attention to three very significant environmental factors (Table 1). The most defining factor this year is expected to be El Niño. El Niño is the accumulation of hot water in the equatorial East Pacific Ocean. Air rises over the hot water and sets up an unusual circulation pattern ‘downstream’ above the tropical North Atlantic Ocean. Accelerated eastward trade winds at low levels rush to fill the gap of the ascending air. At the same time, winds blow in the opposite direction aloft, and the combination of those two winds ends up causing strong wind shear. Currently there is a 98% chance that El Niño will have developed by the middle of the North Atlantic hurricane season. I’m not sure I’ve ever seen such high confidence in the state of the tropical Pacific so early in the year. Forecasts over the past few months have consistently predicted El Niño, and the most recent forecasts have converged and point towards a strong El Niño event.

Meanwhile the tropical North Atlantic Ocean is cooler than it has been in recent years. With surface temperatures closer to normal (but still warm compared to the long-term average), the North Atlantic will provide the baseline level of support required for hurricane formation (but not more than that). If the expected wind shear should slacken off briefly, fleeting windows of opportunity would open for hurricanes to form and intensify as they tap the fuel provided by the summertime North Atlantic.

Of course, climate change always offers the possibility to rewrite many rules of thumb when it comes to seasonal hurricane activity. This ‘chaos effect’ can, at least potentially, include the opposing relationship between El Niño and hurricane frequency. The year 2023 put this to the test. A strong El Niño did battle with a record-hot North Atlantic. The North Atlantic prevailed and the end result was in the fourth most active hurricane season on record. This year, however, I’m not expecting too much disruption to the overall character of the North Atlantic hurricane season. Because North Atlantic Ocean temperatures are generally unremarkable, I have confidence that El Niño will exert its usual suppressing effects on hurricane activity.

Many people, including me, are fascinated by hurricanes. But if we can’t forecast the weather more than two weeks in advance, how is it possible to predict an entire hurricane season? The scientific basis for being able to predict a hurricane season is rooted in relationships between hurricanes and their environments. And to anticipate how the components of North Atlantic Basin will combine to encourage or suppress hurricane activity, we have developed two quite different approaches.

The first method is built upon a foundation of observed hurricane behavior in the historical record. That behavior is fed into statistical models that use current environmental factors as inputs and then predict seasonal hurricane activity. These statistical models either use as inputs the environmental factors observed at the beginning of the season or environmental factors expected to be present throughout the hurricane season. A related method leans on the ‘analog’ approach to compare current conditions against hurricane activity in prior years with similar environments.

The other path to hurricane prediction is based on process-based models. These models are a collection of mathematical equations that encapsulate known physical laws of how the atmosphere behaves. In my research, I provide these models with a snapshot of what the atmosphere looks like today and solving the equations tells me the future state of the atmosphere. Many forecasting centers use collections of these process-based models to forecast future environments or, if conducted in sufficient detail, even forecast future hurricanes themselves.

Increasingly, artificial intelligence and machine learning methods are being applied to seasonal hurricane forecasting. Machine learning allows us to interrogate the historical record in new ways and reveal previously unknown patterns of behavior. The hurricane forecast community now uses AI to emulate our process-based models at lightning speed, which allows us to run many more forecasts and scan a more complete range of possible outcomes.

The price tag on a hurricane varies dramatically depending on whether and where it makes landfall. Given that we are expecting below normal basin-wide activity, what does this mean for landfall rates or locations?

We are all now familiar with the standard caveat that it only takes one storm hitting exactly the wrong place to produce a high-loss year for insurers. The reason this caveat is repeated constantly is because of a geographic mismatch between hurricane forecasts and hurricane impacts. Seasonal hurricane forecasts are issued for the entire North Atlantic basin but for impacts we care about much more local scales. I think scientists and forecasters can do better.

In some ways, hurricanes are like corks in a flow (or for A.A Milne fans, like Poohsticks). They tend to follow the environmental winds. This control is so strong that even major hurricanes with 110mph+ winds can be pushed around by relatively weak environmental winds. It was amazing to me that Category 5 Hurricane Dorian was churning away over the Bahamas in August 2019 — just a stone’s throw from Miami —yet the National Hurricane Center issued no warnings for Florida. That lack of concern (and correct forecast!) speaks to the power of steering flow. As Cameron Rye and Jessica Boyd showed in their 2022 analysis, without that influence, if Dorian had instead made a direct hit on Miami, insured losses could have plausibly exceeded $250 billion.

Knowing these environmental steering winds is therefore crucial to understanding whether a given hurricane may impact land. Steering flow is perhaps the most impact-relevant environmental factor. Why don’t we read much about it?

Many forecasting centers shy away from making statements about steering flow. And for good reason. Steering flow varies from day-to-day and it’s not possible to know steering patterns on a given day in advance of the season. However, we do know that the seasonal average steering flow varies from one year to the next. The broad region of high pressure over the subtropical North Atlantic is the dominant feature controlling steering flow. It varies in strength and westward extent, controlling seasonal steering flow either towards or away from the Caribbean and the United States.

My ongoing collaboration with the Willis Research Network is squarely aimed at the connection between basin-wide storm activity and hurricane landfall. By applying the NCAR hurricane model, which uses physics to reproduce the behavior of actual tropical cyclones, we’ll create many simulations of North Atlantic hurricanes under current environmental conditions (solving the problem of our too-short historical record of real storms). Those synthetic hurricanes will allow us to test how above-normal or below-normal hurricane activity over the North Atlantic basin translates into real risks for the United States – for the whole country, individual states, or specific locations. I’m optimistic that research will help make future hurricane forecasts more relevant to insurance companies and other decision makers concerned about the risks posed by tropical cyclones.