11.03 - March 2003
envisions a more scientific solution. Over the past decade, researchers
studying hurricanes have abandoned so-called black-box regression
analysis in favor of computer simulations. Ammann wants to do
the same thing for avalanches: build a complete, seamless model
that encompasses the entire gestation of an avalanche, from new-fallen
flakes to mutating layers to thundering slabs. Although he has
the processing power to handle the number crunching, he's still
awaiting the scientific principles that would drive it.
shot-by-shot view of an onrushing avalanche in Vallée
de la Sionne, Switzerland
find out how a slope becomes a powder keg, the SLF's Martin
Schneebeli spends most of his time watching snow melt - literally.
Using a tabletop device called a micro-computer tomograph, he
suspends lipstick-sized samples of snow inside a cylinder the
shape of a soup can. The sample rotates slowly along its vertical
axis as the machine takes grain-thin (25- to 80-micrometer)
density readings. Scanning the sample takes eight hours. Then
Schneebeli tweaks the heat source and repeats the entire process;
often, he spends 30 days observing the same piece of snow. "We
always knew tiny differences in weather make huge differences
in snowpack stability, but we could never pinpoint what happened,"
he says. "Now we can actually see the bonds changing between
individual grains." Schneebeli (whose name loosely translates
from German as "little snow man") also studies larger samples,
collected in the backcountry and infused with an acid that preserves
the snow's microstructure. Sliced hair-thin, these segments
are photographed by a machine designed for large-scale medical
biopsies; combining the images yields a 3-D rendering down to
the individual bonds. His work is the closest thing to pure
science at the SLF - at the earliest, it will start to reveal
functioning physical laws within five years.
Schneebeli's results will be factored into Snowpack, an SLF
software program that simulates how packed layers change during
the course of a winter. "Snow is an extremely unpleasant material
for modeling," says Michael Lehning, who spearheads the project.
"There's settling through condensation, there's recrystallization,
there's downward movement of water after melting. In the future,
I'd like to add factors like wind-drifted snow and underlying
terrain." Pulling up his digitally rendered cross-section of
a snowpack's predicted composition, the angular German points
to blue streaks representing buried hoarfrost: a crystalline
layer that forms at the surface when cold nights follow sunny
days. Hoarfrost becomes a time bomb under freshly fallen powder,
a slick plane that slabs can slide on. Today, Lehning says,
there's a 90 percent chance that Snowpack will nail the location
of submerged hoarfrost. But that's just a first step. "We're
good now at predicting the types of grains that formed within
individual layers," he says. "But the big challenge is relating
that knowledge to stability within the snowpack."
the only way to determine how an avalanche works is to set one
off yourself. In 1997, the SLF built a private detonation site
in western Switzerland's Vallée de la Sionne. Its bunker there
has four concrete hatches that open toward the avalanche slope,
revealing digital video cameras, a massive Doppler radar dish,
and a large metal tube designed to capture airborne snow particles.
Halfway up the mountain, a 70-foot pylon covered in sensors
measures the force of descending snow. After a blizzard, a technician
detonates explosives high on the slope. The area around the
blast shatters like plate glass. The pylon disappears inside
a towering flume of powder. At the last minute, the concrete
hatches slam shut to protect the instruments. "What happens
inside a smaller avalanche does not matter to me," says Betty
Sovilla, an effervescent Italian who manages the site. "I'm
only interested in the most extreme cases."
experiments at Vallée de la Sionne began, Sovilla has had to
reconsider what "extreme" really means: The SLF's avalanches
displace up to 500,000 cubic meters of snow and ice - the volume
of a 25-story building with a footprint the size of a football
field. That's roughly five times more than predicted, because
scientists and engineers had always underestimated entrainment,
the domino effect through which descending avalanches swell
in size by tearing up and absorbing all that's underneath.
data is factored into simulation programs like AVAL-1D, a software
package released in 1999 that is now used by countries all over
the world, including Chile, Iceland, and the US. The concept
is simple: Key in the profile and maximum snow possible for
a given slope, then watch a pixelized avalanche sweep onto the
terrain below. "Commercializing our simulation programs forces
us to turn research findings into something operational," Ammann
explains. "We need to be very confident in the simulations.
This is life-or-death software." This year, the SLF will release
the next version, NewMix, which factors in the greater entrainment
and higher avalanche speeds from the Vallée de la Sionne experiments
- and will likely enrage alpine property developers as a result.
"Not all the old hazard predictions are wrong based on our new
data," says Sovilla. "Maybe just 10 percent. But in that 10
percent, people can die."
SLF's multiple avenues of research might seem only loosely
related, but to Walter Ammann they are all elements of a grand
design. Unfortunately, that's unattainable until Schneebeli
devises the laws to drive it. "Until we understand how grains
change and how bonds form," Schneebeli explains, "we won't truly
be able to predict the development of the critical weak areas."
meantime, riders in Switzerland's teeming backcountry depend
on the SLF's daily avalanche bulletin, generated by a team of
mountaineering experts using automated weather forecasts and
reports from 80 local experts. Last winter, the bulletin drew
more than 1.6 million hits online, served up 31,000 faxes, and
pushed 30,000 text messages to mobile phones. The bulletin's
hardly slope-specific, however, so riders still need a lot of
mountain experience to avoid death off-piste.
take years - a decade, maybe two - before Schneebeli's days
of watching snow melt pay off and Ammann has his seamless model.
Perfect avalanche forecasting will require flawless weather
prediction. Still, Ammann predicts that "nowcasting" - reading
the current safety of a slope - can reach 95 percent accuracy.
Standing at the top of a steep descent, snowboarders would be
able to enter GPS coordinates and rapidly receive a risk analysis
far more likely to save their lives than anything available
today. "Even when the people who died were reckless, I can't
blame them," Ammann says. "I always ask myself, 'Could we have
warned them better?' To me, any avalanche victim is one too
many. Even after 10 years, the news hurts every time."
Marc Spiegler (email@example.com)
is a freelance journalist based in Zurich.
Copyright © 1993-2003 The Condé Nast Publications
Inc. All rights reserved.
© 1994-2003 Wired Digital, Inc. All rights reserved.