« PIOMAS July 2016 | Main | Iced lightning »


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

Robert S

I'm afraid I take a more pessimistic view of the ice trajectory. 2012's major melt pond anomalies show up in an area which has sustained pretty much annual ice loss every year since then - in other words, these anomalies may build multi-year momentum. Given the poor winter freezing conditions last winter, and the condition of the ice in those areas, I'm not holding my breath for ice retention there. Meanwhile the low compaction is providing another major avenue for heat absorption, and warm air in July continues to dominate over precisely the area where the multi-year ice has been best retained - just north of the CAA. While I'm not convinced that a new record will be set, I would be entirely unsurprised if September ice extent wasn't close to 2012, and clearly in second place.

Cato Uticensis

Great update Neven!

To a 'beginner' like me it's just stunning how many variables come into play for the determination of the September low. As my competences are restricted to meteorology only, I can just keep following the synoptic configuration on the Arctic which actually keeps being characterised by the persistence of LP systems.

These first 10 days of July have been quite different from 2015 and the next 10 days are expected to confirm the current trend: plenty of clouds, prevalence of LP and associated precipitations. From a mere meteorological point of view conditions do not look much conductive to melting also in July, after a generous May and June.

Of course all the other variables that you have mentioned remain in place and this is just what makes it so exciting to follow the ice melt season, in my opinion.


Thanks for bringing this information to us. I subscribe to Robert's comment 100%.

John Christensen

That's a great update, thanks Neven!

With 5.2, Dr. Schroder is now on par with NASA, which until recently was at the high end of Sept average extent predictions.

It seems to confirm what I also said recently that top-3 for Sept minimum extent average is becoming less probable.


Slater's 925mb Arctic Ocean area temps went way above normal in July after a cool June. This is going to be a very interesting summer.

Chris Reynolds recently cited a paper that made a very interesting point. When the sea ice loses its integrity the atmosphere is warming the ocean under it so there's a lot more heat uptake from the atmosphere under broken up thin ice that there is under continuous thick ice under June conditions.

Right now heat and clouds coming off of the Asian monsoon are flowing into the Arctic at high levels. Right now Wunderground shows that the coastal village of Billings, Russia, on the boundary between the Chuchki sea and the East Siberian sea is 67 deg F a fraction under 20C with offshore southwest winds. The air aloft at 850mb is about 15C. These conditions will melt ice like a sauna.

The ice is in horrid shape across much of the Arctic but there isn't going to be enough time for this year to challenge 2012 for the record unless something wild, crazy and unexpected happens. However, Slater's forecast shows a precipitous drop in extent in August in response to the heating of the last 2 weeks. His forecast curve keeps 2016 in the running for the top 3 years in September.


John, the heat spike in July has pushed Slater's forecast down to the low 4 level at the end of August. 5.2 reflects the cool June but don't forget that Jan - June was ridiculously warm. Now July is way above normal.

This year a huge amount of ocean water was heated up because ice extent was at record lows. The open water takes up heat to greater depth than ice covered water because open water is better mixed.

I'm very curious to see how this season plays out. I think we might have a few surprises.



Considering sst found 0 C around the Pole, the standard thinking of predicting events on experiences, such as 2007 style compaction not happening, is obsolete like a nice compact uniform Arctic Ocean pack in July. The sea ice is morphing beyond conservative projection practices. And that will be the lesson learned.

O C sst around the Pole is the warmest water imaginable with the ruble pack extending way beyond wide open water. Last we read, 2012F buoy had sst's much colder to allow a complete melt. 0 C sea water melts every bit of sea ice, especially the top part desalinated by slow brine rejection.


The dipole everyone counts on for predicting a massive melt, the very process causing it, moved further Northwards. It is not conducive to compact sea ice when so. But it does it stop it from moving around. And a probable flip by ECMWF (predicting a massive Cyclone yesterday) calls for a standard dipole forming in a week or so,
perhaps then the traditionalists will change their minds again, since the weather is in line with their thinking.

By the way June 2016 had very wide open water in Barents, strange melt ponds are placed there. Did the anomaly map above depict shrunken ice with fresh and salt water on top of it in the very small pack which vanished by July 1?


And a probable flip by ECMWF (predicting a massive Cyclone yesterday) calls for a standard dipole forming in a week or so, But it does not stop it from moving around.


The ruble is depreciating as we speak and there's nothing Putin can do to stop it. Yes, Wayne, the water in the Arctic has taken up huge amounts of heat for months which will keep the ice melting into at least mid-September. Don't expect this melt season to end early.




I do appreciate seeing some model displays, this helps narrowing what they are potentially doing wrong. In the examples above, the "caveat" melt ponds are truly bizarre, I would say they give "parallel" world calculations:


Finding the cause of model -not- displaying melt ponds, when we saw them, is a very good start to fix the problem.

I also de-magic the greatest issued illusion sea ice has ever played on many fans. The greater gyre is still roaring, the outlook of all this years melt is quite familiar....

Rob Dekker

Hi Neven,
I understand that you have some preference to highlight Dr. Schröder's model prediction (of 5.2 M km^2 in September).

But considering that his SIPN July projection last year was about 700 k km^2 too high, and considering that according to Wipneus' AMSR2 extent ratings, 2015 is still running some 600 k km^2 below 2015, what makes you think this year his projection will be any better ?

Rob Dekker

That should read : 2016 is still running some 600 k km^2 below 2015.
That makes for a 1.3 M km^2 difference that somehow would have to be made up by significant stalls in the 2016 melting season for Dr. Schröder's projection to become reality.

Personally I attach more value to Dr. Slater's prediction method.
Not just because he includes all of "ice concentration" numbers and not just the "melting ponds" section of it, but also because his method has a better track record than Dr. Schröder's model.

And Dr. Slater's model suggest that we will go into September with some 4.3 M km^2 of ice left :
which is hinting at a 4.1-4.2 September minimum, which is in line with my own projection (of 4.1).

In other words : I will personally eat my hat if we hit 5.2 in September.

I understand that you have some preference to highlight Dr. Schröder's model prediction (of 5.2 M km^2 in September).

But considering that his SIPN July projection last year was about 700 k km^2 too high, and considering that according to Wipneus' AMSR2 extent ratings, 2015 is still running some 600 k km^2 below 2015, what makes you think this year his projection will be any better ?

Rob, I see it as a very useful piece of information on melting momentum. As you know, I think this phenomenon has a big influence on the final outcome of the melting season, extent-wise.

I don't think it's all-important, however. There are many other factors, like sea ice volume and ocean heat flux, for instance. But I do think, that at current volume levels, a lack of melting momentum can thwart record attempts. And in the case of this year, a top 3 ranking.

As you can read in this blog post, I personally believe the September average will end up somewhere in the lower range of Schröder's prediction. I think 5.2 million is too high as well, but that's what his simulation produces.

The main reason I highlight it, is because I think it's important research (and public attention may help securing grants for further research) and I'm grateful to Dr Schröder for sharing data in the shape of melt pond fraction maps, as he's probably quite busy.

But again, I think melting momentum is important, and this information is a very useful piece of the puzzle. I hope that in the near future we can add observational data as well (like the stuff Anja Rösel was working on a few years ago).

Rob Dekker

Thank you for clarifying your perspective, Neven.
I just think that there are other perspectives out there (like Dr. Slater's model) that deserve attention.

Dr. Schröder has done marvelous work on melting ponds, and I understand that his work has been incorporated into GCMs even.
But melting ponds are not the whole story of providing melting momentum.

Ice concentration, and land snow cover are equally important, and as Dr. Slater's and my own work have shown in SIPN projections in the past.

Maybe in the end, this melting season is a test of the influence of "albedo" feedback in general against the effect of "melting ponds".

Albedo (snow cover, and ice area) are quite far in the lead for 2016, as work by Tealight and others on the forum have shown.
That heat went somewhere, and despite the lows in June, statistical averaging suggest that that heat WILL come and bite us at some point.

I just think that there are other perspectives out there (like Dr. Slater's model) that deserve attention.

You're right. I will try to mention it in upcoming posts, as the model is getting more interesting/accurate as the melting season progresses. I just have to make some effort to try to understand how the model works and what anomaly persistence is. ;-)

That heat went somewhere, and despite the lows in June, statistical averaging suggest that that heat WILL come and bite us at some point.

Oh, it will bite us at some point, I'm just not sure if that bite will be (clearly) reflected in September extent numbers.

Maybe in the end, this melting season is a test of the influence of "albedo" feedback in general against the effect of "melting ponds".

Definitely. That's why I ended this blog post with: "I have a feeling this melting season will teach us some more valuable lessons."

The thing with melt ponds is that they're situated right there in parts of the ice pack that are tough to melt out in the final phase of the melting season, as they are so far North, often thicker multi-year ice. Melt ponds are a direct conduit for solar radiation into the ice pack, unlike open water in the periphery, or land masses even further away.

Things have simply been too cloudy over too much of the Arctic for too long (even as we speak, and forecasts aren't showing any big changes), although I will admit that this has been a crazy melting season for the greatest part of the year, and I'm not at all sure that even a non-top 5 ranking will mean the ice has been let off the hook.

There are things going on that aren't boding well at all in the long term, like the massively dispersed ice front on the Pacific side of the Arctic (especially Beaufort) and the massive heat and open skies over the CAA and oldest ice region just above it.

And all the heat - SSTs are really high all over the place - that doesn't go into ice melting, will be released into the atmosphere come September and October, messing up weather patterns further.

There's September extent, and then there's long-term changes.

Rob Dekker

Not to mention the collective intelligence of your audience :

John Christensen

A factor often not being considered is the other expansive Arctic ice area - the Greenland ice sheet.

Under the conditions in 2012, very high temperatures were created at altitude on the ice dome, which also caused extreme fohn being created, whenever some of this heated air was moved down from the dome towards sea level and the remaining sea ice.

This year, temperatures in Greenland have been average and even below average in some areas after a warm spring, and higher than average precipitation has caused the accumulated mass balance to be above the 1990-2013 average:


It seems to me therefore that Greenland is one of the factors required to create record-level arctic sea ice melting, and the Greenland weather this year is not contributing significantly.


"But considering that his SIPN July projection last year was about 700 k km^2 too high, and considering that according to Wipneus' AMSR2 extent ratings, 2015 is still running some 600 k km^2 below 2015, what makes you think this year his projection will be any better ?"

Spot on Rob!

If you look at their Melt pond model presentation for June, you will easily find missing water ponds over wide areas. It is not a hard search, I cited one example


but there are others. There is a fundamental flaw in their models. And they keep relying on it. It is standard practice many often experience, when a model predicts rain in your neighbourhood, but it did not happen for instance. In this case, the neighbours on sea ice are Polar Bears , Walruses, Foxes and Seals, sparsely populated area of the planet, less people than Antarctica.

Those relying on flawed models are disciples of the scientific method, even when their computer software is wrong wrong, eventually the models will be better. But they clearly don't realize the flaw in their system, until some of us points it out.


Perhaps the melt pond anomaly model is clever enough to distinguish melted water under snow on top of sea ice? Baffin Bay sea ice is going away real fast now. Yet the anomaly model suggests it was cold there. Yet Baffin Bay sea ices practically gone.... That is quite familiar


Barrow strait ice did the same disappearing act, was once strong and now is gone in a racy few days:


How does Baffin Bay ice disappear so quickly, when the fraction of melt Ponds indicate it was colder? brrrrrrrr or water under snow? That is the question....



Remember the old saying:

All models are wrong; some are slightly less horrible than others.



It is a true saying, but we can make them better, only if we are curious!

Baffin Bay sea ice is going away real fast now. Yet the anomaly model suggests it was cold there.

It doesn't suggest that. The blue means there were less melt ponds there, compared to the average for the last 10 years. That doesn't mean the model is predicting that Baffin Bay won't melt out.

And in fact, just like last year, Baffin Bay is melting slower than many a year, because it was relatively cloudy in the past few weeks.

But anyway, I'll repeat, from the blog post:

Caveat: This is a model result, and so the distribution of melt ponds doesn't necessarily reflect reality.

And from the 2014 Schröder et al Nature paper:

The CICE simulation in this study uses NCEP_Reanalysis-2 data13 for the atmospheric forcing which are available with a delay of less than one day.

Don't bash the model if you don't even know exactly what it is doing. And if you can, point me to a better model or observational method that tells me something about melt pond fraction/distribution.


Hi Neven,

I know what it is not doing! If there is more snow, there will be a lag in the complete melt, like Barrow Strait, the ice appears solid , but it fools, even numeric models, the melt ponds don't show when there should be water. Any other effort, to find holes in this product would not take much time or effort. Bashing is too strong a word, it needs some work, nothing wrong with criticizing. Have you considered finding more errors or imperfections? If not, you wont be convinced.... The state of sea ice is never what it seems without an holistic approach. Besides have you looked at the map they use? The geography seems from another planet :).


Baffin Bay is more melted now than 2010 and 2015 :

"And in fact, just like last year, Baffin Bay is melting slower than many a year, because it was relatively cloudy in the past few weeks."

What about 2010? No one likes a critic :)


This is meant as a reminder and nothing else.

It is always important to remember that models are models and that all models have serious limitations based on their development. They are built with best intent and generally diligent effort, but they are still limited based on limited understanding and limited data.

Models and data should never be conflated or confused with one another.

Data can have problems, and often does and most often from missing data or acquisition issues.

Too often models become accepted as revealed truth forgetting both their origins, their weaknesses, their limitations, and the errors inherent in both the data used in their creation and their own embedded errors and error margins, both from model development, calculational errors and bounds, and most importantly based on missing information and knowledge in their creation.

Far too often models are extrapolated far outside their valid bounds without recognition that this Is even the case. Models generally provide users no warnings about such cases or their limitations.

In the case of climate change it is clear that the models omit huge complications that make the result worse (more rapid and more severe) than the models suggest. This is principally due to missing elements in the knowledge when the models were created.

It is vital to always bear these things in mind.

For what we talk about here, this is most evident in the ice thickness models. They clearly disagree with one another by huge margins.

Portrayed graphically they often do not even look similar.

Using any of those to project very far into the future clearly goes outside the reasonable bounds for the models use.

At the same time, the data and the observational information show clearly that things are worse than they have ever been, worse than any of these models project, and are deteriorating very rapidly.


I'm not interested in 100% accuracy, I'm interested in comparisons between years. If melt pond fraction is simulated in the same way every year, I can use this information to get an idea of the range of possibilities, which makes it easier to write.

And as far as models go, I think PIOMAS is quite kick-ass.

There is always room for improvement, but as things stand, we have a pretty amazing set of tools at our amateur disposal.


Neven, I mean no offense.

PIOMAS may be correct, or not. The US Navy Hycom may be correct, or not. The other models may be correct, or not. Unfortunately we have little in the way of ground truthing for any of the models. And we know they disagree wildly.

The Navy has strong reason for wanting a correct model. They also have the ability to ground truth the model. And they may well have done so whether or not that is ever revealed or acknowledged. Or they may not have.

The only take away that I can make is that there is strong divergence in at least three well developed model sets. That leads me to distrust all of them and to seek other information (like the shattering if the ice sheet) and to place greater reliance on those direct observations than upon any model.

In general I consider all models like I would taking a long road trip with a bear cub curled up asleep in my passenger seat. Sure it is cute and looks cuddly and tame. And it might well be all of those things. But it is more likely that at some point it is going to awaken and I am going to get a very rude surprise.

If I am fortunate I will only have to deal with cleaning bear scat off the seat. It likely isn't going to be that easy or nice.

With models, it is usual and customary to develop an inappropriate degree of confidence in them and to use them until well past their sell by date.

I don't know how things will turn out for PIOMAS, though my caution remains the same. And it is the same caution former US President Robald Reagan was so found of:

Trust but verify.

Sadly, we have little means to verify other than visual observations via satellite. And those images are looking increasingly distressing.


Question regarding the following ice concentration products:



Both are Sea ice concentration maps, but the navy product is showing 50-40% ice concentrations in the center and dropping every day, where AMSR2 is showing close to 100% in the center of the pack, any reason for this?


The Uni Bremen SIC map is observational data (passive microwave), the Navy product is model-based and has a history of errors, glitches and artifacts.

Coincidentally, there's a topic on the ASIF about this very subject.


"I'm not interested in 100% accuracy"

We the Borg, are always gunning for perfection. 100% accuracy is what we live for, those who don't adhere will be assimilated.... But resistance is welcomed, as long as it is polite like most Canadians.

PIOMAS lacks in situ verification, its always good to be skeptical about it. So June was a bad month yet volume dropped quite a lot.
I can be confused, skeptical and appreciative at the same time. It is a nice effort.

I am like a Polar Bear now, which claw shall I use to open this melt pond tent?

I pick 2012. Yes, strange to see that the greatest anomaly was on the other side of the Pole, where all the action was mainly North American sector? Did we loose or collective memory? Let me help:



Yes the bulk of the action was where there was no + anomalous melt pond activity. Why is that? I suggest todays snow layout:


100 + cm of snow. In same sector where usually most of the real melt action occurs. Melt Pond model meaning is thus in question, and needs me clarified and especially refined.

Neven, you obviously did not read my latest piece last night, and I respect that you respect numeric models and especially the people who work for them. But we won't be helpful, if we merely accepted everything they present at face value. If you have read my latest piece written before your criticism of my criticism:


You would have known exactly how I feel about your efforts..... :(


Thanks, Wayne. I'm not ruling out anything you're saying, but this blog post simply reflects how I'm seeing things as of now. I will keep watching, of course, and adjust my views when needed (do that every day anyway ;-) ). I'm doing a new ASI update this weekend.

We the Borg, are always gunning for perfection. 100% accuracy is what we live for, those who don't adhere will be assimilated.... But resistance is welcomed, as long as it is polite like most Canadians.

PIOMAS lacks in situ verification, its always good to be skeptical about it.

Actually, speaking of assimilation, that's the A in PIOMAS (Pan-Arctic Ice-Ocean Modeling and Assimilation System). As it says on the UW Polar Science Centre website:


Sea ice volume is an important climate indicator. It depends on both ice thickness and extent and therefore more directly tied to climate forcing than extent alone. However, Arctic sea ice volume cannot currently be observed continuously. Observations from satellites, Navy submarines, moorings, and field measurements are all limited in space and time. The assimilation of observations into numerical models currently provides one way of estimating sea ice volume changes on a continuous basis. Volume estimates using age of sea ice as a proxy for ice thickness are another useful method (see here and here). Comparisons of the model estimates of the ice thickness with observations help test our understanding of the processes represented in the model that are important for sea ice formation and melt.

Things like IceBridge and CryoSat-2 data are incorporated as much as possible. No model is perfect, but PIOMAS really is kick-ass and an invaluable tool.


ευχαριστίες για όλες τις λαμπρό έργο που επιτέλεσε

Rob Dekker

Regarding Dr. Schröder's model of melting ponds as a predictor of September extent, this paper by Liu et al (2015) casts doubt on their predictions :

Specifically :

Here we show that satellite observations show no evidence of predictive skill in May.


a persistent strong relationship only occurring after late July.

Which suggests that Dr. Schroder's method works best late in the season. But there is more :
satellite observations indicate a much higher percentage of melt pond formation in May than does the aforementioned model simulation

Which suggests that Dr. Schröder still has work to do.

None of this is intended to down-play the work that Dr. Schröder has done. In fact, his work is important in adjusting the IPCC models that we use in forecasting changes in the Arctic and globally.

But for short-term predictions (like from June to September), the performance of statistical models like Dr. Slater's still outperforms Dr. Schröder's model.

Regarding Dr. Slater's model (here) :

Neven said :

just have to make some effort to try to understand how the model works and what anomaly persistence is.

Dr. Slater presents two models in SIPN : a "persistence" model and a "forecast" model.

Note that the 'persistence' model has no skill. Dr. Slater explains this quite clearly in his SIPN report :

None of these methods have true skill at this long lead time.

However, his "forecast" model DOES have skill over 50 days ahead, and even up to 85 days, so that June data is sufficient to make skill full predictions. And last year it was spot-on :
And yes, the "forecast" model is appealing since it is "daily" so gets better the closer to September we get.

It is not entirely clear how Dr. Slater's forecast model works, but from what I understand is that he uses "ice concentration" (a gridded 2D form) as his input variable, so that he also can make forecasts on a regional basis. Here is his gridded forecast for the end of August :

In summary, I think that Dr. Schröder's model (using modeled melting ponds only) is looking at the trees, and fails because it does not use the forest to make September predictions.

Dr. Slater's model uses ice concentration (which includes melting-ponds but also leads, and ice edge fragmentation) and captures more of the overall picture than Dr. Schröder's model.

Now if either one of these scientists would include land snow cover than I would be a happy camper :o)

Nuf said for today.


The relationship is more important to establish than 'the constant'!

The rest is just details... but humanity has already established the relationship is there!

Has it not? Yes, otherwise electric storage would never be seriously contemplated!!


JAXA is back, domo arigatou gozalimas, the latest numbers are big. To be successful, NASA, great outfit, with some of the best scientists and a huge track record of success, needs an average drop of 40,000 km2 a day from now till minima. Unless they know something we don't, a sudden Polar cooling in the cards? They are out of the race. There is no record of 40,000 km2 a day drop in history. But they are good by technique, they have a stiff upper lip, hold on target, don't change our mind technique, however they will have to re-evaluate how they came up with these conservative numbers.

I wish to cheer up Jim and Chris, the Met Office looks real good, hip hip hurrah! Would have been better staying with ECMWF group, most accurate GCM model yet, bravo les Europeens! But their 'new' ? Double 0 Top secret model (not even Jim has access?) is smack in the middle of the current bell curve. So cheer up mates, British science still 'kicks ass' , oh sorry, kicks bottom!

The all time Jaxa July 14 to minima average record is about 62,000 km2, the lowest was 2004 53,819 km2 a day, the highest being 2012 75,295 km2.

Rule Brittania! In science of course....


@wayne: In Blighty, bottoms and monkeys get spanked while asses, donkeys and horses get flogged and sheep and carpets provide a good shag but it's an arse that gets kicked, especially a lazy one. ;-)



Arses it is ! What a pity, kind of like bottoms....

Jim Hunt

Wayne - I'm afraid I posses no inside knowledge on what the Hadley Centre intend to do with their shiny new supercomputer. However I do have a weather forecast of my very own.

Wind and waves in the Beaufort Sea at the weekend:

Time will tell if it proves to be any more accurate than some of the UKMO's recent 3 day forecasts for their own vicinity!


I knew it Jim!

They are keeping things secret, if they would share outputs we would correct things faster than the Cray....
However they are better, so I'll have a special drink in their honour.
Waves, but of course, so many things to consider....


Neven, you are doing a service to us and science by posting Dr. Schröder analysis. Real time examination of how a forecast is made and how it compares to evolving observations pulls us all into the science.

"Based on May/June melt pond fraction, we predict the September ice extent 2016 to be larger than last year: 5.2 +/- 0.44 million km2.

In spite of the mild Arctic winter resulting in thinner ice and an exceptional low May ice extent, the atmospheric conditions during June did not allow much melting. Therefore, our simulated June melt pond fraction is clearly below average, lower than in any other year since 2001. Taking into account our prediction from last month and our current prediction, we expect the September ice extent 2016 to be between 4.5 and 5.2 million km2."

Clearly there is a problem with the model when it shows no melt ponds when anyone on the internet using NASA's web sites can see melt ponds where the model says there are none. No model is perfect but this problem gives all objective analysts a reason to expect the outcome of this model to predict a higher September value than will likely be observed.

There's another, more insidious problem with this melt pond model. At some point ice becomes too thin, too fragmented, too warm and too weak to hold melt ponds. The surface temperature of this ice will appear to be lower than more solid ice covered with ponds because the sea water/ice interface temperature is around negative 1.5 C. The difference between the melting point of ice and fresh water vs ice and salt water can bee seen in some of the thermal imagery such as the JAXA RGB images. Arctic ocean slush may show up as pink or purple on JAXA just before it melts into sea water.

Thus as Arctic ice thins and weakens actual melt ponds may indicate where the ice is in better shape, not worse shape.

The critique cited by Rob in the peer reviewed literature of the melt pond model shows that scientists working in the field of sea ice have identified the need for the model to be corrected. It's not just an unruly mob of "citizen scientists" who are criticizing this model.

Models need to reflect observation and they may need to be adapted to reflect the changing conditions of the Arctic. This year will be a strong test of models.


Jim Hunt

P.S. More on the Met Office Cray:


Sharing the supercomputer will enable collaborative research such as a UK-wide research project to create a next-generation climate model (known as an Earth System Model) which captures all major aspects of the Earth's climate system (oceans, atmosphere, atmospheric chemistry, terrestrial carbon cycle and ocean biogeochemistry).

For some strange reason they neglect to mention the cryosphere.

You can watch it being assembled on YouTube:




NASA has always been quite helpful for us, always providing nicest products, opened our minds to the universe and more. Lets say hello and return the favour. Can you calculate the statistical probability of their projection being successful? 5.2 million at minima, given that it will take 40K loss of sea ice a day from now on. Something that has never been recorded in recent history... Is it accurate to calculate 0% chance or is there a more appropriate calculation? If so, NASA should know.

John Christensen

Agreed on that last part wayne, it does seem odd that NASA can be at one extreme, and so far from the Met Office..

Colorado Bob

The extraordinary years have become the normal years’: Scientists survey radical Arctic melt


Clearly there is a problem with the model when it shows no melt ponds when anyone on the internet using NASA's web sites can see melt ponds where the model says there are none.

Fish, I stated why I find the model results interesting (as a comparison to previous years), but that doesn't mean I'm going to defend the model as being the best thing since sliced bread. No one is saying this, least of all the people from CPOM.

However, I'm going to have to repeat the caveat that "the distribution of melt ponds doesn't necessarily reflect reality." And also that the maps you're seeing show anomaly from the 10-year average, not the existence/absence of melt ponds per se. If it's blue it means there's less melt ponds than on average (white is the same, red is more melt ponds). This doesn't mean there aren't any melt ponds there.

Bill Fothergill

The new Prime Minister of the (not so) United Kingdom has, in her wisdom, scrapped the Department for Energy and Climate Change.

So that must mean that, rather fortuitously, there is no need to worry about the accuracy, or otherwise, of climate models.

The creatures at the GWPF are probably toasting her right now with glasses of Moët & Chandon.

Rob Dekker

wayne said :

Can you calculate the statistical probability of their projection being successful? 5.2 million at minima, given that it will take 40K loss of sea ice a day from now on.

That depends on some factors.

First off is where "average" melt will lead to.

I think you made a point above that average melt from now until halfway September is 60 k/day. If that is the case, the September minimum with "average" weather from now on will lead us to something like 4.05 M km^2 minimum.
That is actually remarkably close to my own projection, so I cannot disagree with that.

Second is the "uncertainty" in that projection.

Now, there are various calculations you can do for that. If you use a straight line decline you end up with something like a 550 k km^2 standard deviation. If that is the case, 5.2 is about 2 sigma's away from the "average" target, which suggests about 2.5 % probability of Arctic sea ice hitting 5.2 or above in September.

If you take my projection method (which includes land snow cover and ice concentration and ice area) then the standard deviation (with June data) is only 340 k km^2. That suggests that 5.2 is about 3 sigmas away, which suggests it is virtually impossible (less than 0.1 %) to hit 5.2 this September. That is why I feel comfortable stating that I will eat my hat if we hit Schroder's 5.2 this year.

More interesting is that given the "average weather" target (about 4.05) and the uncertainties from the statistical method I use, there is a 92 % chance that 2016 will be in the top-3. There is where I differ in opinion with Neven. It really does not take "exceptional" weather in July and August to end up in the top-3. To the contrary : It will take "exceptional" cool weather NOT to end up in the top-3.

One final note on that "uncertainty" : These probabilities assume that the residual is Gaussian "normal", which is not necessarily the case.
After all, I only use 24 data points (the past 24 years), which means that speculating on any probability resolution below 1/24= 4% is questionable.


Much appreciated work, hope NASA reads your statistical dissertation Rob.

At best 2.5% is not good odds. Here is what matters most: How did their model calculate such erroneous monster? All we can do is tell them they are way off, and they will certainly find the reason, at least I wish they do.

Rob Dekker

Wayne, let the ice decide for itself before we start judging.



Would like to make two suggestions to DMI, if you are the man with the hotline to them (if they don't read this blog I would be surprised, but they do outputs I review every day):

1- add a Model display of daily average surface temperatures North of 84N just like they do to 80N.

2- Have the real Arctic Oscillation, call it anything you wish, like AO65+ , which is Pressure calculation index North of 65 N. This one would really be helpful for sea ice projections.

You have a good idea, it needs a little refining.


Another huge 146K JAXA drop yesterday, 3 in a row, from now it will take a daily drop of 69,500 km2 a day for 2016 to become #1. First place is more and more within reach every day there is a huge drop. I suspect I will make a well in advance announcement after the mini or large Dipole synergizes further the larger unstoppably clockwise Arctic Ocean Gyre .



But of course, that is the scientific way. Although It wont be long before they do realize a glare in the works. Many thanks again.

Artful Dodger

Hi folks,

It won't be long before all the sea ice in peripheral seas (ie: Hudson, Baffin, Kara) has melted.

Then, with 2012 AMSR2 data coming online July 23, the REAL race can begin.

Right now, it's still too close to call. With 2012's sunnier June and 2016's warmer Winter, it may just come down to August weather.


P.S. anybody watching Siberian WX/wildfires this Summer? I suspect the 30C temps there in 2012 contributed greatly to the formation, duration, and intensity of GAC2012.

Holy Moses!

Jim Hunt

PS Lodger:


Jim Hunt

Surface and bottom melt have now begun in earnest at the one remaining CRREL/ERDC Ice Mass Balance Buoy - 2015F:

Current Buoy Data (07/15/2016)

Pos: 82.89 N, 137.34 W
Air Temp: -0.76 C
Air Pres: 1011.16 mb
Surface melt: 8 cm
Bottom melt: 11 cm (albeit variable!)
Ice thickness: 185 cm

(Click the image for a larger version)

Jim Hunt


The US Navy Hycom may be correct, or not.

ACNFS forecasts are evidently incorrect at present. The Kara Sea is not covered in sea ice this morning:




I don't think the Navy's model is likely any better or worse than any of the others. My points are these:

These are each models. Models are fraught with limitations. These models are not exceptions to that. Models are all too often relied on to degrees not actually justified by what they are. We tend to forget all of these things. Forgetting these things can lead us astray precisely at the time we most need to not be led astray.

These models disagree wildly with one another and with reality at this point. That disagreement is telling us important information if we will only listen. And this though applied here as a specific case is a general lesson about models that applies wherever and whenever models are used.

What I am NOT doing is arguing that they don't serve useful purposes. They do. We just rely on them more than we should. And that too is usual, expected and all too common. I am NOT arguing that any particular model is better or worse than any other. They likely each excel in some ways, while failing miserably in others. And again, this is usual, common and expected.

The area and extent models suffer similar problems. The thresholds for counting were designed to deal with the inherent difficulties in analyzing the edges of the areas. These never considered the case where the sheet itself is wholly shattered. As a result, as we near the breakup of the whole sheet, these indicators point us in directions that mislead more than enlighten.

This too is usual and expected. Models fail worst when pressed into areas outside the bounds of conditions used or considered in their development.

We humans tend to grab onto specifics. In soing so, we often miss the big picture precisely when that is where we should be focused.

Jim Hunt

The current ECMWF prognosis for July 17th seems to support yesterday's surf forecast:


Nice work Jim!

I am still thinking about wave implications.

IN support of US NAVY much appreciated displayed work, their model and assimilation of data system may be in for a confusing period:


I think there is "breathing" of sea ice on the North Atlantic front. Due to very Rapid melting.


"It won't be long before all the sea ice in peripheral seas (ie: Hudson, Baffin, Kara) has melted."

Not quite Artful,

Look at how fast the Arctic Ocean Gyre is moving, melting and compacting sea ice in the link just given. This is reminiscent of Barrow Strait rapid disintegration of sea ice a few weeks ago.


Loved the sketch though....

We got a thousand points of light
For the homeless man
"We got a kinder, gentler,
Machine gun hand
We got department stores
and toilet paper
Got styrofoam boxes
for the ozone layer
Got a man of the people,
says keep hope alive
Got fuel to burn,
got roads to drive."
Neil Young

John Christensen

Hi Jim,

I believe both ARC ice images are incorrect, but the new DMI ice thickness chart seems to be more accurate, where both Kara and coastal areas of Beaufort are cleared:



Is there a map/model showing the total ice amount, or would such a map be useful?

I'm thinking we look at concentration and know there is ~50% Ice in the pixel, we look at thickness and see there is 2m thick ice in the pixel. Would a map that showed that pixel as 1m total ice in that pixel make sense?

I understand things are not that simple and maybe adjustments should be made depending on how much open water or reflective ice there is as well.

Right now when you consider how fast an area is melting out I feel like both volume and concentration are misleading. You look at concentration and it 40%, but it will take longer time to melt if there is 5m thick ice there or 1m thick ice. Same is true for the thickness graph with regard to concentration.



Did a little birdy tell me you're now anticipating a September low under 1 M km *km ?

Susan Anderson

North America is in a huge hot heat dome, the northern edge up by Barrow was in the 80s (26C+). Re Siberian fires, this was a couple of weeks ago.
I think I saw something about Alaska as well but can't find it, here's what I did locate: http://smoke.arsc.edu/current_fires.html

Crazy circulation up there too, looking lively:

Rob Dekker

Not to change the topic or so, but I would just like to express my disappointment with Cryosphere Today.

In their failure to transfer from the F-17 to the F-18 satellite data source, they are missing out on one of the most interesting melting seasons in history.

And compliments to Wipneus for doing the work for them.

Jim Hunt

AIG - No. Unless it was a remarkably ill informed birdy! In actual fact my money is (somewhat tongue in cheek) on 2022 at the latest:


Of course the "perfect Arctic storm" might occur well before then, but it doesn't currently look as though 2016 will be that year.

Rob - I agree with you about "One of the most interesting melting seasons in history". Iced lightning for example:


My compliments to Wipneus too. His latest work:


dominik lenné

There is an interesting analysis at taminos blog:
in case no one other has noticed it.

John Christensen

wayne provided suggestions for DMI:

1- add a Model display of daily average surface temperatures North of 84N just like they do to 80N.

2- Have the real Arctic Oscillation, call it anything you wish, like AO65+ , which is Pressure calculation index North of 65 N. This one would really be helpful for sea ice projections.

For the 84N vs. 80N, I am not sure DMI would want to add another metric, as this is already a simplified model bringing just the average temperature for each degree latitude and with the same weight for each latitude. Therefore, IMO, the 80N temp charts are really only meaningful from an inter-year comparison perspective and not particularly helpful when comparing with buoy data.

I agree on the second suggestion, which would provide a useful metric for the level of radiation reaching the surface at different times of the year.

I will check with DMI, who are in the process of making some enhancements to the daily ice volume chart and other items on polarportal.dk, so maybe they could add more data, provided they have this available.

The comments to this entry are closed.