Cookies   I display ads to cover the expenses. See the privacy policy for more information. You can keep or reject the ads.

Video thumbnail
Hi everybody. Today I'm visiting the University of Oxford and I'm here to
talk to Professor Tim Palmer about climate models. Tim, on the weekend I read
this article in the New York Times. It was titled "How scientists got climate
change so wrong" and it was mostly about weather extremes and said that climate
change has been much more abrupt than climate scientists expected. And I was
wondering if that's really the whole story. Because I kind of remember that
people were talking about tipping points and unstable equilibria already in the
80s. So, I was a little bit surprised about this and I thought maybe I should
I should ask you what do you think about this.
Well, the first thing to say is it's
kind of interesting that now for decades having been attacked by the climate
skeptic stroke denier community that the models are somehow useless, now the
attack is coming from the other side saying the models are somehow too
conservative and not telling us enough about the extremes. Now I guess there's a
couple of points to make here. The first thing is from a scientific point of view
most of the focus of the model predictions has been on global mean
temperature and the reason for that is sort of you know because that's you know
that's that's the basic thing that increased carbon dioxide is doing to the
to the atmosphere, it's increasing the surface temperature and by measuring the
the global average of this what you're actually doing is is is measuring, or
predicting a quantity where the signal-to-noise is maximal, the
signal being the effect of the carbon dioxide forcing that we're putting into
the atmosphere and the noise being the internal variability of climate, the
fluctuations that really have nothing to do with increasing levels of co2 but
just arise from the the natural chaotic variability of the
atmosphere. So when we go on to global scales this the chaotic variability is
actually at a minimum and the impact of the co2 forcing is at a maximum so from
a from a science point of view that actually is a very kind of robust
indicator of how carbon dioxide is changing the climate, it's warming the
global temperature. And actually I would say that the models from the 20th
century through to the present day have been remarkably accurate in predicting
the rise in global temperature. So, from that point of view I don't I don't
think the models have been you know under estimating the effect of carbon
dioxide on global warming. However, when we come to talk about more regional
extremes so things like particular heat waves - could be over Europe or United
States - or you know flooding events or intense hurricanes or tropical cyclones
or indeed as you say kind of tipping point types of phenomena, then you're
dealing with a situation where the internal variability of the atmosphere
is much greater and the signal therefore relatively speaking is smaller, so it
becomes the more difficult statistical exercise. But on the other hand this is
exactly what people want to know. I mean nobody physically is affected by global
mean temperature but they are affected by extremes of weather.
And I think basically what this is the article is is correctly pointing to is
the need you know now we've established beyond doubt I would say that humankind
is warming the planet we need to think in much more detail and much more your
much more accuracy what this implies for regional extremes of temporal extremes
of weather and climate so there are two issues here one is
you know one is is developing this sort of statistical techniques where we can
be confident in saying that such and such a weather event or climatic event
had a had an anthropogenic if had an anthropogenic component in other words
part of it was due to the fact that we are increasing co2 levels but it also
puts for very much an onus on climate models to be able to simulate these
extremes well and actually that's an area where I think we can still improve
things considerably so I think that the article I think it got a it it sort of
exaggerated some aspects of the issue particularly in relation to global mean
temperature but it correctly drew attention to the fact that we do need to
focus much more on our ability to simulate and predict and assess how
extremes of weather and climate are being affected by climate change forcing
so speaking about the quality of the predictions you told me something about
this figure 9.8 in the IPCC report and this took me forever to understand but I
will I will try to summarize what's in the figure and had you told me if that's
if that's about correct so what you see in the figure is the temperature anomaly
for a period of years and so the temperature anomaly is the it's the
global temperature basically up to a reference value and this reference value
on this figure is is in the yellow region which is from the years 61 to
1990 and so all the thin squiggly lines are the predictions from the different
models and the red line is the average from the models and the three black
lines are the data from different organizations and so what I what I
didn't understand for mr. Craig it's not the data from different organizations is
how different organizations analyze the data to the
the common data sets to produce estimates of global temperature yeah so
so what what I didn't understand forever was what's this little bar on the right
side where it says mean temperature so that's the actual temperature the
absolute temperature that these models predict in this region from 61 to 1990
so basically tells you that the spread in the in the absolute temperatures is
much larger than the uncertainty you know the little squiggles in these
models in in the whole region where they have data right so I think what this
figure tells us is that the models all agree that there is a certain trend you
know looks pretty good you know in terms of forecasting but it also tells me that
you know the models have some difficulty getting they are full of temperatures
right yeah it just highlights the fact I think I think perhaps it might be worth
backing off a bit here and saying that these models are attempt to represent
the climate of the earth from pretty much first principles you know from the
basic laws of physics so from Newton's laws of motion as expressed in you know
the what are called the navier-stokes equations of fluid mechanics to
equations which basically represent the laws of thermodynamics in a in a sort of
in terms of differential equations coupled together with laws which express
more quantum mechanical laws which express how photons from the Sun are
absorbed by different molecules and so on in the atmosphere and re-radiated
back to space within the infrared so these are all very basic sort of
equations they're not you know it's not that we're just kind of guessing
empirically how we think the the world works you
know by just sort of drawing equations out of a hat and putting them into a
computer these are the basic laws of physics now if you look at it from that
point of view trying to get exactly the right you know good to simulate exactly
the right surface temperature of the earth which you have to remember you
know over the oceans the surface temperature is is a very sort of
complicated balance between you know regions of weather whether the ocean is
ocean water is sinking and other areas where it's rising to the surface and
regions where the Sun warms the surface other regions where you know you're
under cloud and there's very little Sun so getting the surface temperature not
only requires for example getting the dynamics of the ocean right it requires
getting all that cloud cover right in the right place so it's a really really
really complicated and difficult thing to get right that's the first thing to
say and the little bar on the right hand side is just pointing out that actually
you know it's a kind of manifestation of that problem because the range of
estimates of global mean temperature from the models actually range over a
few degrees which is which is much larger than this trend in temperature
that we've seen over the last you know 6070 years or so now I don't think this
particularly undermines there where it doesn't undermine the projections of
global temperature it doesn't it doesn't I don't think it casts any doubt that
the trends in in temperature that we've seen over the the last 70 years are
indeed directly associated with human emission of carbon dioxide but what it
indicates is that you know we still have some way to go before we can say we have
simulated the climate system to the extent that you really can't tell now if
I you know if I show you output from a climate
model you can't tell whether you're looking at a model or the real world we
still have some way to go to do that and that's particularly becomes particularly
important and more at the regional level
for example you know we talked a little bit earlier about tipping points and
these are these are kind of what you might call very nonlinear transitions
sudden transitions in the in the climate system getting these right actually does
depend on getting the actual absolute values of temperature right so for
example if you take the melting of you know I mean there's a concern that the
melting of Greenland ice and actually the sort of disintegration of the
Greenland ice core caused by you know the lubrication of the surface the
bedrock from from melting water I mean that requires models to get that
absolute that the temperature right because water fresh water at these
freezes at 0 degrees so if you have a two degree bias or something I mean that
you're going to get that process wrong another example is is possible tipping
points for the bias here where you know either due to heat or or combination of
heat and and an availability of moisture of rainfall you know a forest can
suddenly become no longer self-sustaining and will collapse but
again that to be able to model that requires getting these temperatures and
rainfall amounts not right in a kind of anomalous sense but getting right
absolutely and all that I think that what that bar
on the right hand side I don't think it should make us doubt at all that the
temperature is warming due to co2 but what it what it's indicative of is the
fact that we you know particularly now as climate really starts to become an
important societal issue we've got to step up a gear in getting our models
bias-free yeah so this bar on the right side this was one of the things that our
dinner around and about this figure the other thing is
that I find it peculiar that you have the prediction from these models but the
predictions don't have any uncertainty attached to them which is what what I
would expect would be the output of such a model so my understanding is that the
the figure that for the projection of the increase in temperature until the
year 2100 or something in the IPCC report it has an uncertainty and that's
basically the spread in the projections from the different models not actually
the uncertainty from the models okay I mean so the first thing to say is that
the whole philosophy underlying the IPCC report is that it's it's an assessment
of the you know of the state of the art of climate science as as determined by
the peer-reviewed publications that exist at the time the report is written
now there are many climate Institute's around the world you know typically
certainly virtually all of the bigger countries of the world have their own
weather or climate Institute's and and they they have their own climate models
they might be they might be literally their own model produced by scientists
in-house or they might have taken the code from from another institution and
maybe done some modifications and you know produced produced results for that
model now many Institute's do not have sufficient computing resources to
actually run the model itself in a kind of ensemble mode where they might
produce you know 50 projections where you try to buy a number of different
possible ways of representing uncertainty in the particularly the so
called sub grid parameterization so that's the most unser
and part of climate models you have to represent processes cloud processes
perhaps the most important uh which are occurring on scales where where the
model can't resolve the grids of spacing between the grid points in the model is
larger than that you know typical size of the cloud I mean many climate models
have grid spacings of many tens that you know you may be up to 100 kilometers or
as individual clouds they're just you know a few kilometres and big ones
induce comatose in the horizontal horizontal sorry in the horizontal yes
it would be less in the vertical so many climate Institute's don't have the
computational resources to to try to explore the uncertainty in the you know
in the very in the in the sub grid parameterizations
so they would they would typically just have one one run or one or two let's say
other other Institute's may have may have multiple ensemble integrations
where they do I mean the Met Office the UK Met Office is a good example where
they produce very large ensembles of climate change integrations where they
try to exactly do what you say try to perturb the uncertain parts of the of
the models maybe using some kind of stochastic process and then run these
but for these IPCC assessments you know to avoid being dominated by you know if
one Institute had a hundred runs and the others only had one you know you'd be
dominated by the center which had a hundred so I suppose you know a way of a
way of dealing that with that is just to make the assumption that the ensemble of
all of these different models is itself a reasonable representation of model
uncertainty now you can argue whether that's true or not and I would argue
certainly on the again coming down to the regional scale that's probably not a
good assumption but I think for these global mean temperatures it's not a bad
so what do you think are the main reasons that the predictions from the
different models diverge well there'll be a little bit of divergence from chaos
if you like that you know if you just started them from infinitesimally
different initial start initial starting conditions the butterfly effect will
actually produce a certain amount of spread but that's probably not the that
isn't really the major contribution to uncertainty it comes from the
uncertainty in in how to represent processes which you know are important
for climate but where you don't have the computational resources to resolve them
so you have to parameterize to use the bit of the jug and parameterize these
sub grid processes with very simple well it's relatively simple anyway formula
which you know which and then so you have a closure formula so there is a
basic assumption somewhere where you would say okay if I know the temperature
in a grid box and I know the humidity in the grid box and you know maybe some
other variables the wind and so on I can predict in a bulk sense what the cloud
the amount of cloud in that grid box you know whether it'll be completely cloud
free or completely cloud covered or you know 50/50 or something like that half
covered and half creek so so there'd be a formula which would be based on these
these resolved scale variables now you know in reality there isn't such a
formula you know it's not like nature you can't go up and you know read of a
textbook on physics and discover what that formula is because there is no
formula like that so different groups may come up with different formulae
different closure schemes
for you know for various reasons they may have some datasets which other
groups don't have which may be supports their formula or whatever it is I mean
my own view is that the only way to deal with this objectively is to express all
of these sub closure schemes in a stochastic way using kind of some ideas
of random variables and just acknowledge that that's actually from from a basic
physics point of view that is the best way to represent uncertain processes but
but in any case the origin of the uncertainty is this sub grid the sub
grid parameterizations and you know of them of all of them the most important
our cloud processes but there's also other things to do with you know how you
represent topography the mountains of the earth you know if you have a very
sharp flow that's blocking so if you have a very sharp barrier that's
blocking some flow the the the width of the barrier might actually be too small
to represent with your grid if your grid you're with your finite grid so you have
to try to represent that blocking effect in a in a more approximate way I mean
that's just one example but and the oceans you know the oceans have what are
called a mesoscale Eddie's which are really important for determining how
strong currents like the Gulf Stream or the Kuroshio Current in the Pacific are
but again you know modern-day climate models the ocean the ocean part of these
climate models is the resolution is I mean we're getting starting to get
getting close to being able to resolve these types of ocean eddies but we're
not really there yet so they have to be parameterized and again that's the
source of uncertainty so I guess this brings up the obvious question what can
be done about it well you see the the the issue you know the the interesting
thing from my point of view is that climate has
to have gone in the last few years from something that it's always been
potentially of societal concern but I think a lot of a lot of scientists felt
that although you know the the societal concerns were important in a way that
the the the tools that they had were if you like primarily being used for
scientific research to really understand
you know the way in which for example co2 in house co2 emissions would impact
on different parts of the climate system it was it was a yeah I mean you know
it's a scientific endeavor but what's literally happened you know in the last
year or two is that it suddenly become this incredibly pressing societal issue
you know we're seeing all around the world these quite devastating weather
events which are you know affecting people's lives and society has got to
know what can they expect in the future and how can they better prepare for the
future what sorts of you know buildings do we need to withstand these extremes
where should we be living to withstand these extremes you know can the can the
human body actually literally exist in parts of the world where temperatures
and humidity --zz become once they reach a certain level so it's kind of gone
from a you know a society important that fundamentally scientific question - one
that really is societally crucial and I think therefore as a result of that
we've got to think much more in a much more pressing way about making these
models fully realistic and accurate and really trying to eliminate where
possible these these parameters which just are just to approximate and
we know in a way the the the bottom line is the resolution of the model in other
words the spacing between the grid points that's what you know that's
having having these grid spacings of many tens or hundreds of kilometres
means that many of these important processes key types of cloud processes
ocean meso-scale eddies, the flow over orography, topography, whatever you want to call it,
have to be parametrized. We know if we can get the grid down to, say,
about one kilometer globally then we can eliminate certain - not all
parameterizations - but probably the most important ones. And I feel, given this new
sort of urgency to try to be able to answer questions which governments
around the world and individuals around the world are asking about and that and
the New York Times article drew attention to about extremes how you know
how I mean we're you know this week for example in the UK there was considerable
flooding in in an area near Doncaster hundreds of people had to leave their
houses I mean that's the kind of key question a government wants to know how
much more frequent will that type of situation occur in the future the you
know poor people that suffered in Mozambique under these tropical, enormously
powerful tropical cyclones. Again we want to know how much more frequent are these
unbelievably intense tropical cyclones going to be. So, we've got to develop
models where these biases and so on you know are eliminated. And we can do that
in principle, but like all scientific big projects it requires a certain amount of
of investment. And it's primarily investment in supercomputing it's to do
with supercomputing. So you know I would certainly agree with
that that this is information that we need so I find it a little bit ironic
that I keep hearing that the science is basically done so I I have another quote
which I found in The Guardian in an article that appeared last week. It says
"For ordinary citizens it is important to recognize that scientists have done their
job. It is up to us to force our leaders to act upon what we know." Ok, I mean that
particular quote was I think referring to the issue of trying to cut our
emissions of carbon dioxide I would agree
I think more or less with that quote if all that we were talking about was do we
have enough evidence to make a decision about cutting our carbon emissions
because in a way you know it's like every decision you make in life you
don't necessarily have to know exactly what's going to happen to make that
decision you have to you have to know the threat that you're facing and
whether the decision is justified given that threat now the one thing that you
know climate models have been quite unequivocal about is actually that as we
if we if we continue to emit co2 as we have as we are doing now and as we have
done then by 2100 although we can say that you know the most likely amount of
global warming might be you know might be systems maybe three or four degrees
we know from these ensembles of integrations that there is a tail which
goes out to more than that could go out to six or seven or or even more than
that degrees now again since then I don't know that sounds a lot but for
anyone who knows their climatology that really is captain
strophic so you know I mean I as a scientist don't want to be I don't want
to say that you know that means that we must cut our emissions immediately
because that's a political statement but I think the politicians in principle
have that have enough information to make that decision so that the fact
there is this threat this risk not only of you know very undesirable levels of
climate change but actually you know catastrophic that was the climate change
the risk is quite clear and the only way to reduce that risk is to reduce our
emissions so from that point of view I agree with the statement but to say that
the science is is kind of all done and dusted and the scientists are not needed
anymore misses the you know the other aspect of the problem which is that you
know even if we cut our emissions to zero tomorrow we've already put into the
climate system a certain amount of climate change that will continue. Now
we're not going to cut our emissions tomorrow, we're going to carry on for
sure for many decades to come, and so we are faced with a change in
climate, and we are faced with decisions on how to make society more resilient to
that change in climate. And I think nowhere is that more important than in
the developing world who after all have had absolutely nothing to do with this
problem, I mean they have not caused it in the most minute way, but in a way
they're suffering they're likely to suffer the most
either from extreme levels of drought to these really occasional but
exceptionally damaging storms or to levels of you know as I was saying to you earlier,
to periods where temperatures and humidity could get so high collectively
that the human body can no longer lose heat either by sweating or any other
means so you know then that becomes an existential threat so that's sort of
what we've got to be do better I think in quantifying and and that that puts
very much the onus of climate change at the regional level not just
the global mean temperature you know my own view is that this is a little bit
like you know the the the famous Marshall Plan for bailing out Europe
after the Second World War where the u.s. pumped you know large amounts of
money into stimulating the European climate or the European economy not
particularly for an altruistic reason but because they feared the spread of
communism and they wanted to stop that now you could very much view that the
the whole investment in climate adaptation in the developing world could
similarly be viewed at a very you know self-interested level in the sense that
we're already seeing you know migration you know in Europe from Africa and the
Middle East in the United States from Central America and South America and
there are certainly aspects of climate change in the reasons why people are
migrating now this is potentially nothing compared to what it could be
like you know in in later in this century and so I think a kind of
modern-day Marshall Plan by the developed world to try to make life just
more bearable in the developing world would would you know like like the
Marshall Plan to stop communist and this would be to try to keep people in place
and say actually you know living where you are is not so bad
but if it becomes unbearable then then the trickle of migration that we're
seeing now will become a torrent and so that's that's where again I think the
climate science is not done and dusted because we don't yet have a good
and I would say reliable picture of how these extremes of climate at the
regional level are changing and whether for example these tipping points as
they're called colloquially but these kind of sudden
rapid changes in climate which cannot be reversed, I mean that's the key point
about tipping point you can't reverse it, once it's flipped into this new state
it's irreversible. This by the way has this this this is actually where
this actually has a kind of a feedback into this question of emissions
reduction because there's certainly a body of political thought which says
well cutting our carbon emissions today is really really difficult for various
political reasons but we don't have to worry because in 50 years time we'll
have developed the technology which will enable us to suck the co2 out you know
will will suck it out of the atmosphere and dump it underground and so we know
we don't have to be too kind of aggressive in our emissions cuts today
because that will be a technology that will be there in the future which will
help us. Now the point about that is if in that period before before the
technology has been developed if it ever can be developed which is a major
question mark I would say if we have undergone these tipping points for
example in Greenland ice or some of the biosphere or indeed at oceans some ocean
dynamics might potentially have that capability then you've gone to a stake
that you can't reverse. So sucking the co2 out of the air once the tipping points
happened won't do any good at all you're not going to recover back to where you
were so again that's an area actually where I'm slightly contradicting myself
because I was saying that we perhaps need all we need we have as much
knowledge as we need to put into place emissions cuts but I think the question
about whether we can delay emissions in the hope that sort of the sucking it out
of the air at some future stage will occur that's going to be totally
ineffective if we actually have crossed some of these tipping point things and
that's much that requires knowledge of the climate system at that much more
regional level and also at a much more detailed level because
invoice is never to involve quite nonlinear processes which are which are
hard to you know simulate accurately and and that's where you need good models I
think that's a good place to stop thank you for your time thanks everybody
for watching see you again next week