Its Not The Heat Its The Tepidity 20150310

Very preliminary draft, please do not quote
ITS NOT THE HEAT, ITS THE TEPIDITY

CLIFFORD ASNESS AND AARON BROWN
VERY PRELIMINARY DRAFT: MARCH 10, 2015
COMMENTS VERY WELCOME
There is a graph youve probably seen a lot of lately. Articles everywhere feature iti, or the data
behind it, or a closely related cousin to either, highlighting temperatures at a new high, a visually
strong long-term trend, and no recent temperature hiatus. This historical series is usually
presented as proof that warming must be quickly and vigorously fought or we will reach
dangerously high temperatures in the next few decades. These data and the implications taken from
them require careful statistical scrutiny. We emphasize statistical at the outset because we are not
challenging climate science; we are not climate scientists. We are, however, pretty experienced at
statistics and inference.
The now famous graph shows global land-ocean annual mean surface temperatures since 1880 (the
chart plots deviations from the 1951 1980 average)ii:
Global temp index (degrees C)
0.8
0.6
0.4
0.2
(0.2)
(0.4)
(0.6)
1880
1900
1920
1940
1960
1980
2000
While visually impressive it's our contention that most wed say all but we havent seen them all
of the recent articles forget to look at the vertical axis and to ask the simple question, are these
numbers big?
This essay is not about the science of climate change, its about what the data say on their own. In
particular, we think it is important to distinguish the level of worry you might have from looking at
this chart, versus the level of worry you might have from complex climate models.
With a chart like this, whether about climate or anything else, a natural thing is to draw a straight
line through the data (we used a linear least square regression, but you get about the same thing
with other statistical techniques or just doing it by eye). This is a guess at what the long-term trend
is, and could be a base case for thinking about how hot things are likely to get in the future
(extrapolating a trend must be done with skepticism but it seems to be a reasonable starting point
for discussion).
Its not the heat its the tepidity 20150310
Page 1
0.8
0.6
0.4
0.2
(0.2)
(0.4)
(0.6)
1880
1900
1920
1940
1960
1980
2000
It shows a warming trend of 0.67 Celsius per century. That doesnt sound too scary, and as well
see that if that remains the trend, according to the experts, its indeed not too scary. Its surprising
that in the myriad of recent articles highlighting this graph, pointing to the trend, the lack of a
hiatus in warming, and the recent highs, none weve seen mention (we apologize to any weve
missed!) that the actual magnitude of 0.67C per century is quite small and itself doesnt take us to a
dangerous level until much farther in the future than climate scientists forecast.
Of course the real but rather small trend doesnt prove that global warming is a minor issue, far
from it. Were just saying the graph taken on its own is actually pretty reassuring, at least compared
to predictions, and declared danger points, of the IPCC and similar groups. If things continue along
the way they have for the last 135 years, the point at which we reach dangerous temperatures is a
very very long time from now. Those predicting that we face a big problem much sooner arent
arguing this from these data, instead they have to be arguing that historical warming trends will
change drastically in the near future; that they will not continue at the trend of the past hundred
years or so. The historical record to date, and in particular this ubiquitous graph, cant be the basis
of an argument that we will hit dangerous levels soon. To argue that we will hit them in this, or even
next century requires us to explain away this graph, to explain why the rate of warming will
increase.
To see this, lets take a step back and use larger axes. Were going to drag the timeline (the
horizontal axis) out another 50 years to 2065, and extend our trend line, and enlarge the scale (the
vertical axis) to go up to the 4C of warming that has been used as a somewhat arbitrary danger
level since the 1970s (we call that the IPCC limit below later, and in our end notes, we discuss
what happens if this danger level is reduced to 2C, a level we also show in the following graphs).
4.0
3.5
IPCC Limit
3.0
2.5
2.0
1.5
1.0
0.5
(0.5)
1880
1900
1920
1940
1960
1980
2000
2020
2040
2060
Page 2

It says yes, over the last 135 years the Earth has warmed, but not nearly to the danger point (the
top of the graph at 4.0 on the y-axis) and if we continue at this pace (the crux of the issue) it wont
become scary until more than 500 years from now (not by the end of the 50 years weve added to
the x-axis). Thats quite a different message from what weve read in the articles accompanying the
original version of this chart.
Of course, this raises the very important issue of whether or not 4C is the right danger line. No one
knows the answer to this question, but 4C seems the most common figure used by the experts. Its
what the IPCCiii uses in its most recent report. No one denies that there are some risks and costs to
any amount of warming, and on the other hand, almost no one is predicting that warming at or
slightly beyond 4C will cause extinction of the human race either. Risks go up with the amount of
warming. We just dont know how fast. Despite the uncertainties, there seems to be a scientific
consensus that less than 1C or 2C of warming would make global warming no more serious than
several other environmental issues, but warming above 4C would likely make global warming a
unique danger. Even if the danger point is 2C this trend doesnt reach it until over 130 years from
now.
Next lets add to the graph the IPCC 1990 business as usual (the IPCCs term for what happens if
no efforts are made to restrain CO2 emissions) predictions for future warming. These are shown in
the next graph as the three green linesiv, which show the expected (middle) case along with the
good (lowest) and bad (highest) cases. The black line is still the linear trend implied by history and
the blue line is still the actual temperature data. We start the green lines in 1990 when the forecasts
were made.
4.0
3.5
IPCC Limit
3.0
2.5
2.0
1.5
1.0
0.5
(0.5)
1880
1900
1920
1940
1960
1980
2000
2020
2040
2060
The rate of warming since these forecasts were made, albeit over a period too short to judge such
things, so far appears to be less than even the lowest IPCC estimate and substantially less than their
higher forecasts. Someone who predicted in 1990 that global temperatures would remain constant
(i.e. no warming) through 2014 would have been closer to the truth than someone who bet on
average warming rate from the IPCC middle prediction (if you take their lower rate and consider
the natural variation the IPCC represents in the wiggles not shown in our graph around that
rate its not far off, but then again its also not much warming). The IPCC gets points for predicting
correctly in 1990 than the subsequent 25 years would warm faster than the long-term trend up to
that date, but even their lowest scenario showed about double the increase that has actually
occurred to date.
Of course, just looking at a chart is not climate science. Rather, its climate-knowledge-free
statistics. Perhaps temperatures will snap back to model predictions as feedback loops kick in? Or
Page 3

perhaps lower amounts of warming than 4C will have catastrophic effects? Again, thats for experts
to investigate, all were saying is that if you look at the chart it suggests global warming is a much
smaller and less immediate threat than the IPCC thought in 1990. And, this is the part we find so
confusing, remember, this is the chart trumpeted everywhere, implicitly and often explicitly, as
verifying that global warming is a huge proximate danger.
The most recent IPCC reportv is far more complicated than the 1990 one. We picked its A1FI
scenario as the best comparison to the 1990 estimatesvi. Here is the graph with those. Below are
the same three green lines, with the same interpretations, with the green lines (IPCC forecasts) now
starting from the end of 2014.
4.0
IPCC Limit
3.5
3.0
2.5
2.0
1.5
1.0
0.5
(0.5)
1880
1900
1920
1940
1960
1980
2000
2020
2040
2060
Again, the IPCC projections could certainly turn out to be correct, but theres little sign of that in the
data (admittedly, again, data analysis, not climate science, is what we do). They are predicting a
sharp increase in the rate of warming that has not yet been observed. Other evidence for warming,
such as decline in Arctic sea ice from 1980 to 2012 or sea level increases since the early 1900s, if
measured accurately, could easily be consequences of the moderate degree of actual warming we
have already observed, not evidence that the rate of warming is about to make a dramatic increase.
While we aren't climate scientists we are reasonably confident that ice melts from actual warming
not future forecasted warming, whether forecasted accurately or not. So things like ice melt must be
evidence of the warming that has occurred, i.e. the rather modest warming we and others graph,
not evidence for or against a model forecasting the future. It is certainly not new evidence
independent of the 0.67 Celsius per century warming trend, but rather the same evidence
repeated. Other consequences of warming, like ice melt, can be reported as confirmation of
warming, perhaps to convince those who doubt the direct temperature data, but not as evidence
that the problem is bigger than the amount of warming weve seen.
In other words, when you see a picture of a polar bear stranded on a tiny piece of rapidly melting
ice, it may indeed be a consequence of the 0.67 per century warming weve seen, and it may indeed
be a serious problem, but it cant be the consequence of, or the proof of, a model for the future, as
forecasts of the future have, as of now, not harmed any polar bears. On the other hand, to be fair, if
we already see consequences of 0.67 per century of warming it seems intuitive that if the models
forecast of much faster warming are correct, the consequences would be severe.
If you want to use the chart and argue for a faster rate of warming, one solution is to fit a curve to
the data instead of a straight line. We accomplish this, again with statistical brute force not climate
science, by adding a square term to the regression and show the graph below.
Page 4
4.0
IPCC Limit
3.5
3.0
2.5
2.0
1.5
1.0
0.5
(0.5)
1880
1900
1920
1940
1960
1980
2000
2020
2040
2060
Extrapolating curves is notoriously unreliable even less reliable than extrapolating straight lines!
But it does seem to fit the historical data reasonably well. The squared term empirically improves
the fit and indeed leads to forecasts of accelerating warming. It predicts roughly twice the amount
of warming by 2065 than the linear extrapolation does, and hits the IPCC limit in 132 years versus
530 for the straight line. While this could be cause for concern, its still obviously much less dire
than the IPCCs models (slower warming over a longer period equals less dire).vii Even looking the
2 amber line the accelerating curve doesnt hit it for 66 years vs. the IPCC middle (expected) case
striking it in 44 years. In fact the accelerating squared fit doesnt catch the IPCC middle (expected)
case until 200 years from now!
What is our conclusion from all of this? If you believe in a significant probability of catastrophic
global warming in the next 50 years, you cannot base this belief on the last 135 years of global
mean temperature data compiled by NASA. But this is not the text usually accompanying this graph.
If you work from just the historical temperature data, even aggressively fitting the data, global
warming appears to be a manageable problem at least until very late in this century, and if 4C is
the true danger point, much later than that. It follows that we cant base our confidence in global
disaster on any consequences of the fact that things are warmer now than they were a century ago,
such as sea level rise or the string of warmest year on record announcements, as those are part
and parcel of the modest warming weve seen.viii
That leaves three ways you might support the argument for impending catastrophe. One is, again, to
rely on climate models and other technical arguments, despite, not because of the data.ix The second
possibility is a non-empirical belief that human impacts are inherently unnatural and threaten a
delicate environmental equilibrium.x The third possibility is that our examining when well hit 4C
understates the danger. Specifically, while 4C might be the approximate danger point, if you wait
until youre near it to take action, you may pass the point where you can prevent going through it.xi
These can all be reasonable positions (although not all people who hold them argue them
reasonably) but none of them can call on the famous chart we study for support an action that,
again, has become perplexingly common.
As we said at the beginning, this is not an essay about the intricacies of climate change, neither the
science nor the policy implications. Its an essay about data analysis and how data gets used and
sometimes abused. Popular coverage of the issue seems deeply confused as journalists, and many
advocates, use the graph we showed or similar data, or straightforward consequences of the
mild warming in the historical data as support for models that predict dramatically different
behavior in the near future (and even in the recent past). You can believe the models if you like, or
Page 5

you can look at the data and assume the most likely future is an extrapolation of the past. What you
cannot do is both.xii
Finally, the data do say a lot of things (we leave any quarrels about the data itself to others). They
say the world has indeed warmed for more than a century. They say there has not been a recent
hiatus as some critics contend. While not statistically distinct they indeed say the most recent
measurements are at a record high. They even suggests that warming has accelerated over time.
However, what we don't find in the data is evidence that warming is dangerously fast, that we are
anywhere near the generally agreed danger point or will be anytime soon, nor that the actual
scenarios forecasted by the models seem to be playing out in the near term. Yet, as proof of these
last three is precisely how most commentators use the data and the famous graph we focus on.
Put simply the history of global temperature, and specifically the recently famous graph, shows
the opposite of what almost all articles presenting it argue. If the next 135 years repeat the
experience since 1880, we are in good shape. How much we should worry about future warming,
and the answer may indeed be a lot, is entirely about how much one believes in the models
forecasts that the future will look very different from the data.
Here is a sampling of recent articles from technical publications, newspaper articles, posts on popular blog,
educational materials and opinion pieces (some use only the data not the graph, or either but use very close
Anup Shah, Anup, Climate Change and Global Warming Introduction.
http://www.globalissues.org/print/article/233.
Brekhus, Keith. Attention Climate Change Deniers: NASA, NOAA Report 2014 Was Warmest Year
Recorded. http://www.politicususa.com/2015/01/16/attention-climate-change-deniers-nasa-noaareport-2014-warmest-year-recorded.html.
DeLong, Brad. Department of "WTF?!": Yes, Clifford Asness's Views on Global Warming Are Insane:
Why Do You Ask? http://www.bradford-delong.com/2014/10/yes-clifford-asnesss-views-on-globalwarming-are-insane-why-do-you-ask.html
Hansen, J.E., R. Ruedy, M. Sato, and K. Lo, NASA GISS Surface Temperature (GISTEMP) Analysis, DOI:
10.3334/CDIAC/cli.001. http://cdiac.ornl.gov/trends/temp/hansen/.
Kinver, Mark. 2014 warmest year on record, say US researchers.
http://www.bbc.com/news/science-environment-30852588
Romm, Joe. 2014 Was The Hottest Year On Record Globally By Far.
http://thinkprogress.org/climate/2015/01/05/3607735/2014-hottest-year-by-far/
Schulman, Jeremy. Was 2014 Really the Warmest Year? Here's Why It Doesn't Matter. Focus on the
trend.
http://www.motherjones.com/environment/2015/02/was-2014-really-warmest-year-nasa-noaa
Global Temperatures: Global Mean Temperatures as an Indicator of Global Climate Change.
http://ete.cet.edu/gcc/?/globaltemp_teacherpage/.
NASA, NOAA, Pope Francis make climate change hottest topic.
http://www.utsandiego.com/news/2015/jan/16/nasa-warmest-year/.
ii NASA-GISS, GLOBAL Land+ocean Temperature Index in degrees Celsius, base period: 1951-1980 using
elimination of outliers and homogeneity adjustment (GHCN-v3 1880-01/2015 + SST: ERSST 1880-01/2015).
There are many variants of this graph, some using land measurements only, some using different adjustment
methodologies, some averaging differently. We picked this one because it seems the most popular in recent
widely read articles, and it also is cited heavily in the technical literature.
iii Quoting from the report, Some risks of climate change are considerable at 1 or 2C above preindustrial
levels. Global climate change risks are high to very high with global mean temperature increase of 4C or
more above preindustrial levels in all reasons for concern, and include severe and widespread impacts on
unique and threatened systems, substantial species extinction, large risks to global and regional food security,
and the combination of high temperature and humidity compromising normal human activities, including
i
Page 6
growing food or working outdoors in some areas for parts of the year. The precise levels of climate change
sufficient to trigger tipping points remain uncertain, but the risk associated with crossing multiple tipping
points in the earth system or in interlinked human and natural systems increases with rising temperature.
The report talks about 4C or more above preindustrial levels, whereas our chart shows warming above the
1950 1980 average. Unfortunately the preindustrial temperature data cannot be measured very precisely,
so its possible we should draw our danger line a few tenths of a degree higher or lower than we did. You
could argue for a few more tenths one way or the other depending on precisely which global temperature
index you choose. None of these adjustments would make much difference to our point. And given the basic
uncertainties in how warming affects the Earth, these minor corrections would be false precision anyway.
iv The IPCC predictions we are using are stated as warming rates without specifying either the initial point or
precise definition of mean global surface temperature. We just added them to the actual temperature at the
end of the year the report was released. Other places in the reports have predictions that do specify
temperature predicted ranges at specific points in the future, and give the definitions. While these are
generally consistent with the predictions of warming rates, they are not tied to them precisely. Therefore,
someone might argue for shifting our IPCC prediction lines up or down by a constant few fractions of a
degree. This would not matter for the purposes of this essay. Also, the IPCC represents these same scenarios
as waves or wiggles around these warming rates, representing stylized natural variation.
v http://www.ipcc.ch/report/ar5/wg2/
vi The 2014 report has too many scenarios and variations to discuss here, but they differ mostly in the shape
of the increase and the secondary and local effects rather than the general average global trend over the next
fifty years. Moreover, A1FI represents a consensus forecast, essentially a median of many models that
commands widespread general agreement among climate scientists, while many of the other scenarios are
the result of individual models or classes of models. Furthermore, in 1990 they produced straight line
projections like we show here, in 2014 they do not, though we have represented them so here based on their
warming rates.
vii Ironically, if we used land only measures not land+ocean, as is the recent norm in climate science, we hit
the dangerous 4C point faster, though still far slower than consensus model forecasts like those from the
IPCC we highlight in the text. Instead of the linear extrapolation of land+ocean temperatures hitting 4C in
530 years it hits in 184 years, and for the quadratic it moves up from 132 to 66 years. The point of this essay,
that the data show warming that is far less worrying than headlines declare, and that worries are based on
models not the commonly trumpeted data which actually argues for less worry, would be unchanged if we
switched back to this land-only measurement rather than land+ocean measurements most currently favor.
viii We can stress test our findings using some other assumptions. We have been fitting our lines and curves
to the full 135 years of data. While those are the data ubiquitously graphed and cited these days, the IPCC
actually highlights the post-1950 period as when humans really began affecting the warming rate, and
estimates that a bit more than 50% of the warming weve seen over this period has been man-made. We go
further by fitting our line over the 1965-2014 period (visually that seems to be about when the slope changed
and indeed estimates of the slope over this period are bigger, and thus more conducive to the warming case,
than from 1950-2014). You do get a bigger slope over this period, 1.57 Celsius per century over 1965present vs. 0.67 Celsius per century over the full period (over 1965-2014 the squared term does not matter
much at all). At this rate we will hit the 4.0 Celsius danger point in just over 200 years instead of the 500+
years we found in our first linear trend exercise. Much faster, but still way slower than the major model
forecasts and well short of a time frame anyone would be sounding the alarm about. But, here, according to
the IPCC, we are actually being too alarmist. The IPCC only assigns a bit more than half the warming since
1950 to man-made causes. We ignored this earlier as we were also including the earlier period where the
argument that temperature was affected by emitted carbon is weaker. We cannot ignore it now. If we take
sixty percent of the 1.57 rate we get 0.94 Celsius per century and dont hit 4.0 until 385 years from now.
Now, perturbing our experiment should go both ways. As we noted earlier, using 4.0 Celsius for the extreme
danger point is certainly not fixed in stone. That is the point noted by the most authoritative sources, but
recently many have been discussing lower danger points, with 2.0 Celsius often being mentioned. Using the
full rate of warming (the slope) from 1965-present we will hit 2.0 Celsius in 84 years, and using sixty percent
of that slope as, again, the man-made part, we hit it in 174 years. All still comfortably longer than models
forecast (comfortably if things continue at the same rate as in the past, not comfortably if the models are
right, a possibility we do not dismiss!). Wed also note that even estimating over this shorter 1965-2014
Page 7
period and haircutting by 40% we get just under 1C of expected warming per century. Even if you ignore the
IPCC and believe that all the empirical warming is man-made, and thus dont haircut the empirical trend by
40%, you get 1.57C per century rate. In contrast the three IPCC scenarios correspond to warming rates per
century of 2C (low, best case), 3C (middle, expected case), and 5C (high, worst case). This again
underscores how you need a model to worry much about the future as the data, no matter how you slice it,
just doesnt get you there.
ix Believing models is fine, but you should acknowledge that youre expecting a sudden major change in the
rate of warming, and you absolutely should not appeal to the long-term rather mild change weve experienced
to buttress your case. You also need to explain why the model forecasts havent been showing up in the
temperature record. You shouldnt lead your article with the now famous chart we study, you should show it
reluctantly as evidence you must overcome and say, Yes, but. . . You should state clearly that you are
choosing a model over the actual temperature measurements, perhaps a model with feedback that threatens
to accelerate the process in the very near term. You should not claim current temperatures as evidence in
your favor but rather an anomaly you need to explain as your model does not, at least as of now. We arent
arguing that you cant or shouldnt build such a model.
x In that case you have to fear catastrophe from any significant anthropogenic effects. Even small amounts of
warming could touch off self-reinforcing reactions, and if not warming, some other effect of CO2 or any other
emission or non-emission environmental impact. Once again, you should be clear that your position is not
based on empirical evidence. It may be true but the chart doesnt make that argument. We have a lot of
sympathy with the view that all-else-equal complex systems should be disturbed as little as possible as
equilibrium can be delicate. It is entirely rational to be concerned that anthropogenic influences could push
things past tipping points. History is full of examples of human activities causing unintended disasters (not to
mention some pretty horrible intended disasters). But theres nothing in the data to suggest this has
happened, as of now, with respect to CO2 and temperature. It might have happened, or it might happen soon,
but you need a model to tell you that as the ubiquitous graph we begin with tells you the opposite. Similarly,
some have argued that severely limiting carbon output is an insurance policy we should take out regardless
of evidence-to-date as, and again we agree, disturbing complex non-linear systems always has some degree of
extreme danger. If the cost of this insurance is very low wed sign on to it without hesitation. If the cost is high
then it comes down to your faith in models as again, and really our only point, thats all we have to go on.
xi That seems to be quite a reasonable fear in general. But, in this case we're talking over 500 years (linear fit)
and over 100 years (squared extrapolation) before we hit 4C. Either is a long time for mitigation, adaptation,
and natural decline in carbon use. Also, there are a lot of current trends that are scary if extrapolated out 100
or 500 years, and anything we do to fight one might exacerbate another, or create new ones. So if you want to
use this argument to support aggressive action on climate change, you need some reason beyond the data to
single out global temperatures from everything else on paths that will become dangerous in the distant
future.
xii We are dying to say you cant have your temperature spike and heat it too even though it doesnt really
make much sense
Page 8

Its Not The Heat Its The Tepidity 20150310

Uploaded by

Copyright:

Available Formats

Its Not The Heat Its The Tepidity 20150310

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Its Not The Heat Its The Tepidity 20150310

Uploaded by

Copyright:

Available Formats

Very preliminary draft, please do not quote