2023 saw a “Battle Royale” commence between two competing forces – a developing El Niño in the Pacific and warm seas in the Atlantic – and we waited to see which of the two would win out. As it happened, an above-average number of tropical storms (20 versus the average of 14 based on 1991-2020 hurricane seasons) meant that it was a busier season than average, even though these twenty storms only led to a single category 3 landfall in Florida, which mercifully produced low single-digit billions losses.

After last year’s tug of war between the El Niño and the warm Atlantic, the forecasts this year are suggesting that both basins are worryingly pulling in the same direction in terms of hurricane formation – and the early hurricane season forecasts are alluding to this. The Atlantic remains warm, but the Pacific is headed towards a rapid-fire swing from the warm seas of an El Niño, to the cool seas of La Niña: that typically act to enhance hurricane activity in the Atlantic Basin.

Atlantic Remaining Warm?

The chart below highlights Inigo’s attempt to synthesize some of the seasonal sea surface temperature (SST) forecast information to understand how warm the “main development region” (MDR) – the area of warm seas through which many landfalling hurricanes pass – may end up by the peak of the hurricane season in August to October.

The diagram highlights the distribution of SSTs in the MDR from the past 20 years in green. The yellow line shows 2023’s remarkably warm MDR and the blue distribution shows Inigo’s attempt to use the forecast information from the Copernicus C3S seasonal forecasting service to provide an estimate of how warm the MDR might become based on each forecasting house’s summer forecasts. It is clear that this summer’s seas have the potential to become just as warm as last year in the MDR.

We can also take a slightly different look at this by understanding how the sea typically warms from now until the middle of the hurricane season and projecting the current sea conditions forward in time using historical behaviour. The chart below takes the maximum, minimum and mean historical change in SSTs (as well as 10th and 90th percentiles) from the 15th April for every day until the end of the year. It puts a historical constraint on how the temperatures should change as the sea warms.

The key takeaway from this work is that, pending a remarkable lack of warming of the Atlantic MDR, the warmth for this summer is already “baked in” given how warm the sea is already. Even the 10th percentile of recorded rate of MDR warming since 1981 would give us likely the 2nd warmest August-September MDR going back to 1981, with 2023 the highest, followed by the notable years of 2010 (12 hurricanes in the basin) and 2005: famous for the four landfalling hurricanes Dennis, Katrina, Rita and Wilma.

Warning Shots Across the Bows

Last year’s hurricane season really was a step into the unknown. If we look at the hurricane seasons since 1950 couched in terms of the warmth of the in the Atlantic and whether the atmosphere is in a state of El Niño (positive x values) or La Niña (negative x values), we can see that 2023 was a complete outlier: a very warm sea along with a fairly strong El Niño. We’ve also put a dashed box in the chart representative of where we feel 2024 could end up.

Notice here, however, that we do have one or two historical seasons that serve as a reminder of what a La Niña coupled with a warm Atlantic MDR can bring. The two closest matches, 2010 and 2020, were seasons that, for differing reasons, ended up with fairly unremarkable insured losses. However, both had enough evidence to suggest that the combination of warm Atlantic seas and La Niña has the potential for busy hurricane seasons and – potentially – notable insured losses.

2010 may have had no US hurricane landfalls, but 12 hurricanes patrolled the basin during the summer months. 2020, in contrast, had 13 hurricanes and remarkably, 6 of these made landfall, but fortuitously tended to avoid areas of high exposure concentrations. But should we take the low losses in these years at face value?

Counterfactual Understanding

With the help of modelling firm Reask, who are developing a dynamical, flexible way of viewing risk, we’ve been able to understand the loss potential of previous seasons by performing a counterfactual analysis – the study of “what might have been”. This is carried out by re-running these hurricane seasons hundreds of times using realistic airflows and thermodynamics that occurred in those years as the basis for generating synthetic seasons of “what might have happened”. We then join up the Reask model with the losses from our in-house vendor catastrophe model from Verisk. The results make compelling reading, and suggest that these seasons should not be ignored in spite of relatively low losses. Each line shows the probability of exceeding the loss shown on the x-axis from 2010 and 2020 based on 2500 re-runs of that season, using realistic weather and sea surface conditions for those seasons. The transparent line on the 2020 curve represents the loss from that season (2010 essentially had zero loss) – and we’ve annotated the mean modelled expected losses from each season: simply the average loss across our 2500 simulations.

2010 had no loss, so the “expected” mean loss of 52bn across the many simulated seasons is an interesting indication of how loss-bearing the season could have been. The reported losses from the 2020 season had nearly a 55% chance of being exceeded, so was towards the lower end of the simulated loss distribution for that season but by no means unexpected for that year. However, the mean losses across these two “2024 analogue” seasons amount to around 60bn – through using this methodology we get a stark reminder of the unfulfilled potential of these two seasons. The mean losses here exceed the median because of the marked increased chance of “blockbuster” seasons in the tail: seasons that would happen rarely, but their chance becomes more likely when La Niña and warm seas are present.

Any Positive Spin?

Is it all guaranteed misery? Quite possibly not. The first chart we showed above is a forecast of the forthcoming season: even though the various forecast models are all tending to agree on a warm MDR and La Niña emerging in the Pacific this summer, they’re not always that accurate, especially this far away from the hurricane season.

Secondly, the chart below shows some simulations from a climate model where we’ve picked years with warm Atlantic seas and a La Niña (in green) and compared this with all years irrespective of sea temperature and La Niña. Yes, you can see that compared to all the simulations, typically the season is expected to be busier when there is warm sea and a La Nina present, but that’s not always the case: the green curve in the chart below still contains around a 20% chance of a below-average season.

The climate model example highlights how when things are set up to be active, sometimes other less forecastable factors or short-term weather impacts can aid a quieter season: both from a landfalling or basin activity perspective.

From a landfalling perspective, 2010 was remarkable in that it was primed for an active season, and indeed we did have 12 hurricanes in the basin, but not one made landfall in the United States. That year the Azores High was strong, which acted to shift storms out into the Atlantic. In 2013, another year primed for a busier-than-usual season with high sea surface temperatures and a relatively neutral ENSO ended up with only two hurricanes forming in the basin partly in thanks to the “Saharan Air Layer” that propagates dry, dusty air into low-levels. This, in accordance with broad-scale descending air at mid-to-upper levels of the atmosphere provided good conditions to inhibit storm formation that might not have been easy to forecast at the start of the season.

So there are things that can scupper what might be lined up as a busy season that may be hard to predict at the start of the season, but we can’t ignore how much the risk in the tail is elevated in these simulations when La Niña and a warm MDR is present.

Things are set up ominously for a busy hurricane season. Historical precedents suggest we may have been lucky in the two most similar historical years to this year. Are we going to dodge another bullet this year?