The numbers are in from the National Snow and Ice Data Center (NSIDC) for Arctic sea ice extent and area in September 2011. The average September extent came in at 4.61 million square kilometers. It was back in October of 2010 that I predicted this year’s September extent would be 4.63 million square kilometers. That prediction turned out to be right on the money.
Before you’re too heartily impressed, bear in mind that my prediction was 4.63 +/- 0.9 million km^2. It was a simple statistical prediction, based on continuing the existing trend into this year. Clearly this year’s evidence indicates: the trend continues. That’s why my prediction was correct. But the error range (the range in which we expected the result probably to fall) was sizeable. The reason my prediction was so close to the final result is: I got lucky.
Or maybe not. It’s good luck that the prediction was so close. It’s bad luck — for all of us — that the trend continues. It’s because of global warming, and that’s bad. For all of us.
In July I updated the forecast to include information from both the extent and area figures for June of this year. That yielded a prediction of 4.66 +/- 0.66 million km^2, not much different from the original prediction. In yet another update in August, a model using past September extent and this year’s extent from July as well as area from June and July predicted 4.22 +/- 0.62 million km^2. However, I also noted that I guessed the odds were about 50-50, and in comments a reader asked whether my actual expectation was therefore 4.3, to which I agreed. In either case the prediction was also correct (its error range included the final result) but didn’t turn out nearly as good as the original trend-only prediction either. That’s statistics for you.
Here’s the final result for this year, which takes over 2nd place for all-time low extent, together with predicted values (from continuing the trend) for the next three years:
The prediction for next year’s NSIDC September average is 4.43 +/- 0.89 million km^2.
This year did not set a new record, the 4.30 million km^2 September average (NSIDC) from 2007 is still the lowest yet observed. Neither did it set the record for daily minimum in the JAXA data set, for which this year came in 2nd as well. But according to data from the University of Bremen this year did break the 2007 record.
As for sea ice area rather than extent, 2011 did break the record according to data from Cryosphere today. Perhaps the most meaningful metric for global warming is sea ice volume, which according to PIOMAS also set a new record low.
In fact, in spite of the quite steady decline of Arctic sea ice extent and area, the volume has declined even more steadily because the ice is thinning as well as covering less area. The trend is evident from a graph of sea ice volume anomaly:
Perhaps most startling is that every day this year has shown a lower volume than the same day for any previous year:
One other fact deserves mention. As pointed out on RealClimate, Arctic sea ice is declining all throughout the year, not just during its summer minimum. This has a profound effect, especially during summer (June/July/August), the months of greatest solar insolation, because what used to be covered with highly reflective ice is now highly absorptive ocean. This increases the heat input to the Arctic (ice-albedo feedback), one of the reasons that the Arctic is warming so much faster than the globe as a whole.
Whatever month one chooses, the trend in Arctic sea ice is one of rapid decline. The latest data indicate that the trend continues. The reason is: global warming.
I had downloaded the extent data from NSIDC about two days ago, before they released their September value. The Jan-Aug time period saw its record in 2011, and with the updated extent the new record holder (of all years for Jan-Sep) is: 2011. But barely, it’s 0.05 million square kilometers less.
The October record for lowest extent is held by 2007 by a sizable amount (just under a whole million square kilometers), though it holds 3rd in November and December each.
I’m personally interested to see how 2011 fairs on an annual average basis.
This past month also set the second place record (albeit barely) for lowest 12-month period of average sea ice extent, the record being Nov 2006-Oct 2007.
A question if I may…
Do your error ranges for the extent predictions of the next 3 years take into consideration volcanic eruptions?
Katla in Iceland seems to be coming to life and if it matches past eruptions it could be quite intense.
I fully realize that even a major eruption would only have a temporary influence on the overall trend.
[Response: No, volcanic activity was not part of the model.]
Watching paint dry — except for the stakes
As anyone who has ever sat in a bar close to closing time will attest, watching ice melt is far more entertaining than watching paint dry.
It’s going to be very interesting to see how well the PIOMAS model fares when it is tested against the Cryosat data.
Thinking ahead – the lack of a similar trend in Antarctic is often explained by changes in wind circulation caused by the Ozone hole. Does anyone have an opinion on when the ozone hole is going to stabilise and we might expect sea-ice decline to kick in for Antarctica too?
A simple extrapolation of that trend doesn’t look physical, though – and not just because there isn’t any physics in it. How far into the future would you feel comfortable drawing it? Given the propensity of some fake sceptics to fit quadratics or higher order polynomials to things, extrapolating and then pretending there’s meaning there, you might want to give some caveats here.
I believe the same was true in 2010, all year. And, just possibly, in 2009. PIOMAS numbers are so alarming that I’m frankly ignoring them.
Out of curiosity – could sea ice retreat during the _winter_ be a negative feedback? Eg, no sun (so the albedo doesn’t matter) and it takes the insulating cover off the ocean so that more heat can radiate to space?
(of course, sea ice retreat during the winter is smaller than during the summer, so I expect that the net effect of annual sea ice retreat is still strongly positive. And of course, sea ice retreat at any season can have negative implications for ecosystems, coastal erosion, possible methane hydrate destabilization, and potential perturbations of northern hemispheric weather systems…)
[Response: I believe that yes, winter sea ice retreat is a negative feedback.]
Sea ice retreat during winter? To me this is the key. It indicates that something happens then.
The simplest explanation is the waters beneath the ice. They are, now, always warmer than than they were 30 years ago. So they allow just that bit less freezing or ‘strength’ of freezing. When warmer air and insolation come along in spring and summer, the not-as-strong-as-before ice is progressively more and more vulnerable to insolation and warmer air above and insolation heated waters at the edges.
And next year the same thing happens all over again. The summer weather may vary in its intensity and attack on the ice so the minimum reached each year varies up and down a bit – on the obvious trajectory. But all day every day, the water is just that little bit warmer than it was 20, 30, 50 years ago. .
(I’d dearly love a better word than ‘warmer’. I live in a place where it never, ever snows – in fact we’ve only once ever had a temperature marginally below 0C. My concept of ‘warmer’ is a very poor match to Arctic waters.)
Well, the surface air temps are warming during the winter, too. Less heat flux from ocean to atmosphere, presumably–unless the ocean is warming faster than the atmosphere, of course.
Or the causality could run the other way–winter air temps could be warming (wholly or in part) because thinner ice and warmer water mean *more* heat flux to the atmosphere.
I suppose reanalysis numbers should be able to shed some light on–(no, I’ll avoid the obvious pun!)
It’s interesting to take the daily ice volume data from PIOMAS, aggregate it to monthly, and then divide it by the monthly sea ice area data from NSIDC to produce a monthly mean arctic sea ice thickness data set.
When you do this you can see that winter ice thickness has been declining very steadily since 1979 (the summer thickness is noisier, but still exhibits a fairly steady decline). This is consistent with the observations that the arctic ice pack is becoming progressively more dominated by first-year ice.
It’s also interesting to look at the progression of annual cycle of ice thickness in a manner similar to the JAXA plot or tamino’s last PIOMAS plot. For the first 30 years there was a regular annual ice thickness pattern that has broken down almost completely in the last 3 years or so. I’m not sure what this means for the future of ice in the arctic but it’s rather striking.
My predictions for NSIDC sea ice extent for the next three years are:
prediction for 2012 = 4.563014 (+/- 1.033720)
prediction for 2013 = 4.396993 (+/- 1.059939)
prediction for 2014 = 4.228526 (+/- 1.090855)
Produced by a different method, but very similar. Not a serious analysis, just the result of brief experimentation with a statistical approach (Gaussian process), so YMMV.
For PIOMASS, the predictions are rather uncertain:
prediction for 2012 = 4.342434 (+/- 2.071584) 10^3km^3
prediction for 2013 = 3.306950 (+/- 2.227936) 10^3km^3
prediction for 2014 = 2.211567 (+/- 2.426253) 10^3km^3
The data are gloomy enough, no real need for a model to suggest what is likely to happen next year.
prediction for 2014 = 4.228526 (+/- 1.090855)
prediction for 2014 = 2.211567 (+/- 2.426253) 10^3km^3
I don’t think I’m going out on a limb here, but your volume prediction for 2014 includes zero, but your extent prediction doesn’t. Which doesn’t make sense to me……
Well, it’s not a physical forecast, but a purely statistical one. Nothing says that volume and extent need to have the same relationship throughout the evolution of the ice state–we just know, as your comment points up, that they must intersect zero at the same time.
The two predictions are based on two different datasets, with different sets of assumptions, with no attempt to enforce any kind of consistency. I suspect if you tried to forecast extent based on volume, as a purely statistical exercise, again you might get extent but no volume or vice versa or other such wierdness. Neither model has any real basis in physics (I find it much easier to have confidence in a physical model than a statistical model – but perhaps that is because I work in a branch of statistics ;o). Tamino’s model is likely to be more reliable, as unlike me, he has (AFAICS) a solid background in time-series analysis.
“No, volcanic activity was not part of the model.”
Well, to the extent that the statistics for the past 20 years include volcanoes (both global – Pinatubo, Chichon) and Icelandic (the unpronounceable volcano from last year) then as long as upcoming volcanoes are similar to those past ones, those volcanoes shouldn’t put next year’s ice out of the error bounds by itself.
Missed a perfect opportunity to gloat again over Anthony Watt’s abyssmal ice forecast.
Science isnt supposed to be about gloating… which is why some previously-mentioned people are fake scientists.
The so-called “warmists” have really thrashed the so-called “skeptics” in terms of Arctic sea ice extent predictions the past two years.
‘The trend continues’ – note the nonlinearity of said ‘trend’.
This is going to a catastrophe, I mean this in the physics-mathematics sense. At some point soon, certainly much sooner than curved trend suggests, summer ice in the region will have disappeared. I reckon: any year now. There is a threshold average thickness which I estimate to be about two feet and I think we are there right now.
@ cRR Kampen
I posted this also over at Skeptical Science a bit ago:
Eisenman and Wettlaufer: “Nonlinear threshold behavior during the loss of Arctic sea ice” (PNAS 2009)
H/T to Artful Dodger
Thank you, Daniel!
AMSR-E is broken!
Ramifications for dataset continuity/quality etc. Grrr…
In the presence of overwhelming evidence of declining sea ice in the Arctic, Watts and Goddard are making a lot of the following chart:
In the last two years 1 year old sea ice has made up less of the total than in previous years while 2 and 3 year ice has increased as a proportion. Unless all of this multi-year ice melts out in the next season, it is likely that 3 year ice will turn into 4 years and 4 years into 5 years and thus multi-year ice is likely to increase as a proportion of the whole (that is of a declining total, of course).
This seems to be caused by distribution of the multi-year ice which is largely clinging to the north Canadian archipelago while the biggest loss in ice area has been in 1 year ice that in the last 3 years has been melting (more) and reforming (less) in the open ocean near Siberia.
At some point, though, there will be another “bad weather” (i.e. warm) year like 2007 which will see a catastrophic loss of multi-year ice. But until then I can see WUWT and Goddard continuing to use this graph to “prove” that multi-year ice is somehow increasing.
Is there a graph that shows 1, 2, 3, 4 and 5 year ice area in absolute terms rather than in relative proportions? That would be a more representative indicator than the NISDC graph which I think is a bit misleading.
Just to save anyone else the trouble, I manually re-created the “September Ice Age 1983-2011” figure. Here is the resulting table (all values in %):
Year 5+ 4 3 2 1
1983 33 16 13.5 14 23.5
1984 39 13.5 12.5 13 22
1985 42.5 13 11 16 17.5
1986 41 11 13 10.5 24.5
1987 32 14.5 11.75 16.5 25.25
1988 32 10.5 13.5 13 31
1989 26 11.5 10 18 34.5
1990 25.5 9.75 15.5 21 28.25
1991 23 13 17 15.5 31.5
1992 21 12 11.5 19.5 36
1993 22 10 15.75 20 32.25
1994 16 12 15.5 15 41.5
1995 20.5 11 13 25 30.5
1996 16.5 8.5 14 11 50
1997 17.5 11.5 10 28 33
1998 19 7.5 19.5 16.5 37.5
1999 20.5 17 14 20 28.5
2000 23 11 14 13 39
2001 19 10.5 10 19.5 41
2002 19 9 15.75 23 33.25
2003 18.5 10.75 15 16 39.75
2004 17 10.75 11 20 41.25
2005 17 8 15.5 21.5 38
2006 13 8 14 18 47
2007 13 8.5 11.5 24.5 42.5
2008 8.5 5 9 10.5 67
2009 8 4 5 20.5 62.5
2010 5 3.5 7.5 30 54
2011 4 2 17 27.5 49.5
If you multiply these by the September ice area data at NSIDC (which I assume is the appropriate variable, but it might not be – someone correct me if I’m wrong, especially if it’s supposed to be % of ice volume) you will see that one year ice has been increasing slowly (but with lots of year-to-year variation), two year ice has been very flat (with noise), while 3, 4, and 5+ year ice has declined steadily. The noise-to-signal ratio (eyeballed) drops noticeably one the ice gets to 3-years of age or more.
You’ll also notice that big changes in 1-year old ice track through to later years. For example, a big spike in 1996 1-year ice can be seen in 1997 2-year ice, 1998, 3-year, … to 2000 5+-year ice.
1-year ice also increased in 2008 and 2009 after the 2007 low and this is reflected in the 2009-10 2-year ice and 2010-11 3-year ice. This could continue into the 4- and 5-year ice over the next couple years but 1-year ice has come down a lot in the last 2 years so it’s probably only a temporary blip.
We’ll see I guess.
Also, for what it’s worth, based on this analysis:
1983, 5+ year September ice area = 1.53 million km2
2011, 5+ year September ice area = 0.116 million km2
i dont know if you re read old posts, but what doe you think/make of current NSDIC’s graphs of extent. For me, they seem to indicate that the ice extent is quite normal , compared to the 1978-2008 range
Looking at the total Arctic sea-ice extent at this time of year tells you nothing about the expected extent in the summer. If anything there is a fractionally negative correlation over the last decade between sea-ice extent on April 20th and the eventual minimum extent later in the year. But this won’t stop certain people getting excited by it.
True, as far as it goes. Of course, not quite two months ago NSIDC extent was just as low as in 2007, so as far as it goes isn’t all that far. It’s not as though the new ice that froze in the last couple of months is going to be as thick as the ice in the same area in a more or less average year from 1978-2008.
Well, let’s think about this. We are still early in Spring, so it still gets cold, right? Even a relatively short cold spell will cause ice to form on open water–thin, rotten ice, but ice nonetheless. And then it gets warmer, ant that ice is gone. So, maybe extent isn’t that informative in the early Spring unless you are interested either in weather or deception.
I think the best description is Hank Roberts’–we had a somewhat similar scenario last spring, and he commented (paraphrased):
When an approach to normal becomes worthy of remark, that’s telling. Logic says that the ice formed late in the season must necessarily be thin and briny (since the salt rejection process is not that rapid.) This ice won’t be durable. How high or low the minimum turns out to be this year is in the lap of the weather gods, as always. Nevertheless, there’s no reason to think the ‘death spiral’ is being seriously interrupted.
> NSDIC’s graphs of extent.
you know the definition for “extent” they use, right?
Are you looking at the picture, or at the data?
“… NSIDC has updated our processing of the daily sea ice extent graph. …
… The value of the trailing mean lags the actual data values, so sea ice values will appear lower when ice extent is increasing, but will appear larger when ice is decreasing. …
… Sea ice data processing methods are described in detail in the Sea Ice Index Documentation.”
Damn, WordPress improved that link so it’s broken.
Trying again: Sea Ice Index Documentation
dagnabbit, I hate software that tries to be clever.
Sea Ice Index Documentation
And that doesnt’ work either. I dunno. Use Google watch the pretty ads tapdancing and farting across the screen and they’ll find it for you eventually.