2011-08-19 03:17:00Why Human Civilization Will End This Century
Daniel Bailey
Daniel Bailey

Why Human Civilization Will End in This Century

Posted on 19 August 2011 by bpl1960

NOTE: This is a guest post by Barton Paul Levenson 

Global Climate Models (GCMs) predict that as global warming goes on, it will bring more drought in continental interiors.  I'll explain in this essay why this will destroy human civilization in the next several decades.

You can't run a civilization--any kind of civilization, feudal, imperial, mercantile, capitalist, socialist, or Utopian--if you can't feed your people.

Is there increasing drought in continental interiors?  Ask the Australians.  Or better yet, ask Drs. Aiguo Dai, Kevin E. Trenberth, and Taotao Qian of the National Center for Atmospheric Research in Boulder, Colorado.  In 2004 they published "A Global Dataset of Palmer Drought Severity Index for 1870–2002: Relationship with Soil Moisture and Effects of Surface Warming" (J. Hydrometeorol. 5, 1117-1130).

Dr. Dai and his colleagues put together historical temperature and precipitation records for most of the globe for the years 1870-2002.  Their investigation covered the area from 77.5° north to 60° south latitude.  That doesn't sound like the whole globe, but in fact it's more than 92% of Earth's surface, and the parts excluded are pretty much covered with ice.  (And yes, we're only talking about land surface--measuring drought in the ocean doesn't really make sense.)  Their grid was broken up into 2.5 by 2.5° squares, which means 7,920 grid squares.

They couldn't find records for all the grid squares in all years.  But from 1948 on the coverage is nearly perfect.

The Palmer Drought Severity Index rates the severity of drought.  Positive numbers are good, negative numbers are bad.  When the PDSI is -3 or below, you have "severe drought."  Dai et al. noted that although the fraction of the globe's land surface in severe drought was 12% from 1948 to 1970, it had risen to 30% by 2002.


Turns out it's a very variable time series, though, so numbers from individual years don't tell you too much.  Dr. Dai has kindly made the dataset publicly available, and has recently revised and extended it up through December 2005--136 years worth of data, or to be more precise, 1,512 months worth.

I did some work on this data myself.  Here's the fraction of Earth's land surface in severe drought from 1948 to 2005:


A lot of variation, a lot of jogging up and down, but overall, wouldn't you say it's going up?

Here's the problem:  The figure was 21% in 2005, after briefly jogging up to 31% in 2003.  When it hits 70% or so, there will be no good agricultural land left anywhere in the world.

When will that happen?

Let's relate the drought fraction to something physical, like temperature.  If I try to explain the severe drought fraction, call it F, in terms of Hadley CRU temperature anomalies (dT) and the value of F the previous year (F1), I get this "regression equation" for 1949-2005:

      F = 0.06158 + 0.5852 F1 + 0.1188 dT         (N = 57, R2 = 0.86)

For you statistics guys, Durbin's h = 1.52, so we can rule out serious residual autocorrelation at 95% confidence.  And p < 2.80 x 10-6, 4.29 x 10-10, and 2.14 x 10-6, for each respective coefficient.

Here are the F values for the whole period:


But how do we know what the temperature will be in the future?  We need that to have the dT which gives us the F.

I used the IPCC's "A2" scenario, the "Business as Usual" Forecast.  Economic growth continues as in the past; population grows as the UN expects, leveling off around nine billion in 2050; and nobody moves away from fossil fuels.

I used the Myhre et al. (1998) equation relating CO2 levels and radiative forcing, and a climate sensitivity λ = 0.75 K W-1 m-2.  This gave me temperature projections for the A2 scenario up to the year 2100.

F reaches 70% in 2056.  The error bars are very small.

Now remember, this little scenario shows civilization collapsing from drought--and drought only.  It assumes no other pressures on the ecosystem we depend on.  How realistic do you think that is?

Can we save the day in time?  Switch away from fossil fuels, go to solar, wind, biomass, and geothermal?  Insulate buildings, use more trains and fewer trucks?

We just elected sixty people to Congress who don't even think global warming is happening, let alone man-made.  So what do you think the chances are?  Seriously?

Suggested Further Reading

The Dai After Tomorrow

2011-08-19 03:35:01
Chris Colose


A bit too alarmist....

2011-08-19 03:56:02


Wow.  Small world.  Barton and I were in a sf/fantasy anthology together back in 2007.  Fancy meeting you here!


Mine was "Bilroy 5".  Now I have to go dig up my copy and read Barton's.

2011-08-19 06:17:47
Mark Richardson

Too alarmist for me. Fitting curves and projecting into the future is useful to <i>highlight</i> problems, but you need a lot more evidence before you can put such weight in this claimed result. If this were to be posted, then it needs a lot more qualifiers or a lot more evidence (peer reviewed, preferably) IMO.



My first reaction to this is to be very skeptical: global average land area affected by drought may not be linearly related to temperature - the extension of drought zones seems largely related to the extent of the Hadley cells. Which is complicated: Earth is a sphere and the relationship between Hadley circulation and temperature is also complex.


Also: this doesn't say much about desalination or GM crops. Either (or both) might be enough to prevent collapse. Or maybe not, but I don't see any evidence or discussion about that in the post...

2011-08-19 06:19:33
Mark Richardson

As an attachment, there is a great post here explaining why curve fitting without more physical considerations can often lead to the wrong result :p

2011-08-19 06:43:37
Dana Nuccitelli

Yeah I'm pretty skeptical about this too.  The error bars are small?  But there are a lot of assumptions (like maintaining Scenario A2 emissions, and most imporantly, assuming the model is correct).  I echo Mark's concerns about curve fitting.  We just published this post criticizing 'skeptics' for drawing incorrect conclusions based on physically unjustified curve fitting, and now we're going to publish a post which is based around curve fitting?  At the very least the regression formula needs some physical justification.  Then the post also needs some uncertainty caveats.

2011-08-19 07:27:56
Rob Painting

Reckon the title needs to be changed for starters. Sure the global economy will collapse this century, perhaps sooner rather than later, but I don't know how that is going to motivate readers. 

And ditto what others have said about curve fitting. Be nice to see a more in-depth treatment of the issue.  


2011-08-19 08:32:41Earth's Land Fraction
John Cook


I'm not convinced of the general tone of the piece. USA is not the world and while USA is seriously lagging behind the rest of the world, plenty of other countries are acting. Deltoid showed this great map of what each country is doing about climate:

SmithSchool rates each county's actions

It's not as fast or widespread as we would like but I don't think A2 will come to pass. So to basically say "we're screwed, pack up the farm" is a counterproductive message. Any fear message MUST be accompanied with hope, otherwise you just paralyse the reader and inhibit action. And that's the last thing we want to do.

Now some technical nits. Firstly, if you were going to go the alarmist route (which I advise against), the headline should be "Human Civilization will end this century if we continue business-as-usual". To say "will end" is to invoke fear without hope. So the framing throughout this article should be "what would happen if we don't change our trajectory?"

Secondly, what about showing F up to 2010 with the temperature projection, with actual F and projected F? Why say it with words when you can say it with visuals?

I concur with other comments about uncertainty caveats - not comfortable with the "error bars are very small" comment. At the very least, I'd take the A2 range, to give your F projection upper and lower bounds. But even that doesn't take into account uncertainties in the relationship between F & T.

Little messaging suggestion - rather than use 'Earth's Land Fraction' in Figure 1, can you change it to "Percentage of Earth's Land Surface" or something like that and use percentages in the y-axis? That will be a lot more meaningful to people than fraction.

2011-08-19 09:40:16
Andy S


I haven't looked at this in any detail but I have to agree with everyone else that this is the kind of curve-fitting that we usually deplore. Drought may correlate with temperature--perhaps both are directly linked to ENSO and other ocean oscillations--over the period studied, but it does not mean that it willcontinue to correlate with a projected trend, and not linearly. Once we get to higher numbers of the drought index and higher temperatures, wouldn't we expect the physical conditions to change so that previous empirical relationships would no longer apply? Surely the only way of predicting what will happen as the climate boundary conditions change drastically is through detailed and physically-based GCMs. 

I don't see that this analysis adds to Dai's work. Lord knows, that was dire enough.

2011-08-19 09:50:45


I have some problems both with the tone, as others already commented, and with the model itself.

In the post it is said that dT is the anomaly. It can't be, otherwise a constant positive/negative anomaly would make F increase/decrease forever; in other words, the choice of the baseline would artificially result in a trend in F. Maybe he means anomaly change.
The model is linear, sooner or later F will go beyond 1, while in the real world the increase of F must slow down and asimptotically reach 1. I'm not against simple linear models, on the contrary I think that they may give important insights. Though, extrapolation is a risky business, when not plainly wrong.
The use of the equilibrium sensitivity is unphysical, though it only changes the regression coefficient of dT. A similar argument holds for F, it is assumed that the response time is less than one year.
As far as I remember, severe drought is for a PSDI index of -4 or lower, not -3.
The 70% threshold appears somewhat arbitrary. Whatever it is, it needs to be based on the amount of crop produced relative to the world population needs.

In summary, I like the model and the agreement with the data is impressive; I find that it has a more direct impact than the original Dai et al maps. The conclusions are overstated. Simply showing the second graph and maybe adding a simple T anomaly vs PSDI plot is scary enough.

2011-08-19 10:06:04
Andy S


Let me put a denier's hat on for a minute:

Grain yields have been increasing over the same time period. See here.

If we were to fit this to temperature, the drought index (or, better, CO2) we'd get a nice correlation, too. The 21st Century projection is obvious.

Does this not mean that we are all going to get to pig out indefinitely?

</deniers hat>

2011-08-19 10:50:22Putting on my 'getting skeptical about skeptics' hat
John Cook


Andy, what happens after 1990 to grain yields?

2011-08-19 10:52:46


I dont think this post is supported by published scientific papers. The CMIP archive has precipitation models under various scenarios and it doesnt look like they support this contention.

2011-08-19 10:59:41
Dana Nuccitelli

Reviews of this post would seem to suggests that we're not all the biased alarmists the "skeptics" think we are :-) 

It's an interesting comparison between the 'skeptic' near-universal acceptance of Loehle and Scafetta's and Spencer's curve fitting, vs. everyone here saying that Barton's analysis needs some better physical justification.  Proves who the real skeptics are.

2011-08-19 11:24:33
Andy S


JC: Andy, what happens after 1990 to grain yields?

Drat, I forgot to hide the decline ;-))

2011-08-19 14:05:46Skeptical SkSers
John Cook


It's great that SkSers are so skeptical and that our peer-review system is robust. When I showed the Scientific Guide to Skepticism to both scientists and SkSers, the best and most rigorous feedback came from the SkSers.

I have to admit though, I'm also conscious of biases that can influence us without us even realising. James Hansen refers to scientific reticence and Freudenberg calls it 'the Asymmetry of Scientific Challenge' (Freudenburg 2010). We're so conscious of the unholy rage that our more alarmist statements will bring down upon us, that we unconsciously tend towards more conservative statements. Freudenburg 2010 suggests this is why the IPCC tends to underestimate climate responses. So have to watch for that bias and not pull our punches if they're evidence based and robust.

2011-08-19 14:08:46
Glenn Tamblyn


I agree this post is a bit too speculative, just projecting one data series out like this.

However a series of posts on how the various pressures we face - AGW and all the rest - can compound each other and produce outcomes that may be worse than any one problem alone might produce.

For example:

- Ocean acidification & Overfishing trends & growth of jellyfish in fishing grounds & expansion of 'dead' zones

- Crop yields & Drought expansion & declines in ground water reserves & Glacier melt & greater climate variability & growing population


Hybrid topics like this could be useful in breaking people out of the reductionist mind set of looking at separate problems in isolation. Also things like the insularity of some countries (we wont name names here) in thinking they will be protrected from problems just because they haven't been affected much yet or are affluent

In this way we can highlight the sort of threat BPL is addressing without having to go out on too much of a limb on any single topic.


As to the need for Hope else we paralyze people. The problem is that the only hope is action. So how to galvanise action as the source of hope without hope paralyzing fear as the spur to action.?

2011-08-19 15:01:03


"We're so conscious of the unholy rage that our more alarmist statements will bring down upon us, that we unconsciously tend towards more conservative statements"

Wouldnt that actually be caring rather a lot about truth rather than alarmism? I dont think any cause is advanced by hyperbole that can be easily popped.

2011-08-19 15:30:16
Andy S


A long time ago an old professor at Glasgow University gave me the advice: "Don't overstate your case, you'll only be making the job of your critics easier".

The Dai study result (see the last one for the end of the century here) is already so desperately bad, showing the great grain belts of the N America, Europe and west Asia to be in a state of extremely severe drought, that it hardly needs to be made worse if our goal is to spur action to prevent it happening. 

If we make a technically weaker case, predicting an even worse outcome, we'll bring criticism to bear on the details of that argument and risk discrediting the whole case that climate change could bring appalling food shortages later this century.


2011-08-19 19:12:10Heads up, Mark et al.


Mark R:  global average land area affected by drought may not be linearly related to temperature

BPL:  Hello?  I just showed that it IS linearly related to temperature.  The linear relation between temperature and F is significant at p < 2 x 10^-6, which means one chance in 500,000 that this is due to sampling error.  Do you understand what a regression analysis is or how it works?

And the idea that this is "curve-fitting" -- you guys should really understand what a term means before using it.  "Regression" is not the same as "curve-fitting."  If I'd fit F to a quadratic of time, that would be "curve-fitting."  Regressing it on physical causes is not the same thing.  If you're disappointed that this isn't a peer-reviewed article, then let me recommend that you start reading peer-reviewed articles on the subject and not blog essays.  I didn't go into how temperature affects outgoing longwave radiation, and details of the hydrological cycle, and how AGW moves the rain, because this is NOT a technical article.  This is an overview for an educated lay audience.

Has anyone here ever cracked a data analysis textbook?  Let me recommend Gujarati's "Applied Econometrics" for a good general overview.

As for being "alarmist"--when the building is on fire, the sane, rational response is to trip the alarm.  I'm not being alarmist.  The situation is alarming.  Try to understand the difference.

2011-08-19 20:44:58Barton
Daniel Bailey
Daniel Bailey

If I may offer up a suggestion:

Is it possible you could post here an internal version of your full paper, rather than the lay synopsis at the top?  SkS Forum members have a wide range of backgrounds; several members I'm sure have the background to give your paper a robust technical assessment, rather than filtering their perception of your paper through the lens of the lay synopsis.  The full paper wold not be made available to the public, as this is a private forum.

2011-08-19 21:00:47
Rob Painting

BPL - spare us the lecture on "curve-fitting" and "cracking open textbooks". I don't see how a casual reader is to learn much from your post in it's current form. The gist is "here's the observations of drought, it got worse as it got warmer, extend that into the future and we're fucked". How is that helpful?  

" F = 0.06158 + 0.5852 F1 + 0.1188 dT (N = 57, R2 = 0.86) For you statistics guys, Durbin's h = 1.52, so we can rule out serious residual autocorrelation at 95% confidence. And p < 2.80 x 10-6, 4.29 x 10-10, and 2.14 x 10-6, for each respective coefficient."

The above is going to mean diddly squat to most readers. I know there's still a bit of to-and-fro between authors about the technical level of posts, but we're trying to make this stuff more accessible to the layperson, not less.

As I said before, the title of your post, and tone, are wrong for SkS. I don't have a problem with what you are trying to explain, just the way you're going about it. Screaming louder isn't going to help. I think your post could be re-written still portraying the alarming message, but with a matter-of-fact tone and a bit more supporting work. 

2011-08-19 21:03:56Suggestion
Daniel Bailey
Daniel Bailey

Perhaps tweaking the title?

Why Human Civilization May End This Century

2011-08-20 04:07:34
Mark Richardson

I think it's important to include the maths and statistics detail. But perhaps as an endnote: in the text you would probably describe it as an equation that says that drought zones expand when it is warmer than average.


But there are serious problems with using this for future projection... you have to explain why you reckon there will be no good agricultural land when regional projections including the water cycle (e.g. Dai) expect lots of rainfall in the equitorial belt and at high latitudes. I suspect again this is to do with the large scale circulation: we expect more total precipitation with warming.



I expect drought will be a huge disaster this century, but I don't think that your analysis shows that the end of civilisation is guaranteed from this. Coastal areas can use desalination and some countries will continue to produce food whilst new areas at high latitudes will be opened to farming.

Your model to observations fit is really interesting, but I think it needs more analysis into why it fits so well and from that you'll have a better idea of whether it'll continue to work in the future, or what modifications it needs.

e.g. is over 110% of land area in drought at +3.5 C anomaly realistic, even though we expect significant intensification of the water cycle?

2011-08-20 04:50:55
Dana Nuccitelli

Your model to observations fit is really interesting, but I think it needs more analysis into why it fits so well and from that you'll have a better idea of whether it'll continue to work in the future, or what modifications it needs.

That's the point I was trying to make as well.  It needs more discussion about why your model is an accurate representation of physical reality.  I would also suggest if possible, running the model for scenarios other than A2.  I understand that choice (right now our emissions are right on track for A2), but it would be helpful to see how much of the drought increase we can avoid by reducing emissions.  That would tie in with John's suggestion to add some discussion about how there's still hope (hopefully!).

Then the tone of the post could change to something more like "if we continue with BAU then civilization may end this century, but if we reduce emissions to [x], we can avoid [y] of this drought expansion."

2011-08-20 05:09:46
Mark Richardson

Using different projections causes a problem.


It hits 100% drought fraction by 3 C. Anything above that and you start getting >100% in drought. Maybe the fit will hold for 3 C warming, but that seems unlikely - what about 2C? 1? 0.5?

2011-08-20 06:37:18
Andy S


I would recommend that we also do a history match/hindcast on this. I suspect that this will reveal the limitations of attempting to project correlations derived over a 50 year time frame into the past and future; like Dana did on the Loehle and Scafetta article.

2011-08-20 15:58:26


i think that there is a problem that posting this would have to anticipate. Either BPL's model is wrong, or the climate models are wrong. I cant see how both can be correct.

2011-08-20 23:28:19
Daniel Bailey
Daniel Bailey

I dunno.  Dr. Dai has already used model ensembles to anticipate future drying:


Figure 1: Current Palmer Drought Severity Index [PDSI] 2000-2009. A reading of -4 or below is considered extreme drought. Regions that are blue or green will likely be at lower risk of drought, while those in the red and purple spectrum could face more unusually extreme drought conditions. (Courtesy University Corporation for Atmospheric Research [UCAR])



Figure 3: PDSI 2060-2069. (Courtesy UCAR)

If the drying is anything resembling that shown in these figures, a very large population will be severely affected in the coming decades over the whole United States, Southern Europe, Southeast Asia, Brazil, Chile, Australia and most of Africa.  That would seem to tie in nicely with what Barton is saying.

I can understand the issues with tone.  I can also understand some of the technical issues that have been shared.  Would it be of help if Barton were to share the more technical version of his paper with some here?  After all, isn't the purpose of this forum to provide constructive feedback to better aid those in constructing blog posts?

2011-08-21 01:28:39
Dana Nuccitelli
Yes, seeing the paper would be helpful. But we are trying to provide constructive criticism - i.e. modify the tone and provide more physical justification for the model. I like the hindcasting suggestion if the data is available to do it, though that may not be the case.
2011-08-21 02:24:52
Tom Curtis


A couple of questions:


1)  Is the palmer drought index an absolute index, or relative to current land use?  From Fig 1 in Daniel's post I would guess relative in that it shows neutral to slightly positive values across large sections of what are essentially desert and semi-desert in central Australia, while showing a dark red in relatively moist areas like the european mediterranian and the northern south island of NZ.


2)  IF the PDI is a relative measure, what is to prevent the shift of land use to adapt.  Across Australia agricultural land is broadly divided into cattle country, beef country, wheat country, and dairy country based on aridity.  A drying of Australia will have the general effect turning beef country into roo country, sheep country into beef country, wheat country into sheep country, and so on.  That means there is a reduction in agricultural productivity, but not a cessation of productivity.  How are these possibilities factored into this scenario?


3)  If you have a PDI index from 1870, and HadCRUT from 1850, why do you only check the correlation from 1949?  Does the same correlation hold prior to 1949, and if not what physical theory explains the fact that it holds after 1949 and not before?  Of course, if it holds prior to 1949, why not show it?


4)  Based on figure 3 from Daniel Bailey's post, some areas will have significantly positive.  Why will new agricultural productivity in those areas, particularly Siberia and Northern Canada not substituted significantly for lost productivity in other areas?

2011-08-25 08:38:19


It's taken me some time to trawl through Dai and those reconstructions, and I agree that I need to withdraw my contention that the BPL model was at odds with climate models. More to drought than just precipitation. I am still concerned that tone is too alarmist and drought level would collapse civilization. Dai et al has some cautions too - actual drought is related to ocean-atmosphere circulation pattern like ESNO. The models are not robust in predicting how these will change. That said, both the Dai model and BPL model are reasons to fear the future under even an A1B scenario. Just dont overstate it.