(AV17952) Demons and Butterflies: Weather Predictability and Predictions

(AV17952) Demons and Butterflies: Weather Predictability and Predictions

November 16, 2019 0 By William Hollis


well good evening I’d like to welcome
everyone to our spring 2013 Sigma Z national lecture I’m Lee Burris and it’s
my distinct privilege to introduce our speaker as you can see from the screen
our speaker is Richard anthis president emeritus of the University Corporation
for Atmospheric Research dr. Anthes is a native of Missouri grew
up spent much of the early life in Virginia he’s a graduate of the
University of wisconsin-madison where he earned his BS MS and PhD each in
meteorology following his his PhD he joined Pennsylvania State University as
a faculty member where in a less than 10 years he was promoted to professor he
stayed there till 1981 when he left Penn State to go to the National Center for
Atmospheric Research in Boulder Colorado and he spent a subsequent time in his
career in Boulder initially with the National Center for Atmospheric Research
and more recently with UCAR the university Corporation for Atmospheric
Research dr. antis as you’d expect as a
distinguished career as with any of this speaker with Sigma’s I has a very
impressive Veta he’s published more than a hundred
papers he’s been very active he’s a fellow in the American Meteorological
Society I’m having a cheat at the moment read things he’s a winner of the
American Meteorological Society Clarence LM singer award for promising
atmospheric scientists especially in researching tropical cyclones
he’s a holder of the Jules G Charney Award for sustained contributions to in
theoretical and modeling studies he’s a fellow in the American Geophysical Union
and perhaps really even more noteworthy award than those is he received in 2003
he received the friendship award from the Chinese national government which is
an extraordinarily competitive have an impressive award so I could
continue with that but you’re really not interested in having me speak about dr.
antis we’re really much more interesting hears him speak so dr. Hamza’s please
the floor is yours okay thanks thanks very much Lee and thanks to a sigma zai
for inviting me given the later start then I usually
give lectures I’ve shortened this to three and a half hours so I promise you
even if I go over a little bit we’ll be done by midnight for sure the title of
my talk is demons and butterflies weather predictability and predictions
and you’ll see what all this will become clear to you if it’s not already I’m
going to give a prologue during which I will say almost nothing pictorial essay
of Hurricane sandy and you will see the relevance of good predictions I think
more forcibly if I just show these slides two seconds apart without any
comment whatsoever I was certainly surprised when I first saw it I follow
hurricanes closely my wife and I were on the East Coast during sandy south of the
storms landfall and I had no idea of the real devastation until I saw this series
of photographs which were put together by my good friend and colleague nausia
Pierrot so I will just be quiet for the next couple of minutes and that’s you in
silence see these incredible pictures of the devastation of Hurricane sandy and
then I think there’s a lot some lessons I will draw from that that are relevant
to predictability and predictions so here we go okay so I think that gives you a feeling
that statistics and words can’t and give this isn’t really the topic of the talk
but I thought it was a cover on the Bloomsburg Business Week and that this
is a whole nother subject which I won’t get into tonight but the few statistics
on Hurricane sandy aka superstorm sandy and frankenstorm
it’s formed in the Western Caribbean on October 22nd very late in the season for
a tropical cyclone made landfall New Jersey at 8 p.m. Monday October 29th as
a week later it was the largest Atlantic hurricane on record not the most intense
but the largest meaning the diameter of the circulation the diameter of the
clouds at least two hundred and fifty-three people were killed and at
least 74 billion dollars in damage which is about equal to the much-discussed
sequester so a huge economic impact in this one storm and I’ll get back to that
253 people killed in a minute how good were the forecasts this is a talk on
forecast predictability and predictions well they were excellent then the
question I have for you is how many more lives would have been lost without the
good forecast and I’ve been have no way of knowing this of course but I’m
guessing it would be thousands rather than 253 people I gave this talk at the
Oregon State University last week a person came up afterwards and said that
she thought that the thousands was probably an underestimate it probably
might have been tens of thousands of people and that’s because so many people
were in harm’s way and were out of harm’s way by the time this thing hit
there was warnings of at least three days and there’s massive evacuations
people moved inland instead of in being in all these these houses I gave a
similar talk at Fairfield University which is in Fairfield Connecticut right
on Long Island and there’s some houses summer houses
that wealthy people go to during the from New York go to during the summer
and these houses are vacant during the whether or not vacant they’re used by
students that rented by students during the winter season winter season being
said you know after Labor Day into until Memorial Day and so they were probably
six hundred students in these these houses you know how students crowded
into places that save rent and it’s right on the on Long Island this this
area was totally evacuated mandatory evacuation they got everybody off out of
these house all the students were taken off some of those houses were dragged
into the ocean totally demolished all were severely flooded
I would guess just this one little settlement alone would have been
probably five hundred young people would have been been washed out to sea so
incredibly important that the forecasts were accurate and people heated them
thousands instead of two hundred and fifty 250 s not great if you’re one of
those family members but it sure beats twenty five thousand or something like
that so forecasts are really important forecasts and these storms happen in
different parts of the country all the time and so that’s just to set the
context how important this is now why were these forecasts so good that’s kind
of a lesson I’m probably repeated several times forecasts were good
because modern meteorology is science not folklore not based on statistics but
based on mathematics physics satellite observations and these systems are
almost all funded by the US government and the the government has funded these
systems and development for over 50 years much of the research was done in
universities including universities like Iowa State and so if you hear people
saying that government is too big it’s worthless it wastes its money think
about hurricane sandy and the thousands of lives were saved by the
government-sponsored research the government-sponsored satellites the
government-sponsored computer models and the research done in in our great
universities around the around the country that’s that’s one kind of
take-home message never before had a hurricane approached the East Coast from
the east and late October that’s just how unusual it was and this
shows the the track and I’m not sure I have a pointer here but there’s a red
button but I’m not sure I want to push it so so you can get the you can get the
idea hurricane sandy has labeled Sandy that’s
the storm that comes up off the coast of Hatteras and then makes that left hook
and goes inland all the other hurricanes of category two are that passed within
200 nautical miles of New York City for the last hundred and fifty years as you
can see we’re moving to the Northeast if they made landfall at all they just
brushed the coast so hurricane sandy just stands out like a sore thumb so any
method based on folklore experience of forecasters statistical methods
empirical methods would have all failed because it had never happened before
and yet the storm was forecast more than a week ahead of time now here’s their
three panels here this is from the European Centre for medium-range weather
forecasting in Europe it’s the best weather forecasting center in the in the
world on the Left shows the probability the forecast probability of a wind storm
nine and a half days before landfall and you can see that already nine and a half
days before a storm it even formed there was a highlighted area of possibility of
heavy wet winds nine and a half days before the landfall and then the middle
panel shows a series of tracks that were forecast by the model and that’s the
so-called ensemble of forecasts you see that many of them if not most did indeed
predict that left hook making landfall along the East Coast and the actual
storm of Sandy is on the right so this was
indication the forecast was made of something like sandy two days before the
storm even formed much less storm had formed and they were just forecasting
the track so a forecast the formation of the storm two days before it occurred
and then the subsequent track ended in a track that had never before been
observed and it’s all based on science basically a hands-off operation based on
computer models and observations that go into these models so that’s the prologue
we thought that was the talk right well now we’ll get into some more history and
some more fanciful I guess that well this one worked so somewhere in a jungle in Brazil this
is going on right now it’s okay so butterflies and demons what’s the
significance of butterflies which you saw in the jungle and Brazil and the
demon that you saw it’s my friend Mel Shapiro who did that a nice photo essay
from you well foretelling the future has always
been a fascination of humanity and profits over the ages have been
worshiped and vilified it’s not just weather forecasting but forecasting
foretelling the future of everything it’s just a fascination of human beings these are all course ways of expressing
how we look at foretelling the future the last one was when kind of fast but
said that all predictions are based on observations and I can say that all good
predictions are based on observations I can’t think of any predictions that are
not something that are scientifically based that are not based in some sense
on observations so what about predictability and prediction well the
idea of prediction and foretelling the future goes back many many hundreds of
years and Gottfried Leibniz and who lived from 1646 to 1716 said that
everything proceeds mathematically if someone could have sufficient insight
into the inner parts of things and in addition had remembered and had
remembrance and intelligence enough to consider all the circumstances and take
them into account he would be a prophet and see the future in the present as in
a mirror now what he’s saying is that if I new completely understood life and
life is chemistry basically biology and if I knew every one of your brains and
how they worked and how you’d react to stimuli of various kinds everybody in
the world all 7 billion and I knew how the butterflies flapped I knew
everything about the world I could predict what’s gonna happen five hundred
years from now convicted Rijn you’re gonna have how many children they’re
gonna have that’s the extreme of what he’s saying but that’s basically what
he’s what are you saying that’s a philosophical question obviously and
it’s a good thing to just have fun talking about how much can we know and
how far will that take us buy into predictions well similar ideas followed
the Marquis de Laplace dreamed of an intelligent being and intellect which he
called an intellect and was later dubbed Laplace is demon who knew the positions
and velocities of every single atom and use Newton’s equation of motion to
predict the future of the entire universe and he said this is a quote we
may regard the present state of the universe as the
effect of its past and the cause of its future and intellect which at any given
moment knew all the forces that animate nature and the mutual positions of the
beings that compose it if this intellect were vast enough to
submit the data to analysis could condense into a single formula the
movement of the greatest bodies of the universe and that of the lightest atom
for such an intellect nothing would be uncertain in the future just like the
past would be present before its eye so this is a physical physical universe
kind of statement it doesn’t get into the prediction of people and animals how
the biology would respond but it’s a it’s a physical one if we knew all the
atoms and could foot and knew the laws that govern all the atoms we could
predict the future of the universe now this is a surprisingly good description
of how numerical weather prediction works because numerical weather
prediction we try to know the positions of all the storms all the high pressure
systems of the entire atmosphere the oceans which interact with the
atmosphere we try to use Newton’s law mathematically to move these systems
ahead in time in forecast their future is really exactly what he’s saying we do
it for weather we don’t go much beyond weather we don’t go into the forecast of
the universe and certainly not forecasts of people but this is pretty much what
we do in in numerical weather prediction so it’s observations knowing the present
and knowing the laws that govern the fluid which we call the atmosphere in
the fluid which you call the oceans and how they interact in a very
deterministic predictive way and I keep coming back to this as I go along but
then there’s some some skeptics niels bohr you can read it here it sounds to
me it’s like something Yogi Berra would have said and he probably might have
said it so more relevant to meteorology one of the greatest
early meteorologists of all-time times Vilhelm björk knees in norwegian said we
said what Laplace said with direct relevance to meteorology he said if it
is true as any scientists believe that subsequent states of the atmosphere
developed from preceding ones according to physical laws one will agree that the
necessary sufficient conditions for a rational solution of the problem of
meteorological prediction are the following one and two one is one has to
know what sufficient accuracy accuracy the state of the atmosphere at a given
time these are the initial conditions what we call the initial conditions and
– one has to know with sufficient accuracy the laws according to which one
state of the atmosphere develops from another so number one is observations we
have to observe the atmosphere at some initial time and number two is solving
the appropriate equations that govern the future behavior of the atmosphere on
very big computers and this is solving a relatively small set of differential
equations and physics such as clouds solar radiation radiation emitted from
the earth and turbulence and their their mathematical and physical laws to
predict the future state of the atmosphere so this is exactly what we’re
doing today as we make these forecasts like hurricane sandy number one is the
satellites the observations that give us the initial state of the atmosphere
number two is the computer models and the computers that are required to solve
these equations on the computer to give the future state of the atmosphere well
perhaps the most famous theoretical meteorologist of all times was Ed Lorenz
and I’m told that he is a meteorologist was a meteorologist at MIT I’m told he
had a big influence on mathematics the field of mathematics because he
developed a very simple set of differential equations that were
nonlinear and showed that from very small differences in initial conditions
differences that were too small to be measured the future evolution
this system would become chaotic and therefore unpredictable so no matter how
well you observed the now atmosphere atmospheric analogy is no matter how
well you observe the atmosphere you’d always miss something and that what you
missed would then eventually cause the atmosphere to behave in a way that you
could not predict no matter how good your computer models were and he
estimated from these very simple models that the limits to predictability which
is the potential ability to predict the weather in a deterministic way was about
two weeks and you may have heard for many years that we can’t we can’t ever
expect to forecast individual weather events beyond two weeks there was a
surprisingly good prediction that we do have some skill in today’s weather
prediction out to two weeks but it’s not very good and it’s not clear that we’re
going to go beyond two weeks anytime soon so that’s the butterfly effect the
butterfly effect meaning we can’t measure every butterfly in Brazil we
can’t measure what that butterfly is doing so therefore there will always be
something we don’t know about the initial state of the atmosphere and that
will cause any computer model no matter how good to lose accuracy to lose skill
in roughly two weeks so I was skeptical or at least had a somewhat more
optimistic point of view in my days at Penn State in the late 1970s I won’t
read you the whole thing but Tom Warner a student of mine was was developing an
early version of a small scale a regional scale meteorological model and
we were noticing that this model was capable appeared to be capable of doing
some amazing things that could could predict features that weren’t even in
the initial conditions that weren’t even observed at the initial time and the
model was somehow creating this information
information that could be observed and so we hypothesized that in in many cases
storms that didn’t even exist at the initial time the formation could be
formed and the subsequent evolution and movement could be could be forecast it
was kind of an optimistic point of view so what has happened in fact that
actually did happen in hurricane sandy as I as I indicated the the model
predicted the formation of Sandy two days before it even even was present
form well this is kind of a complicated chart but what I want you to notice this
the left side is 1981 the right side is 2011 so this is a map of forecast skill
from the the model that we just heard about that fork did so well in
forecasting sandy and the these are basically correlation coefficients so
100 would be a perfect correlation that would be pretty much at the top of the
curve and then we see four sets of curves all increasing which means the
forecasts are getting better the top one the blue envelope is the three-day
forecast the next one below that the red is the five-day forecast the greenish
one is the seven-day forecast and the bottom one the yellow one is a ten-day
forecast so what it shows is the three-day forecast is on the average
better than five days which is on the average better than seven days which is
on the average better than ten days the two parts of each envelope are the
Northern Hemisphere and the southern hemisphere so for example the top
envelope the blue ones the best scores the northern hemisphere is the top dark
blue curve and the southern hemisphere is the lower lighter blue curve and what
it shows is 1981 the northern hemisphere forecast was much better than the
southern hemisphere forecast but over time the southern hemisphere forecast
has caught up to the northern humans for hemisphere forecast
and by 2011 they’re virtually identical the same thing happens at day five day
seven and day ten well why were the forecast in the northern hemisphere
better much better than the forecast in the southern hemisphere in 1981 not many
observations in the southern hemisphere all the much more land many more people
in the northern hemisphere and only Australia New Zealand had weather
observations in the southern hemisphere but as time went on we got more and more
satellites which measured the whole earth whole atmosphere and so they
measure the southern hemisphere just as well as they measure the northern
hemisphere and so with time the influence of satellite observations
became greater and the southern hemisphere forecasts have now become
just about as good as the northern hemisphere forecast and that’s true of
day five day seven and day 15 there are other aspects of this this diagram you
might and that’s that the the models are getting better in addition to the
observations in one of the ways the models are getting better is in the
horizontal resolution and what the these little squares that are going to come up
are going to show is the average horizontal resolution of this model at
the different times so in 1981 this is a the pixels of the model if you want to
think of them like a camera was we’re 200 210 kilometers apart and what is
that feature that’s being shown there you have no idea right and why do you
see this as we increase the number of pixels which means increasing the
resolution and you will see what this just turns into so these are the
resolutions as time goes on and in 1984 125 still can’t really tell what it is
63 kilometers in 1990 39 kilometers in it is 2,000 now you can sort of begin to
see what what it is and then by the time you get to 16 kilometers you can tell
what it is right it’s a tropical storm or hurricane now the net the very last
one is going to be the actual picture of that storm and these other
representations photographs if you will were constructed by just successively
removing the pixels degrading the resolution of the camera if you will
until the storm was unrecognizable so a lot of lessons here you need good
observations but you also need a lot of pixels in your model a lot of pixels in
your model means a lot of calculations need bigger and bigger computers this
shows another way of looking at how realistic computer models that become
the left-hand image is from meteos at which is a European satellite showing
the evolution of cloud patterns over a day or so I guess and the on the right
so that’s truth that’s what the satellite is seeing and then on the
right is the model forecasts of these same cloud systems and if you look at
these many times you start seeing a lot of features that are the same that are
in common and so the models are doing a darn good job of representing a nature
at least out for a few days so before Hurricane sandy there of course made
other major hurricanes this is Hurricane Katrina you can see right this is a
satellite photograph this is not a model and it shows the storm moving northward
from the central Gulf of Mexico and right over New Orleans and in the
photographs from New Orleans I think many of you have seen different city
different kind of bathymetry different kind of
situation but the devastation was huge and this storm was also forecast well
three days in advance not as far ahead as sandy was but it was still forecast
pretty well in advance now I’m gonna show you it’s just work she may not work
I’ll have to click on this Hurricane Katrina Katrina this is
actually a much later generation version of the model that Tom Werner and I
worked on in them in the mid 70s it come a long way through the research of many
students in many universities so let’s look at how hurricane forecasting has
improved over the ages this is this graph goes back to a 1970 and in this
case this is the track error the forecast error by the National Hurricane
Center the official forecast body of the National Weather Service in the US these
are Atlantic hurricanes and tropical storms and on this graph the the graph
is forecast error at different times with the red curve the lowest curve
being a 24 hour forecast the next one up in 48 hour forecast and I can’t quite
see that at 72 our forecast is the yellow and then over on the right just a
short record is the what is that 120 hour forecast so what you see here in
this case lower is better smaller errors zero is at the bottom so zero would be a
perfect forecast you see the forecast getting better at all these different
times one day two days three days and four days but there’s a lot of variation
from year to year some years the forecasts are better than others but
overall the trends are are getting significantly better over the over the
years and so this is really important you need to know you need to restrict
the warning area as small as you can so that you don’t unduly over warned a lot
of people and that people get in cynical about the forecast and then refuse to
evacuate because there’s so many false alarms so this is very important to get
the track prediction correctly so this raises the question how much more
accurate can her in forecast become can they be forecast
a month in advance month is 30 days so the wrens the butterflies would say no
that’s beyond two weeks so we really can’t expect hurricanes to be forecast a
month in advance well let’s look at a very interesting not a month in advance
but five days in advance about yet another storm in a different part of the
world with a different type of graphics if you don’t want to understand this you
can just kind of look at the colors because they’re very beautiful what
you’re seeing here is a attempt to depict the flow at different levels in
the atmosphere and the the reddish colors are the high level winds and the
blue colors are the low level winds and you’ll kind of see what what happens as
we go through this but look in the in the middle of the Indian Ocean this is a
I don’t have a pointer here but the Indian subcontinent is on the upper left
the Equator is pretty much in the middle of this figure and Myanmar or Burma is
on the ISM as the landmass on the right hand side so in the middle of these two
jet streams the pink the pink colors you see a region of calm winds of blue very
light winds at all levels as we go into this animation you will see as if by
magic a storm the cyclone that hit Myanmar form in this in this region
where there is no absolutely no trace of the storm at this initial time so we’ll
see if this plays if not I’ll have to go wake it up again I think you just have to click on Italy
move the cursor on the thing okay you know the point of view kind of
moves around looking at things and so you can see the winds blowing in the
northern branch and the southern branch and right where he left the cursor it’s
kind of close to where this this storm forms do you notice there’s just no
evidence of it for the first couple of days of this this forecast so this is a
an example of the large-scale features of the atmosphere somehow forming a very
small scale but intense cyclone which wasn’t even present in the initial
conditions now you can see the swirl of the storm right and there it is my magic
has just appeared in the right place at the right time and moves in the right
direction well that’s not really supposed to be possible because there
are no observations of that storm at the initial time yet the model the equations
produce that storm and there it is full hurricane force you don’t have to
understand exactly what the colors mean but you can see a vortex there right
everybody can see that wasn’t there at the beginning so I’m going to kind of
wrap up here I’m not going to definitively tell you anything but about
about what’s really what’s really possible but let me just summarize that
there is evidence that greatly improved forecasts of tropical cyclone tracks and
intensity are predictable days in advance we’ve seen a couple examples
we’ve seen we saw hurricane sandy nine and a half days in advance to Myanmar
cyclone five days in advance so we have at least some cases in which this is
true but this is the science part of this talk
realizing this potential requires high-resolution global models I’ve
Illustrated that improved physics in the model interactive ocean-atmosphere
models improved initial conditions in atmospheric temperature water vapor and
wind satellite observations improve data assimilation techniques you have to have
really fast computers to do one all this stuff I also didn’t put on here but it
requires scientists putting all this stuff together and making it work you
can do all of these things sequentially or in parallel and unless somebody some
really smart people put it all together it’s not gonna mean anything and so Pat
yourselves on the back if you if you’re a meteorologist if you work it on
understanding of tropical cyclones if you work on meteorology if you work on
satellite observations radar observations and and you try to develop
and use numerical models it’s a hundreds of people like you largely supported by
the feds because the private sector is not going to invest money with a 50-year
payoff and no matter how important it is it’s got to be public source of funds
it’s a public good it’s saving us saving all of our lives and our children’s
lives I want to kind of close with another set a real satellite view of
hurricane sandy and just mention that a new satellite program that I’ve been
involved in called cosmyk2 is under development which we think will improve
the initial observations of all weather systems globally and will produce in
particular better hurricane forecasts and so what this will show well deliver so we’re just you can see
the the Sun go by every once awhile the the picture gets light and it gets dark
again here comes the Sun and so this is a several day period in October the week
period actually and these dots that pop up go away our individual happiness
Furyk measurements that are made by this satellite system so you get an idea of
the global coverage and the frequency of these satellites all around the storm
which would feed information into these models and cause the predictions to be
better again these are actual this is the actual seven-day satellite loop of
Hurricane sandy and you can see it as it moves up the coast of Florida and then
does what no storm had ever done before and made that left turn very late in the
in the fall almost winter time into New Jersey and again very well forecast
thousands of people had warnings huge property damage but a relatively few
lives lost due to due to the hurricane noted by the time the storm actually
made landfall it had kind of been transformed into sort of sort of a
hybrid storm with a lot of extra tropical characteristics that’s because
a lot of cold air came in from the US but notice the size of that storm the
diameter being as I say the biggest Atlantic hurricane that ever has existed
a very unusual storm it’s cold enough so that much of the
precipitation that fell in West Virginia was actually snow okay so in the big picture who wins
butterflies of the demon well there’s a difference between what is theoretically
possible that’s predictability and what can actually ever be done and that’s
predictions the demon may be theoretically possible that’s a question
for philosophers probably a bottle of wine wouldn’t hurt the argument but the
butterflies will ultimately win meteorology is ultimately going to be
unpredictable whether it’s two weeks three weeks four weeks or whatever but
there’s no reason not to help the demon a little in such beneficial activities
such as weather prediction so thank you very much I don’t think so oh there you go okay thank you very much
so you didn’t really touch on the difference between weather and climate
but there’s lots of concern about changing climates and if we can’t
predict weather in two weeks what’s the prognosis for predicting the climate in
50 years great question glad you asked that climate prediction
is not deterministic in the same way weather prediction is but you can
predict climate very well without even the convenor computer model you can
predict that July 5th is going to be hotter and more humid in Ames than April
5th right and and you know guess what that’s kind of trivial but the reason
you can do that and you’re predicting the average July 5th there will be some
July 5th it might actually be cooler than on April 5th but on the average you
can predict very well that climate you’re in advance and that’s because
climate is forced the force system it’s not dependent on the initial conditions
the initial observations it depends on the angle of the Sun primarily and
that’s an orbital characteristic of the earth going around the Sun and then we
know that in July the northern hemisphere is facing more directly
toward the Sun than in April and certainly January so that that forcing
that radiation forcing from the Sun determines the climate is the biggest
factor that determines our climate there are other factors too but that’s a
climate is a force system and you’re not predicting the weather on July 5th
whether it’s gonna rain or not you’re predicting that on the average July
fizzle is going to be hotter and more humid than April 5th and you will be
right 95% of the time 99% of the time if you say June 1st versus May 25th
you’ll be right more than often than not but it’s so close the forcing is so
close you’ll won’t have as much skill so that’s what climate models are doing
they’re forecasting the radiation balance of earth and as the radiation
balance Changez we can say with some certainty I
think that with more greenhouse gases we are going to have a warmer climate just
that’s pretty sure and observations showed that sea levels rising oceans are
getting warmer ice is melting all over there it’s another talk
so my it’s one of my signature I talks but yeah it’s a great question and I
hear it all the time we can’t predict the weather beyond two weeks how are you
ever going to predict I’m at 50 years it’s a different system a different kind
of prediction we’re predicting the statistics of weather 50 years from now
not the weather itself so what would it take for the European
senator to go from nine and a half day forecast to a ten and a half day
forecast with skill and secondly what will it take the u.s. to catch up with
ECMWF well there’s been a lot of discussion of the second question is
actually probably easier to answer than the first European Center for one thing
has a has a much faster computer and they can do more runs at a higher
resolution and they can have better ways of assimilating the data that’s one big
reason it’s not the only one they’re also focused on one mission and that’s
better global weather predictions they don’t have a lot of other missions like
the weather service which has to deal with tornadoes and short-term
forecasting and so forth there’s also I’m sorry to say a cultural difference
the Europeans a European center from the very beginning has been extremely open
to new ideas and they have the best minds from all over Europe coming there
for short periods of time three to five years focused on this one goal of being
the best model in the world we have a lot of people the smart people in the US
but they don’t go to the Weather Service they don’t spend three to five years
focused on that mission there’s not nearly as much fresh blood moving
through the system it’s kind of a cultural thing so there’s those are
three reasons oh now what’s it going to take I hate to say this but it’s more of
the same but to go from nine and a half days to ten and a half days faster
computers better observations better use of observations smart people the same
thing that’s carried them from 1981 to 2011 going from you know going from a
forecast now at it’s seven days it’s just as good as the forecast was in a
two days so that same kind of thing is proven it’s hard work it’s grunt work
but we know how to do it and unless there’s some wildly out-of-the-box thing
that nobody’s thought of it’s going to be more of the same kind of boring but
it’s a good model proven yes sir there was a claim made I think on Clifford
masses blog that the United States is spending more computational resources on
climate change research as opposed to operational weather forecasting is this
true if it is why and does that make any difference well it’s true and I think
it’s it’s terrible that The Weather Service doesn’t have more computing
power I think the climate problem is very
important and it’s but we should have a weather service that has access to much
much faster computers than it does have and that’s an immediate important
importance immediate importance so I agree with cliff he’s got a good point
there I think that and we’re talking about like to really upgrade the
National Weather Service something like a hundred million dollars which sounds
like a lot but remember sandy was seventy four billion so a hundred
million over five years that’s not that much when you look at the live saved by
you by these computers so I think Weather Service should get a hundred
million dollars for a new computer that would tie put them up where the European
Center is and they would be making much better
excuse me much better forecast I’ve heard it said at mathematical meetings
on dynamical systems that in certain sets of conditions weather prediction is
possible for much longer times than under other conditions you didn’t
mention anything about that in your talk is that true and to what extent it’s
true now some cases the butterflies don’t
make much impact for it takes a longer time for the butterflies to make an
impact some cases are more predictable than others that’s that’s just basically
a fact and some in mathematically that means the
addictions are less sensitive to unknowns in the initial conditions than
in other conditions and what it happens if you’re in a situation where there are
very strong instabilities in the atmosphere you have to measure that
region very carefully otherwise the instability goes off in one direction
and so it’s related to instabilities in the atmosphere the atmosphere kind of
quiescent not large qvv instability then the forecast can be better predicted so
there’s actually a science of predicting the predictability just from the initial
conditions looking at it and saying this is a more predictable case than another
sometimes even TV weather forecasts get that that you know this is a really
tough thing to forecast so don’t rely too much on my forecast tonight
and that’s based in science yes sir I’m gonna use the mic I know
that you said that hurricanes aren’t really gonna be any more predictable
then then they probably are now or just a little bit better but what about the
number of possible storms or depressions that become hurricanes does that ever do
you think that I would ever be possible to predict just the number of them per
season yeah I said well I didn’t mean to leave the impression that hurricane
forecasting is not going to improve in fact I think it is going to improve
especially the intensity forecasts in the next five years I see a big room for
improvement there in terms of how many in the next season I think that’s a
worthy goal that gets into climate prediction using such prediction
indicators as El Nino La Nina various kinds of oscillations and that’s a
tougher problem I think just a little bit of skill in predicting the number of
storms now say in April or May for that season but not a lot I’m fairly
optimistic I think that there will be some improvement how far that’ll go thank you