Last year (2017) will not be the hottest year on record; it’s likely to be 2nd-hottest. But it will be the hottest which was not enhanced by el Niño conditions.
El Niño is the “warm phase” of a natural oscillation of wind patterns over the Pacific ocean, which increases the amount of heat transferred from ocean to atmosphere. When that happens, there’s more heat in the atmosphere so surface temperatures increase (actually, temperature of the air near the surface, which is what matters for land-based living things). When El Niño subsides, the warming subsides too. The opposite face of the coin is la Niña, when heat tends to go from atmosphere to ocean and surface temperatures tend to be cooler. Together, they make up the el Niño southern oscillation (ENSO).
Those effects are only temporary because they don’t represent more or less heat accumulating on Earth, just the re-distribution of heat that’s already here. It’s one of the factors behind all the fluctuations in global temperature, those never-ending ups and downs and downs and ups that make global temperature wiggle around a lot, but never really get anywhere — el Niño doesn’t cause a trend in temperature, just fluctuations.
Those fluctuations can be considerable, which is why record-hot years tend to happen when a strong el Niño arrives (actually, shortly after — the global warming effect lags behind the el Niño itself). Back in 1998 Earth set a new record for hottest year globally, following the monster el Niño of 1997-1998. Likewise, 2016 smashed the hottest-year record after 2015-2016 brought another strong el Niño.
El Niño isn’t the only thing which causes fluctuations but not trends in global temperature. The best-known example is prbably big volcanic eruptions, they can throw a lot of junk into the atmosphere, including sulfates which end up as sulfate aerosols, which tend to scatter incoming sunlight back to space, reducing the incoming heat and therefore cooling Earth. If we want to know how Earth’s temperature is changing apart from those fluctuation factors, we can estimate how strongly they affect temperature, then remove their estimated influence. Let’s do so, using the method of Foster & Rahmstorf.
I’ll use the global temperature data from NASA GISS (the Goddard Institute for Space Studies) from 1970 to the present. Here’s the original data in black, and the “adjusted” data (with known fluctuation factors removed) in red:
These are monthly data; here are annual averages (again, original in black, adjusted in red):
The final year (2017) isn’t complete, because the numbers for December aren’t in yet, but the average based on 11 months’ data rather than 12 is still sure to be quite close.
There’s still fluctuation; we don’t know all the fluctuation factors and probably never will, and our estimates of the influence of the factors we do know about is bound to be imperfect. But we’ve accounted for the biggest known factors, making whatever trend exists clearer. Two of the biggest differences between original and adjusted data are in 1998 and 2016, which is no surprise; those years were affected not just by el Niño, but by big el Niño events.
Let’s look at just the adjusted data:
From this it’s clear that 2017 is destined to be the hottest year on record after the estimated influence of el Niño is removed. Not by a little. We don’t have to remove fluctuating factors to demonstrate that global warming didn’t “pause” or “slow” recently, statistics alone is enough to do that — but the adjusted data make it not just mathematically clear, but visually as well.
The most worrisome aspect is just how hot 2017 was even without a heat boost from el Niño. This will surely bother the climate deniers who have been trying so hard to blame 2016’s record heat on el Niño. It’s obvious that 2017’s extreme heat isn’t “just because of el Niño” — it’s in spite of the lack of el Niño.
But the reason it’s so worrisome isn’t that it will annoy climate deniers (that’s just an added bonus). The problem it highlights is that global warming continues apace, that there’s no valid evidence (and never has been) of its abating. The problems climate change brings are going to get worse. And worse.
While there’s no evidence (and never really was) of a slowdown in global warming, what about a speed-up? There are many ways to look for it, probably the simplest being to fit a curve to the data, a quadratic curve which, unlike a straight line, allows for a change in the rate of increase. Here’s such a fit to the data:
The p-value for the quadratic term is 0.046, so we could claim that there is statistically significant evidence of a change in the global warming rate since 1970 — not a “slowdown” or “pause” as climate deniers claim, but acceleration.
Personally, I’m not convinced. The test is close enough to the “significance level” of 0.05 that maybe the fit is accidental. There are many sources of uncertainty in such a calculation, more than are accounted for so I believe the uncertainty is bigger than what the bare numbers show. And the time at which such a study should start can affect the results — should we really be talking about “since 1975” or “since 1980” rather than “since 1970”? The test is so close, the caveats so many, and it depends on removing the influence of fluctuations when we know the removal to be imperfect, the method used to account for autocorrelation is only one of many and is itself uncertain, that I’m not ready to declare acceleration yet. But there is at least some evidence of it.
If the quadratic trend so estimated persists throughout the century, by 2100 we’ll be warming at 0.036 deg.C/yr, about twice as fast as we are now. But, to my mind, it’s folly to forecast far into the future by extrapolating a simple curve fit. Simple estimates like that are fine for the very short term — but expecting them to persist, in this case much longer than the data on which the curve is based, is “simple” minded.
I’ll close by showing the adjusted data together with a smooth fit, in a form I find both entertaining and informative:
The take-home message is clear: global warming continues, 2017 reinforces that conclusion strongly.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.
Open Mind 
It’s cold in the Midwest, so deniers are doing just dandy. Sad but true.
Excellent article. How warm would 2018 have to be to put more curve in the curve? Would an annual GISS mean of 83 ℃ strengthen the acceleration?
Why start in 1970?
[Response: It is approximately the time at which the rate of global temperature change increased, rather dramatically from no increase or even slight decline, to the rapid warming witnessed over the last 40 years or so. The time at which the transition occurred is uncertain, which is why I specifically mentioned my doubts about what was the best choice of starting time.]
Curvature signal to noise ratio should increase at the forth power of the length of observation. Hence, if the signal is real it should be obvious within 5 years,
Does the curvature signal to noise ratio always increase at the fourth power? I am especially interested in the rate of sea level increase (although temperature is the final data point).
I get the feeling I missed some humour/reference. Why is the last graph with red, blue and yellow error bars ‘entertaining’ as well as imformative?
People still in denial about global warming are really pissing me off these days. Not even admitting there is a problem at this stage just makes my blood boil. It amounts to ‘please have your world fucked up because some people are bloody-minded’.
On the other hand sensible people (which is pretty much everyone in charge, except the embarrassment which is the US govt) can pretty much ignore those people now and just get on with it, and I see increasing signs of that.
What happened in 1974-1976? Volcano?
[Response: Explosion of the Mt. Agung volcano in Indonesia.]
“Explosion of the Mt. Agung volcano”
In ’74-76? Not that I can remember. 1963, yes.
There were three VEI index 4 eruptions over the period ’74-76, which might have done it. Mt St. Helens in 1980 was a 5, and El Chichon in 2982 was a 5 (with a couple of other 4s in 1981).
https://en.wikipedia.org/wiki/List_of_large_volcanic_eruptions_of_the_20th_century
[Response: I stand corrected. I hope it teaches me not to rely on faulty memory.
Looking at MEI (the multivariate el Niño index), there was reasonably strong and quite long-lived la Niña conditions at that time.]
I just noticed a typo in my response above: El Chichon was 1982,not 2982…
When I looked up the VEI, I was a bit surprised to see Mt. St. Helens and El Chichon rated the same. Mt. St. Helens did not (IIRC, although memories can be faulty!) eject as much into the stratosphere, as it blew out sideways. I also didn’t remember any of the 4-rated ones in the 1970s and 1980s, even though I was doing undergrad and grad work in climatology during the period… Agung was the biggest in the recent record at the time.
Tamino, as you say, ENSO, decadal changes in solar activity and the like have caused temperature fluctuation but no trend. What are the strengths and weaknesses of simply averaging GISTEMP anomaly numbers across rolling 10 year periods (say)? Such a plot appears to be highly linear, with only the heaviest ENSO periods visible as minor deviations.
Such a plot appears to be highly liner since ~1970…
I did just that recently, using a retrospective 10-mean and comparing early and recent HadCRUT4 data. For the recent period, it is indeed pretty linear. http://i1108.photobucket.com/albums/h402/brassdoc/Early%20vs.%20modern%20HadCRUT4%20raw%20%2010-year%20retrospective.png
El Niño increases the amount of heat transferred from ocean to atmosphere. When that happens, there’s more heat in the atmosphere so surface temperatures increase
Those effects are only temporary because they don’t represent more or less heat accumulating on Earth
When El Niño subsides, the warming subsides too.
–
This has always been one of my comprehension difficulties so thanks for making it a little clearer but more difficult. You do however seem to suggest that the temperature of the atmosphere is not a reliable measure of global heat increase while at the same time saying it is [as you point out the upward trend].
–
Last year (2017) will be the hottest which was not enhanced by el Niño conditions.
Yet— the global warming effect lags behind the el Niño itself). So?
–
If I am allowed to comment on the significance of a non El Nino year, I would point out that we had 3 years in which close to El Nino conditions were active, followed by a dip or short La Nina and then a return for at least 5 months of 2017 of pre El Nino conditions.
There is a big difference between neutral years, average of temps close to zero anomaly with no preceding El Nino and ones which had a large temperature burden [global warming effect lags] carrying over into the year.
Given this this year should prove your narrative, or give pause for a small rethink.
Thanks for making the El Nino effect more intuitive.
Tamino, what would happen if we were to apply the Foster &Rahmstorf method for filtering out the effect of GHGs? Would we expect a line with a slope not significantly different from zero?
[Response: Perhaps. One might also use climate forcing rather than just GHGs, but if you do that bear in mind that some of the exogenous factors (volcanic aerosols e.g.) are already included in estimates of climate forcing. I also caution that climate forcing estimates are less precise the further back in time you go.
If so, I would caution against regarding it as “proof” of man-made climate change. It would only establish that the trend in global temperature was correlated with the increase of GHGs, and correlation is not causation. But it would be evidence.]
Tamino, could you publish the R code for the figures, especially Fig. 4?
[Response: R code for the method of Foster & Rahmstorf is here. As for the figures … if you use R, you should be able to handle that yourself.]