« Open Thread 3 | Main | Open Thread 5 »


Feed You can follow this conversation by subscribing to the comment feed for this post.



I thought "doughnut" was catchier, even if "tonsure" fits the original metaphor. Unless you live next to a monastery, tonsures are so MWP.

Plus, it allows for the classification of smaller areas of drifting calved ice as "sprinkles".

Mmmmmm, sprinkles....


Interesting paper by Hansen & Sato, here:


Makes some of us here seem like optimists.



I've read similar stuff to what idunno is saying in general media, but I have my doubts. The thing that is supposed to upset the Gulf Stream is changed volumes of cold water going deep off the polar regions. But the whole circulation operates on a multi-century basis (1600 years to complete a circuit?) so it seems to me we won't see changes in the Gulf Stream due to THC changes for many centuries yet.

Atmospheric circulation operates on shorter timescales, and seems more relevant to the funky winter weather. In any case, that extreme weather has only covered a few weeks, not the whole of winter. So the recent cold spells in a few areas don't even establish anything that could be considered "seasonal" let alone "climate".

Evile reductionist - some of the heat is "local" from summer warming being re-released in winter, as you infer, but a lot is flowing in from temperate / tropical latitudes and is unaffected by how much sea ice cover there is in any given winter. While the Arctic may be losing more heat this winter than last, there is more heat to lose as long as ocean currents keep pumping fresh supplies into the Arctic from further south.

Kevin McKinney

FrankD's point makes sense to me. Also, while the ice cover may indeed mean more heat loss, we're obviously starting from an ample supply of heat given the very elevated temperature anomaly of the surface waters in the most affected areas, so there's some margin--so to speak.

That said, I'm guessing the La Nina conditions mean that we won't get quite the same slug of warm water coming into the Arctic from the Pacific that we did in 2010.

Patrice Monroe Pustavrh

Neven, I think PIPS is not very useful for determing thickness, just check 2001 prediction:

Gas Glo

Andyborst wrote "Did you read ...

That seems perfectly plausible and is different from the MOC general media scare stories that I and FrankD have mentioned. (Media don't use term MOC as public haven't heard of it but I think it would be useful to use it (esp places like here) to separate MOC from gulf stream issues.)

It was only the looking for GS changes in two severe winters that seemed inappropriate. That doesn't mean there isn't a longer term change happening as supported by the above report.

If there are changes to GS from 2 winters that is good news weather as it might mean this was just unusual and thinning of arctic ice has chance to recover when weather returns back to more normal. Unfortunately I don't think that is fully believable and there are serious possibilities that the unusual weather is consequence of climate changes not just random variability.

The surface flow through Nares strait is predominatly from Arctic, but is there considerable subsurface flow to Arctic?

Gas Glo

FrankD wrote

"The thing that is supposed to upset the Gulf Stream is changed volumes of cold water going deep off the polar regions. But the whole circulation operates on a multi-century basis (1600 years to complete a circuit?) so it seems to me we won't see changes in the Gulf Stream due to THC changes for many centuries yet."

Sticking to my stated preference for labeling MOC issues as such, "changed volumes of cold water going deep off the polar regions" pretty well defines MOC changes.

Just because it takes over 1000 years for a circuit, does this mean that such changes take centuries? Is the whole thing is mainly driven by gravity acting on waters that suddenly become denser as they lose heat or lose fresh water that freezes? If you make changes to what is driving the whole thing would it take centuries?

This seems like asking if you turn off a pump does all the water in the swimming pool flow though the pump before the circulation ceases? Of course that is not perfect as we are not turning off the pump more making slow changes to where the sinking occurs. For that we need models and the models seem to suggest only slow changes in response.

(Not that this stops the media selling a good scare story.)

Gas Glo

The amount by which the IJIS extent inceases from 20th Jan to Maximum seems to show a bit of a dicotomy:
2003 952k
2004 973k
2007 842k
2009 974k

2005 1228k
2006 1256k
2008 1386k
2010 1294k

(The gap range 254k is nearly as large as the two ranges 128k + 158k.)

This could of course just be random variation, but before dismissing it as random, what should we try to correlate it to?

I don't see obvious correlation with ENSO or AO.


Gas Glo,
Yes I see your point. I guess it depends on how much "elasticity" there is in the system (which I don't know). So "many centuries" probably was an overstatement on my part, but still, I don't expect to see changes in the Gulf Stream part of the THC for a long time after all the ice is gone. In fact, I don't expect to be around to see them at all.

However I believe surface winds also play their part in driving the Gulf Stream, and I can credit shorter term fluctuations there.

Kevin McKinney

Gas glo, that's a provocation observation about the bi-modal distribution. But I have to say that I'm struggling a bit to relate your numbers to my eyeball assessment of the graph. For instance, comparing 2007 to 2006, it appears that the number for 2007 should be the greater, yet you have 2007 at 842k and 2006 at 1256k. Similarly 2003 and 2004.

Perhaps I'm misunderstanding something here? (Don't have time to check further, I MUST get to actual work. . . )

Gas Glo

Does a little table help?
20-Jan Max Diff
2003 13892344 14844063 951719
2004 13387656 14360313 972657
2005 12870938 14098906 1227968
2006 12526563 13782344 1255781
2007 13103438 13945625 842187
2008 13130938 14516875 1385937
2009 13439219 14412813 973594
2010 13112969 14407344 1294375


From the table Gas Glo, it seems there is a pretty good correlation between Jan 20 under 1.31 m and growth of greater than 1 million km2. That would make sense that years that start out exceptionally slow, tend to catch up to the average. This does seem like a small sample to draw conclusions from. 2007 is the odd year out, but then, that was an exceptional year in many respects in terms of extent.

Wayne Kernochan

Hi all,

I am deundeundelurking again, because I would like to raise the question of what is going on with Antarctic sea ice, and this seems to be the best place for doing it.

I have been following NSIDC's representation of extent (with deviation from a late-1900s "norm") and concentration, since October. What was striking in October was that, although extent was slightly above normal, much of the ice was "pea soup". Given the clockwise circumpolar current, that raised the possibility that the "soup" ice would be more easily swept north, especially along the edges. If this was atypical, then extent would be greater than normal for a while, but as large chunks were detached and melted as they moved north, extent would finally dive significantly below normal.

Over the last three months, that appears to be what has happened. According to NSIDC's daily extent graph, extent was well above the "norm" as well as last year's extent until the last month, at which point it dived below both, and has just reached a point nearly equal to the typical minimum extent, with perhaps 1 1/2 months of further melting to go.

Likewise, the pattern of the melt conformed to a "more pea soup"/current-driven model. The first anomaly was the farthest north, right after the SA/Antarctic Peninsula "choke point", where you would expect a strong current to carry all the ice further north and melt. That happened around late November-early December, and the ice remaining tucked into the Peninsula's eastern "armpit" is still well below average.

The second anomaly, especially over the last six weeks, was much further south, in the West Antarctica bay. Ordinarily, in the later stages of melt, a channel opens up to the shore of the bay. This year, the large set of ice on the left of the channel going south slowly separated from the rest, floated north, and is nearly finished melting. The result is an ice-free shore in the West Antarctic bay that is perhaps four times the extent that it is normally.

Two more anomalies now appear likely to occur. The easternmost part of that shore still has ice clinging to it, perhaps half of that in pea-soup state. With no ice to block the current, it seems likely that this ice will melt/be carried north. At the same time, there is a large crack in the ice to the right of the channel, right near the shore, and ice north of the crack appears to be turning to pea soup rapidly. Again, even a weak current plus melting may very well eliminate most of that ice.

As a result, I am willing to stick my neck out and predict that both anomalies will occur, and the result will be a minimum Antarctic sea ice extent well below last year and the "norm".

It appears that the practical result of this exceptionally low minimum extent is that for a far greater part of the year, a far larger part of the West Antarctic shore (and a significantly larger part of the East Antarctic shore) will be ice free. That includes not only the western Peninsula but also all of the West Antarctic Bay, much of the shore opposite Australia, and some of the shore opposite Africa. In fact, at minimum extent, pretty much all of the shore except the eastern and western Peninsula "armpits" and a part opposite India may be ice-free. That, in turn, should mean greater glacier flow into the sea and melting, primarily in West Antarctica, but also to some extent in East Antarctica, which has far more land ice. And that, in turn, should mean a more significant contribution to sea-level rise from Antarctica.

The reasons for this, afaik, remain a bit unclear. According to an article published in Climate Progress, researchers have found that warmer south atlantic/pacific waters are moving further south, to the point where they are reaching land and melting the ice anchored on the sea floor. However, the graphic indicates that this warm water has typically not moved further south than the tip of the Peninsula; so that doesn't really explain West Antarctic bay melting. I understand from Peter Ward's book that researchers who have visited Antarctica for 30-40 years are seeing higher "summer" air temperatures and a longer summer season, but, again, the researchers appear to be somewhere on or near the Peninsula, i.e., further north than some of the unusual melting going on.

Both Ward and Hansen raise the specter of a sudden "break-off" of West Antarctic land ice, as has apparently happened in the past. However, Ward, I think, suggests that such a break-off would happen only 700-800 years from now -- and I believe that he is assuming a somewhat linear melt of WA land ice. If, in fact, as Hansen et all suggest in a recent paper is happening to Greenland land ice, the rate is doubling every decade, due to ice-free shores a greater and greater percentage of the year, then such a break-off is much more likely to happen in the next 300-400 years -- which would mean a rapid rise in sea level of 30 feet or more.

Just as alarming is the possibility that East Antarctica land ice melt may be accelerating. Both Ward and Hansen appear to assume that most EA land ice will not start to melt until WA land ice is mostly melted. While I'm not sure why they assume this (possibly because much of this ice is farther from a shore), there is far more land ice in EA -- and if both go, and Greenland goes, we are well above 200 feet of sea level rise.

Could someone please provide some counter-arguments, or context, to explain what's going on -- and possibly why I shouldn't be worried just yet?

Kevin McKinney

Gas glo, your table definitely allows me to verify that your numbers are correct. I don't know if my issues eyeballing the graph reflect trouble with with curves or my eyesight!

As to correlations, I don't see anything immediately. I wondered if there would be a relation to date of the maximum, but apparently not.

Greg Wellman

Wayne, I've also been noticing the steep decline in Antarctic sea ice over the past couple of weeks. I think you might be pushing the conclusions a little too far though. Here's my take:

1. Over the last couple of decades non-temperature changes have tended to increase the extent of sea ice formed in the Antarctic winter. Those non-temperature changes are increased precipitation (from moisture carried aloft from the warmer mid lattitudes) and (maybe) the increase in strength of the polar vortex related to the ozone hole.

2. At the same time year-round ocean temperatures have been increasing, and summer air temperatures have been increasing. Effects of that have already been seen with the ice shelves (breakups) and land ice (e.g. increasing speed of some Antarctic ice streams).

3. This Antarctic summer looks like the first time that the temperature effects in (2) are dominating the non-temperature effects in (1) with respect to sea ice. Of course some of that may be natural variation, but I would expect the trend of greater than average Antarctic sea ice extent around June/July combined with below average Jan/Feb extent to continue .. with the latter becoming more prominent over time. This could take another decade to become obvious in the data though.

4. I disagree that more of the East Antarctic coast is ice free than normal. Looks like a toss-up to me. But yes, the Western coast is far more ice free than normal. I don't think sea ice offers any significant back-pressure (unlike ice shelves) so the physical absence of the sea ice is not promoting land ice flow. But it also means that those warmer currents are not being cooled as much as they approach the Western coast - and that could definitely accelerate the undercutting of the WAIS. Such undercutting would cause the ice above to settle, either slowly increasing sea level, or in a more sudden fracturing. I haven't read Ward's book and I'm not qualified to agree or dispute with Ward or Hansen :-)

5. I suspect that the land ice is plastic enough to mostly react to undercutting by settling. This would be measurable by Cryosat (I think) and over a period of a decade should become noticeable in the sea level record. When the WAIS contribution to global sea level is quantifiable, we'll have a sense of what the century-scale acceleration of that contribution is. Of course by that time we may have locked in 10 meters of rise (WAIS + Greenland, ~80% loss) barring a massive CO2 air-capture plus sequestration plan. OTOH, maybe the certainty of 10 meters in a known time frame would shock people into such a plan using biochar. My personal guess is that 10 meters is possible in 500 years.

6. The arguments for why the EAIS is far more stable than the WAIS are compelling. But there has to be some overlap - e.g. by the time the WAIS is 80% gone, the EAIS would be following but maybe only 2% gone. Of course the EAIS is approximately 10x the volume of the WAIS, so 2% of it is a lot!

7. The time scales here are such that someone born today will see some coastline reduction, but nothing particularly serious in the developed world. Their children or grandchildren will live in a world of perpetually retreating coastlines though. Fat chance of getting most politicians (in mid-life, retiring in a couple of decades or less) to get serious about mitigation. I suspect a non sea level event such as a series of correlated weather-related crop failures are more likely to finally push AGW to the top of the policy agenda.

I guess I'm not really arguing with you, just suggesting slightly different emphasis. Other factors will likely be far more important before we get to 200 meters.

Gas Glo


"tend to catch up to the average" seems to get me thinking of a greater variation at 20 Jan than at maximum. There is indeed a noticable difference range of 1366k at 20 Jan and only 1062k at maximum. But it still feels odd to me. Why are there no medium changes? (particularly?? for years that start close to threshold value or perhaps particularly for the 3 years that maintain their relative ratio in the range.)

There seems a variety of things happening: 2003 5 & 6 maintain their position in the range. 2004 and 9 start high in the range and go lower. 2008 and 10 start below average in the range and go above average. 2007 as you said is an exception to 'tend to move towards average' starting low and going lower.

If you want the figures removing the effect of the larger range at 20 Jan, they can be rejigged to between 0 and 1 for lowest and highest point in range.

2003 1.000 1.000 0.201
2004 0.630 0.544 0.240
2005 0.252 0.298 0.709
2006 0.000 0.000 0.761
2007 0.422 0.154 0.000
2008 0.443 0.692 1.000
2009 0.668 0.594 0.242
2010 0.429 0.589 0.832

The big gap from .242 to .709 showing it is not due to the larger range at 20 Jan despite your "tend to catch up to the average" making me notice the greater variation at 20 Jan than at maximum.

Perhaps all this just suggests I am not getting my head around it properly. It might easily just be mildly odd and the sort of thing that can happen with a very small sample.

Kevin McKinney

". . . the sort of thing that can happen with a very small sample." I'm tending to think that, too. But I also agree that it remains a striking pattern--and a provocative one.

Andrew Xnn

Cyrosphere Today breaks the Arctic into 14 basins.
These can be grouped as either Marginal (5) or Core Basins (9).
Marginal basins do not freeze solid during the winter because they are bordered by open water.
Core basins freeze solid during the winter because they have fixed boundaries that freeze every winter.

7 of the basins are currently Seasonally Ice Free (SIF). The duration of these ice free periods vary, but are trending towards longer ice free seasons. It is only natural that as the ice free season grows, eventually these basins will experience ice free winters. The St. Lawrence is first in line with the Sea of Okhotsk and Bering Sea next. Unfortunately, we can expect that eventually all basins will revert to an SIF pattern.

Interestingly, the Greenland Sea, a marginal basin, is displaying the least anomaly of all basins. This may be a reflection that it is more of a conduit for ice flows than a basin. Anyhow, here is my grouping and ranking of basins as to where they fall on the SIF spectrum:

St. Lawrence (6+months SIF; minimal winter ice)
Bering Sea (4months SIF, positive winters)
Sea of Okhotsk (4month SIF, anomalous winters with late “recovery”)
Baffin/NewFoundland (1 Month SIF; consistent anomalous winters)
Greenland Sea (year round ice; slight anomaly without pattern)

Barents Sea (2+month SIF, spring/fall anomaly, brief winter)
Hudson Bay (2+month SIF; spring/fall anomaly, solid winter)
Chukchi Sea (2 month SIF, summer anomaly with long solid winters)
Kara Sea (almost SIF, spring/fall anomaly with brief solid winters)
Laptev Sea (almost SIF, spring/fall anomaly with long solid winters)
E Siberian Sea (almost SIF, summer anomaly with long solid winters)
Beaufort Sea (almost SIF, summer anomaly with long solid winters)
Canadian Archipelago (almost SIF, summer anomaly with long solid winters)
Central Arctic Basin (summer anomaly with long solid winter)



Nice summary. You are correct about the Greenland Sea. It is (semi-) constantly topped up via the Fram Strait in all seasons, and while anomalies occur, they are not seasonal. I suspect you will also see the Canadian Archipelago hold at a minimum level for a while rather than become SIF, but we have little history to go on there.

I'd probably place the Barents with the marginal seas, since the area designated by CT could never fill with sea ice, so it can't be considered as being physically constrained like the other core areas. It is set out like this to include drifting pack and the White Sea, but it is probably never more than ~60% full.

Just a few additional remarks - in some cases a bit more perspective, in others, I don't totally agree with your observations. "Average" is 1979-2009 average, "historical" means broadly pre-2000, "low ice" generally means around 50-100K sq kms:

Bering Sea - SIF period is increasing - 2010 was 6 weeks longer than average.
Okhotsk - freeze up later than average over the last couple of years (hence anomaly+recovery)
Baffin - historically 1-2 months with low (but not zero ice). Recently, low ice period has increased (3 months in 2010) or got to actually ice free (2009, due to Nares Strait being blocked). Recent winters have been low as well, (due to WACC'y weather).

Barents & Hudson - really only 6 weeks or so ice free on average, but extended to 3 months in 2010
Chukchi and Kara - Historically not ice free at all, about 6 weeks (4 for Kara) of low ice area. Now ~3 months at that level, and 2 months ice free last year.
Laptev, E Siberian, Beaufort and Archipelago - historically about 30-40% full in summer, now briefly almost SIF (except Archipelago, which bottoms out at ~20% at minimum).

The spring / fall anomaly just means earlier melt / later freeze. The only reason there is a recovery in summer is because you can't go lower than zero. I'm guessing for this year we will see very strong spring anomalies in Hudson, Baffin, Okhotsk (maybe Bering and Barents too) which will produce a fairly dramatic overall anomaly in May/June. This will probably recover somewhat in July/August, in what Wall Street call a "dead cat bounce".

If that sounds like what happened in 2010, its because it is. That massive drop we saw in June was mostly due to Hudson Bay melting out three weeks early. The anomaly in both volume and area recovered a little after that, but only because of areas like Hudson Bay where the historical average reached the zero that had already been reached by current reality weeks earlier.


I noticed something strange about IJIS data the other day. I have built a table of the monthly change in extent. I took the average for each month, and then calculated the anomaly for each month. Then I strung it into a single series of anomalies, and put on a rolling average.
(graph starts at July 2002 and the year labels correspond to Jan and Jul for the relevant year).

I found an apparent cycle which made me look a bit closer at the data:
2002 - 06 saw above average extent changes between May and October and below average extent changes in December to April. 2007 broke the apparent cycle, and after that date, the pattern reverses - below average extent changes between May and October and above average in December to April. 2007 itself is a mess with a steep rise in March and a fall in June - inserting 1.5 "cycles" into the space usually occupied by 1.

This is remarkably consistent. Every year saw an above average and a below average period (with fairly consistent amplitude, other than 2007). Within the two periods every year showed exactly the same timing for the peak and the trough. In 2007 the amplitude of the annual seasonal cycle increased markedly.

Looking at the raw monthly data, I'm struck by the very strong oscillation between positive and negative anomalies - strong positive anomalies only ever last a month or two before being balanced by strong negative anomalies. Finally I also notice the amplitude of the swing between postive and negative also seems cyclical. In even numbered years in autumn, the amplitude of the anomaly changes reaches a maximum, then diminishes again so that in odd numbered years, while still swinging from positive to negative, the autumn amplitude is minimal.

I have no idea what it means, just thought I'd put it out there for the brains trust. Fourier analysis, anyone?

Disclaimer - the human brain is brilliant at finding patterns even when they don't exist, which might be whats happening here. But still, seems odd....

Gas Glo

The ranges at mid May and early November are small compared to other times of year. That there are low anomalies associated with these times should be no surprise. But it is surprising that they represent changes every time bar one in 1997 where a negative anomaly in early 1997 is followed by negative anomaly in late 1997. As a first estimate, the chances of having one or no excursions in same direction in 15 is (1+15)/2^15 or 1 in 2048. (Much less likely the bi-modal distribution I mentioned for which I calculated 0.6^6= 4.7% probability.)

With these calculations, I think it is important to point out we didn't specify what we were looking for before we looked and therefore could find lots of different things unusual. However even if there are 50 unusual patterns that you could have found that still leaves only 1 in 40 chance. So I think it reasonable to rule out the null hypothesis of random direction after returning to near normal at mid May or early November at the 95% level.

I think that brings us to concluding this is reasonable evidence for Evilreductionist's
"With all of the warmth from the arctic ocean pumping into the atmosphere and slowing the freeze, my thinking is that this will mean that melting this season will be slower, simply due to there being less heat."

as a possible causation for anomaly swings in the opposite direction to the last anomaly swing.

So perhaps this was mentioned before looking for this evidence and takes us back to 1 in 2048 level of evidence. Having said this, I am not sure I am keen on betting on a positive anomaly for March 2011.

Peter Ellis

Um, surely the reason for this is pretty trivial? As you point out, the total year-to-year range in mid-May is low, as is the November range. This means that any year with a high winter maximum must by definition have higher than usual melt in the first half of the season, in order to get down to the (quasi-fixed) mid-May value. Conversely, any year with a low summer minimum must have greater than usual melt during the second half of the season. The same applies reciprocally and symmetrically to the re-freeze.

So, all you're saying is that 2002-06 had higher winter maxima and higher summer minima, while 2007-2010 had lower winter minima and lower summer minima. I suspect we all knew that already!

The real question is why the mid-May and November figures show so little year-on-year variance. This is also comparatively straightforward: it's due to the shape of the Arctic ocean and in particular the Bering strait. The melt moves inwards from the edge of the pack through the melt season, and once the edge gets up to the Bering Strait, there's very little sea area at that latitude, it's mainly land. There will always be a "choke-point" in the melt at the point where the peripheral basins are mostly empty and the central basins mostly full.

In coming years, I suspect this will change somewhat, with early melt in Hudson Bay leading to May values below the current "choke point".

Andrew Xnn

The melting season has not slowed down. However, with the delayed start to the freeze period, there is thinner ice than in the past. This results in a less solid pack that is more free to move around and expand. The pack expansion, can give the illusion of a recovery in 2 dimensions. However, since ice is a 3 dimensions, there has been no sustained recovery and we are witnessing a death spiral.


Thanks for your insights. I know you have more precise numbers and I agree that the Barents should be considered a marginal basin. Also, didn't realize that the Hudson Bay was responsible for so much of last June's anomaly. Going forward, "normal" and "average" are going to have to be re-defined on a regular basis.

Peter Ellis

The start of the melt season (i.e. melt in peripheral basins) *has* slowed down slightly, but only because there's less ice in those basins to lose!

Gas Glo

>"So, all you're saying is that 2002-06 had higher winter maxima and higher summer minima, while 2007-2010 had lower winter minima and lower summer minima. I suspect we all knew that already!"

Er, no read FrankD's post again: 2008-2010 had low minimum and high maximum while 2004-2006 had high minimum and low maximum.

I realise I have made a mistake in that 2003 had high minimum and high maximum so there are two occasions where the next anomaly swing is in the same direction but still 14-2 is a very wide difference.

>"This means that any year with a high winter maximum must by definition have higher than usual melt in the first half of the season"
True but we are comapring anomaly to anomaly not anomaly to melt or melt to anomaly.

Why aren't there more occasions when the two anomaly swings (6 months apart) are in the same direction?

Did you look at FrankD's graph?
I presume from the graph that the anomalies are relative to trend not to a fixed average. Otherwise the graph would look very different with later years having the negative anomalies. I haven't checked the maths of it.

Also I attempted to see if I could find the pattern in a longer record
I haven't done the maths but roughly estimating I got
p->n 11
n->p 9
p->p 3
n->n 9
others were marked as unclear

That isn't nearly so extreme but I was doing that by eye. I should try again using NSIDC monthly averages.

Gas Glo

D'oh try again 13-2 (i.e. 13 going to opposite anomaly versus 2 anomaly pairs where the change is in same direction) is still a wide difference.


100+ answers again, not so bad for hibernators !

Gavin posted http://www.realclimate.org/index.php/archives/2011/01/2010-updates-to-model-data-comparisons/ recently, including a nice updated graph from Marika Holland about Arctic melt.

To be frank, I had kept in memory that the models were "softer", predicting an arctic decline over a much longer period, while reality was steeper - this graph shows that those models are not so far from the recent events. Perhaps will Arctic ice survive until 2050 after all !!? (even if I'm sure that problem is not surface anymore, but thickness)


Hi everybody,

Sorry, I've been asleep. Hibernating.

I have a new theory to share:

The Gulf Stream is in a state of energy imbalance. This involves an awful lot of energy and an awful lot of heated seawater. Compare the North of Scandinavia, and the North of Alaska. Same latitude. One is in the GS, one is not; net result being lots less ice.

More heat is being put into the system in the tropics, because of an extra dose of solar radiation being reflected back by CO2, etc...

First result of this is that less heat leaves at the destination. Because it is hotter, it is harder for the air to cool it enough to send it down to the depths of the ocean.

Overall, more heat in, less heat out.

The result being an expanding, and possibly now deepening pool of anomalously hot water in all of the perennially ice free Arctic and all of the Northern Atlantic to the North of a line between approx Newfoundland and approx Ireland.

ALL of that area of the ocean was at least 0.5°C warmer, averaged over the last decade than a twentieth Century baseline (I checked this out over at GISS).

I think that that sort of anomaly over a whole decade is a lot closer to climate, not weather, and that it could be behind several of the observable oddities that the ice has been up to in recent years.

So over to you guys. Any comments?

Peter Ellis

@GasGlo: Hmm yeah, I really didn't do a good job of explaining that one. Let's have another go. We're trying to explain two observations:

1) There seems to be a month-by-month oscillation in the data, with high anomalies followed immediately by lows.

2) There is a shift in the longer-term anomaly pattern between years 2002-2006 and 2007-2010.

The first one I don't think needs a lot of explanation. Although there is an ongoing loss of ice extent from year to year, the loss is small in comparison to the total seasonal cycle. This means that the anomalies across the course of the year as a whole must sum to (nearly) zero. Every low month must be counterbalanced by a high month. Physically, months with lower-than-average melt may represent especially warm months (weather) or abnormal wind patterns that spread the ice out. These are weather effects which will even out over quite a short term, hence the rapid switching between high and low anomalies.

The second observation says that even after you average out the monthly "noise", there is still some systematic change in the shape of the melt curve. What is this? Let me point out two factors:

a) The black line on the graph is a lagging average of the preceding 5 months. At the point where it crosses the zero line, it means that the preceding 5 months melted/grew ice at exactly the same rate as the average trend. That's not particularly interesting. What we are interested in is not the crossing points, but the positions of peaks and troughs. These indicate timepoints where the preceding 5 months gained or lost ice significantly faster or slower than average.

b) We are looking at anomalies from the average trend. These must sum to zero, by definition! i.e. if you average together the August figure (or the March figure, or the November figure, or any particular month...) for all years in the dataset, you'll always get zero. So, if one year has a particularly drastic summer ice melt, it will show up as a downward deviation from the trend - a negative anomaly. All the other years will show a positive deviation from the trend: the average itself gets pulled downwards by the bad year, leaving the "normal" years with a positive anomaly.

With those facts in mind, let's look at the 5-month averaged graph again.

Recent years show positive anomalies in the first half of the year, peaking around March, while earlier years show negative anomalies in the first half of the year, again peaking around March. This physically means that November through March saw more ice growth in recent years and less ice growth in earlier years.

Conversely, recent years show negative anomalies in the second half of the year, with the trough in August/September, while earlier years show positive anomalies during this period. This physically means that April/May through to August/September saw more melt in recent years and less melt in earlier years.

These are not particularly surprising observations! All it boils down to is the fact that recent years have lower summer minima, while the rest of the year is comparatively unchanged.

Gas Glo

>"These are not particularly surprising observations! All it boils down to is the fact that recent years have lower summer minima, while the rest of the year is comparatively unchanged"

Is it that simple? A downward trend is effectively removed and a downward acceleration would I believe be likely to manifest itself with summer minimums having a pattern of: neg neg pos pos pos neg neg neg as you get if you fit a linear trend to downward accelerating numbers.

What we have is:
Winter p n n n n p p p
Summer p p p p n n n n

That winter pattern is not too surprising - it is almost the p p n n n p p p that you would expect if summer was accelerating downward and winter not changing much. Only one changed.

ppppnnnn is more different from nnpppnnn with 3 changed but it would be suprising if it perfectly agreed.

It is not! those patterns that are odd! At least that isn't the surprising feature I am talking about.

Stretching the anomalies into one chain:
p p n p n p n p n n p n p n p n

Why does it change from p to n and n to p so regularly? If it was random negative anomalies would follow a previous negative anomaly quite frequently. Note these are fixed time periods so it isn't a case of a negative anomaly lasting until it is over then you get a positive anomaly. In a random sequence with fixed time periods you would expect pp, nn, pn and np each to occur equally frequently. But there is only 1 pp and 1 nn.


Of topic but....

The global sea ice area at Cryosphere Today is trying very hard to set a record minimum.

Has that been mentioned already?


Andrew Xnn

An interesting paper on the cryosphere albedo feedback between 1979 to 2008 was published last week:


and here:


Notice the strong feedbacks in the Hudson and Baffin Bays,Greenland, Barents, Kara, Chukchi, East Siberian, Beaufort Seas and Sea of Okhotsk. Only areas that seem to be missing out are the Bering Sea and Central Basin. In fact, the central basin looks to have gone the other way. Could this be from greater snowfall?

Peter Ellis


Consider a simple declining trend: 10,9,8,7,6,5. The average is 7.5, the first three all have a positive anomaly relative to that, and the last three all have a negative anomaly. In the IJIS record, years 2007/2008/2009/2010 are the four lowest summer minima in the record, while years 2002/2003/2004/2005/2005 are the five highest in the record. The former cannot [i]but[/i] show a negative anomaly, and the latter a positive anomaly.

Averaging together monthly decline values is (I'm afraid) a shell game which serves to obscure where the numbers are coming from. The average of five consecutive monthly declines over a 5-month period is simply 1/5 of the [i]total[/i] decline across that period. When considering the August figure on the 5-month smoothed graph, what that means is that you're looking at (endAugust-startApril) / 5. Since the startApril figure is comparatively static, the variance in the smoothed value is dominated by the variance in the endAugust value: i.e. variance in the summer minimum (or close to it).

Thus, the recent negative anomaly "dip" in the 5-months smoothed values, with a trough around August is caused by the recent lower summer minima. Conversely, the positive anomaly "peak" in earlier years is a consequence of the higher summer minima in those years.

The same argument applies reciprocally to the freeze-up, causing the inverse pattern seen there with peaks/troughs in March.

Daniel Bailey

@ Andrew Xnn

Skeptical Science just looked at the Flanner study here

The Yooper

Artful Dodger

Sea Ice melted today in Ungava Bay and Hudson Strait as a low pressure system over the Labrador coast moved warmer water into the area.

Gas Glo

>"The former cannot [i]but[/i] show a negative anomaly, and the latter a positive anomaly."

But that isn't what FrankD graph shows. The trend though 10,9,8,7,6,5 is 10,9,8,7,6,5 leaving anomalies of 0,0,0,0,0. Instead you need to consider accelerating downward values like 10,9,8,6,1,0 this has a linear trend of 11.1, 8.9, 6.8, 4.6, 2.4, 0.2 and anomalies of -1.1, +.1, +1.2, +1.4, -1.4, -.2 ie start negative goes positive and back to negative. Our data only has one change.

There is little point in us repeating what we are saying like this ad nausium especially when neither of us thinks what is shown at annual intervals is particularly odd.

Why do so few years have the same direction anomaly at winter as in summer?

Gas Glo

>"Although there is an ongoing loss of ice extent from year to year, the loss is small in comparison to the total seasonal cycle. This means that the anomalies across the course of the year as a whole must sum to (nearly) zero. Every low month must be counterbalanced by a high month"

Every low month must be counterbalanced by a high month is certainly true, but it need not be in the same year. We have seen anomalies are low in May and November. Does this mean that high anomalies in this period must be balanced by low anomalies? of couse not it can be low or high for whole of that period. Try testing your middle sentence above to 2007 (or 2003).

Peter Ellis

Why do so few years have the same direction anomaly at winter as in summer?
Because the year-to-year variability in winter maximum is much less than the year-to-year variability in summer minimum. Any year with a negative summer anomaly (i.e. faster melt than average) must therefore have a positive winter anomaly (i.e. faster re-freeze than average) in order to get back up to the comparatively unchanging winter values.


Firstly with regard to the smoothed data, yes, I considered the "well of course the anomalies sum to the average" point that you make. I still found it odd that there was such a clear pattern - even given the long term trend, there is a lot of noise in the signal, and I was surprised that it displayed such consistent pattern-switch-pattern.

However, I understand your point, and suspected a simple explanation (hence the disclaimer). Posted hoping for some clearer minded individual to explain, so thanks for that. So the "cycle" simply reflects that there is a greater variation between minimum and maximum than there used to be. That comes as no surprise to anyone who has looked at this.

But even with a limited data set, I would have expected more variation from the year to year data, especially given that it does not treat the data in a binary maximum / minimum fashion and the "signal" becomes less obvious with a 6-month running average. Visually, there is little to pick between 2005 and 2009 on the IJIS graph, yet the patterns on the anomaly from monthly average is quite different. Conversely 2007 and 2009 are quite different with the extents diverging and converging a couple of times, yet the overall pattern is consistent, so I still thought it was kind of interesting, even allowing for the points you make.

The fluctuations in the month-to-month data, as Gas Glo remarks, is less obviously explainable. And the apparent beat over longer periods, where amplitude increases and decreases over a couple of years, also seems non-random and without the obvious explanation you have provided for the 5 month running averages.

Peter Ellis


I don't have a good explanation for fluctuations in the month-to-month data other than to say that "noise" fluctuations around a trend will necessarily be quasi-alternating.

The "beat" over a period of several years is explained by the pattern of change in the summer minimum over several years. We had a massive drop in 2007 and an equally massive refreeze, followed by a couple of years of slight recovery (i.e. less variation from minimum to maximum). The amplitude of the seasonal wiggle is largest for the 2007 re-freeze (seen as the positive peak in ~Feb 2008, since this is a lagging average that covers the end of 2007 in the smoothed data), and drops thereafter. It's slightly surprising that the negative excursion with a trough is August 2007 isn't larger - I think this is because 2007 melt season started from quite a low early-season maximum.

Let's also look at the direct comparison of 2005 and 2009. In the 5-month lagging average, the anomalies plotted in Feb/Mar of these years actually cover the 2004/ 2008 re-freeze seasons. 2008 had substantially more rapid re-freeze than 2004, hence the 5-month smoothed anomaly at the start of 2009 being higher than the 2005 figure.

Turning to the second half of these years, you're right that 2005 and 2009 look reasonably similar on the IJIS plot, except for one crucial point: the melt started earlier in 2005 than in 2009. This means that during April/May, the 2009 line is substantially above the 2005 line. So, when you get to August/September, and the 2005/2009 lines are very similar, it means that the 2009 line had further to drop to get there: i.e. negative anomaly in the 5-monthed smoothed figure for 2009, and positive anomaly for 2005.

So, why does the signal become less pronounced in a 6-month average? Because the peak melt / refreeze seasons are only about 4.5 months long. Essentially you have 4.5 months of melt, 1 month of bumping around the bottom, 4.5 months of refreeze, and then 2 months of bumping around the top. The boundaries between these phases don't line up perfectly with calendar month boundaries either. You'd probably see the clearest signal if you did a ~4 month rolling average rate-of-melt anomaly centred on the date in question.


Cheers again, Peter,

Just to re-emphasise (for the "home viewer") that the blue monthly data points are not cumulative, but independent of their neighbours; it is the loss/gain anomaly in each month and is not directly affected by what happened in the previous month. To consider 2010, June is a strong negative because a lot more ice than average was lost in June, it's not a carryover from the fact that an above average amount of ice had already been lost in May (a rare instance of two consecutive high loss months). Of course the running 5 month total is cumulative, and your observations are pertinent to that.

WRT 2007, the month with the big loss was July. August ticked up (still negative), before dropping again in September.

The signal is clearest on the 5 month average, which is why I included that on the original. It is somewhat clear on the 4 month average, but noisier, so less obvious to the naked eye. Your comments on the 4.5 month season make sense, but the fact that the signal diminishes quickly suggests that there is a lot of variability in that season - if there was more consistency, the signal would be more obvious in 6 or 7 month running averages than it is. Two years with similar amounts of melt can get there by very different paths - 2008 and 2010 are similar, but (as we know) 2010 lost a huge amount in May-Jun with little after, while 2008 started slow and finished with a bang.

I didn't really think anything in here was too revolutionary, just odd and worth reflecting on. Really, I suppose it just reiterates how exceptional 2007 was - I'm sure we'll see a new record minimum extent soon, but even after it loses its record, 2007 will still stand out as the year the game changed.

Gas Glo

Trying a quadilateral fit through NSIDC extent monthly values, I got anomalies:
Mar Anom Aug anom Same direction?
1979 0.1815 0.0191 1 0
1980 -0.0212 0.0689 0 0
1981 -0.2397 -0.0391 1 0
1982 0.1799 0.2970 1 0
1983 -0.0104 0.2394 0 0
1984 -0.3086 -0.2540 1 0
1985 0.0873 -0.3272 0 0
1986 0.0294 0.1477 1 1
1987 0.2175 -0.0231 0 0
1988 0.1313 -0.0077 0 1
1989 -0.2578 0.0798 0 0
1990 -0.0672 -0.5704 1 1
1991 -0.1686 -0.2624 1 1
1992 -0.1318 0.3337 0 1
1993 0.2631 -0.2359 0 0
1994 0.1601 0.2147 1 0
1995 -0.1748 -0.5506 1 1
1996 -0.3395 0.6064 0 1
1997 0.0399 -0.0424 0 0
1998 0.3734 0.0470 1 1
1999 0.2750 0.1345 1 1
2000 0.0047 0.1183 1 1
2001 0.2666 0.1683 1 1
2002 0.1025 0.0124 1 1
2003 0.1506 0.1888 1 0
2004 -0.2212 0.2074 0 0
2005 -0.3548 -0.0739 1 1
2006 -0.4684 0.0271 0 0
2007 -0.2918 -0.7497 1 0
2008 0.1849 0.0896 1 1
2009 0.2637 0.1652 1 1
2010 0.1446 -0.0290 0
19 15
32 31

34 out of 63 changes in anomalies are in the same direction as the anomaly from the period before. This seems pretty random.

Frank, I do not know if it may be worth trying 12 linear fits, one for each month then calculate anomalies from those trend figure. If the anomalies are generally smaller, then that may be a better technique and maybe the oddities will be less puzzling.

Peter Ellis

I appreciate that you've calculated one month independently from the next, but the reality is that physically they can't be independent. The melt that occurs during a given month will be strongly influenced by the condition of the ice at the start of the month - which will depend on what happened the previous month.

To give a couple of concrete examples - 2010 had record high melt in June, but low melt in July. This is because Hudson Bay melted out early - it inflated the figure for June while meaning there was nothing left at lower latitudes to melt out in July. Similarly, in 2010 we had near-record melt in April and May. This was because of the unusual "ice factory" set up in the Bering sea during February and March. The ice extent kept on increasing long past the normal winter maximum, but it was thin ice which was rapidly lost at the start of the melt season.

Or, to cast it in anomaly terms, the positive anomalies in Feb/March 2010 set up the record negative anomalies in April/May, while the record positive anomaly in June set up the negative anomaly in July.

This is I believe the root cause for the oscillation between positive and negative anomalies every 1-2 months - that being the duration of an average spell of abnormal weather.


The global sea ice area at Cryosphere Today is trying very hard to set a record minimum.

Derek Moran, I mentioned it a while back in Open Thread 1 (4 months ago) when the global anomaly was really low, about a million lower than now.

The anomaly is still pretty big as we speak (mostly due to the Antarctic sea ice melting particularly fast this year). I don't think it will go much lower though, because the total Arctic sea ice will start to move more and more towards the average. That's because in the Arctic there are land masses that constrain total sea ice extent/area.


Please continue in Open Thread 5

Peter Ellis

"while the record positive anomaly in June set up the negative anomaly in July."

Should say "negative anomaly in June set up the positive anomaly in July."

Artful Dodger

Congratulations, lads! You've laboured hard, and discovered Bounded Mean Oscillation.

Predicting 1st year sea ice is weather forecasting. Predicting MY sea ice is Climate science.

Peter Ellis

Dammit, who let the mathematician in? Now we'll have to lock up the whisky.

Artful Dodger

Haha, single malt?


First visible breakup Antarctic summer? On 23 januari two parts of the Pine Island Glacier SW front took off. Not much, probably totaling 5-6 km2.

The comments to this entry are closed.