Without that, we haven’t got much hope of
predicting what it’s going to be doing in the
future, so live observations are key.
“It’s crucial for metrology to be part
of the global community, and we do that
through many routes, but one of the key
ways is through the World Meteorological
Organization. Based in Geneva,
Switzerland, it’s an apolitical community
where all countries can contribute
observations in order to help each other.
The weather doesn’t have any boundaries,
and we really do need to know what’s
happening right the way around the globe
to be able to use our models effectively.”
Having gained a good starting point for
the model from all the live observations
coming in from around the world, the
Met Office puts its Crays into action.
“The physical equations are run at the
grid box corners, and we run what we
call ‘ensembles’, where we run the
model more than once,” says Roberts.
“At the moment we use 1 2-member
ensembles, where we slightly tweak the
starting conditions. As brilliant as the
observations we get are, we haven’t got
them from every single point across the
entire globe, so slightly tweaking the
starting conditions of the model will give
us a different forecast.”
“If all 12 in the ensemble are saying
a similar thing, that gives us a high
confidence in the forecast the model is
predicting. If nine of them are giving one
sort of solution and three are giving a
different one, then there’s uncertainty and
we start to bring probabilities into play.”
Which sounds a bit like Minority Report,
only with more people in the swimming
pool. Rather than just shooting those they
disagree with, however, the Met Office has
ways of dealing with split opinions: “When
we get an answer, or a variety of answers,
from the supercomputer, some of the data
goes straight to our customers, and some
is further processed using algorithms and
machine learning,” says Roberts. “Or it
could be done by human meteorologists,
and that is done regularly because our
models aren’t perfect, and until they are,
it’s really useful to have people in the loop
to see how well the models are performing
and to tweak them as and when required.”
GLOBAL RESOLUTION
Being UK-based, the part of the computer
model that sits over the British Isles uses
smaller boxes for a higher resolution
forecast in that area. Roberts looks
forward to the delivery of the new
supercomputer so that resolutions can
be tightened elsewhere too, and more
models can be run at the same time
as part of an ensemble. “By increasing
our supercomputing capability, we can
increase the resolution of our global
models and run them more frequently,
which is beneficial. As you can imagine,
ensembles take up a lot of processing
power, so having more members in
ensembles and being able to run more of
them at a higher definition gives us more
information and more accurate too.”
This description makes it sound like
there’s one climate model to rule them all,
but that’s not the case. Even within the Met
Office, they run different models together
as part of the ensemble approach. “We
have one type of model that’s dedicated to
the oceans, and one that’s focused on the
atmosphere,” says Roberts. “Increasingly,
what we’re trying to do is to combine
those models, and this is where increased
computing power comes in. The ultimate
would be a digital twin [of the Earth] where
everything interacts with everything
else because, as you can imagine, the
atmosphere and predicting climate
change is so incredibly complicated.
There are feedback mechanisms that
are difficult to model, and it’s difficult to
understand how they will play out. The
interactions between things like nitrogen
and the uptake of carbon into the land are
a small example of mechanisms that we
don’t necessarily understand yet.”
“Something our model has historically
struggled with is cloud types. What
we don’t know fully yet is whether high
clouds, such as cirrus clouds, will be more
or less frequent in a warming climate,
and that has big implications for how
much of the Sun’s radiation is reflected
back, or reaches the Earth’s surface.
All these tiny interactions can have huge
implications for what will actually happen
to the climate, so the more we can link the
different models together, the better.”
NOAA
All the different models have their own
strengths and weaknesses, and many
other national weather services around
The Met Office’s three Cray XC 40
supercomputers are to be replaced
HOL 2021 MAXIMUMPC 39
©^
PR
ES
S^ I
MA
GE
/M
ET
O
FF
ICE
