Storm modeling, weather modification and changing claims patterns
When thunderstorms form, the atmosphere mixes the warm air just above the earth’s surface with the cooler air above it. As the atmosphere warms, that first layer of air will get even hotter, and instability will increase. “So we’re pretty confident that there are going to be more storms capable of producing severe weather in the future,” says John Allen, a climate research scientist at Columbia University in New York City. Thanks to climate change, we’re going to see more intense and severe weather events, and whether those are tornadoes or windstorms or hailstorms isn’t yet certain. And the insurance industry has never liked uncertain. Enter forecasting companies that draw connections between a catastrophe, insured property and a company’s claims history.
“We’re not simulating 100,000 years into the future,” says Scott Stransky, manager and principal scientist at catastrophe modeling firm AIR Worldwide in Boston. Using data from historical weather events, “we’re simulating 100,000 plausible versions of what could happen next year. And each of those years will have real events in it. So some years will have big tornadoes that hit Calgary. Some years will not have much activity at all. Maybe one rare year we’ll have a big tornado that impacts downtown Toronto.”
Afterwards, AIR’s engineering team takes over to determine how the weather events in each simulated year will affect buildings in a given region. “So if you put in a wood frame building with 150 kilometre per hour, 200 kilometre per hour tornadic winds, it’s probably not going to do very well,” says Stransky. “And engineers provide what we call the damage function, which helps convert the wind speeds or the hail sizes into an actual damage.”
Finally, AIR uses claims data from each insurer they work with to calibrate and validate a hypothetical year’s financial effect on the company.
But at the end of the day, Stransky says, no matter how useful catastrophe models are, insurance companies need to take ownership of the risks they cover.
That can be done incorrectly when an insurer determines pricing from their own loss history and ignores meteorological data. “My impression out of the U.S. is that lots of times [insurance companies] prefer to use their historical losses as the driver of their pricing structure,” says Harold Brooks, a senior research scientist at the National Severe Storms Laboratory in Norman, Oklahoma. That method will often expose companies to rare disasters they couldn’t possibly have studied and account for—because they predate the company itself.
Catastrophes like the severe tornado that ripped through the southern and eastern United States in April 2011 “probably happened six times in the last 130 years,” says Brooks. “You could have been in business for almost 80 years, and had one event like that.”
Most environmental rather than historical data comes from Doppler radar and post-disaster groundwork. After big disasters, Stransky and his colleagues “go out to the location that was hit and survey the damage. So we’ll look at a wood frame home. We know what the winds were approximately there. And we see what type of damage the home suffered, and we compare it to a brick home in the same area or a high rise building or whatever it happens to be.”
Over the past decade, “there haven’t been really that many changes in the [number of] stronger tornadoes,” says John Allen, the Columbia researcher, “but there’s been a rapid increase in the number of weak tornadoes.” But that kind of thing depends on who’s looking—if Doppler radars suggest winds in one area looked like a tornado, amateur storm chasers will investigate, and if they find damage, Allen says, “they go, ‘Okay, well, there was a tornado here.’”
While storm chasing may strike folks as a highly dangerous profession (and the movie, Twister, didn’t help with the image), there’s value in the hunt. “So the radars we normally have in a fixed location would give us an idea of what’s going on with weather,” explains Allen.
But these have their limits—they can’t confirm when a tornado touches down, for example. Storm chasers equipped with mobile Doppler radars can “try and understand things on the local scale, because the bigger radars scan too high in the storm. You can find out things about tornado dynamics,” such as how each tornado actually forms. That data may help extend the 13-minute warning time researchers can currently give before a tornado hits.
While 13 minutes or even 13 hours can save countless lives, it’s often not enough time to prevent serious property damage. But insurers, at least, believe there are other ways to minimize catastrophe damage. Enter Weather Modification Inc., a North Dakota-based company that tries to change weather patterns, from making it rain in drought-stricken regions to minimizing the affects of hail on the prairies.
One of WMI’s Canadian clients is the Alberta Severe Weather Management Society, a non-profit created by the 20 largest P&C insurers in the province. “There is no office, there’s no webpage and just one person doing administration,” says Terrence Krauss, the Society’s project director and a former research scientist at WMI. He explains that “it was created as a legal entity to enter into the contract with a private contractor to do this service.”
That service is weather modification; specifically, cloud seeding. Aircraft send flares of silver iodide into storm clouds. Krauss’s research has shown “Nature provides typically only one ice crystal per litre of cloudy air at -20°C,” meaning clouds hold a lot of liquid water even at freezing temperatures. The lone ice crystal in each cloud, “held aloft by the strong buoyancy in a thunderstorm,” grows enormously because of all that supercooled water, and will eventually fall to earth as a large hailstone.
Silver iodide particles cause those supercooled water droplets to freeze sooner than they normally would, creating many more— but much smaller—hailstones. WMI aims to make 1,000 ice crystals in each litre of cloudy air instead of one. “I don’t mind seeing pictures where the road is covered with hail,” Krauss says, “because that always is pea-sized hail, and pea-sized hail doesn’t damage roofs or dent cars or damage siding.”
During this year’s hail season—June 1 through September 15—WMI seeded 79 storms with 355 kilograms of silver iodide in 9,200 flares. The season’s work cost the consortium of insurers $4.2 million, but even a small reduction in hail-related claims would cover that price tag. The August 2014 storm in Airdrie, Alta., caused $560 million in damage, says Krauss, “so a one percent reduction on that day more than pays for our program.”
The Alberta government first started the cloud seeding program in 1974 to minimize damage to crops. But in 1985, the government “said the money would be better spent if it just went to crop insurance,” recalls Krauss, who worked on the Alberta Research Council, which administered the program, from the mid-1970s to the mid-1980s.
The group of private insurers began paying for seeding in 1996, focusing on the densely populated Calgary and Red Deer areas instead of on farmland. “Priority is assigned to storms depending on their severity and the size of the community,” a WMI brochure reads. “Only those storms threatening populated areas are seeded.”
Unfortunately, there’s no scientific consensus yet on whether this million-dollar plan reduces hail damage at all.
“Hailstorms have a lot of energy,” says Gerhard Reuter, an earth and atmospheric sciences professor at the University of Alberta in Edmonton. He’s not convinced cloud seeding works for hail and is more confident of its effects on changing the amount of snowfall. “…Therefore the amount of changes you can do is very limited. You almost have to put an atomic bomb there to change the hailstorm.”
Other climate scientists are also dubious, simply because it’s almost impossible to test cloud seeding’s effectiveness with fair and conclusive results. “Theoretically, it should work,” says John Hanesiak, an environment and geography professor at the University of Manitoba in Winnipeg. But in practice, it’s much harder, because “you could have two storms side by side and you could seed one and not the other and see what the difference is, but there could be differences in environment that [lead to] differences in hail size… in the storm.”
And Krauss doesn’t like the idea of randomized studies. “…You don’t get identical twin [hailstorms],” he says. “…It’s not like testing a drug, where you can have people of the same age, and then you inject half of them, but not the other half and compare. Because [with cloud seeding], even the injection process has a lot of variability, not only in the dose, but where do you put the needle, in other words.”
Scott Stransky from AIR Worldwide is also unsure whether or not cloud seeding works.“A lot of our clients, the insurers, do support this. They pay a lot of money for cloud seeding. We’re not going to tell them to stop.”
But from his perspective, it doesn’t really matter. “Because we build our models on historical data, if cloud seeding works, that’s reflected in the historical data we use to build the model. If cloud seeding doesn’t work, the fact that it doesn’t work is also reflected in the historical data we use to build the model.”
Until the 2013 floods, eight of the 10 biggest insurance claim events in Calgary had been hailstorms, says Krauss. So even if cloud seeding reduces claims just a little bit, it’ll be worthwhile. “It’s the dollar amount that is the motivation.”
Copyright 2015 Rogers Publishing Ltd. This article first appeared in the November 2015 edition of Canadian Insurance Top Broker magazine