Wednesday, January 06, 2016

Lower Troposphere "Pause" is Purely an Artifact of 1997-98 El Nino

With each month's measurements of the temperature anomaly of the lower troposphere comes the by-now standard claim that "there's been no warming for 18 years and X months." X varies a little -- but it doesn't seem to get bigger with time.

It's straightforward to show that this "pause" in the LT is purely a product of the 1997-98 El Nino -- that is, of natural variability. It would not exist without it, if everything else stays the same.

In the following graph I plot what I call the "reverse trend" -- the trend from any year on the x-axis to today. So, for example, the graph shows that the trend from 1983 to 2015 is 0.12°C/decade, and from 2003 to today is 0.04°C/decade. (There are error bars, but the point made here doesn't depend on them so I'll leave them off.)

The red line is the actual measured temperature anomalies. (I used annual averages to keep this simple.) The blue line assumes simply there was no big El Nino spike and that 1998's anomaly was the average of 1997's and 1999's.


As you can see, the "pause" for the actual data -- the red line's dip to slight negativity in 1998 -- depends closely on choosing 1998 as one's starting point. That's precisely the definition of a "cherry pick" -- choosing the starting point to give the result one wants, regardless of its scientific legitimacy or whether it is long enough to represent climate, and not natural variability.

The blue line, with no 1997-98 El Nino, shows no pause at all at that point. In fact, it's never negative for any starting point before 2015.

As expected, the trend over climatically representative intervals, such as the 30 year period 1985-2015, is essentially identical to that which includes the natural variability of that big '97-98 El Nino. El Ninos and La Ninas balance out over the long-term. (They're not the only natural factors that can have an influence, of course -- volcanoes and changes in solar irradiance do too -- but no need to include them here to make the point.)

And shorter trends, of about 10 years or less, aren't statistically significant, and usually not even close to being statistically significant.

The lower troposphere "pause" over 18 yrs X months is purely an artifact of a big El Nino -- that is, of natural variability.

Note added 9:00 pm: A commenter asked a good question: what about the other El Ninos that have occurred since? The average of the monthly MEI (Multivariate Enso Index), an ENSO proxy, from Jan 1999 to present is 0.04, very close to neutral. (For example, 1997's MEI was 1.58. 2015's is 1.59.

In other words, La Ninas and El Ninos have tended to cancel out since 1999. So it's not a bad assumption to take the data as it is, excepting only the 1997-98 El Nino. But it's not a perfect assumption, as one would still need to account for ENSOs before 1999 -- the average MEI from January 1979 to December 1996 is 0.48.

41 comments:

David in Cal said...

Nice analysis, David. One quibble. You correctly pulled out the 1997-1998 El Nino, but I believe you left in the 2015-2016 El Nino. If you also pulled out the current El Nino, say by ending the chart a few months earlier, I think you'd come closer to a zero slope for the period starting around 2002.

BTW the long-term ex-El Nino rate of warming of 0.11 casts some doubt on climate sensitivity being as high as 1.5 to 4.5 deg C, or even higher as some claim. A warming of 1.1 deg C per century would mean that we have lots of time to figure out how to deal with global warming.

Cheers

David Appell said...

David: Yes, you're right. Ideally I would have pulled out all El Ninos and La Ninas, or at least the 2010 and 2015 El Ninos, say by some function of an ENSO index like the MEI or ONI.

MEI http://www.esrl.noaa.gov/psd/enso/mei/table.html
ONI http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml
MEI extension: http://www.esrl.noaa.gov/psd/enso/mei.ext/table.ext.html

Maybe another time. (I'd be surprised if no one has already done that, like Nick Stokes.) For now all I can say is that if the 1997-98 El Nino spike had never happened but everything else stayed the same, there'd be no "pause."

David Appell said...

David: I added a note to this post, at the end. Have a look.

Lars Karlsson said...

David Apell: " Ideally I would have pulled out all El Ninos and La Ninas, or at least the 2010 and 2015 El Ninos, say by some function of an ENSO index like the MEI or ONI."

There are already methods for doing that.

JoeT said...

David in Cal,

One of the problems I have with your posts is the imprecision in your language. I see David's blog as more or less a science blog that looks at data, even if every now and then he ventures into other topics like Delingpole. David addressed your first point, but I would still quibble with your language because while David calls it the 97/98 El Nino, he only averaged out the 1998 temperature. Similarly, when you refer to the 2015/2016 El Nino, which do you think is the year he should average out -- 2015 or 2016? Or to put it another way, if you look at the RSS or UAH data, what month do you think corresponds to the impact of the El Nino on the data?

But I have even more problems with your second comment. You claim that a warming rate of 0.11C/decade (actually David said 0.12C/decade) casts doubt on the climate sensitivity. But when you quote the sensitivity you say it's 1.5 to 4.5 C. What you left out is that this is the equilibrium climate sensitivity (ECS) per doubling of CO2. There is no way you make a statement about the sensitivity unless you also took into account the change in CO2 concentration. So given a warming rate of 0.12 C/decade, what is the transient climate response (TCR)? NASA GISS gives a surface warming from 1970 of around 0.17C/decade. What TCR do you get then? Given that the warming also depends on cooling factors such as aerosols, do you think this calculated TCR would be an upper or a lower limit?

And finally, Kate Marvel just published a fairly comprehensive paper looking at both TCR and ECS. What numbers did she get?

David in Cal said...

David -- In linear regression, the slope depends disproportinatedly on data points near either end. Even though La Ninas and El Ninos have tended to be equal in total magnitude since 1999, the current El Nino has a bigger impact on the slope than the earlier La Ninas.

Cheers

David in Cal said...

Joe, the .12 was starting in 1983. The chart, by my eye, shows .11 from the earliest date shown.

I'll try to get back to your point about CO2 concentrations and sensitivities when I have more time. I have seen exhibits pointing out that the actual temperature increase in troposphere was far below what the IPCC sensitivities are consistent with. I can get a hold of the growth in CO2 from 1979 and check how big an temperature increase would be consistent with various sensitivities.

Cheers

David in Cal said...

Suppose climate sensitivity were 3.0 deg C -- the midpoint of IPCC's highly likely range. From 1979 to today, CO2 in the atmosphere rose from about 335 to 405 ppm. http://www.esrl.noaa.gov/gmd/ccgg/trends So, temperature should have risen by 3*log(base 2) (405/335) = 1.04 deg C.

However, at a slope of .11 deg C/decade, the warming for the period of troposphere temperatures being available (1979 - 2015) would be only 3.6*0.11 = 0.4 deg C.

The difference might even be a bit greater for two reasons. Having an El Nino at the end point may overstate the trend in the temperature data. And, the planet has been warming ever since the end of the Little Ice Age. So, part of that 0.4 warming may have been natural, rather than anthropogenic.

Now, these figures aren't precise, since they ignore the difference between Equilibrium Climate Sensitivity and Transient Climate Response. However, the difference between the 1.04 projected and the 0.4 actual is so great that I feel comfortable saying that the actual troposphere temperature rise, since these temperatures became available, looks nothing like the catastrophists' projections.

David Appell said...

David in Cal: The term "climate sensitivity" pertains to surface temperatures, and while it pertains to CO2, there are other factors influencing temperatures (CH4, black carbon, aerosols, volcanic eruptions, etc).

As Joe T wrote in an adjacent thread, there is no reason to expect the surface and lower troposphere to warm at the same rate, or even be proportional.

David Appell said...

Lars Karlsson said...
"David Apell: " Ideally I would have pulled out all El Ninos and La Ninas, or at least the 2010 and 2015 El Ninos, say by some function of an ENSO index like the MEI or ONI."
There are already methods for doing that."

and he pointed to: http://www.skepticalscience.com/foster-and-rahmstorf-measure-global-warming-signal.html

Thanks for that, Lars.

David in Cal said...

David -- In 2007 the IPCC said that physical expectations and most model results were that the troposphere would warm faster than the surface.

The IPCC 4th Assessment Report said,

For global observations since the late 1950s, the most recent versions of all available data sets show that the troposphere has warmed at a slightly greater rate than the surface, while the stratosphere has cooled markedly since 1979. This is in accord with physical expectations and most model results, which demonstrate the role of increasing greenhouse gases in tropospheric warming and stratospheric cooling; ozone depletion also contributes substantially to stratospheric cooling.

http://www.ipcc.ch/publications_and_data/ar4/wg1/en/faq-3-1.html

Cheers

David Appell said...

David in Cal:

Actually your quote says that the troposphere *DID* warm at a slightly greater rate than the surface.

David Appell said...

David in Cal wrote:
"In linear regression, the slope depends disproportinatedly on data points near either end."

No, it does not.

The equation for the slope "b" is given on this page, equation #14:

http://mathworld.wolfram.com/LeastSquaresFitting.html

It does not weight any points more than other points.

David Appell said...

PS: Why do you keep writing false things?
And then never acknowledging when you are wrong?

JoeT said...

David in Cal

Some comments on your post

1- Good for you, your post is already much better since you've put some numbers in.
2- The slope in linear regression does not depend disproportionately on the endpoints. The linear regression minimizes the variance. If each point has the same uncertainty than over fitting the endpoints just pushes up the residual somewhere else.
3- I get 3*log2(405/335) = 0.82
4- For RSS from 1979-2015 I get 0.12C/decade, UAH(version 5.6) = 0.14C/decade. GISS for that same period is 0.16C/decade.
5- For an ECS = 3.0, the TCR is around 1.9. How does that compare to your calculation now?
6- We haven't even considered aerosols yet. Does that give you a lower or an upper limit on your estimated TCR?
7- You keep bringing up that the troposphere should be warming faster than the surface. As we've been discussing on the Delingpole post, it's the tropical troposphere that should be warming faster. Do you understand the reason why?
8- Although this isn't a firm conclusion yet for me, I'm of the opinion that satellite measurements of microwave brightness can't see the tropical hot spot because it's not in the TLT range but closer to the stratosphere. In this range, the measurements are contaminated with temperature from the stratosphere.
9- On the other hand, the RATPAC radiosonde data, agrees with the model prediction of the extra warming in the tropical troposphere versus the surface.
10- All of the warming since 1970 is due to greenhouse gases. The warming right after the Little Ice Age is due to the cessation of volcanic activity. If the planet warms it has to be due to a forcing. All of the natural (and some anthropogenic) forcings since 1970 or so are cooling.

JoeT said...

Here's a link to the radiosonde data
http://i51.tinypic.com/35mgjr6.png

Layzej said...

Further to Joe's point 10 (All of the natural forcings are cooling), here is the TSI since 1979: http://woodfortrees.org/plot/pmod/mean:120

Here is the PDO:
http://woodfortrees.org/plot/jisao-pdo/mean:12/from:1979/plot/jisao-pdo/from:1979/trend

TheMostCasualObserver said...

Tamino has a discussion of RSS vs RATPAC radiosonde data here:
https://tamino.wordpress.com/2015/12/11/ted-cruz-just-plain-wrong/

He shows that RSS and RATPAC have been diverging since 2000.

Link to plot
https://tamino.files.wordpress.com/2015/12/compare_overlap.jpeg?w=500&h=332

Link to divergence plot
https://tamino.files.wordpress.com/2015/12/diff.jpeg

Layzej said...

David in Cal suggested: "I have seen exhibits pointing out that the actual temperature increase in troposphere was far below what the IPCC sensitivities are consistent with."

Here are the satellite models compared to projections: https://andthentheresphysics.files.wordpress.com/2015/12/cv-xxg7w4aabg3-jpg_large.jpeg

Not inconsistent, although veering towards the low side at about the same time that the satellite models diverge from the RATPAC (radiosonde) measurements: https://tamino.files.wordpress.com/2015/12/diff.jpeg

It would be interesting to see how much more closely the measured tropospheric temperatures align with the projections.

David in Cal said...

David -- maybe you and I are using terms differently. The following examples illustrate the point I was making about linear regression. The following three series are time series values corresponding to time = 1,2,3,4.5.

A. 5, 5, 5, 5, 5 Mean = 5; Slope of fitted line = 0;
B. 5, 5, 5, 4, 6 Mean = 5; Slope of fitted line = 0.1
C. 5. 5. 5. 6. 4 Mean = 5; Slope of fitted line = -0.1

Two observations:
1. Revising the values at times 4 and 5 doesn't change the mean, but the revision changes the slope.
2. The value at time 5 has a much bigger impact on the slope than the value at time 4.

Cheers

David in Cal said...

Layzej - I am not sure I know what the model numbers represent. Am I correct that this chart is based on the 5th IPCC report? Are the model figures a mix of fits to actual past data and predictions of future data? Having models fit known results is no trick. John Von Neumann supposedly said, "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk."

The test is how well the models predicted unknown future results. Do you know what was the latest year of actual data available when this report was done? The match isn't too bad for early years, but the difference from 2011 on is large and all in one direction. Were years through 2010 known when the models were created?

cheers

Layzej said...

If you are worried about an El Nino at either end of the series skewing the results (which I agree is a concern) then your best bet is to subtract ENSO. The result is a fairly straight line: https://tamino.files.wordpress.com/2011/12/figure05.jpg?w=500&h=499 .

This makes it quite clear, (as our host points out) that the lower troposphere pause is purely an artifact of the 1997-98 El Nino. Although there does appear to be some other bias in the satellite models that becomes obvious once you compare them to the actual tropospheric temperature measurements: https://tamino.files.wordpress.com/2015/12/diff.jpeg

David in Cal said...

Joe T

2. See my reponse to David Appell above
3. You are correct. Still a slope of .82 is far above the actual fitted slope of .4
7. See quote from the 4th IPCC above. It doesn't say tropical troposphere. It says that the troposphere warming faster than the suface is consistent with the science.
10. Perhaps all the warming since 1970 may be due to greenhouse gases. Perhaps the end of the Little Ice Age was due to changes in volcanic activity. But, I don't believe these are established facts or consensuses. Indeed, since the IPCC cannot put a lower bound on the true value of climate sensitivity, it would seem impossible for anyone to be sure that the warming since 1970 had no natural component.

Regarding some of your other questions, I don't understand the models well enough to address the difference between TCR and ECS, the impact of aerosols, etc.

Cheers

Layzej said...

This is the CMIP 5 95% ensemble: http://www.climate-lab-book.ac.uk/comparing-cmip5-observations/

It is interesting that the satellite models start to veer towards the low side at about the same point that the satellite models also diverge from the actual tropospheric temperature measurements. Further work to explain the discrepancy between the satellite temperature models and radiosonde temperature measurements is warranted.

Layzej said...

DiC: "Indeed, since the IPCC cannot put a lower bound on the true value of climate sensitivity, it would seem impossible for anyone to be sure that the warming since 1970 had no natural component."

If you want to base policy on what we know then 3+/-1.5 is a good measure. sensitivity is very likely to be within that range. If you want to base policy on what we don't know then we should be planning for much worse (and hoping for much better). You assume that uncertainty will always work in our favour. This is not likely to be the case.

David in Cal said...
This comment has been removed by the author.
David in Cal said...

Layzej, the world is not now planning for a climate senstivity of 3. The practical impact of the Paris agreement and Obama/EPA proposals are relatively small reductions in CO2 emissions. These plans will result in the continued rapid growth of atmospheric CO2.

I don't think 3+/-1.5 is a good measure of sensitivity. I think the IPCC models are not reliable enough to make predictions. I think a better estimate is simply the actual rate of warming since 1970 or so, when CO2 became a big factor. Ocean, surface and satellite temperature records are all flawed measures to some degree. I think the satellite measures are the best, so the best estimate of long-term warming is around 1.0 to 1.5 deg per century. If you use the surface records, then you'd get a somewhat higher figure. In any event. the world currently lacks the technology and political will to stop the increase in atmospheric CO2.

Layzej said...

DiC: I don't think 3+/-1.5 is a good measure of sensitivity.

Needless to say, the worlds scientists disagree with you.

DiC: Obama/EPA proposals are relatively small reductions in CO2 emissions.

Inefficient too. A market based approach would be much preferable. Hopefully more libertarians and fiscal conservatives will step up to the table.

David in Cal said...

You mean, some of the world's scientists disagree with me. BTW the word "good" is pretty vague. It's not a scientific term.

I don't know what kind of market-based approach you have in mind Layzej. Don't forget that that global warming, is, well, global. A fixed tax amount on each person sufficient to stop the increase in atmospheric CO2 would mean starvation for most of the world's poor. A tax or other financial burden on the world's power suppliers would be passed on to the users, with a similar impact. I don't believe any mechanism would work. The world isn't willing to make the kind of Draconian cuts in energy that would be needed to stop the growth of atmospheric CO2.

IMHO we're not going to stop increasing atmospheric CO2 until there's some non-carbon energy that can replace coal and oil and at a comparable cost.

Or, conceivably someone might invent a geo-engineering approach that ends global warming without reducing CO2 emissions. That approach sounds dangerous to me, but if climate sensitivity really is 3 deg C or more, it might be better than frying.

Cheers

Layzej said...

I'm glad you asked. I'm in favour of the solution proposed by Preston Manning, the leader of the Canadian Reform party. His party came in to the right of the Progressive Conservative (PC) party. They are somewhat akin to the Tea Party in the U.S. They ended up consuming the PC party and leading the newly branded Canadian Conservative party to victory.

Reformers believe we should reduce or eliminate sales and income tax. These are activities that we should be encouraging rather than taxing. A modest revenue neutral carbon tax would allow us to reduce income tax and it would steer the market towards a new energy economy.

As it turns out, economists also agree that this approach is both necessary and optimal:
Necessary: http://policyintegrity.org/files/publications/ExpertConsensusReport.pdf

Optimal: http://www.economist.com/blogs/freeexchange/2011/09/climate-policy

JoeT said...

David in Cal,

It's worth going over again the back-of-the-envelope estimate of the transient climate response because it's critical to this discussion. I'll do it somewhat differently than what you did.

As I said above, for RSS from 1979-2015 I get 0.12C/decade, UAH(version 5.6) = 0.14C/decade. GISS for that same period is 0.16C/decade. From ftp://aftp.cmdl.noaa.gov/products/trends/co2/co2_annmean_mlo.txt I get that CO2 concentrations in 1979 were 337 ppm and in 2015 they were at 401. Then I get:
RSS 3.6*0.12*log2(401/337) = 1.7 C per doubling of CO2
UAH: 2.0 GISS: 2.3
Even if I use your numbers I get 3.6*0.11/log2(405/335) = 1.6 C per doubling of CO2

So here we have a range of numbers from 1.6 to 2.3, but let's just call it something on the order of 2. Now the transient climate response has a specific definition, namely that it is the warming after 70 years when the CO2 concentrations rise at 1% per year. So what I did above isn't technically the TCR, but we're only trying to do rough estimates here.

The latest paper by Marvel et al., is discussed by Gavin Schmidt over at realclimate.org seeks to narrow the constraints on climate sensitivities. In their paper they show TCR values on the order of 2 and ECS on the order of 3. My b-o-e estimates are not out of line with Marvel's calculations.

You write, "it would seem impossible for anyone to be sure that the warming since 1970 had no natural component." However, the consensus opinion, once again, is that all of the warming in recent years is anthropogenic. Based on the IPCC report, Gavin Schmidt derives a probability distribution function of human influenced warming that peaks at 110% (because of the influence of aerosols): https://twitter.com/ClimateOfGavin/status/674347988937007104

From this one concludes that it makes no sense at all to just look at the warming rate per century, which comes from an assumption that the warming is natural. Rather one has to look at emission rates of greenhouse gases which is what the IPCC does in its RCP pathways.

David in Cal said...

problematic survey.

1. Treats the decision as binary, that is, action or inaction. In fact there's an enormous range of conceivable actions. With specifying magnitudes, there's no real agreement. A tax that adds 25 cents to a gallon of gasoline is qualitatively different from a tax that adds $20 to a gallon.

2. Presumably applies to the United States only, or perhaps US and western nations only.

3. Opinions are like a**holes. Everybody has one.

4. It's not surprising that a survey of economists would like an economic approach.

Cheers

Layzej said...

1) True, although if economists think that a modest tax could solve this issue then so much the better.

2) I suppose each nation and state will need to determine their own strategy. That is as it should be.

3) That is why we look to expert opinions.

4) I suppose not, but what is more important than the economy?

David Appell said...

David in Cal wrote:
"Suppose climate sensitivity were 3.0 deg C -- the midpoint of IPCC's highly likely range. From 1979 to today, CO2 in the atmosphere rose from about 335 to 405 ppm. http://www.esrl.noaa.gov/gmd/ccgg/trends So, temperature should have risen by 3*log(base 2) (405/335) = 1.04 deg C."

You continue to misunderstand and misrepresent climate sensitivity.

1) ECS is the warming AFTER EQUILIBRIUM HAS BEEN RESTORED. That won't be for millennia.

2) There are many other forcings in play besides CO2 -- such as methane, N2O, land use changes, black carbon, other GHGs, and (cooling) aerosols.

David Appell said...

David in Cal wrote:
"The test is how well the models predicted unknown future results."

Models cannot predict the future. Period.

Unless you can tell a modeler exactly what GHGs will be emitted between now and then, as a function of time, and what other GHGs, and what aerosols will be emitted, and where and when, and what volcanoes will erupt, and how big, and how the sun's irradiance will vary, a modeler can't predict the future.

Even if a modeler knew all this, they still couldn't predict the future, because no one know the initial state, not even today's. That would require knowing all air currents and all ocean currents as a function of f(x,y,z,t). We do not have sufficient measuring tools to do this; not even close.

Models can never predict. Ever. They can only project, given stated assumptions, And even then they are wrong, because they are only an approximate to reality. Always.

David Appell said...

David in Cal wrote:
"David -- maybe you and I are using terms differently. The following examples illustrate the point I was making about linear regression. The following three series are time series values corresponding to time = 1,2,3,4.5.

A. 5, 5, 5, 5, 5 Mean = 5; Slope of fitted line = 0;
B. 5, 5, 5, 4, 6 Mean = 5; Slope of fitted line = 0.1
C. 5. 5. 5. 6. 4 Mean = 5; Slope of fitted line = -0.1

Two observations:
1. Revising the values at times 4 and 5 doesn't change the mean, but the revision changes the slope.
2. The value at time 5 has a much bigger impact on the slope than the value at time 4."

I don't understand your point at all. In linear regression, all points are weighted in the same manner. No point has more a priori weight than any other, and certainly not the endpoints.

David Appell said...

David in Cal wrote:
"Layzej, the world is not now planning for a climate senstivity of 3."

The world isn't planning for ANY value of ECS.

It's planning according to the climate-carbon response (CCR), which Matthews et al 2009 put at 1.5 deg C/trillion tons of carbon emitted (+/- 1/3rd, about).

There might have been a slightly different CCR used at Paris, by a tenth of a degree C or two, but not by much.

David Appell said...

David in Cal wrote:
"I think a better estimate is simply the actual rate of warming since 1970 or so, when CO2 became a big factor."

This simply isn't true.

NOAA publishes the net GHG forcing (calculated) here, since 1979:

http://www.esrl.noaa.gov/gmd/aggi/aggi.html

It's quite close to a straight line. That fit says that GHG radiative forcing of 1970 was already 1.5 W/m2.

David Appell said...

And if you extrapolate that line back to 1970, about 60% of this radiative forcing was due to CO2.

David Appell said...

David in Cal wrote:
"A fixed tax amount on each person sufficient to stop the increase in atmospheric CO2 would mean starvation for most of the world's poor."

Nothing says the poor have to pay the same carbon tax as the rich. It would be unfair if that was so. The US has already had an outsized impact on global warming -- with about 2.5x the cumulative emissions of China, and about 10x on a per capita basis -- and should morally pay for that climate damage.

"A tax or other financial burden on the world's power suppliers would be passed on to the users, with a similar impact."

Yes, taxes get passed to consumers -- so what? A revenue-neutral carbon tax is decreasing emissions in British Columbia. It would do the same here.

David Appell said...

David in Cal:

Once again, you did not respond to the majority of my comments in reply.

So, no progress ever gets made.

From now on, I'm going to assume you agree with me if you do not respond to a comment I made in response to one of your's.

David