History plays its part in climate modelling

Catastrophe models are built on permutations of historical events and based on a stochastic catalogue of hundreds of thousands of hypothetical events based on what nature could produce.

That is how Liz Harris, vice president of modelling and research at Ariel Re, described the traditional, probabilistic approach of re/insurers to predict extreme weather. In an interview with Bermuda:Re+ILS, she said that as climate change becomes better understood, cat models are evolving.

Ariel Re underwrites a global portfolio of reinsurance products in Bermuda, London, and Hong Kong via its Lloyd’s platform, Syndicate 1910. Through its link to Lloyd’s, Ariel worked with the Prudential Regulation Authority in the UK to design its climate change stress tests.

In addition, vendors have created a “suite of frequencies”, Harris said, based on different time horizons and Representative Concentration Pathway (RCP) scenarios. RCP is a greenhouse gas concentration (not emissions) trajectory adopted by the UN Intergovernmental Panel on Climate Change.

“To each historical event we assign a probability, which we split up into ‘more likely’ or ‘less likely’ events by their intensity, and we then benchmark our stochastic catalogues against what we have seen. That’s our only real data point,” she said.

“The main thing is that we’re largely writing annual contracts, so that’s just for the risk this year. The most important aspect of climate change for us to capture is that it has occurred during the course of our historical record, which we use to create our cat models.

“We do that by doing permutations of historical events by moving them north, south, east or west; or by making them larger or smaller. In short, cat models are trained on what we have seen before—for example, the frequency of a cat 3 over the last 100 years.”

Hurricanes are measured on the Saffir-Simpson Hurricane Wind Scale, which runs from category 1 up to category 5. In a category 1 hurricane, winds range from 74 to 95 mph, are very dangerous and will produce some damage. In a category 5 hurricane, the winds are 157 mph or higher and will produce near total destruction. There is no category 6 hurricane yet, but Hurricane Irma in August 2019 had maximum wind speeds of 185 mph.

“While science has come to the conclusion that there is a high level of confidence around the materiality of climate change impact to some natural catastrophes, such as wildfire and flood, the Bermuda market is heavily driven by hurricane risk where there’s still some uncertainty around the impact that climate change has had to date on extreme winds, although it’s clear that we’re in a regime of warmer-than-average sea temperatures in the North Atlantic, partly from climate change, and partly through natural variability,” Harris explained.

What ‘100 years’ means

How does a “one-in-100-years” event remain relevant, given the increasing frequency of extreme weather?

“Reinsurance is built around a framework of what could potentially happen in 100 years, but that’s assuming that all 100 years are the same. So, it’s less about what happened in the last 100 years, and more about what would happen in 100 simulations of next year,” Harris said.

“A hurricane such as Irma or Ida is a major catastrophe but those are not the kind of things that keep us up at night. We know they are going to happen once every five or 10 years, and our models are built and informed by catastrophes like that. What we truly worry about is the ‘$150 billion event in Miami’. That’s more like a 100-year event and it’s one that has a major impact.”

In forecasting what might happen this summer, modellers know that Atlantic sea temperatures are above average and that the tropical Eastern Pacific is cooler than usual, which results in less wind shear over the tropical North Atlantic. That produces a higher likelihood of hurricanes in the Atlantic Basin this year, Harris said.

“To create a market-moving event, there needs to be a major hurricane hitting one of five major cities: Houston, as in Harvey; New Orleans, similar to an Ida or a Katrina; Tampa, which is a bit harder to hit; Miami; and the Northeast, a Sandy-type event.

“Anywhere in between you can get a situation like a Michael in 2018, which was a cat 5 on the Florida Panhandle which hit a much smaller town.”

The extremity of an event is one thing, but a surprise event can be much worse, she said.

“The US wildfires in 2016 and 2017 were unexpected in their magnitude, and I don’t think the industry had realised the extent to which there were a lot of small towns built right into the wildland-urban interface in California, which is very susceptible to wildfire risk,” she added.

Frequency and severity are not necessarily separate phenomena in extreme weather events, she said, because cat modellers assign a frequency to different severities.

“Hurricane winds in the US are very much the driver of losses for the Bermuda market, and we are expecting that the total hurricane frequency may actually decrease with climate change, as we have stronger upper-level winds.

“When you do get an event, however, the strongest events could be even stronger. The frequency of cat 4 and 5 hurricanes has been seen to be increasing in the Atlantic Basin.”

How technology is helping

Harris was a weather forecaster for eight years and has been a cat modeller for 15 years. In that time, the workflow for cat modellers has become far more efficient, she said. This means they can spend more time focusing on the science and the numbers and less time on managing the simulations.

Technology, such as cheaper storage, is helping with the processing of historical data, since cat modellers are able to “simulate” more years than before. This means the pricing workflow will be further automated, so that analysts can spend “more time analysing and less time wrangling with software”, she added.

Ariel has invested heavily, she said, in an in-house portfolio management system, called ARPS (Ariel Re Portfolio System), which: “we believe gives us an edge over the industry”.

“The developers are in-house, so if we need to make a change to the system, we can do that fairly quickly, and they work very closely with our modelling, underwriting and exposure management teams to make the system as useful as possible.

“Not only is the platform very fast—major modules can be rolled out in a matter of months—the pricing is very fast, too, meaning our underwriters can give a new quote in seconds while on the phone with brokers.”

Understanding the science

Having spent her whole life on the Island, Harris has anecdotal evidence of hurricanes.

“In Bermuda, our top natural catastrophe is hurricanes, and I experienced my first one in 1987—that was Emily, a cat 2, and the eye went over the Island. That was a very dramatic event: the electricity transmission distribution system at that time was quite weak, and people lost power for weeks.

“There wasn’t another major event that impacted us in the same way until 2003 but, since then, we’ve had several damaging hurricanes pass directly over the Island. In my personal experience therefore, I have noticed an uptick in severe hurricanes in this one spot,” she related.

Some changes may be related to, but are not necessarily driven by, climate change, she stressed.

“Sea temperatures in the North Atlantic have been anomalously warm since 1995 and, not only that, have been warmer than ever in the last few years,” she said.

Science is still trying to get to the bottom of what caused this trend.

“Is it the Atlantic multidecadal oscillation? We have a 70-year time period, potentially, of cooler sea temperatures, for 20 or 30 years, and then warmer-than-average sea temperatures for 20 or 30 years. We had a cool period in the 1970s and 1980s and then, since the mid-1990s, we’ve been in a warm period. There are tree ring studies that document this change, but as far as actual sea temperature records go, we have only one-and-a-half periods that we’ve recorded.

“There’s been a lot of study into why the cool period might have happened and it’s come to light that a lot of it could have been driven by cleaning up the atmosphere—we no longer had this ‘dirty shield’ stopping sunlight getting to the ocean and warming it up, which we might have had in the 1970s and 1980s,” she explained.

“It’s warmer now that we have a cleaner atmosphere due to industrial practices in the US and Western Europe. How much is coming from that? It seems like a good thing, cleaning up pollution, definitely, but it may have had unexpected effects.

“How much is coming from the blanketing effect of carbon dioxide and other greenhouse gases? How much is coming from changes in the speed of ocean currents?”

When there is a faster overturning circulation, the ocean has less time to build up heat in the tropical Atlantic, but when it slows down, there is more time to accumulate in the main development of the Atlantic region. This gives modellers even more to think about when they analyse the risk of hurricanes on the US and Caribbean coastline today, Harris said.

She attended Bermuda’s inaugural Climate Summit in May, and said she was particularly struck by the presentation by Michael Crow, the president of Arizona State University, about its partnership with the Bermuda Institute of Ocean Sciences. Crow said this reflects the growing awareness of the central role ocean health plays in the planet’s ability to cope with rising carbon dioxide levels and other human impacts.