When the tropics and extra-tropics collide: revisited for 2019
A couple of years ago I wrote a blog post wondering how easy it might be to propel a hurricane into Europe, based on extremes of sea surface temperature and tropical cyclone forward speeds. Not long after this, Hurricane Ophelia came within 12 hours of landfall from Ireland as a tropical system.
Two years on, there is the potential of another system - Lorenzo - anomalous in its intensity and location for this time of year - being accelerated towards NW Europe in the jet-stream, such that it could re-intensify, laden with tropical moisture. Although it's unlikely to still be classified as a tropical system by the time it reaches the latitudes of NW Europe, hard on the heels of Ophelia's anomalous strength in the eastern Atlantic from two years ago it's worth wondering what might be going on.
Some work I've been doing at the University of Reading has been looking at 6000 years of data from a high resolution climate model. This model re-simulates 1951 to 2010 one hundred times over to give a fairly realistic windstorm climate of north-west Europe. Although in our work we've spent most of the time concentrating on windstorm activity in the main thrust of the windstorm season from November to February, we also were able to use the dataset to look at the frequency of windstorm events on the fringes of the hurricane season, especially in that crossover period overlapping the hurricane in September and October.
To set the scene, the chart shown below shows the frequency, in each decade from the 1950s to the 2000s of an event that, in terms of its damage potential (not loss amounts, I should add) is equivalent to around a 2-year return period across the whole 6000-year simulation.
You can see that this strength of event becomes more likely towards the end of the simulation: The level of damage potential chosen has a 2-year return period across the entire simulation, but you can see that its chance of occurrence is slightly higher towards the end of the simulation, at 1.8 years for 1980-2000 versus 2.2 years for 1950-1970.
It's not the point of this blog post to discuss reasons why, but it would be fair to say the changing climate might be, at least, partly responsible for the increase in frequency (and thus decrease in return period) of this particular chosen strength of event.
Now, what we can also do is take the same approach and see the change in frequency of an event of this strength falling in each month, in the same style of chart:
Clearly December and January, in the middle of the windstorm season, represent the months where the chance of this strength of event happening is highest (and therefore has the lowest return period). Like the earlier example, in all of the months, the chance of this strength of windstorm occurring increases (and the return period of this event decreases) as we move towards the latter part of the 20th century.
However, the month in which this type of event appears to have the most of a jump is in September, where in the early part of the simulation, the risk of an event is fairly rare, but there is a sharp jump up in the last decade of the simulations.
Now - we can't necessarily point the finger here directly at a change in the tropical storm influence on windstorms, but the increase of events in September in NW Europe in the simulations is clearly an interesting shift of risk given the time of year - and also that the changes are more abrupt in this month and don't mirror the gentle upward shift in other months.
Naturally you might argue it's only a change in a small number of events here but this is part of the intrigue for me. We're seeing a shift of events that may not necessarily have been experienced in the historical record given the long return period: but are potentially becoming more likely. Climate models have the potential to reveal this sort of detail where the historical record simply doesn't.
We saw above the example of Lorenzo being so anomalously far east and anomalously strong in the Atlantic Basin: this could, potentially, be an example of a rare event becoming more likely owing to the shift in climate, something that attribution studies may prove in due course. And of course, it follows fairly hard on the heels of Ophelia in 2017.
Just to demonstrate this, below is a storm that is the 6th-ranked storm in terms of damage potential (not loss) for France across the entire 6000 year simulation: so, crudely (statisticians look away now), a 1000-year return period event for France.
What made this particular event more startling is that it's not a common-or-garden winter windstorm. It occurs between 24th and 26th Sept 2005 in this particular simulation. You can see from footprint of the system that it originates from south-west of Europe, moving north-east. It appears to hit Portugal with a narrow wind footprint, suggestive - maybe - of a decaying tropical origin, before re-intensifying and broadening across France as an extra-tropical windstorm, potentially with the added "strength" from all the warm air the tropical system possessed. This sort of event, until it happens, is hidden from us in the historical record.
By the time you've finished reading this, there's a chance that Hurricane Lorenzo may have re-intensified as an extra-tropical windstorm close to our shores. We've had remnants of hurricanes that have re-intensified across NW Europe, such as ex-Hurricane Lili in 1996, without bringing notable insurance losses, so this sort of event isn't entirely unheralded.
However, as the seas warm, bringing potentially longer hurricane seasons and allowing larger areas of the Atlantic to support hurricane activity, should we really be surprised to see the likes of Ophelia and Lorenzo?
And should the tails of our windstorm loss curves contain losses from landfalling ex-tropical cyclones?