“Aftershocks drop off over time,” she
continues, “so if there are 100 earthquakes
the first day, there will be 50 earthquakes
the second day, 30 -something on the third
day, it’s a regular relationship. You can
say the same about where they occur,
most aftershocks occur close to their
mainshock but peter out with distance.
“The thing we can’t predict is the size of
an earthquake, it’s just basically random
sizes. So, given an earthquake, there’s
about a five percent chance it’s going to
trigger something bigger than itself.”
And just as with predicting the weather,
there’s lots of computer modeling going
on, but in a way that you might not expect.
“You put out this initial forecast based
on ‘an earthquake happened’ or ‘a big
earthquake happened’, then we look at
past sequences in that region and similar
tectonic regions worldwide,” says Page.
“Once the sequence starts to progress,
we can see if our model is predicting too
many or too few aftershocks, and we can
adjust it based on how many aftershocks
are detected. As time progresses, we can
get a better handle on how accurate the
sequences are and we refine the model.”
LAPTOP PREDICTIONS
Sounds like the USGS must have a
pretty big supercomputer then? “We use
Amazon Cloud for a lot of our forecasts,”
says Page. “Some are simple enough that
we can run them on our own laptops. The
kind of modeling we do is similar to the
epidemiological modeling that’s used for
a disease like COVID. In the same way that
one person can infect a number of people,
and those people go on to infect others,
and there’s a cascade of contagion.
Earthquakes are like that, in that one
earthquake has aftershocks, and each of
those has its own aftershocks, so you get
another generation of aftershocks, and a
cascade of earthquakes.
“The best models [for earthquakes]
are more empirical than weather,” Page
continues. “You’re really just basing
it on things that have happened in the
Future computers may predict storms, such as this one in Broken Bow, Nebraska in 20 13.
Dr Morgan Page is a statistical seismologist
working at the US Geological Survey.
The Denali supercomputer at the USGS
headquarters in Pasadena, California.
WHERE
ARE THE
SATELLITES?
Much of the data for our weather
forecasts comes from polar-
orbiting satellites, those that orbit
the Earth north-to-south and pass
roughly over the poles. They are
generally at low altitudes between
125 and 600 miles up, though more
data comes from geostationary
satellites 22,00 0 miles above the
Earth, and yet more from deep-
space satellites, such as the
National Oceanic and Atmospheric
Administration’s Deep Space
Climate Observatory that sits a
million miles out.
A particular type of polar orbit,
known as a sun-synchronous orbit,
means the satellite passes points on
the Earth’s surface at the same time
every day—for example passing
over New York at noon each day. By
taking pictures and measurements
as the satellite passes over the
same spot at the same time day
after day, patterns of weather
can be studied, and the conditions
before dangerous events, such
as floods or forest fires, can be
identified so that they can be more
easily predicted in future.
A geostationary satellite, being
much farther away and not moving
relative to the Earth’s surface (a
satellite in a 500 - mile orbit travels
at 4. 5 miles per second) can take
in a much broader view, often able
to image an entire hemisphere
of the Earth at once. These are
often the images you see in TV
weather broadcasts, with cloud
cover imaged in visible light and
the near infra-red, while infra-red
imaging is used to see water vapor
in the atmosphere, as well as the
movements of ocean currents.
HOL 2021 MAXIMUMPC 41
©^
KE
VIN
SC
HA
FE
R/
GE
TT
Y^ IM
AG
ES
,^ W
WW
.JP
SS
.NO
AA
.G
OV
/SA
TE
LL
ITE
GA
LL
ER
Y,^
JO
HN
FI
NN
EY
PH
OT
OG
RA
PH
Y/G
ET
TY
IM
AG
ES
,^ U
SG
S/
ST
EV
EN
YO
UN
G
