Death Spiral

We may have already passed the minimum in arctic sea ice extent for 2010, according to data from JAXA. The minimum this year (on Sept. 10) was 4,952,813 km^2.

That’s pretty close to my prediction of 4.78 million km^2, based on a simple trend analysis of September extent since the beginning of satellite observations. I was correct that 2010 came in 3rd-lowest overall, less than the value observed in 2009. Predictions that this year’s extent would be higher than 2009’s were nothing more than wishful thinking by those who want to deny the reality of global warming.

Here’s the past data, together with a quadratic trend (red line), my prediction (red dot), and the observed minimum (black “x”):

Anyone who isn’t blind can see that this year’s minimum extent falls very close to the existing trend. In fact let’s be very clear: this year showed a continuation of the existing trend. And the existing trend is: death spiral.

Anyone who calls it a “recovery” is lying, either to himself or to you, and is not to be trusted about matters relating to global warming.


92 responses to “Death Spiral

  1. Tamino, were your original data and estimate for IJIS daily minimum — which at the moment does appear to be 4.95 — or for NSIDC September mean? The NSIDC monthly mean is not set yet but might be lower than 4.95, because NSIDC smoothed daily values have been running lower than IJIS. The mean depends on what happens for the rest of the month, of course.

    [Response: They were for NSIDC September mean.]

  2. Good forecast Tamino! But you got lucky, the minimum extent could have been a whole lot lower had the dipole anomaly not broken down for most of July ;)

    I too suspect the final NSIDC min. will be lower than 4.95….we’ll have to wait patiently and see.

  3. Harald Korneliussen

    There should be a collection of the various short-term predictions by denialists. At time scales like this, by luck they will strike home from time to time, and then it’s important to show just how many bullets they fired at the target…

  4. You should put the linear trend on the graph too.

  5. William, a linear trend would still fall within the error bars.

    • Ben, a single point along the linear trend would still fall within the error bars. I take it this is what you mean? If Tamino is showing a quadratic rather than a linear trend you can be reasonably confident you are dealing with 95% confidence.

  6. Here is a graph comparing quadratic with linear forecasts, and also with the 1979-2009 mean which might stand for a “recovery” target.

  7. This year’s September average will lie below the linear trend.

  8. What about sea ice volume trends?

    • Keep an eye on:

      Arctic Sea Ice Volume Anomaly

      Current as of 2010-08-31. Of course the results of PIOMAS aren’t to be trusted, at least according to Goddard — since they didn’t show the same “recovery” since last year they didn’t show the same recovery that sea ice extent has.

      • Oops! Cognitive stutter. Or perhaps its something like that black cat in The Matrix.

      • Thats Straw Man fallacy, Timothy. He doesnt trust in PIOMAS because of the NAVY PIPS and ice concentration vs 2000, 1990 and 1980 for example. All of those including ice extent from the most recend years added to the Lindsay et al 2008 conclusions speak against PIOMAS.

        [Response: No they don’t.]

        Tamino, why did you choose JAXA? It’s results are the lowest for sept. For example DMI was around 5,2E6km2 being a total of 10% more than your forecast.

        [Response: DMI doesn’t define extent by 15% concentration but by 30% concentration — which is why their figures are so much higher. The 15% concentration is much more standard (it’s the same as used by NSIDC).

        JAXA is an easily available updated-daily figure based on 15% concentration. It’s also the same measure referred to by Steve Goddard in his forecast. And I made it clear that my forecast was for JAXA — if I’d been forecasting DMI I would have stated a different figure — so your “10% more than your forecast” statement is idiotic.]

      • PIPS2 is based on the thickest ice, not an average of ice thickness. This means it is useful to submarines and icebreakers, but not so much use if you want to know the total volume of ice.

        Seriously, “realist”: did you even try?

      • I had written:

        Current as of 2010-08-31. Of course the results of PIOMAS aren’t to be trusted, at least according to Goddard – since they didn’t show the same “recovery” since last year they didn’t show the same recovery that sea ice extent has.

        sir realist responded:

        Thats Straw Man fallacy, Timothy. He doesnt trust in PIOMAS because of the NAVY PIPS and ….

        I’m glad you called me on that. It wasn’t Stan but Oliver that I was thinking of:

        Obviously they have run their models post 2007, so why did they stop updating their verification? The image below gives a clue. I mapped NSIDC November extent (blue) on top of their verification graph, and something stands out like a sore thumb. Ice extent jumped back up after 2007, but their volume model didn’t.

        This in itself doesn’t prove anything, because it is possible that extent increased while volume decreased. (Not likely, but possible.)

        Does PIOMASS verify?
        Posted on May 28, 2010 by Anthony Watts
        Archive Copy:

  9. Thanks for coming back to this subject, Tamino. I hope you’ll keep doing that.

    I’ve linked to this article in my latest Sea ice extent update.

  10. My prediction was 4.9 mln as i remember (bet on Blackboard).

  11. During global cooling there will still be lots of ice loss and lots of examples of hottest years on record.

    • Daniel "The Yooper" Bailey

      Re: JCH

      “During global cooling there will still be lots of ice loss and lots of examples of hottest years on record.”

      It’s good to know we’ll still have functioning icemakers for our mixed drinks several thousand years from now. Because in a generation, year-round ice in the Arctic will only be a memory. The day may even come in a couple more generations that winter ice may be a no-show there. Many centuries must elapse for long-term sequestration of enough carbon to begin the long, slow decline in global temperatures.

      Of course, that assumes we finally stop emitting more carbon than the sinks can handle. Until then, it’s gonna keep warming up (especially once the methane hydrates/clathrates do their thing). A lot.

      The Yooper

  12. I am not sure that betting is such a good idea since AGW is such a concern. I for one like the post too but not the title “death sprial.” It certainly is not a good thing but that hardly means all is done with the world now either.

    • You may or may not already be aware of this, Jacob, but the “death spiral” title alludes to comments made a few years back by Mark Serreze, now Director of NSIDC. They’ve been the butt of some mockery from the denialosphere, but don’t look quite so mockable at the moment, as Tamino’s post testifies.

  13. Horatio Algeranon

    Anyone who calls it a “recovery” is lying…

    Sea ice is a little like the economy. If it drops enough in one year (2007 for sea ice and 2008 for economy), any positive change (no matter how small) will look like a “recovery”.

    • Well, if we ignore the ‘outlier’ years of 2006, 2004, 2003-1979 inclusive, 2005 and 2010, then there is clearly a strong recovery trend in progress. Indeed, extrapolation of this trend shows that by the year 3000, the entire surface area of the planet will be frozen.

      (I’ve heard that hanging around denialist websites can cause atrophy of numerical reasoning skills.. but I can’t see any evidence for that..)

  14. Talk about leaping to conclusions regarding this year’s minimum!

    (today’s JAXA extent has dropped to 4,941,094 km^2)

    The graph, with its two-day smoothing, doesn’t show the present being lower than the previous possible minimum, and of course that number above is preliminary and will be adjusted up or down a bit, perhaps even exceeding the minimum that triggered this post.

    But I couldn’t help tweaking our host just a bit! :)

    • Yes, the number of commenters reporting that number (myself included) on Neven’s sea ice blog strongly suggests that there was a moment of GCJD–that’s “global collective jaw drop”–just after 11 PM Eastern last night.

    • It’s back up to 4,948,438 after the daily adjustment.

      Still below the figure quoted by Tamino, though.

      Interestingly, NSIDC has reported the end of the melt season … prematurely? Their numbers show a much sharper uptick than JAXA, though …

      • NSIDC’s call is always provisional; there’s a caveat at the bottom of the page about that. Your point about the difference between JAXA and NSIDC numbers is interesting, though; I guess time will tell on that score.

  15. Dang, let’s try that link again:

    Neven’s sea ice blog

    I suspect I typed a “+” for an “=”. . .

  16. “Assuming that we have indeed reached the seasonal minimum extent, 2010 would have the shortest melt season in the satellite record, spanning 163 days between the seasonal maximum and minimum ice extents.”

  17. I learned a lot by pole watching this year, using a combination of a number of different daily extent charts (here, here, here and here), the north pole cam (a wicked cool feature of the 21st century), and most importantly the 30 day animations and side by side previous year comparisons at cryosphere today, along with the daily temperatures from AMSU.

    It was a rather sick, twisted form of daily entertainment.

    But the animations really teach a lot, particular how winds and currents affect the process, and how it involves a combination of motion and melting.

    One additional anomaly which I noticed by looking at the animations, which is not readily apparent by looking at mere extent graphs, is that the ice in the area between Greenland and Ostov thinned in a manner never seen before, even in 2007, with sea ice concentrations dropping below 60% in very large areas. Only 1984, 1991 and 1996 show thinning down to 80%… normally it’s at 95% plus.

    I have a niggling fear that this will mean something more dramatic for next year, but even if not, it’s another nasty sign that things are not just yo-yo-ing from winter to summer, but down right melting.

  18. Quality Hank, not quantity.

  19. Data Analysis of Recent Warming Pattern in the Arctic

    • Arctic Ice Part 2: A Review of Factors Contributing to the Recent Decline in Arctic Ice

      Arbitrary indices of Northern high latitude atmospheric patterns Golubeva 2009, Wu 2010 such as the North Atlantic Oscillation Strong 2010, the related annular Arctic Oscillation Ohashi 2010, Ogi 2010a, the Pacific-North American pattern L’Heureux 2008, and the recently prevalent Arctic Dipole pattern Wu 2006, Wang 2009, have been correlated to varying degrees with variations in Arctic ice extent. When looking over multidecadal periods the correlations weaken, as sea ice has continued to decline throughout regimes of positive and negative atmospheric patterns Cohen 2005, Overland 2005, Maslanik 2007a, Deser 2008, but some statistically significant links remain Strong 2009. Indeed year on year variations in wind speed correlate well with ice extent changes Ogi 2010b, but what about the longer term?
      Looking at the longer term larger picture, none of the factors examined here are truly independent of wider climate change or temperature rise. Arctic air temperature and Ocean temperature are driven by not only regional, – but through teleconnections and coupled atmospheric and oceanic transport mechanisms, global radiative forcings. Therefore it is likely that ongoing greenhouse gas driven increases in global heat content and transport will increasingly affect the Arctic climate, and based on current trends we run risks of losing the significant regulatory effect of the permanent Arctic Ice, with negative repercussions for the cold productive Arctic waters and regional ecology MacDonald 2010. Given that this process is already underway, we may well be witnessing the start of what in geological timescales would be viewed as a step function increase in atmospheric CO2, ice loss, freshwater release McPhee 2009, and permafrost melt, leading to an impulsive release of significant reservoirs of stored methane, all in a period covering a mere handful of human generations.

    • Hard to tell just from the abstract, but it sounds as if they go a “bridge too far” WRT causation: they establish association/correlation of warming with Arctic Oscillation, then argue that since AO is “recognized” as natural variability, then hey presto! human influence vanishes.

      But the possibility that warming affects AO incidence/configuration–analogously to the suspicions currently under investigation that changes in ENSO pattern might be caused/affected by anthropogenic warming–is not excluded. (Or, as far as I can tell, investigated at all.)

      And that’s presuming that their AO “data analysis” holds up, which is never a given in any particular paper.

    • Kevin, September 17, 2010 at 11:16 am

      On AO, some months ago read that it had an ”on the ”record” record index value of -3.4 but can’t seem to locate where that was. It mentioned “at least for the past 50 years” as in the record did not go further back. Looking forward to the post mortem assessment for September and what it has to say about the total of the melt season.

  20. IJIS preliminary extent for 9/16: 4,892,813 km2. Down 56k from yesterday. Of course, it’s subject to adjustment, but…once again, wow!

  21. Another 55K square km extent decrease has just been reported by IJIS. There’s still a lot of compaction potential in the pack and a patch of ice in the East Siberian Sea that might disappear altogether. Weather forecast is not looking too great, but two more days of extent decrease should be possible. Amazing melting season, this year.

  22. Bob, I’m struggling to find Ostov on a map. Do you mean ostrov Kolguev, (Kolguev Island, Russia)? Yes, ostrov means “island”.

    At first, I thought you meant the Nares Strait, between Greenland and Ellesmere Island. The sorry state of the ice there is explained by the missing ice arches – the strait is effectively unblocked.

    • Sorry, yes, I actually meant Ostrov Oktyabr’skoy Revolyutsii, Russia.

      I didn’t realize “Ostrov” simply meant Island, and I didn’t want to bother with the whole long nameski. My bad.

  23. Has anyone come across rational explanations for the higher Arctic Sea-Ice maxima we’ve seen since 2007?

    I ‘ve not been able to find anything rational about this (just denialists clogging up the web). My initial thought was that first year sea-ice has a faster growth rate than multiyear, but this can’t be correct as the loss of multi-year is in the Arctic Ocean, but the maximum extent is set in the regions outside the Arctic Ocean.

    These higher maxima also set the areal melt for the last few years as notably high – now I’m sure the excess of first year ice has a key role in that (melting in the Arctic Ocean – which is largely denuded of multi-year ice).

    I know this is a bit like pondering Christmas in June, but I was wondering…

    • Chris, I haven’t really read a serious analysis of this, but my impression that the issue is basically one of weather. For the times around the maxima of 2008 & 2009 we had La Nina conditions, and for both of those years the annual anomaly in the Arctic featured some sizeable areas of average-to-cooler-than-average values.

      For a discussion along these lines (but a couple of orders of magnitude better informed), I’d check out the updates NSIDC puts out for the maxima in question. Here’s the discussion of the 2008 maximum, to get you started.

    • I suggest googling “regression toward the mean” to understand the fake “recoveries” often trumpeted by denialists.

      Basically, any extreme value at variance with the long-term trend is more than likely to be followed by a less extreme one, as an inevitable characteristic of a Gaussian distribution.

  24. I was just thinking… IJIS showed extent bottoming out on 2010-09-10 at 4,952,813 km2 but extent started dropping again on 2010-09-14 and dropped below what we originally took to be the minimum by the next day. And we are still seeing it drop. This demonstrates how nowadays extent masks just how much ice loss there has been.

    This is something people should be reminded of when the minimum for a given year isn’t as low as the previous year or for that matter the record. Extent nowadays is much more of a function of the weather — or for that matter, which way the wind blows — because of how thin it has become.

  25. Anyone seen a volume figure later than 8/31? That’s the last one I see at

    • Hank, it seems from observing that the frequency is twice monthly and there being a good week delay before the next end and mid month chart shows.

  26. Kevin,
    thanks. No I was not aware of this. I am of course a fan of Tamino’s work after spending some time reading his blog. I just have a conservative view on how stories and titles should be presented, not to be confused with conservative denial of AGW and real consequences. The uninformed ‘environmentalists’ in my humble opinion do us more harm than those who deny AGW. In this case, however, it was I who was uninformed.

  27. Seems the melt has finished. Cryosphere Today have reported a few consecutive days of growth.

    (Up from minimum of 3.077 to 3.213 million sq/km on 16 September)

    But weather has been compacting the ice in the same period, leading to a current extent minimum for this year of 4.832 million sq/km, down another 50K sq/km from yesterday. (September 17)

  28. OTOH, it may be spreading, rather than compacting that accounts for the recent drop in extent. Dunno if it’s possible to check this in real-time on the intertubes.

  29. Hello Kevin,

    Thanks, I’ve read the NSIDC Sea Ice News pages, but note that on the page you linked to in the last paragraph it gives a nod to the fast time constant of thin ice growth. However from Cryosphere Today’s regional anomalies it’s apparent that in the Arctic Ocean itself anolamlies at maximum are zero – due to the geogrpahical limitations. This suggests to me that the substantial loss of multiyear ice cannot be a factor in the maximum.

    My thought had been weather but we’ve had three years maxima at around the same level and above the previous levels. Perhaps it’s just a coincidence. I just keep wondering if my sub-graduate physics means I’m missing a factor here. The area of melt for the last 3 years is greater than any year since 1979, however much of this is due to the low minima in the Arctic Ocean itself. As we move into the seasonally ice-free state we should see the amplitude of the annual swings grow.

    AFAIK there are no other source for sea-ice volume apart from the PIOMAS model. In terms of metrices that can be confirmed with measurements PIOMAS seems the most accurate sea-ice model available (IMHO).

  30. Does anyone know if CryoSat data will become publically available in the same way as the sources mentioned in this thread?

  31. > CryoSat data
    for Thomas–watch this site:

    Last news posted said in part:
    “… Around 150 scientists from about 40 research institutes now have access to the data. As part of the calibration and validation procedure, it is their job to help ensure these measurements meet the mission’s exacting standards before the data are released to the wider scientific community later this year….
    … the calibration and validation team play an important role by carefully checking the CryoSat data products before they are released to the scientific community” …
    … validation campaigns, in the Arctic and Antarctic, involve simultaneous measurements on the ground and from aircraft as CryoSat-2 orbits above. By comparing ground readings with airborne measurements and then the airborne measurements with those from the satellite, the accuracy of the data from CryoSat-2 can be assessed.
    … “… the next campaign will be particularly exciting as it’ll be the first opportunity we have to make a direct comparison between the satellite data and airborne measurements.”

  32. Current JAXA extent = 4,798,750 km2. Nice prediction, although I guess there might be a few days to go.

  33. Hmm, the minimum arctic ice extent will have to be rewritten:

  34. Current JAXA extent = 4,798,750 km2. Nice prediction, although I guess there might be a few days to go.

    Tamino’s prediction was for the September average, rather than the outright minimum. Looks like it will end up around 5 million km2 for the month.

    [Response: Not so. My prediction was based on September average data from NSIDC, but was specifically a prediction for the outright minimum from JAXA. They’re surprisingly close.]

  35. Good call Tamino – though possibly it could yet fall below your prediction!

    What’s Goddard’s latest thoughts on this – I recall he was going for >5 million sq km. Just wondered :)

    Cheers – John

  36. What’s Goddard’s latest thoughts on this – I recall he was going for >5 million sq km. Just wondered :)

    Goddard (with Watts) started the season claiming a “recovery” to roughly the 2006 level – 5.9 million km^2 (JAXA).

    Then when it became obvious the melt was very high, he made his “only numeric” prediction of 5.5 million km^2, claiming that “recovery to 2006” wasn’t a “numeric prediction”.

    And two or three weeks ago, he adjusted it again to 5.1 million km^2.

    I think we should hold him to his “recovery to 2006” proclamation …

  37. Since the 13th we have been losing roughly 42,000 Km2 per day with a loss of 43281 km2 occurring from the 17th to the 18th (the most recently updated date) (According to Jaxa). The sea ice extent area for September 18th was 4798750 km2 which is 90937 km2 above the 2008 minimum. Therefore if the sea ice extent losses continue over the next few days it is a certainty that we will be at the 2nd lowest ice extent in recorded history. Cryosphere Today shows an area of low concentration ice which has shifted away from the main pack (middle of screen) and has descended to lower latitudes near Russia.

    It looks as though the ice losses could continue with the piece of low concentration sea ice being extended to lower latitudes.

    Estimates of Sea ice for this year appear to be pretty accurate with Zhang and Gauthier et al. being the closest

    My Prediction will see the 2010 value be the SECOND lowest on record.

  38. SEARCH predictions are for average September extent, not minimum extent. (Except for Goddard’s late “public contribution” – he has made it clear on many occasions that he doesn’t understand the difference, and his SEARCH submission repeatedly mentions the minimum but never September.)

    Also, it is worth looking at how the predictions changed. Zhang is probably the best prediction, changing from 4.7 in June to 4.8 in August. Zhang started with an excellent prediction from a long way out, then improved it. Pokrovsky has the most improved prediction (5.5 to 4.9) – his technique is presumably better in short term forecasting, and his final estimate seems most likely to hit the nail on the head for September average.

    Morrison and Untersteiner get special mention for altering their bad prediction to a worse one (5.3 to 5.6).

  39. SEARCH predictions are for average September extent, not minimum extent. (Except for Goddard’s late “public contribution” – he has made it clear on many occasions that he doesn’t understand the difference, and his SEARCH submission repeatedly mentions the minimum but never September.)

    It’s better, even – his submission’s based on JAXA, not NSIDC (nevermind the average vs. minimum bit)

  40. but i thought NSIDC had confirmed the WUWT ice forecast?

    [Response: Just a reminder of what they were saying — In Watts’ own words:

    Prediction: Arctic Ice Will Continue to Recover This Summer
    Posted on February 9, 2010 by Anthony Watts

    Steven Goddard writes below that he agrees with the prediction I made in late 2009 that we’d see another 500,000 km2 of Arctic sea ice recovery in 2010. The Arctic Oscillation seems to be negative again, and according to NSIDC, this figures greatly in making thicker ice thus lowering summer losses. – Anthony


    • Thank you for the link. I love the fact that the charts are linked to the sites that show the current sea ice extents and that the pictures and descriptions do not match in the least.

  41. [Response: Not so. My prediction was based on September average data from NSIDC, but was specifically a prediction for the outright minimum from JAXA. They’re surprisingly close.]

    Ah yes, I see that after reading the original post again. But it is surprisingly easy to confuse the two if it’s not phrased carefully. Anyway – well predicted!

    Cryosphere Today and ROOS show growth in sea ice area over the last few days, but JAXA is showing melt for the same period, keeping pace with their extent data. With extent recently decreasing for all but ROOS, I’m less certain now that the melt has quite finished.

  42. Both NSIDC and Jaxa have been going down again and are showing new minima as of September 18. The NSIDC September 18 value looks to be about 4.6 million km^2 (JAXA a little higher as per usual).

    Bad news for Goddard.

  43. Like I stated in my last SIE update, the melting season looks to be over (I base this on the current weather forecast): 4,813,594 square km. I congratulate you on your prediction being so close, Tamino. It will do your reputation a lot of good (not that it needed it). I might do a post on how predictions played out and won’t forget to mention you.

  44. Hold on, I thought you were predicting IJIS absolute (daily) minimum, but your prediction of 4.78 million square km was for NSIDC mean September extent, right?

    [Response: My prediction was based on NSIDC mean September data, but was for IJIS absolute minimum. In other words, I used NSIDC September mean as a proxy for IJIS absolute minimum (because it gives me a longer time series). However, it could just as well be for NSIDC mean September extent; the two are quite close, much closer than the uncertainty range of the prediction.]

  45. Right, the JAXA daily numbers are somewhat higher than the NSIDC ones, and so (by coincidence, it seems) the NSIDC September mean number comes out very close to the JAXA daily minimum.

  46. Regarding Stoat, it’s a sawoff if the final NSIDC value is gte 4.74 million km^2 and lte 4.93 million km^2.

    Since the NSIDC started the month at about 5.1 million km^2 and was at about 4.6 million km^2 on the 18th, this looks to be a very likely outcome at this point.

  47. One of my commenters sent me an interesting image of sea ice concentration in 2010 and 2007. It appears 2010 had less high-concentration ice than 2007, and it’s probably thinner too, because 2007 was a much bigger compaction event.

  48. According to JAXA, new minimum on Sept. 18, and rising since:

    But according to NSIDC, Sept. 20, the latest date available, is the minimum so far. It looks to be just above 4.5M km^2, pretty much the the same as the 2008 minimum. (Still, the 2010 September average will very likely be above 2008).

    NSIDC also added this update:
    Update: 21 September 2010

    Although ice extent appeared to reach a minimum on September 10, rising afterwards for three straight days, it has subsequently declined even further. NSIDC scientists are closely monitoring the ice extent and will provide another update on the data, as conditions develop.

  49. I just posted this over at The Oil Drum. Thought readers here might be interested….

    As some bloggers continue to claim the Arctic is recovering, those with real interests are making plans. Russia is serious about the Arctic. From BBC News:

    This strange construction, part ship, part platform, is unique and lies at the heart of Russia’s grand ambitions for the Arctic.

    When it is completed in 2012, it will be the first of eight floating nuclear power stations which the government wants to place along Russia’s north coast, well within the Arctic Circle.

    The idea is the nuclear reactors will provide the power for Russia’s planned push to the North Pole.

  50. Mr. Goddard has extended his research into limnology and Robert Grumbine has a quick examination of the post here.

  51. Tamino

    Thanks for getting me ready for September, 2011. My 2011 forecast, based on your method, is here.

  52. Tamino and others,

    I’ve just posted a blog featuring Kelly’s forecast for the September 2011 SIE minimum and referring to this blog post as well: First forecasts.

    [Response: I think I was first to predict the 2011 value. My prediction was made at the end of this post, back in October of 2010:


  53. Well, I’ll be damned, so you did. Sorry, Tamino. I’ll write an update to that post.

  54. Done. Thanks, Tamino.