« Greenland 2013 in review | Main | PIOMAS June 2014 »


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


I have put up polls on the ASIF for predicting the minimum: NSIDC SIE monthly/September average minimum and CT SIA daily minimum. Mind the differences before voting.

Bill Fothergill

Hi Neven,

In your post, you state that...
"The monthly average is third lowest after 2006 and 2011."

I know the comparison window you use (2005 - 14) is clearly marked, but May is an unfortunate month. On just about any dataset, the average May value for extent is still going to show 2004 as being lowest.

However, by 31st May, the daily figure on IJIS has already dropped to 7th lowest. On the NDIDC monthlies, 2004 will drop to 10th for June, with none of the remaining (Jul- Dec) months featuring in the lowest 10.

Bill the pedant

Bill Fothergill


Fat Finger Syndrome. Too many beers last night, methinks.

Given the subtitle on Neven's post is "here comes the sun", it's a spooky coincidence that I was off to see a Beatles tribute band.

Anyway, "NDIDC" in my above comment should have been NSIDC.


Thanks, Neven!
I have a couple of 'nuances', mainly on temps/DMI and the relevance of melt ponds.

This is on temps:

DMI +80dN mean temps don’t tell the whole story. During May, on average the sea ice is spread out over the whole Arctic Ocean. DMI temp is relevant for app. 3.3 Mkm2 ice cover in the essential part of the CAB. The part which was thought to have at least not lost more of its supposed remaining coherence last summer.

For the vast part of the Arctic Ocean, some 10 Mkm2, DMI mean temp doesn’t say anything. The lot of it will melt out anyway. As, in my opinion, ‘winter power’ hasn’t been conducive for ice growth in the whole peripheral circle Beaufort Sea-Laptev Sea, there is not much chance for even a 2013-like summer to let ice remain like it did in the Beaufort and East Sib Sea last year.

When NCEP/NCAR is analyzed for the Arctic Ocean proper over May, there’s not that much difference with the ’12 mean temps. The difference with ’13 is large. And ’14 was lots warmer than ’07.
Charts/maps would be great here, but that’s better on the thread.


On melt ponds, I never had the impression that May would show the first important growth of these.
Ususally they form during June, remember FI the timing visible on the Polar cams. Where the first signs show is usually Bering (ice already gone now), Mackenzie Gulf, Amundsen Gulf, the fastice against N Siberia.
In my opinion, melt ponding is a significant starter of the process in the Arctic Basin and its adjacent parts of Beaufort, Chukchi and East Sib Seas, about 5 Mkm2.
For the whole rest, getting ‘grey and blue’ is just a stage in the seasonal swing to open ocean. On the relevant part, melt ponding will start around third week of June, when conditions are right. It corresponds with the ‘June cliff’ in volume during the period ’07-’12 (cf. Chris Reynolds).

Jim Hunt

The sun has come to the shores of Siberia:

Click through to see the full size image then click the date to zoom in and have a good look around on WorldView.


Some last muses...

So there it is. I’m starting to look for ice quality difference (as I have lots of MODIS material saved from June-Sep ’10-’13). There’s no objective measure on that approach.
Even last summer it was my own personal impression that this quality did steadily become worse. Even though the numerical approaches looked like ‘rebound’. I haven’t disputed that, because it is like arguing on different views.

Whatever this years outcome may be.

The buffering capacity of all biospheric systems on the planet is collapsing. The already built-up ‘bank-account’ and the relentless GHG-output growth will have rapid consequences.
Modelling can be done until the bucket is kicked. Zooming in on just counting pixels too. It’s easy to miss the big picture like that.
I think FI Wadham's view on first ice-free minimum is still relevant.


@ Werther
Yes, the big wild card in the near term is ice pond area. The multi-year ice in the Beaufort should be less conducive to melt pond formation. But if the forecast proves out, a few days later warmth and sunshine extend to the CAB, where the (presumably flatter) ice will form larger ponds. And that's at higher latitude, which for the next month or two means greater insolation as a multiplier effect to lower albedo.

"On melt ponds, I never had the impression that May would show the first important growth of these."
I've seen differing data. The CICE simulation shows a peak area in mid-July, though trending earlier in recent years. Rosel & Kaleschke figure 3b indicates earlier, as does inference from CAPIE. There's a recent paper by Marko Mäkynen using data from the late lamented ENVISAT that might clarify timing.

What matters for the total heat budget, though, is not so much the timing of the peak in pond area as the integral over the cycle. So a late start can still retard the entire season - I think Schroeder found May data to be the most predictive (Fig. 3 in September Arctic sea-ice minimum predicted by spring melt-pond fraction; see also discussion in More on melt ponds blog post.) It would take quite a spike in the next week or two for 2014 to catch up on this count.


Iceman, hi,
Why would the sea ice in the Beaufort Sea be less conducive to melt pond formation now? As far as I know, the big intrusion of MYI into there was during winter '12-'13. Still, by 17 June last year the whole stretch Banks-Taymir was coverd by blue- and greyish hues indicating extensive melt pond formation.
Given the flow last winter, I don't expect much more MYI to have gathered up there.

For the prognosis, sure, as June is summer, large swaths of the Arctic Basin regularly get melt temperatures. Expect the whole same stretch to get ´blue´ next week.
I can´t see more than Rosel´s abstract, but that lead suits my own expectations.
If a weather pattern producing compaction and insolation would have lasted for two weeks in July/August last year, it would have produced what most of us expected.

Finally, I managed to do a basic CAD supported count on temp anomaly for May ´12, ´13 and ´14. The correlation (NCEP-NCAR): +1.6dC/+0.1dC/+1.0dC.
Starting from there, I see a lot of support for SIE/SIA minimum getting close to a second place this year.


Bill, thanks for the extra info. I forgot about 2004, even though Espen mentions it on the IJIS thread on the forum every day.

Werther, I agree with what you said about the DMI 80N temp map, but a lot of people watch it, and so I make mention of it every now and then.

Why would the sea ice in the Beaufort Sea be less conducive to melt pond formation now?

The theory is that MYI has more ridges, etc, making it more difficult for large melt ponds to form. I thought that the Beaufort Sea was crucial wrt melt ponds and their influence on the melting season outcome, but maybe not.

Jai Mitchell


you said, "Whatever this years outcome may be.

The buffering capacity of all biospheric systems on the planet is collapsing. The already built-up ‘bank-account’ and the relentless GHG-output growth will have rapid consequences.
. . .
I think FI Wadham's view on first ice-free minimum is still relevant.

My initial response was, "yeah, 2012 DID happen" Because it was such a shock when it did. I expect that this will continue.

The idea that warmer ocean currents from the pacific is driving further melting is a growing concern. The recent north pacific sea surface temperature anomalies will continue to make this an issue.

Even though current temperature anomalies at the straights are cooler than normal, the massive body of 5C+ waters in the North Pacific is a significant threat to arctic ice.


compare with Alaskan stream and the Anadyr current: http://www-tc.pbs.org/harriman/images/log/lectures/alexander_bering/bering_circ_lg.gif

A strong high pressure ridge has also set up again in the north pacific. The Ridiculously Resilient Ridge has reformed (for now)

This high pressure zone has been stationary for 10 days and is not forecast to move in the next 10.

This will contribute again to high regional surface temperatures.


The melt Pond signature likely signifies a shift from partial underside melting (the greater part of the Arctic Ocean now) to overall net bottom sea ice melting. Since I devised a method in determining overall net underside melt, its time to put it to the test. When someone has a picture of melt ponds will look for the correlation taken from outer space. http://eh2r.blogspot.ca/

I'll have more about this , including melt ponds soon.

Rob Dekker

Neven, what a great, balanced, analysis. Thank you. This is one overview that reflects the data, which suggests that the "jury is still out there" for 2014's melting season.

Regarding melting ponds, I would like to 'second' the opinion by "iceman", above :

What matters for the total heat budget, though, is not so much the timing of the peak in pond area as the integral over the cycle.

with the side-note that early heat absorption is more effective than later heat intake. And earlier heat best correlates with snow cover (as I showed before).

So far, what we have seen so far is that April showed a rather negative Northern Hemisphere snow cover, similar to 2010, notably MUCH lower than 2007 and just slightly above 2007 and 2012 :

which suggested that by the end of April, the early projections for 2014's minimum point at somewhere near the 2007 minimum.


Now in May, correlation data suggests that melting ponds start to rival snow cover as an early predictor. At least with my "poor man's" estimate of melting ponds being "Extent minus Area", using 1995-2012 data.

Also, there is a pretty strong correlation between Rutgers' snow data and NSIDC extent and area numbers for May, although snow cover is still a better predictor. Which suggests that the 'weather' in May in the Arctic may be important for setting snow cover as well as melting ponds, but snow cover is still a bit more important for determining the outcome in September.

I will give a full correlation update when the final snow/area/extent numbers are in for May, which will be at the end of the week.

Rob Dekker

Oops. "MUCH lower than 2007" should read "MUCH lower than 2013".


@ Werther "Why would the sea ice in the Beaufort Sea be less conducive to melt pond formation now?"
I haven't compared current conditions, but in Neven's 2013/2014 Winter Analysis post last month, March/April 2014 shows older, thicker, and more deformed Beaufort ice than in 2013.

I've also seen recent mention on the forum of snow cover on the Beaufort ice. Can anyone point to analysis of this factor? My understanding is that snow cover is conducive to melt pond formation under some conditions (flatter ice surface, warmer temperatures) but not others.


Hi Werther,
Long time lurker/learner with a question.

Can the melt pooling fraction be determined only for those higher-latitude areas that aren't destined to melt out anyway?
And, if so, is is likely to provide a better forecast of Sept min?

Thanks in advance


Been lurking lots here and very interesting as always.

On the melt-pond thing, isn't the melt area integrated throughout May an indication of the natural variance aspects of the sea ice minimum, i.e. its predicts how far away from, the current norm that end of September sea ice will, but does not take into account the continued downward trend is sea ice as they de-trended the data.

Therefore the question is what is the new norm for 2012-2015, and the melt pond anomaly will indicate how far below of above that September will be.

as the sea area is rapidly trending downwards for the same melt pond extent 2014 would end up lower than 2012, therefore there is a little extra melt in 2014 due to general sea ice loss trends for September. Therefore to get to the same actual ice loss as 2012, 2014 melt pond amount need not be quite as low (needs to very close though) as 2012, for same September low, if melt pond area same should go lower and if melt pond area larger should go much lower.

I suppose the point is, is that the melt pond extent predicts the natural variation around the mean, but does not tell anything about the negative (and accelerating) trend for the September sea ice loss over the last decades.

2013 had a very low melt pond and probably represents the upper end of natural variability for sea ice loss around the current mean. The question is is where does 2012 lie, is that the new mean, the lowest possible, or just slightly low?
Whatever the answer as the downwards trend should continue it shouldn't be long before a 2013 May set up will have a lower ice extent in September anyway, but at that point a 2012 melt pond set up will probably the first ice free Arctic summer?

Anyway we'll see what happens but one for sure is that the downward trend is likely to continue and the melt season will continue to extend as we warm and therefore what ever the melt pond amount ice free summers are long away.

What does an El-Nino do to arctic sea ice?

As 1997 and 2014 similar in Nino data for date in year, will 1997's sea ice variance from the mean then, be an indication of what we can expect this year?


Neven , there is one way to study melt ponds, in Austria even! Provided you have cold enough winter and a lake. Water ie ice has physical properties which replicate from a huge to very small scale.
May be some of us have seen the show where a giant ice dam broke open and the physical attributes were repeated in a large lab: http://www.pbs.org/wgbh/nova/earth/megafloods-of-the-ice-age.html , NOVA on PBS did a wonderful summation. With water and land, the similarities from large to small are repeatable to a certain extent.
One must think in terms of temporal causalities, a melting ice sheet on a lake happens in much smaller time frame, but the transformations are undeniably identical:



There's quite a bit of break-up in the Lincoln Sea but no sign of the Nares Strait unjamming yet...



It would be nice to have a place were projection by professional are regrouped.

For example, Adrew Slater algorithm produce daily predictions.


In the past, I used to open a german page (I think), which also produced daily projection using an simple statistical argument.

Hans Gunnstaddar

On the Arctic sea ice extent graph it shows 2014 currently close to matching the same position as in 2012. However from there 2012 took a dive, and we'll see if this year does anything similar.

Rob Dekker

Ranyl, thank you for your thoughts.
When looking at the decline in Arctic sea ice in September, such as by NSIDC here :


it is tempting to see this development as seasonal noise (due to weather) superimposed on a continued downward trend, which adjusts the norm.

However, from a physical point of view, no "continued downward trend" without a cause. After all, the planet (and the Arctic itself) does not know about downward trends that have to be followed as a new "norm".

So, to explain the downward trend, or when building a model that attempts to make predictions of September sea ice extent, we can use physical variables ONLY.

Take these melting ponds in May for example. We know that more melting ponds will absorb more heat, which will cause more melting (and thus reduce ice extent) later on.

Melting ponds in May thus CAUSE reduced extent later in the season, and thus that variable CAN be used in a physical (probabilistic) model.

But you cannot just assume that because there is a downward trend that therefor melting ponds in May 2014 have more influence than same sized melting ponds in 2012.

Extrapolation of a trend in not physical, and thus should 'time' should not be used in any predictive models.

In fact,

Rob Dekker

YvanDutil, thank you for the link to Andrew Slater's 50-day probabilistic model.

I just checked his poster of the method he uses :

and I think this is very interesting.
Basically he attempts to predict ice extent 50 days in advance by looking at the ice concentration in each area (pixel).

Looking at his first graph on that poster, it seems to me that his method is most sensitive for the areas where ice concentration is between 30 - 80 %. That is where he observes the most 'decline' in ice extent 50 days later.

This suggests that his method works best for the ice - margin. And of course that makes physical sense : if an area has low ice concentration in May, it has a higher chance of melting out later.

Note that his method is NOT so sensitive for melting ponds in the main pack (which typically have 80-90% concentration in May), even though these melting ponds may be important FURTHER out than 5 days, as Schroder et al suggests.

But it is interesting that he gets very nice correlation numbers for his 50 day model and also that his model DOES predict a rather steep drop in the next couple of weeks :


Jim Hunt

@Rob/@Yvan - For the Slater Probabilistic Ice Extent and other slightly out of the ordinary Arctic sea ice "graphs" please see also:



We are unlikely to see a significant drop over the next couple of weeks. Based on Chris Reynolds analysis of the PIOMAS data, in 2012 there was a significant area of thin ice, approx (1400 Km2) ready to melt out at the start of June. This year there appears to be little more than half that. http://dosbat.blogspot.com.au/2014/05/piomas-april-2014-gridded-data.html.
It looks as though we will stick near 2013 figures until about July 15th, when in stock market terms, the bottom falls out of the market.
My calculations suggest we lose about 50 cm of ice thickness up until the beginning of June, 80 cm in June, another 50 in July and then 20-50 after that. This year we need about 210 total to break the 2012 record, which required 230 cm of melt. 2013 had only 190 cm of melt as did 2011.

Jim Hunt

@David - Are you willing and able to share your calculations?

BTW, it has recently been reported that CT Area dropped 286,729 km2 today:



Hi Rob,
"However, from a physical point of view, no "continued downward trend" without a cause."

Isn't Arctic warming is the cause for the downward trend in mean September sea ice extent?

And yes predicting the trend for September needs a physical value, however they de-trended the data, both sides, melt pond and summer sea ice extent.

And thus aren't they looking at the natural variability left after the warming forcing, that is primary cause of the melt situation, is removed.

Happy with the melt pond rationale for increasing etc, however that isn't the warming trend in the arctic and the warming forcing will increase melt ponds I'd of thought and why they detrended to get rid of that influence.

I'm not really sure what you think I'm saying here.

I'm not saying Arctic isn't melting and this is all natural variation in anyway at all, the arctic sea summer sea ice is already a goner some time soon directly due to the warming it is experiencing which is mainly from GHG warming (the WPAC has heated to highest levels for some reason too!)

However the Nature paper is describing the influence of melt pond extent on September sea ice amount around the running mean, and thus a good parameter to define what can be expected that September for a given September mean. If I'd said 2024 instead when the running mean would be much lower, the melt pond extent % will still be a predictor for the ice loss around that new lower mean. And the lowering of the mean does have a direct mechanism driving it, the accelerated Arctic warming due to the huge impulse of GHG gases, indeed if don't stop using all fossil fuels by 2020 humanities civilizations future is then just a mistaken theory.

However can anyone provide a link or insights into what effect the el-nino brewing may have?


As we learned from 2007, big ice thickness is quite insignificant compared to consistent compaction and or direct high sun exposure.
Compaction is already on, quite different than 2013, as I expected.
The direct sun bit ..... May be not, it is quite cloudy, El-Nino has a role with that, however temperatures are key, if there is a significant increase in surface temperatures, this will cause a diurnal cloud effect, the most ideal melt by which the noon sun would reach the surface while the low midnight sun rays would be deflected. The thicker ice zone seems to help generate a dipole, I am quite pleased that this was a correct projection.

Chris Biscan

The ice is already -600K below 2013 on CT and -300K below 2013 on Jaxa.


This was the web page I used to follow:


Unfortunately, they stopped produce prediction in 2012. Maybe someone, could rebuild their model however because this was quite simple.


If you compare CT for 6/1/2014 with the same day in 2007 the difference in the condition of the ice is startling.


Obviously melt ponds have been forming since time immemorial, but what is making the melt ponds in the last few years so deadly to the ice?

Is it a change in weather patterns, something fundamental to the ice, or both?

Is it possible that freezing under slightly warmer conditions creates more solid, brittle ice than would form under colder conditions? And that instead of leaching through and purifying the ice, liquid water is becoming trapped on the surface?


Jim, Chris, yes, that is an enormous dip on CT-SIA. I have no time to check this thoroughly in CAD.
Anyway, from what I did pick up on MODIS, the steep drop does make some sense.
In a mere two days the area with melt ponds and extensive snow melt has 'exploded', FI over Hudson Bay and - Strait, Foxe Basin, Baffin Bay. But gaining traction in the Beaufort, Chukchi and on the East Siberian landfast ice, too.
On top of that, even the Central Arctic Basin shows incredible lead-extension. The Frantsa Yosefa region is splintering up just like last year.

Some effects of cloud moisture cannot be ruled out, like the melt-pond-fooling.
But it is exciting to watch. Even the Nares-cork seems to be under pression...

Obviously melt ponds have been forming since time immemorial, but what is making the melt ponds in the last few years so deadly to the ice?

Is it a change in weather patterns, something fundamental to the ice, or both?

There is a lot more first-year ice than there was before, Rlkittiwake. And this ice is relatively flat (unlike multi-year ice that has lots of ridges due to compaction etc), which means melt ponds form easier on it. It's a positive feedback in a sense.


@ Rlkittiwake

If the CICE simulation is correct, the trend in recent years has been toward earlier pond formation and greater surface area (the latter relates partly to Neven's point about FYI). While the CICE graph doesn't show latitudinal distribution, it's likely a good bit of the increased ponding is at higher latitude: see Werther's mention of CAB among other regions. That's three multiplier effects in the positive feedback.


HI and very interesting as usual Werther "In a mere two days the area with melt ponds and extensive snow melt has 'exploded', FI over Hudson Bay and - Strait, Foxe Basin, Baffin Bay."

If you read to bottom of my article:


you will see Hudson Bay change from a Ta=Ti , its envelopment of warmer air from above and warm sea from below. Started about May 25, 1 week later melt ponds formed...


for some strange reason my formula was truncated

from average Air temperature colder than sea ice to air temperature average greater or equal to sea ice:

Ta < Ti -> Ta>=Ti

Eric Orr

I feel like people are ignoring or forgetting the other big hammer that dropped in 2012 for the Beaufort. The McKenzie was stuck behind the land fast ice and when it broke the massive heat pulse of warm water melted out the highly fragmented sea ice. The gyre had ground up the surface ice and allowed the water to flow between and further into the Beaufort.

The McKenzie was really cold this year and the thicker ice made me worry about an ice jam building up a warm water.

There was some paper I read about ice jam floods in Siberia that suggested there was a critical ice thickness that had to be reached before a serious flood could occur. It was from the 90s and someone was looking at the historic Lena floods of the 60s when in 1963 and 1967 the Lena exceeded the discharge of the Amazon.

Since the Soviets and now Russia monitor the floods and start bombing them from planes to break up potential ice jams.

The big 3 Russian rivers have a lot more ice in their mouths than at the same time in 2012 from my perusal of MODIS.

I expect the Lena to break through its land fast barrier this week on the very North where it has been turning much darker over the last week. I was kinda surprised as the water is backed up on the delta and is flowing North over the thinnest part of the land fast ice.

Considering the high that is expected over that area this week the clear skies should start to allow some serious melting in the Kara as the continental heat from Siberia is transferred down the Lena.

The On is slowly working its way to the arctic but the Yenisey is quote a ways from any serious discharge.

I wonder if the late melting might be worse for the ice. The water that gets backed up is a lot shallower than the arctic ocean and builds up a lot more heat in the south. Once the land fast ice breaks the water that has warmed in the stronger Siberian sun is going to flow right into the Russian side of the arctic and melt a lot of ice before it can mix with the colder waters already in the arctic.

June is when it all starts to happen and I'm really curious how this year will turn up.


Robert Scribbler is keeping a close eye on the deltas. Note: [url=http://robertscribbler.wordpress.com/tag/arctic/]Summer Sea Ice Melt and The Storms of 2014[/url] and [url=https://robertscribbler.wordpress.com/tag/extreme-weather/]Arctic Sea Ice Loss Goes Vertical[/url].
All the deltas appear to be set up for the flash melting. It just comes down to timing (which should be very soon) and a case that the ice dams have to give way al at once without a chance of redamming before the delta empties totally.


I've posted an explanation of my calculations here.
It suggests that in the next 6 weeks the extent will decline more slowly than the last three years but will then catch up in the last two weeks of July.


Note: the calculations are quite crude but display a definite pattern that will feed into what we see.

Chris Biscan

-180K today on CT area.

Essentially tied with 2010 and 2012 for dead last.

Jai Mitchell

Am I the only one who finds that the Cryosphere Today website is impossible to navigate?

where is the direct link to daily Area values?

Jim Hunt

@David - Thanks. I will take a look tomorrow.

@Jai - Try the misleadingly named:


Jai Mitchell

oh, I found it, though it is not labeled and the dates are in 1,000th of a year!

Thanks Neven your Arctic Ice Graphs page is invaluable!

Rob Dekker

DavidR, these are great calculations, and if you are right they are a bit scary if I may say so (projections of 2 M km^2 by September if weather will turn out to follow summer of 2012).

One question : the thickness distribution graph (courtesy of Chris Reynolds from PIOMAS) gives a good overview ice thickness in the Arctic, but to determine which part will melt out in the next couple of weeks, don't you need to know the spacial distribution as well ?

Rob Dekker

Ranyl, I misunderstood your comment.

Yes, you are right that the Nature article (Schroder et al 2014) de-trended their data, so that they could find out the immediate (seasonal) effect of (correlation between) May melting ponds on September sea ice extent.

That way, they could concentrate on the seasonal (albedo amplification) effect of melt ponds (in May) on Sept minimum without being bugged down into a discussion about where the long term down trend really comes from.

However, since they created a predictive model (they even made a prediction for 2013 minimum) I'm actually no longer sure what they did exactly to create that (absolute) predictive model based on de-trended data from melting ponds in May.

Jim Hunt

The Mackenzie River is flowing, and hence the Beaufort Sea is turning brown:


Rob Dekker,
The calculations are too general too identify where the ice will melt. The original idea was just to identify the extent of ice at each original thickness and then calculate the area left as each 10 cm of thickness melted out.
However mapping the estimated area loss against the actual losses showed that the rate of thickness loss was surprisingly consistent out to July 30.
For example 2012 and 2013 had both lost on average between 1.75 m and 1.85 m by July 30th.
That's quite a different impression from the general impression of the differences between the two years.

Rob Dekker

DavidR said

For example 2012 and 2013 had both lost on average between 1.75 m and 1.85 m by July 30th. That's quite a different impression from the general impression of the differences between the two years.

It sure is..
You may be on to something here, and I think it deserves a bit more investigation.

If I understand your calculations correctly, they suggest that the Arctic summer 'melts it's way' through the thickness distribution in an almost linear fashion, with weather playing less of an influence.

And, as a result, almost the same amount of volume melts out in the ice margin (at least until July 30).

If that's true, then this tells that the thickness distribution can serve as a predictor for later ice extent, and that is interesting.

Chris Biscan

Are we talking about modeled loss?

That seems pretty crazy given the amazing difference in total surface insolation by July 30th.


If my predictions pan out I'll look at doing a more accurate study and putting up a prediction for next year.
With warming we should expect a gradual increase in average melt loss. Covering a longer period may enable me to identify an increase in melt rate as well.
I suspect the melt after July 30th may just be too affected by short term variables to be predictable.


"However, since they created a predictive model (they even made a prediction for 2013 minimum) I'm actually no longer sure what they did exactly to create that (absolute) predictive model based on de-trended data from melting ponds in May."

Hi Rob,

I think they took the correlation factors between sea ice extent and melt ponds in May from the year or years preceding the forecast year and forecast using that, so to do the forecast you need the CICE model May melt pond extent and the previous few years correlations I think?

I presume there ins't a good enough correlation between the peak melt pond fraction to the September sea min. as it doesn't capture the early start as well and why they used integrated time and area extent series values.

I also wonder what effect earlier forest and peat fires in Siberia and Alaska have on melt pond formation as a secondary driver akin to Greenland melt?


The loss was modeled and then compared with the actual figures to identify the similarity.
See my explanation here.
The question is how much of the melt is caused by increasing sea temperatures and how much by atmospheric and other variables.

Chris Biscan

In the next 4-5 days at least the Laptev and ESS get totally smashed. The GFS smashes them up hardcore for the next 7-10 days.

The Euro is less invasive but messes them up something fierce.

Lot's of off shore flow with highs in the 60s to near 80F in places with lows in the 40s and 50s.

I guess that would be like highs in the 17-25C range and lows in the 5-10C range.

The open water area which is already huge will get larger the fast ice will turn blue and start showing fractures and stuff.

Or just get really dark with the melt being so off the charts.

Rob Dekker

A bit late, but as promised, an update on the correlation analysis I've been doing using snowcover, ice area and extent as variables to make predictions, now that the May data is in.

I'm using NSIDC 1995-2012 monthly average ice extent and area numbers, and for snow cover using monthly average Rutgers snow lab numbers.

For starters, it is still very hard to make statistical predictions about Sept sea ice extent as far back as April and May, since none of the physical formulas of the three variables gets much below 500 k km^2 standard deviation, which is about the same uncertainty as using an (unphysical) linear extrapolation of a trend over the 95-2012 period.

With that disclaimer out of the way, let's analyze the May data.

May Extent came in at 12.78, Area at 10.66 (both similar to 2011), and snow cover at 17.0 M km^2 (which is more that the past 4 years, and similar to 2007).

A straight analysis using the best correlating physical formula for extent area and snowcover points at a Sept 2014 EXTENT of 4.62 M km^2 (with Standard Deviation 511 k km^2) and AREA of 3.05 M km^2 (with standard deviation of 421 k km^2).

These numbers are higher than the 4.39 extent and 2.84 area numbers I estimated from the April numbers.

The difference is caused mostly by the slowdown of snow melt in the Northern Hemisphere during May.
Snowmelt still has the strongest correlation (R=0.73 in April and R=0.81 in May) with September sea ice area.

In contrast, my poor-man's metric for "melting ponds and polynia" (using simply EXTENT minus AREA) still correlates rather poorly with September sea ice area (R=0.25 in April and R=0.53 in May).

I know that there is a lot of interest in melting ponds in late spring, but either my poor-man's metric EXTENT minus AREA is not a good indicator of melting ponds and polynia, or variability in melting ponds in May is simply not as important as the albedo-changing land snow cover.

Also, I just started experimenting with 'integrating' snow cover and ice area and extent over spring, (combining March, April and May data) and that appears to get me a bit smaller standard deviation for prediction. That suggest that the 4.62 prediction for Sept 2014 is on the high end.

2014 still looks like it going to be coming in low (pointing at 2011 or 2007), but probably not low enough to break the 2012 record.
And current data suggests that it is unlikely we will top 2013's high numbers.

There simply is more energy in the system than in 2013 and less than in 2012.

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been posted. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment