A conversation with RMS’s head scientist about why water is the new wind when it comes to hurricane risk
This week we spoke to RMS’s Chief Research Officer, Robert Muir-Wood, about the latest advances in storm surge models and why it’s crucial for reinsurers and ILS investors to ensure that they fully understand the potential impact of storm surge on the risk that they are assuming.
How has our understanding of storm surge developed in recent years?
If we look back a bit, it’s clear that in the 1990s people tended to assume that Hurricane Andrew (1992) was a typical hurricane. However, Andrew was small, intensified rapidly just before landfall and it arrived over deep water – just offshore of Miami. As a result, it had a relatively small storm surge and the surge only contributed 1% or 2% of the total loss. In retrospect it is clear that, throughout the 90s, there was a general belief that if you model the windfield of hurricanes you were going to capture the majority of the loss impact.
Since 2000 we’ve seen a number of events which have highlighted the opposite case – events like Katrina (2005), Ike (2008) and Sandy (2012). In each of these storms, the storm surge was much bigger than would be assumed if you simply looked at the intensity of the storm at landfall.
It turns out the size of the storm surge at landfall is determined by what the hurricane has been doing over two days or more before landfall, not simply what the storm is at landfall. This mound of water is pushed in front of the storm and accumulates in a compounding way through time.
Catastrophe models were using a simple approach to relate the height of the storm surge to the intensity of the storm at landfall, but this turns out to be overly simplistic. If you are on a coastline where hurricanes typically weaken before landfall, then you are going to be significantly understating the impact of the surge.
In the two years leading up to the release of RiskLink11 we employed our in-house supercomputing capability to run the wind field over the sea for each of tens of thousands of synthetic hurricanes for the entire period before the storm made landfall, to generate the storm surge all the way through to the extent of coastal flooding. When we ran all these events through the catastrophe loss model we found that the contribution of storm surge to hurricane loss is much higher than people thought. In terms of economic impact, closer to 40% of the loss is caused by storm surge, including hurricanes like Katrina and Sandy where a majority of the loss was actually caused by the storm surge – in Sandy even a majority of the privately insured loss.
How does storm surge risk vary along the US coastline?
Coastal flood can drive a high percentage of overall event losses. This image depicts the contribution (%) of storm surge to overall insured losses at the 100-year return period for commercial lines of business across select U.S. cities, as calculated using RMS modelling technology.
Hurricanes along the northern Gulf coast tend to weaken before landfall as they tend to pass over slightly cooler water than is typically found around the Loop Current in the centre of the gulf. The same applies for storms making landfall along the East Coast and up in the North East. The only place in the USA where it tends to be the other way around is where hurricanes cross the Gulf Stream adjacent to southern Florida, where, as they head into Florida they actually tend to intensify. Hurricanes that make landfall in Southern Florida, like Andrew, therefore tend to be more intense than their accompanying storm surge.
Storm surge is also very sensitive to the offshore bathymetry. If the sea is very shallow the water pushed in front of the storm has nowhere to escape downwards to get out of the way and so tends to pile up ahead of the storm. If the water is deeper the mound will tend to flow away underneath the storm.
The largest area of shallow water off the US coast is immediately adjacent to the Mississippi Delta. The sea has been filled up with sediment coming down the Mississippi. So the coast of Mississippi is the most hazardous place for storm surges in the US because of this large area of shallow water that tends to amplify any arriving storm surge. The deepest water next to land in the US is close to Miami. As a result Miami has relatively muted storm surges.
Bathymetry of Northern Gulf of Mexico and the Atlantic Ocean east of Florida. Shallow seabeds are depicted in red and act to increase the impacts of storm surge, while deeper seabeds are depicted in yellow and green and act to reduce the impacts of storm surge.
Did you find any other factors that can lead to disproportionate storm surge losses?
The physical size of the hurricane is very important. A bigger hurricane pushes a broader front of water in front of it and again there is less chance for that dome to drain away. So, very big storms like Katrina or Sandy have a much bigger storm surge than you might otherwise have expected, given their intensity.
How has this new scientific understanding been used in practice?
This is something we feel is actually very important to be properly captured in the models. I think we’re the only modeller who has this full time-stepping coupled ocean atmosphere storm surge catastrophe modelling capability. We used our storm surge modelling in the issuance of the cat bonds for the Metropolitan Transit Authority in New York, itself a parametric deal based on tide gauge measurements of the water levels outside New York Harbour. The capital markets seem to agree we have the best capability for looking at storm surge impacts.
There has also been a project underway for the past year called Risky Business which is launching its findings on the 24th of this month. It looks at the future costs of climate change in the US and was set up by Michael Bloomberg, Hank Paulson and others to try and come up with a non-partisan, business minded perspective on expected climate change costs through the 21st century.
We were commissioned to do all the modelling work for coastal risks for this project. This has involved using our storm surge modelling but with adjusted local sea levels for expected changes in water level throughout the 21st century. We have also modelled various perspectives on expected changes in hurricane activity from work done by Kerry Emanuel and Tom Knutson, who are advisors on the project. So the RiskLink11 storm surge catastrophe model has also attracted a lot of interest around exploring how the impacts of hurricanes and in particular storm surge losses are expected to change into the future.
Posted: Monday, June 16th, 2014