The Future of Weather Forecasts

Why does it take so much computing power to forecast the weather? And how could a weather study help one billion people?

Learn more about your ad-choices at https://www.iheartpodcastnetwork.com

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:00):
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking. Hey there, and welcome to Forward Thinking, the
podcast that looks at the future and says, it's like
rain on your wedding day. I'm Jonathan Strickland, I'm La,

(00:22):
and I'm Joe McCormick. And today you don't need a
weatherman to know which way the wind sucks, because we
are going to be talking about predictive modeling of weather,
weather forecasting. Yeah, we've talked in the past a lot
about weather and sometimes when I wasn't here. Yes, we
had a two parter about the potential future of weather

(00:43):
control with special guest Julie Douglas back in February. Yeah,
And one of the interesting things about that episode is
I think in the end we decided, after all of
our research that really the best avenue for humans to
sort of get a grip on the weather is not
to try to control it, because in many ways that
is a fool's are end, it's physically impossible, Yeah, but

(01:05):
to instead try to understand it, just to have a
better better idea of what's coming your way and win right.
The further out and the more accurately you can forecast
the weather, the better prepared you are for the various
eventualities that will unfold, things like flooding. Like if you
know ahead of time that flooding is is almost certainly

(01:27):
going to affect a certain region, you can start to
take steps to protect people and property in that area.
Sandbags are an amazingly effective low tech solution to things
or maybe out there. They don't do much good if
they're not there, Yes that's true. If they're they're somewhere else.
If they're in a warehouse, that warehouse maybe nice and dry,
but the area that you were hoping to say, will

(01:47):
be rather squishy. Uh same same sort of thing that
if you're talking about, like you're looking ahead at a
very long term forecast and you were to say, oh,
it looks like there's not going to be any rainfall
for quite some time, you can start to make plans
for that so that you're not stuck in a situation
where it happened but you weren't aware that that was

(02:08):
going to you know that was going to be the case.
Uh So, in other words, we don't necessarily try and
control it. We just get a better idea of what
is going to happen, so we're more prepared for that. Yeah,
And we glanced across that topic in those Future of
Weather Control episodes. But uh yeah, so we wanted to
talk about that today. And we were also inspired by

(02:28):
a video episode that you did, Jonathan about Bubble Yeah,
the Bay of Bengal Boundary Layer Experiment or Bubble Yes.
I I said in the video that I consider myself
a bubblehead because I'm a huge fan of this project.
The video and that just came out this week. You
can check it out on YouTube this very day if
you would like to, or on fw thinking dot com.

(02:51):
But specifically, Bobble is a very particular regional weather predicting project. Yeah.
It's a study of how a number of complex factors
in the Bay of Bengal come together to create a
monsoon season of heavy rains in northern India every year.
And it's a particular interest to researchers because that monsoon

(03:11):
season drives the agriculture and the water supply and the
energy supply for about a billion people, so seven of
the world's population, no big uh, And and clearly variations
in the seasonal norm of rainfall either too wet or
too dry reek havoc on this region. So what if
we could predict those variations before they happen, disaster could

(03:33):
hypothetically be if not prevented, then then perhaps mitigated. Uh Okay.
And so besides being a project that could hypothetically change
the lives of a billion people, Bobble is really cool
because it's kind of a microcosm of weather prediction research
in general, because it's it's so multi disciplinary. You've got
ships and satellites making classic observations in the bay. You've

(03:54):
got robotic submarines that are checking out the situation under
the surface. You've got researchers designing did little simulations to
crunch the data, and they'll compare their models to the
actual season's results to see where they went right and
where they need to make improvements. Right. So we're going
to talk more about Bubble in detail a little bit later,
But first, as per our usual m O, we like

(04:16):
to go back and look at how we got to
where we are now. Like, like, obviously, when you look
back to the ways humans tried to forecast the weather
centuries ago, they're supercomputers were sorely underperforming. Yeah, so those
advocacies didn't process it quite the same thing exactly. You

(04:39):
have all of your little scribes working in parallel, attempting
in vain to simulate weather well, and the thing they
were trying to calculate was how angry the god was.
So there were several steps along they were going, going
off the path in a few different ways. So let's
let's talk about let's talk about you know, kind of

(04:59):
the a shoudn't approach to forecasting weather and work our
way up to what we tend to do today. Okay, well,
joking aside, there were of course lots of just straight
up magical thoughts about how to control the weather or
predict the weather originally, and so that's you know, that
goes that's a tradition that goes way way back into
the ancient world. Has to do a lot with that

(05:20):
with astrology, um and as kind of an offshoot of astronomy,
but mostly it was astrological yeah, um, But those those
sort of magical predictive interpretations. Aside, there were actually throughout
history plenty of weather superstitions and sort of rules of
thumb that actually do have grains of truth to them. Yeah,

(05:45):
there's a really great article on how stuff works Dot
com about this, and I did a what the Stuff
video about it once and and it's it's interesting, how
many of them really do hold water? Yeah? Yeah, Well
it makes sense because you figure people are paying attention
to what has happened, and they realize that there's a
pattern where when once a circumstances happened, then typically you

(06:10):
might get a lot of rain. And so you start
to make a rule about that, and you know it's
in some cases it can be you can be completely
off base. It's just coincidence, or you do what I
like to call you. It's it's called, you know, a
confirmation bias, but I would call it. I would say
the van is always at the corner, which is where
whenever there's a van parked at the corner, you notice it.
Whenever there's not a van parked at the corner, you

(06:32):
don't register it. So to you, the van is always
at the corner. Uh. In those cases, obviously it may
be that you've made an observation, but it's a faulty one. However,
there's some that are at least somewhat you know, reliable. Yeah,
here's one. You've probably heard some version of this weather
prediction before and often in couplet form about red sky

(06:54):
at morning, sailor take warning, red sky at night, sailor's delight.
I feel this. I feel badly for that one sailor
right like it's just like like, oh man is going
to be awful because it says sailor was singular, so
it's one guy. Well, that's the way I always heard that.
There are other versions. But this is old, old, old,

(07:16):
It goes way back. People have been using this forecasting
rule for at least a couple of thousand years. We
know because it shows up in the Bible, shows up
in the Gospel of Matthew chapter sixteen, where uh it
says quote the in RSV, the Pharisees and sad Juicees
came and to test Jesus. They asked him to show
them a sign from heaven, and he answered them, when

(07:38):
it is evening, you say it will be fair weather,
for the sky is red, and in the morning it
will be stormy today for the sky is red and threatening.
You know how to interpret the appearance of the sky,
but you cannot interpret the signs of the times. So
obviously that they're trying to make a spiritual or religious
point there, but but just incidentally in the narrative, at

(07:59):
least we know that some people back then we're saying
this rule. Um so, so the author of this passage
had heard of this before, and crazily enough, it is
partially true. So what's the scientific basis for this? Why
would the color of the sky at sunset or sunrise
have anything to do with the weather? Well, strongly tinted

(08:21):
red light at sunrise and sunset actually tells you something
about the contents of the atmosphere between you and the sun.
So specifically, it tends to indicate dry air filled with
dust and solid particles which we would call aerosols. So
these particles in the air are the cause of the
reddening of the light because dust and aerosols in the

(08:43):
atmosphere scatter visible light in a way that makes the
light turn red. Uh And in turn, this dry, dusty
air tends to indicate that you're in a high pressure region,
which means less cloud formation and less likelihood of a storm.
A low pressure region, on the other hand, would mean
that that they're tended to be more cloud formation and
more storms. So if you are looking through red tinted

(09:07):
atmosphere to see the sun you're looking through a high
pressure region that's less likely to rain on you. Uh
And and the thing about the atmosphere is that it
travels in in the same direction. Well yeah, and that's
what this rule doesn't work everywhere, because while the Sun's
path is unidirectional around the Earth, of course it's actually

(09:27):
the Earth's rotation, but metaphorically, the Sun's path is unidirectional.
I'm gonna need to see a site for that. The
weather tends to travel in different directions depending on where
you live. So if you're in the Arctic or the Antarctic,
or in the tropics, the sort of three extreme bands,
weather patterns more often move east to west, and this

(09:48):
rule doesn't apply, or in fact, actually I guess the
opposite would apply, right, But for the mid latitudes, you know,
sort of the temperate zones between the tropics and the
Arctic or the Antarctic, this is actually more often true
because the weather patterns more often moved from west to east.
And what that means is, if you look towards the sunset,

(10:08):
you're looking west at the weather that's probably coming your way.
And if a red light scattering patch is to the
west of you, that's a high pressure area. Probably that's
probably headed your way, meaning the weather will probably be fine. Uh.
And of course why would red sky at morning be
a problem. Well, that's because if you're in the mid
latitudes again looking east toward a sunrise, you're seeing the

(10:30):
weather that has probably already passed by you. And high
and low pressure systems often do trade off in cycles.
But you may have noticed that I kept saying the
word probably over and over again there, And that's because
like all weather prediction, this is probabilistic. Using the system,
you can predict the weather better than random guessing, meaning

(10:50):
better than with fifty percent accuracy, but still not anywhere
near ad accuracy. Right, So there could be some mornings
where you see a red sky and every it's perfectly fine,
it's beautiful weather. And there might be some evenings where
you see red sky and the next morning you're soaking
in it. Yeah, So the weather, the weather is just
very complex. It's it's difficult to it's difficult to protict

(11:12):
with accuracy even now using the supercomputers and everything that
we have involved in all the data we have, but
this one piece of folk science and weather forecasting, it's
not the only one that turns out to have some
basis in truth. Right. Yeah, a few others that I
wanted to touch on because they're they're kind of a favorite. Uh.
Ring around the moon rain real soon. Have you guys

(11:34):
ever heard this? This is the thing that you've heard. No, no,
not at all, but I believe you. Yeah. Uh, there's
there's another kind of version of it that goes when
a halo rings the moon or sun rains approach and
on the run. I love that sounds like something from
one of their songs, and and and the thing that's
going on here it is it does hold true more

(11:56):
than fifty percent at the time. I think it's it's
a similar probabilistic concept to to the red red sky
at night Sailor's Delight sort of thing. But so, so
what's going on here is that, um, when you've got
a halo that frames the moon or the sun, it's
produced by by moonlight or sunlight refracting through high whispy

(12:17):
clouds that are made of ice crystals, and uh and
those those ice crystals. That type of weather pattern typically
occurs in siro stratus clouds that often move in ahead
of weather fronts, where where temperature differentials are going to
cause warm air to move upward, deensing moisture and potentially
forming rain clouds potentially, So science science thumbs up on

(12:41):
that one. And still not the only Moon related weather,
you know, kind of folklore, right, sure, Sure there's also
clear moon frost soon, Yeah, which which makes perfect sense
because because clear nights do often mean that cold weather
is on the way, Because as far as the planet

(13:02):
is concerned, a cloudless sky is sort of like having
a bed without blankets. Uh. You know, During the day,
the Earth absorbs sunlight and and can converts it into
into heat that we all appreciate to certain degrees um.
When when the sun sets, the surface begins radiating that
heat back out, and lacking clouds to capture the heat

(13:24):
and snuggle it in all all tight and close, the
surface and the lower atmosphere grow increasingly cold. In fact,
I think in a tech stuff episode I talked about
this as a means of creating ice in certain regions,
where you'd leave out a pan a shallow pan of
water outside because the heat radiates out and it actually
becomes ice that way in certain regions of the world.

(13:45):
That's how it was done before refrigeration reached those areas,
so it's kind of neat. Yeah. My favorite one though,
has to do with cows. Of course, there is there's
folklore about uh or not, like a folk saying, but yeah,
that cows will lie down when it's about to rain,

(14:05):
mm hmm. And and I will, I will admit that
cows lie down for probably many reasons, like they're tired.
But um, but but this one, but this one might
be due to to body heat. Okay, cows tend to
stand more often when they're overheating, you know, in order

(14:25):
to breathe everything out right. Sure, yeah, so so as
seated cow could arguably I mean that the weather is
cooling down and therefore a storm is a bruin. I
also like in the notes you have, this one may
have a leg to stand on. There. There are so
many puns in this in this house stuff works article.

(14:46):
And I yeah, I didn't write it, no, oddly enough,
Yeah it was not it was not, I but but there.
But there are definitely some some more systematic approaches that
people have come up with over the years, sure, apart
from just sayings in folk wisdom. One big one through

(15:06):
in history is Aristotle's Meteorologica. That's Aristotle's hugely influential treatise
on winds, water, weather, and some other stuff like earthquakes.
Like much of Aristotle, it is both startling lye intelligent
and hilariously wrong about lots of things. I enjoyed the
section on how earthquakes are caused by evaporation of rains

(15:28):
that have soaked into the earth and exhalations of breath
from the ground. But until a few hundred years ago,
I think the Aristotle's works were sort of the Western
world's gold standard for knowledge about the causes of weather.
And it wasn't until you know, fairly recent times that
we started being able to do much better. Yeah. I

(15:49):
mean generally speaking, you started getting into like the mid
to late Renaissance, and you start seeing some other thinkers
propose alternatives to some Aristotle's ideas. But yeah, his his
approach or his his observations and his his uh writings
held sway for centuries. Yeah yeah, um, And and some

(16:11):
of those new ideas came about alongside changes in concepts
about physics and also about astronomy, like like greater knowledge
of astronomy um up to and including the publication of almanacs,
which were very very popular publications back in the day.
Apparently the only thing that outsold almanacs in the seventeenth
century in England was the Bible, so lots of people

(16:33):
were purchasing these things. Um. And back in the late
seventeen hundreds and early eighteen hundreds, a couple different mathematicians
slash astronomers started publishing yearly farmers almanacs here in the
in the States and what would be the United States
later on the North America continent. Yes. Um. The formulae,

(16:55):
the formulas that they use in order to make these
predictions are to this day guarded as family or company secrets.
It turns out like it it could be something like
consulting the family cat. We don't know, Yeah, and like
intensely guarded. I love I I love stories about old
farmers are almanac and UH and the Farmers Almanac, both

(17:18):
of which are punctuated slightly differently in terms of the
possessive s, but just the lower around all of this
is is delightful. In the case of one of the
two almanacs, I forget which one. UH, there is a
Caleb Weatherbe who's sort of like the James Bond of
of this of this company. Because Caleb Weatherby is not
his real name, I'm not sure if it's a dude.

(17:40):
Uh I there have been this series of Caleb Weatherby's
who have been the one entrusted with the knowledge of
how the of how the almanac does it stuff. It's
like cecil atoms, yes, of straight of the straight dope. Yeah,
there have been many cecil atoms. Yeah so, but so
no one. No one knows exactly how they make their predictions,
but supposedly take stuff like planetary positions and sun spots

(18:04):
and lunar cycles and title patterns all into account, and
I get the distinct idea reading stories about this that
meteorologists find find almanax like this rather quaint. Uh what
One researcher who looked into the accuracy of these kind
of things found that they get their long ranging predictions
because they make predictions a year or two out correct

(18:24):
about of the time. Is that a high number or alone?
Like how much variability is there and what they could
be predicting? You can't because you wouldn't say, like is
that better than chance? Because it's hard to say without
knowing all the variables. Oh, sure, I'm not sure. They
claim to get it right about eight percent of the time,
and and that is that is sore a gap. Yes,

(18:48):
but luckily we didn't. We we haven't had to continue
relying just on stuff like this forever because eventually, UH physics, Yeah,
people started figuring out how hydro dynamics therm thermodynamics both work,
and once humanity got a really good grip on these concepts.
Strangely enough, around the same time that the American farmers

(19:09):
almanacs started publication, the science of meteorology could take off,
and by the early nineteen hundreds, a Norwegian physicist by
the name Wilhelm Erknus devised the first known seven equation
formula for for using observations of existing weather conditions to
solve for future conditions. Taking taking into consideration like like

(19:31):
pressure and temperature and humidity and then three aspects of
atmospheric motion. That forms the foundation. Definitely, I mean, the
more information we have, obviously, the better picture picture we
have what's going on right now, and the more um
the more accurate we can make a forecast for the future.
Of course, the further out you go from the current

(19:53):
UH scenario, the current the current condition. Small differences in
in what you've predicted versus what actually happened add up tremendous. Yes, yeah, well,
I mean it's a it's a sort of principle of
physics that you can extrapolate on a very simple scale
or on a very huge scale. On the simple scale,
imagine aiming an arrow at a target. If you shift

(20:13):
your aim a millimeter over and the targets a foot
of way a foot away, it's not gonna make much
of difference. If the targets a hundred feet away, it
will make a difference, right, So, same sort of idea
is that you know the the temporal distance as opposed
to physical distance, it does make a big difference. But
of course, once you get into the modern history of
our technological and scientific capabilities for predicting whether one big difference,

(20:37):
of course is just going to be the scale of
of observation, increasing the number and accuracy of observational platforms
to collect data about the weather, so we have more
information to work with, uh, And that's pretty easy. But
another thing is that we can sometimes overlook the simple
ways that common technological innovations help us in specific ways,

(20:59):
And one would be communication technology such as the telegraph
originally and then like the telephone facts and uh and
the Internet, and these have allowed people to better understand
global weather patterns in real time by rapidly sharing and
comparing information about local weather. Yeah. Computer science also allowed
prediction to to greatly advanced, starting in the fifties and

(21:21):
sixties and really ramping up over the past say like
twenty to thirty years, along with the rate of our
processing power. So I mean, perhaps obviously, as our computational
ability and our observational ability have increased, so has our
forecast accuracy. There was an analysis that was published in
Nature in and according to that, the forecast accuracy for

(21:42):
the next three to ten days of weather has improved
by about a day per decade um, meaning that right
now our ten day forecasts are as accurate as nine
day forecasts were in the early odts. So, in other words,
every decade we go by, we're getting one day better. Yeah.
I like it. So if I can figure out whether

(22:03):
or not I need to carry an umbrella with me
on Friday when it's Monday, and and be reasonably certain
that that is in fact the right answer, the better
because I'm not carrying it. If I don't have to write.
In a decade from now, you'll you'll be able to
know pretty well on Tuesday. I'm looking forward to that.

(22:24):
So my suggestion, Jonathan, is that you need to get
a cooler umbrella that you feel better about carrying all
the time, Like maybe like a penguin's umbrella, you know
that shoots machine machine gun fire or has a big
sword that comes out the end of it. I have
a blade runner umbrella that's great glowing. Yeah, I've got

(22:45):
one of those. Um So, who is really in charge
of gathering and crunching all this data? I mean, I'm
assuming when I turn on the local news and I
see the local weather corresponded on the news, that person
hasn't personally been responsible for gathering and analyzing all that information.

(23:06):
He has no, no, no, no. The guy I'm imagine
very specific, that guy launched the satellite uh and has
collected the data. He built all of the computers himself.
Uh No. Modernly, weather prediction is a joint public like
governmental and private industry type of business because that the satellites,

(23:30):
the computers, the software, and the the human compilation of
all of this data that go into it is each
each of those separately are huge expensive arms of the venture.
So and and going into it, you know, like, of
course you've got local news stations, which are private companies
that are reporting on whether but it's also a public service.

(23:52):
It's it's not just about personal convenience. It's absolutely a
very critical public service about getting information about big storms, danger, tornadoes, hurricane,
stuff like that out to the public um And it's
also partially a a tool for commerce. The more that
companies can learn about what the weather is going to do,
the better that they can adjust whatever it is that

(24:13):
they need to adjust depending on what's sure. Like if
if you're part of the shipping company, whether you're shipping
stuff across land or see you need to know these
sort of things because that can have a real impact
on everything from a delivery date to the safety of
the people and the products that you're moving. Weather is important,
I mean, it's important to have this as accurate a

(24:35):
picture of what's going to happen. And of course the
further out you can do that, the more beneficial it
is for everybody. So that kind of leads us over
into the discussion of some of the current attempts to
get an even deeper, more keen understanding of the factors
that influence whether UM and that kind of brings us

(24:56):
also to Bobble, to that project we were talking about
off the coast of India. So Bobble is pretty cool
in that it's it's relying upon multiple sources to gather
information UM also that we can get a better understanding
of the monsoon season in India. So that includes satellite data,

(25:17):
atmospheric measurements courtesy of an f A a M aircraft
and I'll go into that in a second, and some
floats that are carrying scientific equipment, as well as those
underwater robots that Lauren mentioned that are incredibly cool. I
was so interested to hear, mostly just about how they
move through the water because it's a brilliant and simple

(25:39):
means of propulsion. But first of all, the project has
a collaboration between India researchers and scientists from the UK,
specifically the University of East Anglia and the University of Reading,
and the research will take place during the two thousand
sixteen monsoon season, which has technically started as we record
this podcast. It's June and July. So the monsoon season

(26:01):
is India's rainy season. India gets a lot of its
rain during the season. Of the rain that falls in
India falls during the monsoon season, and there is a
lot of Yeah, we're talking ten ms annually of rain.
Ten ms, it's thirty three ft or so. In some
places it's up to eleven ms. It depends on the
region of India. Um. So the project's goal is to

(26:24):
gain a deeper understanding of the factors that influence this
monsoon season and that way we can make better predictive
models of what areas of India are going to get
what amount of rain, and that will help subsistence farmers
plan out there they're farming to make certain that they
take the best advantage of that. It also will help
in the case of figuring out this particular region might

(26:44):
be very susceptible to flooding and we need to take
measures to protect the people who live there. Right, So
there's there stands to be a really incredible benefit too.
Like we said earlier, up to a billion people to
to cracking this code, to figuring out better how it
works and therefore how to predict it. Right. So first

(27:05):
step of course is you gotta get the data right.
You have to collect the data before you can do
anything with it, and that's where all of that equipment
I mentioned comes into play. So first we have the
f A a M aircraft. F a a M stands
for a Facility for Airborne Atmospheric Measurements, So it's flying
through the atmosphere gathering data on the atmosphere as it

(27:26):
moves through. It's pretty uh interesting. You need to take
a little look at the picture of of these things
as a special refitted B a E Systems aircraft out
of the UK and uh it's the result of a
collaboration between the Natural Environmental Research Council and the Met
Office in the United Kingdom. Now, the f a a

(27:46):
M has a collection of sophisticated instrumentation aboard it. They
can those instruments can measure everything from radiative transfer so
essentially the way heat is moving through the troposphere, the
chemical composition of the atmosphere, humidity, tem sure turbulence, cloud
physics and more that turbulence in the cloud physics that's
really important. Things like vertical sheer that has a huge

(28:07):
impact on weather patterns and it's one of those things
that we need to have a lot of data on
in order to really understand what's happening, and the team
will actually compare the data gathered by the aircraft to
that from the other sources the floats, the weather satellites
and underwater robots to get a complete picture of what's
happening in the bay during the monsoon season. Uh So

(28:29):
some of that other equipment that the ARGO floats. Now,
ARGO floats are deployed all around the world, not just
off the coast of India. In fact, there are more
than three thousand of them floating in the oceans, and
they measure temperature, ocean velocity, so the actual velocity of
the water, the salinity of the upper two thousand meters

(28:50):
of the ocean. Scientists primarily use ARGO to monitor climate change,
so they're doing it to see how conditions are changing
over time to get a better idea of what is
the x will practical effect of climate change. The data
data gathered by ARGO is publicly available within a few
hours of its collection, so um, the scientists on this

(29:11):
project are going to rely on obviously on the ones
that are specifically off the coast of India. Then you've
got those underwater robots, they're called sea gliders. They look
kind of like um, almost like a torpedo shape. Some
sometimes they're referred to as like an robotic dolphin, which
is odd because they don't really have like they're not
jointed where you have a t now they've got they've

(29:34):
got a pair of wings that can tilt. But they
use changes in buoyancy and those wings to create forward
momentum so they can move through the water. And they
have a battery inside of them that can actually shift
around as ballast, and that will allow them to change
their pitch and roll so they can dive down. They
can they can move through the water. They do so

(29:55):
very slowly compared to say a propeller, But unlike a propeller,
it's in incredibly energy efficient. Yeah, it doesn't have to
use a lot of energy to change. Uh, it's it's
position because of the buoyancy and use of its own
battery is ballast. So therefore, if it's energy efficient, that
means that it can travel quite a great distance, probably
on a single charge, without having to go back to

(30:15):
home base and h and be juice up again. Exactly,
it can stay under water for a long time and
can travel a great distance. Really essentially only has to
surface if you do have to recharge it or for
it to beam the data back. It's got a radio
antenna at the tip of it that will poke out
the water beams that information and the team can gather it. Uh.

(30:37):
It's really a neat looking device and there are videos
online that you can watch of it in action. Um.
They're they're a little expensive there, about a hundred fifty
pounds sterling each. Uh. The University of East Anglia used
to have six of them and then lost two of them.
One of them got run over by a boat. What

(30:58):
a ship really well, because these things tend to stay
fairly close to the surface in order to beam information
back and one and they don't move very quickly and
they're hard to see. They're not huge right there, about
the size of a person, but if you're operating a
large ship like a cargo vessel, you may not see it.
And a cargo vessel collided with one and destroyed it.

(31:21):
The second one was lost in Arctic ice. I believe so.
But there are actually seven of them in operation for
the Bubble project. UM, so really interesting. They also can
hold lots of different types of sensors, not just ones
to measure the various factors in the ocean, but others

(31:42):
as well, for for things like marine biology. Now, of course,
in the case of bubble marine biology was not really
one of the things they were necessarily concerned with. So
that's not the that's not in the instrumentation um for
those particular seed gliders. Instead, they're looking at sensors that
are going to measure stuff like the turbidity of the water,

(32:03):
the temperature, salinity, and the oxygen content. Now you collect
all this data with the floats, the robots, the satellites,
the aircraft, and now you know everything. Now you gotta
do stuff with it. That's the problem is like like
for one thing, like you know, just just that information
alone is incredibly valuable, but without knowing how it all

(32:25):
interacts with one another, which factors are more important, which
ones are really impacting the monsoon season the most, which
are causitive versus just correlative, Right, Like, there may be
some things that change. Maybe they're changed because the monsoons
are moving through, not because they change and then cause
the monsoon. Right, So you've got you've got to determine

(32:47):
all this. You have to crunch all that information, and
that's gonna be the next big challenges grabbing all that
data and doing something useful with it so that then
you can take that knowledge and communicate it to people,
so that you can make actual, uh, real world actions
based upon that data. And this is where we start

(33:10):
to shift over to a very important tool in weather
forecasting and weather modeling and climate science supercomputers. Yeah, because
if you haven't cotton on yet, the problem of weather
is is a big data problem. Yes, it's a it's
a huge data problem because we know lots of different

(33:30):
variables affect weather. We know those variables change greatly over
spans of time. Right, So you've got a lot of
information and that information is constantly in flux. So how
do you process that in a reasonable way. Supercomputers have
proven to be a really important element of this analysis.

(33:52):
So part of understanding this is knowing what a super
supercomputer really is. It's not just a really beefed up PC. Right,
It's not a beefed up Mac. It's not a beefed
up Mac either um also known as big Mac. Yeah,
it's none of those things. Although I mean, Mr Hodgeman,

(34:17):
if you're listening, you don't need to beef up. We
like you the way you are. So supercomputers tend to
be organized in a way where you've got nodes, which
is essentially either a CPU or a GPU um and
those are organized into blades. Those blades are further organized
into racks, which are cooled in some interesting way, usually

(34:37):
water cooled. Because you get that many processors in a
place together, they generate a lot of heat. Heat and
electronics over the long term are not good friends with
one another. So the in effect is you've got a
supercomputer that acts kind of like a multi core processor.
So if you have a multi core processor, you might wonder, well,

(34:58):
how does this make my computer faster? Well, it works
really well for certain types of computational problems. Those will
be problems that could be broken up into smaller bits.
It works less well for problems where you have to
solve one problem before you can start on the next problem. Right,
So if you were to have the first type where
you have a problem, you can split up into little bit.

(35:19):
So you can think of that as imagine you've got
uh A, I like to use this analogy. You've got
a math class, and in that math class is a
math genius, and then you've got a bunch of decent
math students, but they're not of genius level. You've got
a math problem that's that first type one that could
be broken up into several smaller problems, and you give

(35:40):
the math genius the full thing, and you give each
of the math the good math students part of that problem.
The group of good math students are more likely going
to finish it before the math genius, even though the
math genius has a grasp of mathematics that far outpaces
that the rest of the class. If the second type

(36:00):
of problem, like you were talking Joe, then the math
genius is more likely to finish it because you can't
divide that problem up and and give each little piece
to all the different math students. So the math students
represent that multi core processor, right with a supercomputer. You've
just got thousands of these processors, like more than eighty
thousand for some big supercomputers. Right, And so you take

(36:26):
this problem, the problem being, here are all these variables
in weather, and I want My solution is I want
to create a weather simulation so that I can forecast
what will happen in the future based upon the current situation. Now,
so that's your first step. You create your model, then
you look and see if your model is any good.

(36:48):
One way you can do this actually is to feed
in data from the past. So let's say that you
have collected a huge amount of information from two weeks ago. Well,
you already know what happened after that because it's in
the past. Sure, so you can you can feed all
of the information from two weeks in the past into
the computer and say, h if I modeled this a

(37:08):
certain way, then do I get like, like, how close
do I get to what actually occurred after that first week? Right?
And if and if it turns out that it didn't
come very close, you start making adjustments. You start saying,
all right, this one factor that I thought was really
important turns out maybe it's not so important. And this
other thing that I kind of overlooked turns out as
much more instrumental than I had anticipated. And and this

(37:31):
is a long process, but you you refine that simulation.
This is a cool way in which weather prediction, I
think has the potential to be a constantly improving science
because unlike some disciplines, uh, this is not a field
in which testing the predictive power of your theory or
in this case, your algorithm is difficult because compared to

(37:52):
something like psychology, where the results of your experiment might
often be very fuzzy and indeterminate, or like particle physics,
where you might have to test the predictions of your
theory by building some giant experimental instrument that operates at
the giga electron volt scale or something like that, the
weather is not like that. We have tons of data
on it, always new data coming in. We've got plenty already,

(38:15):
and we have lots of good ways of measuring it already. Yeah,
And the problem is really that we have a wealth.
We have to we have we are we are befuddled
by our wealth of information, right, Yeah, No, I just
like I just like the like we have plenty of
it already. Like I was just thinking, like, well, not
much weather today. I had a lot of weather yesterday,
which uh oh there it's from the Mystery Science Theater

(38:37):
episode of pod People Were the Best. One of the
characters asks, uh, do you think the weather all hold in?
One of the viewers comments, no, I think it's just
gonna stop. That was Tom Servo who said that. I
remember that. Yeah, No, that's a fantastic episode. So Tangent
go watch that episode of MST three K. It's one
of the best ones they ever did. Back to back

(38:59):
to the their forecasting. So, according to Science Daily, supercomputers
spend about equal amount of time running their simulations to
assimilating new real world data into the models. So, in
other words, half the time you're simulating whether the other
halftime you're adjusting that simulation so that it more accurately

(39:20):
reflects the real world. And as we get a better
understanding of the things that affect whether, we can refine
that um. A study in Japan ran a global atmosphere
simulation and found that a weather event event in one
part of the world can affect other weather events thousands
of kilometers away. And so it starts to dawn on

(39:42):
you that in order for you to accurately forecast a
local weather system, you have to actually look well beyond
the immediate region, because there are factors that will affect
that local weather system that are happening really far away.
And it may be that it's it's something that's not
i mean, gonna instantaneously affect your local weather, but it

(40:05):
will have an impact. So maybe something that would have
normally been a rainstorm, but that's it could potentially turn
into something much more severe like tornadoes. So it was
really interesting. And the study included ten thousand, two hundred
forty simulations, and they divided the global model into twelve
kilometer sectors, so like a grid of a hundred twelve kilometers.

(40:28):
Now that's also important because the smaller those squares are
in the grid, the more data you're feeding into the simulation,
and the more powerful the supercomputer has to be. Yeah,
and of course we're always expanding our hardware and software capability.
So in January, the n o a A announced a

(40:48):
major upgrade and its Weather and Climate Operational supercomputer system. Uh,
and this was interesting. The two computers they have are
called Luna and Surge Urge not like the soda, like
a wave. Well, yeah, like the soda. Sorry. The Luna

(41:08):
and Surge are based in Florida and Virginia and each
one runs at two point eight nine pedal flops for
a combined five point seven eight pedal flops of computing capacity.
And that is up from the system's capacity of just
seven seventy six terra flops. Nothing to sniff at, but
significantly lower last year. Flops, by the way, stands for

(41:29):
floating floating point operations per second exactly. So in the
press release, the you know, administrator Dr Catherine Sullivan said
that this upgrade would help the organization deal with quote
the tidal wave of data that new observing platforms will generate.
Just once again, I think we've sort of said this before,
but uh, indicating that the problem in weather prediction these

(41:52):
days is not a data problem, but it's an analysis problem.
It's the what we do with the data that's where
the bottleneck is. Right. So also from n o A
A NOAH, the National Oceanic and Atmosphere Administration. In other words, uh,
they're running fifteen hour forecasts using something called the high
resolution Rapid Refresh model, also known as the H triple

(42:14):
R in meteorological circles. So if you have a meteorologist
in your family, just ask them how the H triple
R is going her her or if you want to
put model in there, it's the HERB. Anyway, the model
divides the map, the global map up into three kilometer sections.
So you remember I was talking about the Japanese study

(42:35):
that was a hunter and twelve kilometers, So this one's
more precise. It's divided the the entire world into smaller sections,
which increases the amount of data significantly that they have
to handle in order to make this fifteen hour forecast.
That's also why it's only fifteen hours out, because to
to extend the forecast further would require even greater processing challenges,

(42:59):
which they're working to overcome and slowly push that number
further and further out. Um. But it's really interesting that
they are looking at the world in three kilometer sections.
It blows my mind because you think how huge uh
an amount of data that must be that they're dealing
with consistently, and they're refreshing this hour by hour to

(43:20):
look another fifteen hours ahead. UM. So, in Europe weather
satellites are actually more advanced than the ones that we're
using here in the United States right now, but that
will change. The US has plans to launch the Geo
Stationary Operational Environmental Satellite are also known as GOES ER.

(43:41):
Did they come in the form of a giant slore
as as GOES are the destructor or? Um? Yeah, I'm
having Ghostbusters flashbacks on that. But it's scheduled to launch
in the fall of this year. It will actually become
the most advanced meteorological satellite in orbit for at least
a short time, finally outpacing the ones that are are

(44:02):
currently over Japan and Europe. Other other recent news involved
IBM spent about two billion dollars acquiring basically everything in
the Weather Company except for the Weather Channel itself. And uh,
and so they're apparently gonna pitt Watson against all that
data and just kind of see what they can do.
Interesting Watson takes it down? What Watson Watson will take

(44:25):
all that data and make yet another bizarre and unimaginable
recipe that involves pot stickers that don't have any of
the ingredients in them that they claimed that is it
going to rain next year? First? Grill you r let us?
Oh man, I still think we have to each take
one of those recipes, make it and bring it in.

(44:46):
We never did do that. We should do a live
show where we subject each other to cooking grill your
perade olives. I think we all we all will need
to have a chef hats and and aprons with humorous
sayings on them. Uh, that's that's what I suggest. All right,
Well, well well we'll work on that at any rate. Let's

(45:07):
look at again kind of further off, like what was
the future going to bring. So once we have these
more advanced satellites, we're constantly working on building better supercomputers,
which often are used for this kind of thing, as
well as other branches of science as well. Uh So,
for one thing, as we get this greater understanding of
the global influences of whether we can we can improve

(45:28):
our forecasting when we understand that an event happening thousands
of miles away will have an impact on the weather
in our area and we have a better way of
of predicting what that impact will be. That's gonna benefit
people in ways that we can't even really get a
grip on right now. Um. One of the other things

(45:49):
we have to remember is that it's a lot easier
to predict weather in general, that is severe weather. Um.
So you'll see this on lots of different sites that
are talking about meteorology. They'll say like, oh, you know,
we can predict general weather systems out maybe as far
as a couple of weeks or further. But when you

(46:09):
start getting into the the the possibility of severe weather,
it's closer to like five days, and each day out
is less accurate than the day before, which means that
when you're looking at the tail end of that forecast,
you have to keep that in mind. Um. I tried
to do that all the time when I'm thinking, like, oh,
I'm going on vacation in two weeks, let me see

(46:30):
what the weather is gonna be like in ten days,
And and often I go in with a false sense
of security, or I'm end up preparing for a rainstorm
that had just doesn't happen. But as we get more information,
we get better at anticipating these things and predicting them accurately. Obviously,

(46:51):
this could help lots in lots of ways, like in
that commerce that we were talking about, or in travel.
Absolutely having better weather prediction could have all kinds of
commercial and environmental bonuses, like imagine being able to reboot
flights around bad weather systems before storms hit, thus preventing

(47:11):
having to sit around at the airport all day, or
or having to have your flight canceled, or even allowing
pilots to save on fuel by plotting better courses. Also,
as as Julie brought up, in our prior weather episodes,
changes in whether change our buying habits, supermarkets could plan
to stock up on those frier chickens or whatever it is,
way more in advance. Apparently, apparently during certain disasters, fried

(47:36):
chicken just flies off the shelves unless which is weird
because chicken rarely flies even when it's not fried. But
also there's the issue here in Atlanta. I made the
joke in our notes that it's not really a joke.
It's actually just a fact that if there's even the
hint of snow, you can expect a run on supermarkets
for all the milk, bread, sometimes bleach bleaches, big yeah,

(47:59):
and then people get home toilet paper. What do you
do with this? Yeah, I never buy this to begin with,
exactly lots of French toast. That's what we're gonna be
having kids. So yeah, But they having those predicting those
better forecast means that you know, you can actually prepare
for that sort of stuff and uh and hopefully not

(48:20):
encounter things like shortages or or or you know, where
people go to a store and then they realize that
they're out of luck because everybody has rushed it. If
you've got more time to prepare for that, then you
can build up your inventory and make better profit and
people can be happy that they can you know, get
their bread and milk and eggs and make that French
toast and then when it doesn't snow, everyone complains about

(48:42):
it the bread and milk and eggs go bad, but
you don't care. You sold them already. Yeah, yeah, capitalism. So, uh,
it was fun to kind of look into this. I always,
I always really enjoy discussing, uh, the idea behind weather science.
I'm not big on talking about the weather in general,
but whether science to me, is really neat because you
start to realize how incredibly complicated it is and how

(49:05):
much energy are the the energy that are that happens
to be in these big weather systems like you know,
we we if you talk about hurricanes, the amount of
energy and a hurricane is phenomenal. Right, as Lauren has
so succinctly put it before, there's more wind than truck
fair enough. So to me, that's why I love talking

(49:25):
about these things and why I felt that it was
fun to to come back and revisit this. Plus I
wasn't in the last couple, so I really wanted to
kind of jump into it. But guys, if you have
any suggestions for future episodes of our podcast, let us know,
send us an email. That address is FW Thinking at
how Stuff Works dot com, or you can drop us
a line on Twitter. The handle there is f W Thinking,

(49:47):
or search f W Thinking and Facebook. Our profile should
pop right up. You can leave us a message there
and we look forward to hearing from you, and we'll
talk to you again really soon. For more on this
topic and the future of technology, visit forward Thinking dot
Com problem brought to you by Toyota. Let's Go Places,