What on earth is going on? There we were, lined up for blockbusters of hurricane seasons post-2004 and 2005. Vendor models were primed for it, but nothing happened. Do we really understand what's going on at the moment, and how much can we use what's happened recently to understand the future landscape of hurricane risk?
The world of cat modelling matures in a scientific sandbox of ideas. The understanding around the risk associated with hurricane landfall (and indeed in other perils/regions) is still in its infancy. So it's almost fair that as risk modelling companies distil the latest science into loss numbers, we should expect a bit of a rough ride in our understanding and how it is translated into that average annual loss or 1-in-200 year number as the science matures. Treating model change as "well that last model was rubbish, wasn't it?" is tantamount to not understanding the relative youth of the field in which we work.
I've been doing some thinking recently (hey - don't look so shocked!). My thoughts below are mere conjecture - nothing more than that - around what the future might look like in terms of hurricane risk and landfall rates. This is in the interest of getting some discussion going on the topic, but borrows from recent hurricane seasons and some of the peer-reviewed science that is out there, so hopefully it's not all hand-wavy crystal ball stuff.
One of the most fascinating pieces of work in recent years is that of Wang et al. (2011). I've tried to summarise their work in the schematic below:
To boil down Wang's work:
1) In warm sea surface temperature (SST) years (that are more prevalent when the Atlantic Multidecadal Oscillation (AMO) is in a positive phase), the area of hurricane-supporting SSTs extends further east across the Atlantic. Also, a response to this larger area of high SSTs is the weakening of the North Atlantic Subtropical High (NASH). The NASH can extend right across the basin and direct hurricanes that form on its southern side westward towards the US. This weaker NASH however allows storms to recurve north on the western side of the NASH away from landfall.
2) Conversely in the cooler SST years (typically more prevalent in a cool AMO phase), the area of hurricane-supporting SSTs is smaller, hurricanes form further west and the NASH tends to be larger, meaning storms have more of a chance of being directed towards the US.
One thing however that I feel we can't escape is that the sea continues to warm, even though there has been recent attention paid to the idea that we are heading towards a cooler phase of the AMO. Which leads me to point of conjecture #1:
Could the broad-scale increase in SSTs mask any drift back to a cooler AMO phase?
Here's the August to October SSTs (from the NCEP reanalysis) averaged across a few regions in the the Atlantic Basin, with a five-year backward-looking average applied to the data (so 2016 refers to 2012-2016).
The main story in the chart above is how we can see the gradual uptick of temperatures to those currently that are around 1/2 a degree higher than during the last warm AMO phase in the 1950s. You can see for each region how the post mid-2000s period has SSTs higher than any of the years prior to 2005 - in any of the regions. As an additional point, it's worth noting that the Gulf, that maybe has less of a connection to the AMO signal, has shown a fairly consistent rise in SSTs since the mid-1980s.
There are however, many seasonal factors that typically modulate our hurricane seasons over and above just the SST magnitudes and their fluctuations. In recent years the North Atlantic Oscillation (NAO), El Nino and Saharan Dust are just some of the shorter-timescale influences thought to have decreased the activity at different points in recent years. Also, the most recent research from Kossin (2016) points to increased wind shear and lower ocean temperatures close to the coast that tend to weaken hurricanes in warm SST years.
There are years, however, such as 2004 and 2005, where the atmospheric setup directed storm after storm into the United States for a period of time within each hurricane season. What we can't escape here is that when the sea is warm, there is greater potential for the development of many tropical systems in a season. Thus, in the years where the atmosphere across the basin has been "primed" to allow storm development and intensification but also direct storms towards landfall, as per 2004/5, it would seem there is greater potential for "blockbuster" multiple event seasons.
I'm basically presenting two competing story-lines here for the Atlantic Basin if the main development region of the Atlantic stays warmer than we have experienced for most of the 20th century:
1) Quieter landfalling seasons (e.g. many of the years from 2006-16) because storms form further east (after the work of Wang) with the possibility that they also die out or weaken before landfall (after Kossin)
2) When shorter-term atmospheric factors come together, given that the sea is warm, there is the potential for seasons with multiple storms directed towards land and high aggregate losses (e.g. 2004, 2005).
So maybe 2006-2016 is a taste of “the norm” and 2004 and 2005 are the shock years that we get occasionally because – inescapably – the sea is warm. So: here comes point of conjecture #2:
Could this be a possible paradigm shift in how we think about hurricane seasons, where typically hurricane activity in the basin remains quite high, hurricane landfalls are fewer, but we see an increased chance of "blockbuster" seasons with multiple landfalls when atmospheric conditions permit?
I'd imagine that, at least from the aggregate loss (AEP) perspective, this type of change in our hurricane seasons would see lower losses at lower return periods, but might push up losses in the tail when these multiple landfalling seasons come along, something I've crudely shown in this diagram:
Ok, so this is a lot of thinking out loud here and the variability across hurricane seasons is vastly more complex than this, but hopefully you're able to see I've used some recent hurricane season experience and the most recent research to suggest what we could see in the future. As I've said before, however, this is nothing more than a hypothesis. I'm definitely not going to bet my house on this but as I've written in an earlier blog post, we don't have the time to wait and see what might happen.
This is another area where high resolution global climate modelling, with its growing ability to resolve hurricane activity and seasonal behaviour, can hopefully help us understand just what our hurricane seasons may look like in the coming years and whether what we've seen recently is the norm.
Rather than trying to rely heavily on the past to inform our view of the near future, climate modelling could hold the key to understanding what’s happening to our hurricane seasons right now.