Forecasting Catastrophes Is No Longer Voodoo Science

With increased wild weather, floods and bushfires, the world insurance and reinsurance industry's financial sustainability is under pressure.

In Australia's northern rivers region of New South Wales, for example, many people who live around the city of Lismore will be unable to afford insurance due to the 2022 floods, which caused an estimated AUD5.7 billion in damages.

In assessing and pricing risk, however, insurers have increasingly relied on ever more sophisticated catastrophe modeling, combining machine learning techniques with an exploding volume of data aggregated from historical sources and in real-time from satellites.

Nick Hassam is the founder of an Adelaide-based startup called Reask, which builds state-of-the-art solutions for the reinsurance industry—which ensures the insurers—manage their exposure to climate risk.

Hassam explains that the discipline and methodologies of catastrophe modeling originated in the U.S., where earthquakes in San Francisco and Los Angeles resulted in many fatalities and billions of dollars in property damage.

"As a result of that, a bunch of very clever scientists working in different parts of the U.S. out of places like Harvard, MIT and Stanford looked at ways of building models that represented these physical phenomena in computers and using them to estimate what their losses might be," explains Hassam, who comes from a background in geospatial technology and previously worked in the insurance industry in the U.K.

This early U.S. response was the genesis of catastrophe modeling. However, it could only rely on historical data, and the computing power was many times less than today.

“There is a problem in that history is now not a very good representation of what is going to happen in the future,” said Hassam.

“So the way these models have been built historically is not effective at capturing the variability of conditions, which are changing as the climate changes.”

Volumes of data

Hassam set up Reask five years ago with a colleague experienced in machine learning techniques to build a new framework for catastrophe risk modeling, which takes global climate models and analyzes vast volumes of information such as sea temperatures and sea level air pressure, as well as upper-level wind shear.

Taking 40 years of data divided up into monthly segments, Reask might be working with up to 200 different parameters.

The framework then converts these insights from machine learning into algorithms that can help build views of risk, which help to determine the frequency and severity of the events.

It can also look at future climate scenarios under different scenarios, such as one or two degrees of warming, to create an understanding of future climate expectations.

"The volume of data is such that with these volumes, we really need high-performance computing to push out these simulations."

Reask uses two supercomputers in Australia, located in Canberra and in Perth, to run its analysis.

“The volume of data is such that with these volumes, we really need high-performance computing to push out these simulations," says Hassam.

"And essentially, the machine learning reduces the insights in that massive volume of data down to something that is more manageable and consumable."

Spreading risk

For clients, the more dynamic modeling means that they can update their view of risk much faster than in the past when they typically would have done so every two or three years, and this is helpful when weather patterns change—as they have recently—from La Nina to El Nino periods.

"What we are trying to do is to change the industry by increasing the availability of the information we have and providing that to insurers to make those decisions much faster," says Hassam.

The Reask business model is “data as a service.” Most clients consume pure data from Reask and then run their own risk and pricing models.

They will also consult a range of sources and opinions because, as insurers, they understand that there is a risk in confining decisions based on only one viewpoint.

“Clients will take our service, and that of a competitor and an actuarial model and interview view, because Australian regulators stipulate that insurance companies need to take multiple views,” says Hassam.

“If you took just one view of risk, you will be building up a significant level of systemic risk.”

Reask’s point of difference, he says, is its methodology based on delivering a global solution and being able to provide dynamic views of risk from analyzing current information.

“We talk about what’s happening now, what’s happening next year and what might happen in 20 years’ time, and we update that more frequently,” says Hassam.

Lachlan Colquhoun is the Australia and New Zealand correspondent for CDOTrends and the NextGenConnectivity editor. He remains fascinated with how businesses reinvent themselves through digital technology to solve existing issues and change their business models. You can reach him at [email protected].

Image credit: iStockphoto/mdesigner125