2023: A hurricane ‘Battle Royale’
As ever, we’ve barely had time to take a breather after the 1/1 and 1/4 renewals and the hurricane season is nearly with us again. This year sees a particularly rare combination of conditions as we approach the key months of August, September, and October: the presence of an El Nino and the potential for some warm seas in the Atlantic. Can we make head or tail of what’s to come by looking at what’s gone before; and what other data can we use to help us understand how the season might pan out?
History: The “Truth”, but with limitations
In framing the forthcoming season, we’re going to talk about two competing factors here:
- The temperature of the “Main Development Region” in the Atlantic, an area of sea over which a large proportion of US landfalling storms will have to pass before hitting the mainland.
- The El Nino Southern Oscillation, or ENSO: where warm (El Nino) or cold (La Nina) waters in central Pacific influence global wind patterns that can suppress or enhance hurricane activity, respectively.
The chart below takes historical data from 1950 to 2019 and shows, for each season, the percentage difference compared to the mean of Atlantic Basin hurricanes (left) and landfalling hurricanes (right) plotted on the chart according to August-to-October ENSO index (x-axis) and the temperature of the Main Development Region (y-axis) relative to the mean: we take the mean here as 1981-2011. Seasons with above-average basin count or landfall count are in red, and below-average are in blue.
What we notice, for historical data, is how the basin count is above average as seas get warmer and a hint that maybe El Nino causes fewer storms: but the latter statement is a pretty sketchy one based on the data.
For landfall counts, it’s just a lot of noise. Landfalling events are much rarer than storms in the basin, and historical data doesn’t really give us any hints towards how warm the Atlantic is or how strong an El Nino or La Nina is might influence landfalling storm risk. There’s aren’t enough years to help us out there – and indeed not enough landfalls to try and make some sensible conclusions.
Climate Models: Imperfect, but useful extensions of history
Now, we could just shrug our shoulders and give up here. But there are climate models – especially those that are run in an “hindcast ensemble” mode where multiple versions of historical years are re-simulated using similar (but not identical) sea surface temperatures and greenhouse gases to previous years.
At Inigo we’ve been investing time into understanding how we can best use climate model data to glean as much information as possible to help us make underwriting and risk management calls. It is vital that we understand how tropical cyclone risk might be shifting where historical data might not inform us satisfactorily. The two main topics of investigation are:
1. A world of seemingly increasingly warm seas
2. Season-to-season “natural variability” like ENSO and how this can impact short-term pricing.
We can plot the same chart as above, but here, for every year from 1951 to 2011, we’ve got 100 complete re-runs of each year, with slightly different atmospheric and sea surface conditions. So: around 6000 years rather than 60 or so years of history.
What emerges from this is again, a little noisy – but we can see some patterns emerging more from the data when compared to the earlier “historical data” plot, with regards how Atlantic MDR sea temperatures and the ENSO phase might dovetail:
Again, we are showing the percentage increase or decrease over the average in basin count (left) or landfalling storm count (right) for various combinations of MDR sea temperature and ENSO index.
The key thing to note here is the chart of both basin activity and landfall activity goes from negative to positive from the bottom-right to the top-left: i.e., the warmer the sea and more the Pacific Ocean is in a La Nina phase, the more there is activity in the basin and commensurately more chance of a season with higher landfall activity. This landfalling trend was not evident from the historical data. Whilst climate models are anything but perfect, here they provide some helpful indicators as to how the risk might shift that you simply cannot see easily from a short historical dataset.
There is a storyline here that points clearly to increased risk – both in the basin and at landfall – as seas warm, but there is still a catch here in that when a strong El Nino is present and the sea is warm (the top-right of the diagram), activity can still be reduced.
2023: An unprecedented ‘Battle Royale’
But how does this year sit in the context of all this data? Just to anchor us historically, we’ve shown the first diagram again, but we’ve annotated a few years on the chart: the busy basin years of 2010 and 2005 – a year to bring out a cold sweat in anyone who worked through it.
The question mark in the top-right hand corner here represents how seasonal forecasts might be calling this season: very warm in the MDR, of the order 2010 warmth, but with the potential for a strong El Nino. We are in uncharted territory here by way of historical data – and indeed most of the climate model data we’ve used.
The nearest analogue one could argue was 2015: comfortably above-average warmth in the Atlantic Basin but with a strong El Nino. This produced an average basin season, with no landfalls. Even if we look back the climate model re-simulations of 1951-2011 in the 2nd chart, our dataset falls short of what we might be seeing this season, but there are hints from this more extensive dataset that even for warm MDR years, a strong El Nino could nullify that warmth’s impact.
So go on, pick a side: I dare you! Is it a busy season because of the warm Atlantic or a damp squib thanks to the suffocating wind shear of an El Nino? Let the battle commence!