Ice Free

Arctic sea ice has been declining so fast that most people in the know expect to see ice-free conditions in the Arctic ocean at its annual minimum before very long. The natural question is, how long?

The description “ice-free” doesn’t necessarily refer to the absence of any ice anywhere. There will likely be small pockets of ice near coastal boundaries, in particular the northernmost parts of Greenland and the Canadian Arctic Archipelago, which will persist even when the Arctic is essentially ice-free. Many researchers have suggested that when the extent of Arctic sea ice at end-of-summer dips as low as one million square kilometers, we’ll effectively have reached “ice-free” conditions.

So … when might that happen? I don’t know the physics of sea ice sufficiently well to make a useful prediction, but so far physical models haven’t done a very good job anticipating Arctic sea ice loss anyway — it has vanished faster than predicted. But we can look at the results of purely statistical predictions, extrapolating existing trends into the future, so long as we keep in mind that extrapolating trends very far into the future cannot be considered reliable. They may well give us some insight into how fast things might happen, but betting on the outcome based on long-term statistical extrapolation is, one might say, a “fool’s errand.” Let’s look at such extrapolations anyway, not to make rigid claims about what will happen, but just to explore the possibilities.

There are three common measures of Arctic sea ice: extent, the area of ocean with at least 15% ice cover; area, the area of ocean covered by ice; and volume, the volume of northern sea ice. I’ll focus on two of those measure, area (from Cryosphere Today) and volume (from PIOMAS). As base data, I took daily estimates of area and volume and determined the annual minimum values for each year.

This enables me to compute some derived statistics, namely latitude, which I’ll define as the latitude to which the ice would extend if it were all northward of a given limiting latitude, and thickness, which I’ll define as the annual minimum volume divided by the annual minimum area.

Without further ado, here’s the area data (annual minimum):

I’ve added a lowess smooth, and as many of you already know my program enables me to record the velocity, i.e. the time rate of change at each moment. If we use the velocity at the most recent year to extend the curve into the future, then it will hit zero in about 15 years. It will reach 1 million km^2 (the “essentially ice-free” level) in about 9 years.

Here’s the limiting latitude over time:

Again using the most recent velocity to extrapolate to the future, it will reach latitude 90N (the pole) in about 20 years. It will reach latitude 84.485N (corresponding to an area of 1 million km^2) in about 8 years.

Here’s the data for annual minimum ice volume:

Volume has been declining more precipitously than area (or extent), so extrapolating the present trend into the future predicts total sea ice loss sooner than using either area or latitude. The extrapolation hits zero in just 5 years:

Volume is equal to area times mean thickness. Hence if volume has decreased faster than area, then thickness must also have declined. Here’s the thickness itself, defined as annual minimum volume divided by annual minimum area (an imperfect but still good definition):

We can see that since 1979 (the satellite era for sea ice data), both area and thickness have reduced to half their starting values, so volume is now only one fourth what it was in 1979. One fourth! That is, quite simply, amazing. Astounding. Anybody who claims that this is just a “natural variation” is full of something other than ice.

Extrapolating the thickness into the future hits the zero point in about 7 years. But I don’t trust this extrapolation at all — not even a little bit. For one thing, for thickness the extrapolation is overly sensitive to the “smoothing time scale” of the smoothed fit. For another thing, thickness declined more slowly until 2009, but in 2010 it “fell off a cliff.” But it hasn’t continued to decline since then (albeit a very brief interval), it has held steady at about 1.5 meters. And, I’d guess that there are fundamental issues of mechanical, structural integrity which come into play when sea ice gets that thin. My guess: thickness may hold steady at this value for many years to come, or it may continue to decline, or we might even see another sudden and dramatic decrease — which is the same as saying that I have no confidence in any prediction of future thickness.

We can see the interplay between changes in area and thickness on a scatterplot of the two:

I’ve labelled a few of the years, including the huge decreases in area during 2007 and 2012, and the big drop in thickness in 2010. It’s interesting that those two changes (large area and thickness decreases) don’t occur together.

So … what’s the forecast? There isn’t one. But I will give my impression of the collection of “at the present rate” scenarios. I emphasize this because in the past, when scientists have said “at the present rate,” certain unscrupulous individuals (like Anthony Watts) have claimed they’re actual predictions and used them for ridicule. That’s basically, fundamentally dishonest.

At the present rate, the different variables indicate an essentially ice-free Arctic in about 8 or 9 years. That’s not the distant future. It’s close to the end of this decade.

It will have consequences — it already has. Loss of sea ice leads to more absorption of solar energy (the “ice albedo” effect). It alters large-scale patterns of circulation in the atmosphere. It affects water vapor content in the far north. It influences our weather, especially in winter, and may well be responsible for changes in wintertime snowfall in the northern hemisphere. Further ice loss — and that’s a prediction — will have further consequences. We won’t like them.

Here’s another prediction: even as sea ice continues to decline, fake “skeptics” will call every little random gain a “recovery,” and every further loss “natural variation.” That’s because they’re in denial.


41 responses to “Ice Free

  1. Horatio Algeranon

    “Eyes Free”
    — by Horatio Algeranon

    Down means “denial”
    Up means “upturn”
    Sea-ice “skeptics”
    Will never learn.

  2. To me, volume is the most important measurement. How widely/thinly the ice might be spread at the start of the year might determine the day on which the ice melts out during the summer thaw but it’s volume that gets melted by the incoming and trapped heat. (Widely spread ice does increase albedo while the ice lasts.)

    If one looks at the annual PIOMAS volume minimum we’re three years from the first summer meltout, assuming a smooth line decline. But look back at some of the annual events. We see two times in the last six years in which massive drops in volume occurred. If the summer of 2013 is like either of them then it happens next year.

    • Indeed.

      The big year-on-year drops in PIOMAS (2006-7 and 2009-10) are just over 2500 km3 and last year’s minimum was c. 3400. So just one of those ‘jumps’ that we have already seen twice in the past decade puts us under 1000km3.

      Has to be pointed out that a volume of 1000km3, given a thickness of 1m and a concentration of 33%, would give an extent of perhaps 3 million km2.

      • I don’t think one should use simple averaging for guesstimating extent. There is an ongoing tendency for the remaining ice to get shoved up against the northern shore of Greenland and the Canadian archipelago.

        Take a look at how the ice gets piled up toward the end of the melt season in the right hand series of images on this page. (And, yes, it is “upside down”.)

        In no way am I predicting a 2013 meltout, but I think we need to recognize that we are now within the range of possible. A significant summer melt along with transporting and concentrating winds could leave only a rim of thick, but broken, ice piled along the Greenland/CA north shore.

      • Yes, I was more trying to sketch out how you could have a volume that was within error of zero, with a still apparently-significant extent.

        I do think that unless there is a big contribution from a cycle of some sort which (PDO? OUMCMMILJLGWO?*) which is going to turn around very soon and allow multi year ice to start rebuilding, we are now in a regime of ‘any year now’. I’d be very surprised if we got as far as 2020 without ice-free conditions.

        *Official Uncertainty Monster Curry Magic Make It Look Just Like Global Warming Oscillator

  3. Thanks for a very interesting read, Tamino.

  4. Although there are serious risks when applying a regression to data with such a patchy distribution it’s still interesting to see that there appears to be a trajectory in the area vs thickness where thickness decreases (increasingly) faster relative to area.

    Thi fact is trivially evident both in terms of physics and when considering other representations of the data, but one can get a hint in the last graph above of how the relationship might progress in the future in terms of continuing areal loss.

  5. Are the volume and thickness minimums concomitant enough so that thickness as you calculated it is a reliable estimation?

    [Response: Yes.]

  6. hm?
    Glacial abrupt climate change as a result of internal oscillations
    Banderas, R.; Álvarez-Solas, J.; Montoya, M.
    EGU General Assembly 2012, held 22-27 April, 2012 in Vienna, Austria., p.2857
    “… Here, the role of CO2 and Southern Ocean winds is investigated using a coupled model of intermediate complexity in an experimental setup designed such that the climate system resides close to a threshold found in previous studies. An abrupt surface air temperature (SAT) increase over the North Atlantic is simulated in response to increasing atmospheric CO2 levels and/or enhancing southern westerlies. The simulated abrupt warming shows a similar pattern and amplitude over Greenland as registered in ice-core records of Dansgaard-Oeschger (D/O) events. This is accompanied by a strong Atlantic meridional overturning circulation (AMOC) intensification. The AMOC strengthening is found to be caused by a northward shift of NADW formation sites into the Nordic Seas as a result of an increase in sea surface salinity in the Northeastern Atlantic. The latter is caused by a northward retreat of the sea-ice front in response to higher temperatures. In this way, a new mechanism that is consistent with proxy data is identified by which abrupt climate change can be promoted.”

  7. David B. Benson

    From my limited understanding of sea ice I would certainly expect there to be a minimum thickness. If that is 150 cm then about 15 cm would be exposed above the water. From various pictures of Arctic sea ice I’ll say that is about right.

  8. Tamino, the graphs on sea ice measurements need be coupled with average sea and surface temperature trends, from there a better rudimentary model may be achieved. Refraction wise, there is very little difference between very thin ice and open water, near surface IR fluxes are similar with either conditions, not accounting for evaporation, Northern Hemisphere weather patterns have already changed and give a glimpse of the weather to come with less ice extent. Because of weirder weather, the larger stunning image of nearly no ice at all will be an after thought.

    • The coupling must be done when surface temperatures are -11 C or warmer, combined with sst’s warmer than -1.8 C, when the 2 are combined it takes a number of days before sea ice at 150 cm melts completely. I know that at 75 North this number of days is roughly 60 with the rising sun at 20 to 30 degrees elevation. But this number varies with ice coverage, some amount of open water is necessary. The number of days shortens with more open water. Due to sea ice insolation the number of days to completely melt sea ice peaks at about 80 degrees North. If there is extensive ice coverage the sea surface temperature is shaded from direct insolation and therefore -2 C is more or less maintained, the ice does not melt as fast as when broken up or laced with leads. So having data where -11 C average is exceeded combined with knowledge of sea surface temperatures gives a good approximation.
      My numbers are very rough, but at 75 North surface T >-11 C and sst >-2 C takes 60 days for 150 cm. As a good example Hudson Bay takes 40 days to loose most of its ice once open water and puddles get established:
      Kara Sea took about 60 days this year:

      At 80 degrees to 90 degrees it should take about 90 days before 150 cm melts completely. Since in June 2012 the average temperature was roughly between -10 and -15 C between 80 and 90 North;
      come mid september there is not enough time for it to melt completely.
      If the average temperature at 80 to 90 degrees North increases by +5 C
      there will be very little sea ice left come summer minima.

      • Along with calculating melt rates one should consider the amount of ice that will get transported out of the Arctic to melt in much warmer water.

        No longer is there very thick hard-frozen ice but thinner, broken ice that is more mobile. As the Arctic warms we might expect stronger winds and more frequent summer storms. This can add up to much more transportation through the Fram and, as we saw at the end of this last melt season, movement of some of the thickest ice into the warmer waters of the Canadian archipelago.

        Ice that leaves the Arctic Ocean is ice that will no longer soak up the summer heat. That makes more heat available to melt what remains behind.

  9. Might be interesting to see what the ecologists are making of this; I recall long ago at RC when Dr. Bitz did her guest topic, that the decline in ice was of considerable interest because the ecology depends on organisms that thrive along _edges_ (with other factors). So as the ice changes so does the food chain starting with the photosynthesizers living along the ice edge and on up to the filter-feeding whales.

  10. Does this precipitous decline in Arctic sea ice tell us that we have already exceeded Holocene maximum global temperature? I read something by James Hansen recently which said sea level is rising faster than at any time in the last 10,000 years, and that this was good evidence that global temperature is already above the Holocene maximum… and of course ice melt and sea level rise are two sides of the same coin.

    • Is this the paper you are referring to: ?

      Hansen and Sato do say they believe we have now equalled or exceed the temperature of the Holocene maximum, which occurred around 8 000 years ago in the early Holocene. More significantly they find evidence for us having the misfortune that the (pre-industrial) Holocene was only just cool enough to avoid a small amount of warming causing a large amount of ice melt and sea level rise.

      • Thanks Ed. I tracked it down to here:

        Click to access 20120917_GroverNorquist.pdf

        Hansen writes (p. 5):

        “We have always kept the base period fixed at 1951-1980, which is a good choice for base period because it was a time of relative climate stability prior to the rapid global warming of the past three decades. Most important, the global temperature of 1951-1980 was probably still within the Holocene temperature range, i.e., the climate that existed during the period when civilization developed and to which humanity and other life on the planet are adapted. In contrast, the past two decades and in all likelihood the period 1979-2003 that Michaels employs are already outside the range of the earlier Holocene. Confirmation that global temperature is now above the prior Holocene range is provided by the fact that ice is melting all over the planet, with both Greenland and Antarctica shedding ice at substantial rates. The current rate of sea level rise, more than 3 meters per millennium is far above the rate of sea level rise in the past several thousand years.”

        So he didn’t *exactly* say what I thought he’d said, but it amounts to the same thing.

  11. There are local effects. Today Narsarsuaq is 46F and Saglek, (a thousand KM due east) is 9F. Nuuk, GL is 37F, and Iqaluit, NU ( 800 KM east) is -9F. If latitude was the prime driver of temperature, these would have more similar temperatures.

    In a time of global warming, ocean currents and atmospheric transport of latent heat become more important in the pole ward transport of heat. Then that heat melts ice from the top down and from the bottom up. The Navy models did a much better job of simulating ice loss because they did a better job of simulating atmospheric transport of latent heat.

    As the ice warms, it loses strength. As the ice thins, it loses strength. When weak the ice breaks up into little bits with greater surface area, and the melt goes very fast. These are all very nonlinear at ~0C. Statistical analysis of data about ice colder than 0C, will not predict the behavior of ice at 0C.

    We need to use data from the system that we are seeking to predict.

    Dr. Barber tells us that today, there is a lot of sea ice at 0C. The problem is that when slush gets cold, the outside freezes, and the interior volume remains at 0C. And given the low tensile strength of ice, the slush with a frozen surface remains weak and subject to break up – exposing the core of slush. It would take years of really cold weather (summer and winter) to freeze all that slush solid. However, given current freeze conditions, a few days of a warm wind can remelt the outer shell of cold ice.

  12. Why not go quadratic with it? I mean, it’s quite clear none of them are behaving linearly. Not that they’re really quadratic either, as I’m quite convinced it’s significantly exponential, but the extrapolations would probably be a little more accurate. Perhaps even more so because the lowess might be underestimating the slope in the end, although that’s a tough call after a record year.

  13. Regarding the minimum ice latitude; it makes no sense to me that we’ll have a free floating ice cap around the north pole at some point in the future (unattached to land). I’ve seen graphics from a model that illustrated such an ice island, but I think the increased currents in the Arctic Ocean caused by the wind blowing across open water would preclude this. I think the next big drop in ice extent/area and volume will occur when the main ice cap detaches from Greenland and is pushed into the Atlantic. This past year it appeared that for the first time it became wholly detached from the Canada archipeligo. Maybe Wayne or someone could confirm that.

  14. Andy S. “This past year it appeared that for the first time it became wholly detached from the Canada archipeligo.”

    It is strictly a matter of winds and how much detached the main body of pack ice becomes. 20 years ago it could get detached by 10 or 20 miles with Southerlies, its the said “big Lead” which was known all the way back to the beginning of the 20th century. The more distant detachments are recent, usually short lived, because sea current and tides relentlessly push the ice towards the archipelago.

    Errata from above
    “Since in June 2012 the average temperature was roughly between -10 and -15 C between 80 and 90 North;”

    Should read Since “Early May” instead of June

  15. Posting in a RealClimate blog on arctic sea ice in which you commented on the folly of fitting a quadratic to sea ice “wili” notes that:”“The reputable climate-statistics blogger Tamino, who is a professional statistician in real life and has published a couple of posts on this topic, puts it bluntly:
    Fitting a quadratic to test for change in the rate of sea-level rise is a fool’s errand.”

    And yet the same reputable Tamino seems to be engaged in a similar ‘fool’s errand’ in trying to predict future Arctic Sea Ice Extent in his recent post”

    Any comment on that?

    [Response: I already replied on RealClimate.]

  16. Since the quadratix fitting is now a hot topic at RC with this blog mentioned, too, I think I should be a little more verbose to avoid misunderstandings. Adding a quadratic term for these extrapolations to see when we’ll go below 1 seems well justified in my opinion. One should of course refrain fron using quadratic fitting when the physical phenomenon is not qquadratic. But as its closer to exponential than linear. Quadratic will sure as hell yield better results than a linear fit. Not that it mattets much, we’re probably talking about only a few years anyway. I just wanted to ask about it since Tamino summarized it as 8-9 years when the piomas volume data screams three to five to me.

    I’m also fully aware that Tamino here was not aiming for the most accurate prediction but extrapolating while at it for different reasons and only to show what that yields.

    I’m not convinced about the sigmoidal fit aside from the little that remains in the CAA, just like Tamino said. And in that I fully concur with Bob Wallace here, again, like I have many times in Neven’s blog. Its gonna go sooner than later.

    • What might be helpful is to consider the physical forces which might operate to slow the melt rate vs. those which might accelerate melt during these final years.

      Possible slowing events:
      1. Unusual/unexpected changes in ocean or atmospheric currents slowing the transport of heat in and/or ice out.
      2. Formation of unusually heavy cloud cover during the summer which would block some incoming heat.
      3. Heavy smoke cover from northern hemisphere fires including permafrost burns.

      Possible accelerating events:
      1. Increased ocean and/or atmospheric temperatures bringing more heat into the Arctic.
      2. Less albedo via ice and ground snow cover reduction.
      3. Stronger storms resulting from water water temperatures.
      4. Earlier breakup of the ice pack due to it being thinner.
      5. Earlier opening of the passageways into the Canadian archipelago.
      6. Increased rates in transportation due to thinner, more broken ice along with stronger winds.
      7. More frequent breakdown of the circular wind pattern which tends to keep cold air bottled up. That would bring about a larger exchange of warmer air from the south and slow freezing/thickening.

      There are probably more, these are what came to mind at the moment. I think we’ve seen all the accelerating events in play and I’m not aware that any of the slowing events have been reported. There may well have been some slowing this last year from the extensive Siberian fires but I’ve seen no one quantify an effect.

      The overall pattern of end of season minimum volume is dropping at an accelerating rate. There are multiple observed factors which could drive the melt faster. To stretch out the melt for several more years would seem to require the emergence of a new, unpredicted phenomenon.

  17. Bit OT but very interesting.

    Stoat mentions
    new study combining many observations says Antarctica is losing mass (losses in west>gains in east)

    “…study brought to an end 20 years of disagreement between different teams.”

    “We can now say for sure that Antarctica is losing ice ” says lead author.

    “We combined an ensemble of satellite altimetry, interferometry, and gravimetry data sets using common geographical regions, time intervals, and models of surface mass balance and glacial isostatic adjustment to estimate the mass balance of Earth’s polar ice sheets… Between 1992 and 2011, the ice sheets of Greenland, East Antarctica, West Antarctica, and the Antarctic Peninsula changed in mass by –142 ± 49, +14 ± 43, –65 ± 26, and –20 ± 14 gigatonnes year−1, respectively.”

    Paywall. Any possibility of a post Tamino?

  18. Dear Tamino,

    just a completely off-topic and perhaps presumptuous question: Do you have an R-script used to perform the analysis of trend differences pre/post 1997 in this excellent takedown of David Rose´s latest rubbish which you would be willing to share?

    I would very much like to use your analysis on several datasets and different timeframes to set a local sceptic parroting Rose straight, but I am unsure about how you corrected properly for autocorrelation.

    Best regards and high respect for your nice work, as always.