Those in denial of global warming and its danger have tried to make a big fuss over what they perceive as a lack of warming recently. How recently depends on just how much they’re willing avoid seeing.
Time and time again we’ve explained that natural variations can mask a trend on short time scales, and the time scales used to deny the danger of climate change are short. Time and time again we’ve illustrated that certain known factors (like the el Nino southern oscillation) have acted to do just that. For those in denial, any such explanation falls on deaf ears. But for the sake of those who are honestly wondering, let’s take a look at what climate data have revealed very recently. In particular, let’s look at the last decade (plus a couple of months) — let’s see what has happened since the start of 2003.
Such a short time span should be a denier’s dream! Ordinarily we wouldn’t expect to find significant climate-related changes over such a brief period. The usual definition of climate is the average and variation of weather over long periods of time, with “long” typically referring to at least 30 years. With only a third of that much time, it would be genuinely alarming to find multiple indicators of significant change. Even if global warming continues apace, we could well still find nothing notable at all. Ten years is just too little time.
Yet some parts of the earth are screaming so loudly that not even limiting time to ten years can drown out the message. Perhaps best-known is the sea ice in the Arctic. Its extent has declined so fast that even the last 10 years show a significant trend (because the annual cycle has changed so strongly, anomalies are computed using “since 2003” as a baseline):
If we focus on the month of September when Arctic sea ice reaches its annual seasonal minimum, we find a truly astounding result: a statistically significant downtrend in spite of only 10 years and a mere 10 data points:
As remarkable as that is, we find even more significant decline in sea ice volume rather than extent:
The annual seasonal minimum of sea ice volume has almost dropped off the face of the earth, reducing by 68% in just the last decade:
Even with just a decade of data, the changes we have wrought are evident.
One decade is also more than enough to see the continued change in sea level:
Despite the mistaken claims of some, the rate of sea level rise in this all-too-brief period is significantly higher than its average rate during the last century.
Sea level rise is principally for two reasons. First, the melting of land ice moves water from land to sea. Second, heating of the oceans themselves causes them to expand.
Speaking of ocean heating, we can see a significant trend in just 10 years of ocean heat content down to a depth of 700 meters:
The trend is even stronger (both in its size and its statistical significance) in a mere 8 years of data for ocean heat content to a depth of 2000 meters:
Data for northern hemisphere snow cover anomaly shows a slight decline since 2003, but it’s not statistically significant:
Looking at individual seasons, however, one season does show a statistically significant trend: summer (June-July-August). And it’s a whopper:
Summer 2012 had, on average, 42% less snow cover than summer 2003.
None of the five main global temperature data sets shows statistically significant trend since 2003. But if they’re adjusted to remove the influence of known extraneous factors — el Nino, aerosols, and solar variation — there is one which does show a statistically significant trend despite being limited to a mere ten years and two months. That’s the lower-troposphere data from UAH:
Its estimated trend since 2003 is 0.018 deg.C/yr, even higher than its estimated trend since these data begin in December 1978.
The plain truth is that in spite of using far too little time, the impact of man-made climate change is still clearly imprinted on the data. And that goes for the last decade, not just the decades preceding it.
The pity is that we have precious little time to take effective action to mitigate future changes. Too little time.
May I ask what the units are on the OHC graphs? Is it 10^21 Joules?
Thanks for cheering me up.
Apologies in advance if I’m mistaken; it’s late and I’m tired.
There appears to be a minor error in the labelling of the vertical axes of 4 of your graphs. The 2nd (NSIDC September), 4th (PIOMAS Annual Minimum), and 9th (Rutgers Global Snow Lab Summer) have the vertical axis labelled as being the anomaly, but the scale given corresponds to the actual value. Conversely, the 5th graph (U. Colorado) should be labelled as anomaly, but is not.
Sorry for the nitpicking, but accuracy and clarity are paramount, and even little mistakes like that could open an unfortunate door for deniers. I can already imagine what they would say: “look, the September sea ice extent, annual minimum sea ice volume, and summer snow cover extent anomalies are all positive, so we’re above the normal values, so now it’s just readjusting back to normal values as part of a natural cycle.” The thought of that makes me cringe and shudder.
That looks like it has to be OHC anomaly in 10^22 Joules. The anomaly baseline looks to be similar to Levitus, graph 2 here: http://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/
I am a non-professional following your blog and detailed explanations for some time now, enjoying what seems a growing understanding for the coherencies but frightened no less.
Before I saw the UAH lower troposphere data that significant summer trend made me think that we will be in for a significant “up” once those known other influences will reverse; to me it feels like prestressing the bow which might shoot us beyond any point-of-no-return in just one go while the world plays Euro or other red herrings.
“The pity is that we have precious little time to take effective action to mitigate future changes. Too little time.” – Now that, I certainly agree with. If only everyone can avoid panic until I finish preparing…
I think it might be useful to give a detailed explanation of how you work out how some of the above trends are statistically significant whereas others you refer to are not.
This is such a run of the mill matter in statistics that I think it tends to get overlooked, but it is obvious from reading comments after less informed blogs that many people haven’t a clue what it means, hence the confusion between “no significant warming” and “no warming” that is such a constant meme of denialism.
So true, so true… but I have one reservation: we actually have data for longer than 10 years. We can use it to look at the variability behaviour at time scales longer than 10 years, and take this into account. This would become significant for “red” data.
(I realize it may be tricky to remove the effect of the AGW trend before doing the statistical analysis on the real data.)
My question is: would synthetic data with the same statistical properties as the real data also for time scales longer than 10 years, behave in the same way, i.e., produce trends like the ones observed in less than 5% of cases?
In a way what you do is a sandbox experiment… we ignore everything else we also know ;-)
Actually i had the same question, but you seem to offer an answer : check that the trend is rare compared to other 10 years trends. But don’t you need a much longer period to do that ? if you want to check that it happens in less than 5% of cases, you need at least 20 * 10 = 200 years, and probably much more to have a good statistics.
Startling. Thanks, Tamino.
Why do so many people including you aggregate global temps at the monthly level? It seems to me that monthly is waaayyy to fine grained. Or why not daily? or even hourly? Personally annually seems the very finest one should aggregate and possibly 2-5 years would be better. This would seem especially true as the shorter the aggregation period the higher the autocorrelation would be expected to be in a system like global temps.
[Response: Monthly certainly doesn’t lose information compared to annual, and it does enable us to study seasonal variations.]
Why not hourly to study daily variations???
Some variations are completely irrelevant. At least that’s how I see it.
[Response: Don’t be so quick to dismiss variations as irrelevant. There have been notable seasonal variations in warming patterns (esp. winter vs summer) and there have been notable diurnal variations as well (esp night vs day).]
OK…But it is also true there is no such thing as global winter or summer nor global day or night.
I guess what you’re really looking at then is what happens as various ratios of land versus sea at various latitudes receive insolation at various times of the day/season/year. An interesting effect, certainly.
But for my part, at least, I see that far more as weather than climate in general (ice minimums being one exception, there are probably others).
Tamino, very nice. But I am afraid your point will be lost on anyone not interesting in getting it. And of course your whole point is that looking at 10-year trends is pretty silly.
To illustrate this silliness, just for fun I took at look at GISS global mean data, and calculated trends from 1900 to 2002 and then from 1900-2012. The trend in the first period is 0.071°C. The trend for the latter period is 0.085°C. Doesn’t that mean global warming has accelerated since 2003??! Yet the trend from 2003-2012 is 0.006°C/decade. Try explaining this apparent contradiction to the folks that claim there has been a “warming pause” (in a way that they can understand it, I mean).
My favorite is 1992-2007 (16 years). Cru4 gives .2996C per decade, which is why fake skeptics were screaming about model failure in 2007. What? They weren’t? Surely at least the inestimable David Rose and James Delingpole were trumpeting runaway warming at the time. No? Hrmmm . . .
I also note that the southern oceans (0-2000m) in particular don’t seem to be getting the whole “hiatus” thing. Must be underwater volcanoes, the UHI of Atlantis, or Tisdale’s hidden ENSO heat engine.
What is the criteria then, to positively identify a variable acceptable to present as significant and informative in a simple 10-year period, while declaring others as needing 17/20/30/100 years of data before any conclusions can be robust? I’m assuming it’s that certain variables have already been shown to be more or less impervious to natural/alternative variation, even in just 10 years..?
It depends on the relative magnitudes of signal and noise. Do this. Open Excel. Generate random numbers where the mean follows a trend, but add gaussian noise about that mean (a Box-Muller transform gives you Gaussian random numbers). Play around with the relative magnitudes of the trend (slope) and the noise. How often do you get fooled. Report back.
Wow, Tamino. I actually agree with you on just about everything, but this seems a bit like cherry-picking. Sea ice extent, sea ice volume, northern hemisphere snowfall, ocean heat content, and sea level: I would think these are fairly closely correlated in the real world, so showing each is a bit like showing the same thing over and over again. Why pick just the month of September, or just the summer, unless the trends more easily prove your point? This isn’t anywhere near what we get from the deniers, but please don’t stoop to their level.
One other point: you’ve adjusted global temperatures to account for known extraneous factors, which is great. Is there any chance someone could perform a similar adjustment of the other data above for these factors (for example, it seems like the sea ice extent trend would look a lot worse if you adjusted for the same extraneous factors).
Re. David C…
Your understnding of “cherrypicking” is limited, it appears.
There are some relationships here, but not so many as you appear to think. For example, what makes you think that the amounts of floating ice have anything whatever to do with sea level? Try the following experiment: put some ice cubes in a glass mostly full of water. Make a line where the water level is. Now wait for it to melt.
If your “cherrypicking” argument is that the oceans are heating and that that leads to sea level rise, well that’s true enough, but then that is hardly cherrypicking so much as multiple confirmation.
“Cherrypicking” is when you _limit_ the amount of data and analysis you report, not when you expand it.
Indeed AGW continues apace. Yet despite that, despite all the evidence presented in this post, the headline mean surface temperature is a bit difficult to ignore. Not the least of this is the volume of noise from them squawking skeptics. What I feel is long overdue is a more robust response to their tiresome squawks.
For instance, the denialists were more than eager to run with that quote from Hansen et al – Global Temperature Update Through 2012. “The 5-year running mean of global temperature has been flat for the past decade.” It of course made “Quote of the day” on Wattsupia.
And, of course, the 5-year running mean was indeed flat 2003-2010.
However, just like “no statistically significant warming in 15 years” doesn’t mean ‘no warming’ or even ‘no significant warming’, (Strange that all the words actually mean something and are there for a reason!), the 5-year running mean being flat over the last 10 years doesn’t mean temperature have been “flat” since the very beginning of that period.
While the 5-year averages for 2003-05 can be considered part of that “flat” as the record shows no subsequent rise, those years are also entirely consistent with being part of a strongly-rising 5-year rolling average 1976-2005. It is only with the 2006 5-year average that there is the slightest sign that there is some sort of “flat.” And the reason for that is solely due to the annual 2008 data. It is here, 5 years ago that the “flat” begins.
Indeed, if you analyse GISS monthly data using least-squares for different periods each starting Jan 1980 and ending at a particular end-month, the slope of that regression continues to increase as that end-month is moved forwards from 2000 and continues to do so up to mid-2007, as graphed here.
() This suggests that over the last 15 years global mean surface temperatures were continuing to warm at an increasing rate prior to mid-2007, which I would describe as significant.
() This suggests global mean surface temperatures were not flat over the last decade but were rising at an accelerating rate up to mid-2007 and it is only since then that they have been in any way “flat.”
If all are happy with talk of a 5 year “flat” in mean global surface temperatures (and why not), that begs two questions? Firstly the science – how do we attribute the post-2007 “flat”?
Secondly, when we hear all that denialist squawking, will there be an adequate response, a call for “Stone the crows!”
Salamano, if warming has paused for the past 15 years, then surely it has paused for the last 10. Otherwise people might be using the word in a way that the unsuspecting public would misunderstand. Nah, couldn’t be!
Only the UAH TLT shows a significant “non-pause” trend over the last 10. Tamino has shown that the forcing that is anthroprogenically derived continues through this period, but I still don’t see how you can say that the last 15 years (or 10 using a number of measures) does not show a ‘pausing’ in the resultant of all forces that represent themselves solely in the global temperature record. I understand this can be declared as mere semantics, but I think the distinction is legitimate.
Obviously we’re one el nino away from a serious bump in global temperature that will wipe away the pause, and fully expecting it to happen. Or, even the present suite of natural varation forcings that are currently counteracting the increasing anthroprogenic forcings have few known reasons to be counted upon to continue to ante with what man is making — but that still doesn’t eliminate the obvious situation of a present stagnation in temoerapture growth that just might dip below the confidence intervals on the IPCC forecast projections– even as we all know the net anthroprogenic forcings continue to rise.
You certainly have to permit yourself asome latitude to rethink the metric of contribueions if anthroprogenically derived forcings continue to rise but yet are continued to be matched by what is currently viewed as ‘natural variations’ and ‘natural forcings’, no? If so, how many years must this scenario persist before you think it’s scientifically appropriate to revisit this– never? I think it’s probably happening right now, and there’s more folks than you think that understand the difference between declaring a ‘pause’ in the global temperautre projection as a resultant of all forcings, vs a ‘pause’ in just the anthroprogenic component. Yes, there are certainly a number that will use the ‘pause’ in temperature to declare a ‘pause’ in AGW forcings, but if you were to name some of the ‘usual suspects’, I bet not all of them are actually in that camp.
Perhaps it’s also true that using the known continued progression of AGW forcings as a way to discount/ignore the relatively notable period of stagnant temperatures can also be a misrepresentation? Is this a case of no one wanting to cede any ground on the subject and hold toward their own semantics, (false equivalence aside)? I don’t think that acknowledging the pause threatens the legitimate existnce of the non-pause of AGW forcings, since there’s currently no scientific expectation that the pattern can continue.
See, the thing about noise is that while it may go down for awhile, it will eventually go up. Look at the factors considered in F&R 2011. While it could be conceivable that climate change could alter the balance of El Nino/La Nina, it can have no effect on volcanism or solar output. These latter two must return to their historical levels, and when the insolation increases, it will be augmented by the added feedback of all the ghgs we’ve dumped into the atmosphere.
And even if ENSO were altered by climate change, and that slowed warming, there is no comfort to be found there, as it means simply that more of the heat must go into the oceans (deep and shallow). Warming will still continue until we reach a temperature where we achieve energy balance at the top of the atmosphere (TOA). And in the mean time, we have added a shiteload of heat that must eventually work its way back out of the system–round and round and round she goes, where she stops, no one knows.
Nature is far too cruel a bitch for me to take comfort in unknown unknowns.
You have to understand that the deep ocean is fluid, very cold, and absorbing heat, which is causing a reorganization of mid and upper level currents. Wild annual and decadal fluctuations at the surface is what we would expect from this scenario, and that is what we are observing.
Deep ocean heat (or lack thereof) is the elephant in the room here.
with regard to Thomas Lee Elfritz’s elephant in the deep ocean – cold waters have surged up in la nino fashion over the past 10 years,and that elephant has been filling his bladder with the displacing warmer water. There’s been no stop to the warming over this span, it’s just not been measurable as increasing surface temp.
However, without doubt the arctic ice is melting, as we speak, in a great warming pool as the elephant does what elephants do.
With all due respect, El Nino, La Nina and ENSO are superficial surface temperature fluctuations transposed upon a deep ocean thermal reservoir. But I agree that melting arctic and antarctic ice is a wonderfully efficient manner in which to muck around with mid ocean currents entering and exiting a deep ocean thermal reservoir that is also steadily gaining energy because of the minor energy imbalance that is readily apparent at the surface. I’m surprised we haven’t seen this apparent minor cooling phase earlier. Don’t worry. It will pass. But it won’t be pretty when it happens.
If anyone is interested there is a recent paper which discusses some of these concepts, although for somewhat different circumstances than we are experiencing today. I’m not sure how to handle links here, though.
Rapid changes in meridional advection of Southern Ocean intermediate waters to the tropical Pacific during the last 30 kyr, L.D. Pena, S.L. Goldstein, S.R. Hemming, K.M. Jones, E. Calvo, C. Pelejero and I. Cacho, Earth and Planetary Science Letters, Volume 368, 15 April 2013, Pages 20–32.
Re: John G.
We’re on the same page here, I believe, but think about this a little more. Your experiment aside, it seems to me that sea ice extent and land ice extent in the northern hemisphere would be fairly well correlated, since most sea ice is in the northern hemisphere. If sea ice is melting due to higher global temperatures, land ice is probably doing the same. If there is less land ice, it seems to follow that sea levels would potentially rise. Am I missing something?
Your definition of cherry-picking is perfect: it’s when you limit data. That goes for when the deniers claim that the last few years of the land-ocean temperature record proves no AGW, but it also applies to graphs that just show one month or season of data for enhanced effect. This, I’m sure, was Tamino’s point: if they can cherry-pick, he can as well. I just would prefer our side stay above this kind of argument.
I would not consider a presentation of the summer minimum sea ice values to be “cherry-picking.” And the same goes for summer Northern Snow Cover. To consider the mounting speed of melting ice and snow in the Northern summer as cherry-picking is like considering a bullet wound in the body of an uninjured man as irrelevant being unrepresentative of his other more extensive injuries.
The summer Northern Snow Cover is actually more stark than graphed in the post above. It is more stark even than Arctic Sea Ice Volume. (The data in the graph here ends in at the end of July 2012.)
“…injured man,” surely…
Yep. “…injured…” He is not in good shape.
What you are missing in your allegations of cherrypicking is physics. Of course sea ice will form in the Arctic in the Winter. Warming doesn’t mean that some regions of Earth will not remain frigging cold. The physics says you expect to see the effect when the temperature is warmest in the Arctic–namely September.
I would also note that Tamino did show the full dataset–not just September extent–and that he has treated seasonal effects in detail in previous posts.
Perhaps you might want to familiarize yourself with all Tamino’s posts before you make accusations.
Thanks! This makes perfect sense. I simply assumed that warming would tend to affect each month in as similar way (for an overall downward trend), but you’re saying the physics suggests warming will actually cause faster melt and faster re-freezing as well; exactly what happened last year. I’ll look into the physics (and Tamino’s posts) a bit more.
No, not faster refreezing. It is just that when the sun vanishes below the horizon, it will get cold and ice will freeze. I strongly recommend looking at the previous posts.
I looked in a number of places, and unfortunately can’t find the “physics” you discuss. Tamino writes about arctic sea ice extent here: https://tamino.wordpress.com/2013/03/12/arctic-sea-ice-loss-part-1/
He shows September has the single highest average decline of any month. However, far from discussing the “physics” as to why this would be the case, he states the following:
“And there are some interesting changes in the timing and shape of the annual cycle. I don’t pretend to understand the physical reasons for these changes.”
So my question is this: if Tamino is emphasizing the trend in the single month showing the highest average decline, and there are no physical reasons or processes at work as to why September should be singled out as opposed to any other, how is this not cherry-picking? Can you (or anyone else) point me to the physics for using this month as opposed to others to show the extent of AGW? If so, I would greatly appreciate it. If not, my point remains: let’s discuss sea ice extent and other impacts of AGW in total (which show a trend that is frightening enough), instead of picking out one particular month for effect.
David C., so let me get this straight. You are looking for a reference that says ice melts when it’s warm and freezes when it’s cold?
David, think about this. Winter. The Arctic. Dark. Cold. Ice freezes.
Now it’s spring, but the Sun has just started to shine. There is thermal inertia. It takes time to warm things up. Long about April, May, things start to melt, slowly at first, but then things go more quickly. The peak melt will happen when the Arctic has had lots of time to warm up–the end of Summer or early Fall, before things start to cool down again.
By October, we’re back into the cooldown period.
The timing changes have to do with when the peak melting is occurring.
You say ” Can you (or anyone else) point me to the physics for using this month as opposed to others to show the extent of AGW?” Are you definite? Do you mean that?
Also, David, look at a map. You’ll see that the arctic sea is bounded by the northern edges of eurasia and north america. The sea freezes up to those boundaries and well, that’s all she wrote. Sea ice all along those boundaries can’t continue to grow once they hit land. Sure, the land gets cold, too, but it ain’t sea, and those areas which get deeply cold in winter ain’t sea ice, it’s ice-cold land.
There are, of course, holes in those boundaries because north america and eurasia don’t form a solid wall around the arctic sea. Just around about 90% of it.
That bounded area will continue to freeze in winter for a long time to come (hopefully, at least). The year-to-year differences, then, are almost entirely only in how far ice extends through the gaps between the two continents.
In summer, the ice is now melting within those boundaries formed by the the two continents. The only limit to melting is total disappearance of the ice.
Since geography limits winter growth of sea ice extent throughout so much of the arctic, extent during those months just isn’t as informative as the minimum extent in late summer.
Depends on what you mean by “peak melt.” For example, last year, day 133 began a streak of 14 days where the 60-day average daily melt for area was over 100k. That had not occurred once prior. I’d call “peak melt” the period from day 130 to day 180. Looking at daily STDs across the instrumental record for area, there is a rather sharply defined plateau the base of which starts on day 155 and finishes on day 310. The peak within the general plateau (the day that has seen the most change over the last 34 years) is rather pointedly day 287 (roughly the 13th of October). That’s also the day that shows the strongest negative 20-year linear trend (-152,472 km2 per year) for area.
Good timing–today my politically conservative business man brother sent his environmentally-concerned older scientist brother (me) a post about how there has not been any global warming in the last 20 years. I replied citing AO, and el Nino, melting ice, warming oceans, decadal temperature averages, etc., and then linked to this very timely post by Tamino. I only hope that he reads and is able to understand some of the message.
Really, only counting new annual records is a little like people who sell off stocks and mutual funds in their retirement accounts at the market bottom and wait for a new record stock high, before reinvesting.
Why only UAH?
I’ve had bad results lately with off-the-cuff responses to questions, but I really think it’s because RSS has been running cooler lately for some reason.
According to Woodfortrees all other temperature data sets (RSS, HadCrut3 and HadCrut4, GISS) except for UAH show negative trends (sure, unadjusted for ENSO) from 2003. That’s why I ask.
Oops… Sorry, my bad. I reread and there is a sentence “None of the five main global temperature data sets shows statistically significant trend since 2003.”
The trends since last night are insignificant as well.
…and it’s cold in my fridge, so there.
That’s not what I had in mind. I’m not as stupid as that. I’m just stupid enough to ask a question an answer to which was clearly in the post.
To the list, you could add the mass balances of Greenland ice sheet. Trend is very significant even on such time scale.
Hmm, has anyone used R Shiny http://www.rstudio.com/shiny/ for graphing climate data? That might help folks fiddle around with the data.
Global warming stopped in 2010. That is, at least until global temperature hit a new record high. You see, it’s a non-stop stopping.
You could certainly add the alpine glacier mass balance record to those that show a trend in the last ten years.
From one glacier walker to another: Yes you can….
Your response above is exactly what I was looking for; it makes perfect sense. Thanks, and I appreciate the time for the response. I’m still learning.
Dsl350’s point is a great one as well; perhaps October is the best month to use to show the downward trend.
Thanks for your analysis. I wish you had been one of my profs two years ago!