RACING A RISING TIDE
132
FORTUNE.COM // NOVEMBER 2019
trophe perils. The effort seeks to anticipate trends in a biblical list
of disasters that, in addition to hurricanes, includes earthquakes,
tornadoes, hail, and floods. “If you have a rational approach,” one
that allows you to understand the climate’s vicissitudes “better
than your competitors, you can be profitable,” Bertogg tells me.
Dozens of models spin out of Swiss Re’s shop, each for a differ-
ent kind of disaster in a different part of the world. By far the most
financially important one is for hurricanes. From refinery com-
plexes along the Gulf of Mexico, to beach condos in Miami, to the
global financial nerve center of lower Manhattan, trillions of dol-
lars in infrastructure lie in the path of increasingly violent storms,
pressuring Swiss Re to continually reassess which properties it can
profitably reinsure, and at what price.
So far, the hit to Swiss Re’s bottom line that can definitively be
chalked up to climate change remains small. In part that’s because
its property policies typically last only a year, giving the company
a chance to raise rates as its models detect rising dangers. But by
mid-century, Swiss Re has come to believe, climate change will be a
major driver of increased risks and losses—which is why the com-
pany feels so under the gun to climate-proof its models.
Right now, Swiss Re’s analysts have high confidence that events
that result directly from higher temperatures—
sea-level rise, for instance, and also storm
surges and landslides—will increase as a result
of climate change. But a worsening of catastro-
phes whose relationship to climate change is
indirect—including hurricanes, whose behavior
depends on the way that higher temperatures
interact with complex systems in the oceans and
the atmosphere—is “really in a low-confidence
area,” explains Michael Gloor, an expert in atmo-
spheric physics who’s on the modeling team. The
disasters that tend to cost Swiss Re the most, in
other words, are precisely the sorts whose trajec-
tory it is least able to divine.
Modeling hurricanes isn’t brain surgery. But
to the uninitiated, it can seem pretty close. To try
to understand it, I join Gloor in a sun-drenched
conference room overlooking Lake Zurich. Like
Corti and Bertogg, he studied at ETH Zurich, a
renowned science university in town.
In most insurance models, past is prologue.
They extrapolate from the behavior of previous
storms to predict the path, fury, and, thus, cost
of future ones. That’s the actuarial worldview:
an assumption that, on average, tomorrow will
be essentially like today. To start the process,
modelers input into their computers publicly
available data on the trajectory of previous hur-
ricanes. Then they write code that causes the
model to alter the behavior of that hurricane
both based on how other prior storms in that
part of the world have played out and by moving
around the actual storm’s trajectory randomly,
based on knowledge of how hurricanes typically
behave. For each actual hurricane they analyze,
the model spits out perhaps 100 or 200 theo-
retical variants of the storm in question.
Graphed on a screen, the actual hurricane
appears as a red line, which Swiss Re calls “the
mother.” Each of the theoretical variants appears
as a black line—called “the daughters,” and also
“spaghettis,” because they vaguely resemble thin
strings of pasta. As Gloor flips through the mod-
el of Hurricane Maria, which in 2017 wreaked
havoc on Dominica, Puerto Rico, and the U.S.
Virgin Islands, he stops on one screen of code.
“Create some spaghettis with empty intensity
measures,” says a note from one of the modelers.
For each spaghetti, the model creates a track
of the area likely to be impacted by the hur-
ricane’s winds —a track known as a “wind field.”
It also adjusts the frequency with which it
expects hurricanes to strike based on sea-surface
temperatures. Then it combines that data with
another trove of information, this one quantify-
ing the value of insured property in a potential
client’s portfolio. From that comes what Swiss
SCIENCE’S LIMITS Chia-Ying Lee and Adam Sobel of Columbia Univer-
sity are working to refine Swiss Re’s climate-prediction models —and
reaching some contradictory conclusions as they do so.
PHOTOGRAPH BY HILARY SWIFT
