« Ocean circulation and Arctic sea ice retreat | Main | ASI 2015 update 4: massive heat »


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


Nice convergence around 5M from Reynolds and Dekker with projections of dedicated research teams. Speaks very well of the knowledge of some people here, they grasp the fundamentals and use very valid simplified arguments. I for one admire them.

The second half of June probably spared 0.5M of final ice extent making the 5 M about right :-)

Is there a sustained weather pattern capable to cause that anticipated big melt of Pacific side AND simultaneous fast retreat of Laptev ice during July? That would tilt final extent toward the side below 5 M ....


So...we can look forward to a plethora of "the Arctic ice recovery strenghthens" articles in denier blogs all fall/winter? Oh joy.


I have no idea how anyone can predict the ice area scientifically if they don't have a clue how thick the ice is. Clearly, to make a prediction one must pick a preferred thickness model or use a statistical approach that uses a distribution of possible thicknesses.

The shift in the atmospheric patterns to heating the Canadian arctic has slowed the rate of ice extent decline over the past 2 weeks. The oldest, thickest ice in the Canadian Arctic has been under heavy attack but that has done little to lessen the sea ice extent. It will be interesting to see what happens when the passages in the Canadian archipelago open up. Open areas have been expanding and the ice has been bright blue on the MODIS images for several weeks in the passages between islands.

The inconsistent wind patterns over the past 2 months have reduced export from the Arctic and lessened the chances of a record low, but I don't understand the predictions that this year will have a higher extent than last year. There has been a lot more heat transported into the Arctic by both the atmosphere and the oceans this year than in 2013 and 2014.

I predict that the extent will rapidly drop over the next month as the large areas of thin ice melt out so this year will challenge the lowest years but it would take an absolute collapse to go lower than 2012. Fortunately, that is not likely.

I see very little good news from this summer's melt season even if September extent is little different from last year because so much of the thickest ice is melting out. There's going to be a big drop in September thickness compared to last year.

Colorado Bob

Rapid Response - LANCE - Gallery - NASA

6 images of smoke and fire

We'll be lucky if Greenland doesn't turn completely black this summer.

Colorado Bob

Seven surprising results from the reduction of Arctic Sea ice cover | David Barber | TEDxUManitoba



I've been watching smoke over the Arctic ice but will ash fall out and cause darkening? This summer is a little different from anything we've seen before. The darkening of the Arctic by fire will be a problem from the Beaufort sea to Greenland this summer.


Bill Fothergill

Bob & George & interested others

RE: Fire in the Arctic

I don't know if you guys are aware of the work of the Paleofire Lab at Exeter University. Here is the link to their website...

cheers bill f

Colorado Bob

Half a Million Acres Burned in Just One Day — Alaska Shatters Record For Worst June Wildfire Outbreak Ever

All throughout the mainstream media last week we heard the same myopic litany — ‘a massive wildfire outbreak ongoing in Alaska is not abnormal.’ Well, today, all pretense that there was anything normal about the 314 wildfires still raging throughout the state has gone up in a cloud of boreal forest, tundra, and thawed permafrost emitted smoke.


Colorado Bob


The Bogus Creek Fire burns in the Yukon Delta National Wildlife Refuge in southwest Alaska, in this Alaska Division of Forestry picture taken June 7, 2015.


FishOutOfWater said:

I have no idea how anyone can predict the ice area scientifically if they don't have a clue how thick the ice is. Clearly, to make a prediction one must pick a preferred thickness model or use a statistical approach that uses a distribution of possible thicknesses.

I have a prediction model for sea ice extent based only on extent and using a time series model. I don't know if you would consider this to be predicting "scientifically", but my predictions last year and this year (as long as I have run the model) have been consistent with the SIPN median.

You could say that I have an implicit thickness model that the distribution of thickness is more or less constant, and therefore I don't need to represent the thickness in my predictions. My way of thinking is that thickness data is not more informative about extent than extent data alone, so there is no need for me to consider thickness.

Vergent Bill

People are fixated on area/extent because edge melt has been the prevailing mode of ice loss. However 2012 demonstrated a new mode: Loss of ice by thinning to zero in parallel, or in sito melting.

In 2013 and 2014 the onset of melt was delayed and the parallel in sito melt did not go to zero. This is the not the case in 2012. The melt as instigated by snow loss is two to three weeks ahead of 2012. There are going to be 20 extra days of high insolation on low albedo, snow free, ice.

I am in agreement with Peter Wadhams. It is a little lonely here, but there is good company. At this point, this prediction still needs cooperation from the weather.


Chris Reynolds


How did your model hindcasts handle the 2013 minimum?

Can your model predict with smaller bounds than mine (+/-0.64) the September minimum using April data?

I ask because I struggle to see how it can work without taking into account volume, when volume goes up, as in 2013, after volume crashed so low in 2012.

Perhaps you could link to where you've explained the model on the forum - I can't recall the details, sorry. But I use Larry Hamilton's extent trend approach as a baseline any method must do better than, and I remember your method as being similar (on which I am very probably wrong!)

In terms of similarity to Zhang's PIOMAS my method is similar, but that isn't really informative as we both will use initial thickness in PIOMAS. After that weather variability is approached differently, although Zhang has a range of 1.2, and I'm 1.28 - not that much worse.

In terms of thickness, the distribution is far from constant over the 1979 to 2015 period. Loss of volume has come from thicker ice, with first year ice volume increasing to make up for the loss of multi-year ice.

FWIW here are my hindcasts from 1979.

Chris Reynolds


Thanks, but speaking for myself, I'm doing nothing but implementing an idea that is quite common amongst the proper scientists.

And whilst I doubt it, I could still be wrong this year.

John Garland,

Denialists? Yawn...


Verg I am with you on this one. I grew up close enough to Lake Erie that many times in my life I have seen it go from thick enough to support ice shanties and snow mobile traffic to spongy ice/open water a week later.

The thinnest ice areas in the Arctic Sea mass now is under 2 meters thick and as you pointed out it lacks snow cover and is getting 24/7 sun at low angles. A sunny period can make that thin dark ice melt out far quicker than intuition leads people to believe.

A lot of people in this discussion both here and in other places I frequent point to 2007 and say it was a perfect melt year weather wise. Indeed it was, but the erosion of hard ice in that banner year has never been replaced. The ice is now much thinner and becomes 'rotten' much more easily because one warm wind storm early in the season sharply reduces the snow cover that insulates the surface from direct sunlight.


Details on my model are scattered here and there on the forum. I was planning on a larger post describing my model and analyzing its performance, but I've been inordinately busy for the past few months and haven't had time to complete it.

Larry Hamilton just fits the data to a sigmoid of some kind, right? That's not what I'm doing at all. I don't have a function I'm fitting the data to. My methodology is closer to describing the data as a random walk with constant drift.

My estimated standard error on predicting September extent from April data is 0.356, or 95% CI of + or - 0.70. (I never report more than one decimal digit.) That's estimated under a normality assumption. One of my goals is to produce a bootstrap estimate, which I expect to be wider due to slightly heavy tails.

My method doesn't explicitly retain predictions several months in advance, recomputing all future predictions every month. So I've had to back out historical predictions from the data. Assuming I've done it correctly, hindcast prediction errors of September extent from April data are listed below.

1980 0.42
1981 -0.22
1982 0.03
1983 0.18
1984 -0.15
1985 -0.33
1986 0.42
1987 0.3
1988 0.3
1989 -0.07
1990 -0.84
1991 -0.33
1992 0.82
1993 -0.42
1994 0.43
1995 -0.62
1996 1.4
1997 -0.076
1998 -0.23
1999 -0.48
2000 -0.19
2001 0.3
2002 -0.45
2003 -0.14
2004 -0.1
2005 -0.51
2006 0.0098
2007 -1.6
2008 -0.81
2009 0.053
2010 -0.39
2011 -0.49
2012 -1.4
2013 0.71
2014 0.56

Obviously I had a big miss in 2012, In 2013, the model predicted a rebound, but underestimated it. In 2014, the model expected a rebound on the rebound, which didn't happen.

Chris Reynolds

Thanks for the added detail OSMM,

I was right, I had totally misremembered what your method was about. But I wonder if a similar mechanism was at play in the 2013 'rebound' as that which would apply to Larry Hamilton's sigmoid fit.

How long is the persistence of data in your model? I am wondering if the rebound found in 2013 was due to a return to 'trend'? I am assuming that the constant offset is negative and would act in a similar manner to the 'trend' of a sigmoid fit.

The fails in 2012 is not surprising, there is a graphic at SIPN/ARCUS that shows similar fails for the suite of models in 2012. But in a rushed search just now I couldn't find it.

I will think about this today, but have to get off to work now.

Nightvid Cole

Chris, I think you might mean this:


Bill Fothergill

@ Verg & Tanada

Whilst I think that Peter Wadhams may well be correct in his assertion that, one day, the ice will simply (and quickly) fade away, the basic thermodynamics of ice melt suggests we have yet to reach that stage.

Although the comparison with the sudden melt out at Lake Eire might seem compelling at first, there are some massive differences to consider.

The first of these is, of course, the widely differing thermal inertias that are involved. The average depth of Eire is something under 20 metres, as opposed to the Arctic's mean depth of just over 1 kilometre. So the thermal inertia of the underlying waters differs by a factor of 40 between the two.

Most, if not all, of Eire lies below the 45th parallel, and that's a bloody long way south of the Arctic Circle. The term "ice free" when applied to Arctic Sea Ice conventionally means that the September average will have dropped to below 1 million sq metres. The entire area lying above 85N is pretty close to 1 million sq km, and, compared to this, Eire's surface area of about 25 thousand sq kms is, quite literally, a drop in the ocean.

Given that the meridional circumference is almost exactly 40,000 kms, we're talking about places that are separated by over 4,400 kms, and so the differing levels of insolation are not to be sniffed at. At any time, and all other things being equal, the insolation is going to be proportional to the Sine of the Solar Elevation Angle.

At local midday on the 85th parallel on midsummer day, the Solar Elevation is going to be 28.4 degrees. Meanwhile, down on the 45th, the elevation will be 68.4 degrees. By way of comparison, the Sines of these elevations are 0.476 and 0.93 respectively.

Obviously, there are compensatory factors, specifically the number of daylight hours. At midsummer on the 85th, there are 24 hours of sunlight, with the elevation dropping to 18.4 by midnight. On the other hand, down at 45N, there are only about 15.5 hours of sunlight.

So, for several months - centred on midsummer - the average daily received power is actually higher at 85N than at 45N.

I'm going to go out on a limb here, but, if and when Peter Wadhams vision of a sudden melt out happens, I think it will take place during the month of August*, rather than September. At really high latitudes, I suspect any significant melting will have had to occur before we get to September. (* But I think that's quite a few years in the future.)

cheers bill f

Chuck Yokota

One thing I would point out is that due to the halocline, only the top 50 meters of the Arctic Ocean is near the freezing temperature; below that it the temperature rises. So only the top 50 meters of water needs to be warmed, giving a difference in the thermal inertia of about a factor of 3 rather than 40.


Hi Chuck,

I've been disappointed how rarely the arctic depths are discussed here. I'm sure the undersea currents of the arctic ocean is no less important than any other factor in determining the state of the ice. Perhaps it is the sheer difficulty of gathering data on the under sea temperatures that causes it to be under stated. I feel like modelling the arctic is like the fable of the blind men touching and describing an elephant. I suspect that Wadhams is qualitatively right, even though it seems implausible that the Arctic sea ice plunges to such a great extent this year. I'm quite sure that most models would fail to predict the precise year of melt out the year before the melt out actually happens. The existing data sets would be barely relevant to the terra incognita of that fateful year.

Bill Fothergill


D'oh! I completely forgot to check the depth of the "barrier" created by the pycnocline - which itself can be either a thermocline or a halocline, or indeed both.

However - and I am obviously prepared to be corrected here - when I look at NASA's "Ocean Motion" website, it seems to suggest that the deep ocean more or less connects all the way to the surface at high latitudes.

Have a look at this link and see if your interpretation is different from mine.

I never did any oceanic stuff at university, so this is all new to me and could therefore easily be misunderstanding something bloody obvious.

However, the North Atlantic Drift Current does eventually sink way up there, and, as it does so, it feeds into and becomes the North Atlantic Deep Water. My (very limited) understanding is that, having travelled up from the Gulf of Mexico, and having been subjected to evaporation for the whole way, the NADC is just about the densest, saltiest surface water anywhere. As it cools, it gets even denser, eventually ending up as NADW.

Therefore, every joule that it picks up from insolation whilst flaffing about up there prior to actually sinking, finds itself getting buried into the deep ocean, hence extracting "heat energy" which would otherwise help to melt the ice from underneath. This, so I understand, is what helps provide the thermal inertia.

This in many ways is a great example of the 2nd Law of Thermodynamics in action, with the concomitant "leakage" of usable energy into a low grade, unusable entropy increase.

But, of course, that could all be utter bollocks.

cheers bill f

Chris Reynolds

Thanks Nightvid,

The image I was thinking of was fig 1b in this article.

Chris Reynolds


I'm pretty sure that when the April thickness is low enough for a <1M km^2 extent for September we will see a June ice loss that will outpace anything we've seen to date, that by July it will be obvious we're in for something big, but that losses in August will probably be less than the current average.


The problem with ocean currents is we have no data that informs us within the melt season. As for longer term and more regular impacts: As an example, try Nghiem 2012 - "Seafloor Control on Sea Ice" In that paper various cases where ocean bathymetry and currents have repeated impacts on ice are discussed, throwing light upon the shape of the ice pack at various times of the year (or maybe it's just at maximum and minimum - not read it for years).


Bill you said [quote]However, the North Atlantic Drift Current does eventually sink way up there, and, as it does so, it feeds into and becomes the North Atlantic Deep Water. My (very limited) understanding is that, having travelled up from the Gulf of Mexico, and having been subjected to evaporation for the whole way, the NADC is just about the densest, saltiest surface water anywhere. As it cools, it gets even denser, eventually ending up as NADW.

Therefore, every joule that it picks up from insolation whilst flaffing about up there prior to actually sinking, finds itself getting buried into the deep ocean, hence extracting "heat energy" which would otherwise help to melt the ice from underneath. This, so I understand, is what helps provide the thermal inertia. [/quote]

In the deep past this was true, however in the current climate regime the North Atlantic Drift does not sink because of higher salinity like the waters of the Mediterranean Sea do. What happens under the current climate is the NADC encounters and mixes with fresher surface waters as it travels north. When it reaches the point where its temperature is low enough to make it denser than the surface waters it sinks, but it only sinks a few hundred feet at first. The salinity at that point is only slightly higher than that of the water at about 100 meter depth so it sinks to about that level. As it gradually cools it sinks further and further until it reaches the sea floor a long distance south of where it initially sank below the surface.

The Arctic Basin is actually fairly sharply stratified by temperature and salinity. So long as the inflowing surface Atlantic Drift gets diluted significantly before sinking the Arctic Bottom water is stable and remains in place in perpetuity. In the distant past this was not the case. When the Arctic Basin was ice free surface evaporation outpaced fresh water influx from rivers and this resulted in highly saline water eventually sinking. Because of the high salinity it would sink all the way to the bottom of the basin displacing the deep Arctic Bottom water and forcing it over the sill to flow out into the Atlantic deep basin. After many years a circulation would be set up like that in the current Mediterranean Sea with relatively warm highly saline water filling the bottom water portion of the basin and then itself exiting over the sill into the Atlantic basin.

Here are a couple links about the stratification as it stands today,


Bill Fothergill


My wording was ambiguous. The intention was to say that, in whatever year it happens, I think the 1 million mark would be passed before September.

Like you, I think that when the ice is truly rotten, the really big losses will happen closer to the insolation peak.

Bill Fothergill


Thanks for the enlightening explanation and links. It's never too late to learn - even for old gits like me.

Have you seen Blaine's comments about AMOC variations on the Ocean Circulation thread?

Chuck Yokota

Here is a chart demonstrating the salinity and temperature profile of the Arctic Ocean. The temperature of the deeper water is warmer than the surface layer, but the salinity gradient keeps the fresher surface water less dense.

Adam Ash

I guess, as a mechanism which could lead to an abrupt Wadham's-style ending, there could be a levelling of the thermal gradient across the ice pack.

I imagine, of old, that ice tended to melt from the edge of the pack inwards towards the pole, and the ice itself was colder the further north it lay. But these days, with melt ponds at 90 North, clearly the whole ice mass is getting warmer through its full thickness, even if it is still below 0C and pretending to hold hands.

Could there come a point where the majority of the ice pack is so close to having absorbed all the latent heat it needs to get it up to the edge of melting, whereby all it needs is the final (massive) thermal shove to complete the job and become liquid.

I know the heat required to get from ice at 0C to water at ditto is a lot, but the more ice that is closer to 0C then the more a decent heat wave could wipe out a huge extent very quickly.

The buoys don't give me enough coverage to check this, but the levelling of the thermal gradient across the latitudes, and in depth, would seem to be a reasonable mechanism to consider as a way to position large extents for sudden extinction without the effect being obvious from just extent or area observations.

Jim Hunt

Adam - Here's the temperature profiles from the furthest north IMB buoy, 2015D at 86.90 N, 2.57 W:

Click the image for a closer look. The core temperature is above -3 now, but still has a way to go before the core temperature of the floe is >= the water beneath it.


July polls on the Arctic Sea Ice Forum are up: NSIDC SIE September average and CT SIA daily minimum.


I should really be posting about this on the forum, but let me say a little more about my extent model. Larry Hamilton fits the data to a sigmoid. When the data comes in well below the fitted curve (in 2012), the prediction for the next year is that the extent will return to the trend curve (rebound in 2013).

The most simple version of my model is a random walk with drift. That is, the year over year change is a random number with a non-zero mean. (In this case, -50 thousand sq km.) By assumption, the mean and variance are constant, and the year over year changes are independent. For predictions, the point prediction is computed from the mean, while confidence intervals are estimated from the variance. This means that the set of future predictions at any time is a straight line.

However, this line should not be interpreted as a trend line, because a trend line implies that if the observed values deviate from the line, they will still return to (or near) the trend line in the future. In the case of a random walk, any deviations reset the y-intercept of the line. This produces a new linear prediction, but there is no expectation that future values will return to the previous prediction line (or any other previous prediction line).

My model finds that the year over year random changes are not independent. They are negatively correlated (rho = -0.49) which means that if one year has greater than predicted loss, the next year is predicted to have less than average loss, or even a gain. (This predicted gain could be called a rebound.) This happened in 2012 and 2013, where the loss was so much larger than predicted in 2012 that my model predicted a gain in 2013, but still underestimated the size of the gain.

Because there was a larger gain than predicted in 2013, my model predicted a larger loss than average in 2014. This did not come to pass, and the loss was below my prediction. As a result, my prediction for this year is again that the loss will be larger than average.

There is no longer term persistence in my model (I don't use changes from two years ago to predict this year's loss). There is not any statistical evidence that considering more than one year would improve prediction accuracy.

Note that even with the rebound effect, this is a random walk model. There is no trend line that the data is expected to follow.

As another complication, my model also accounts for monthly changes, in the sense that if this month's (year over year) change is larger than average, next month's change is also predicted to be larger than average. A physical interpretation is that if this month ends with less ice than expected, next month effectively has a head start on melting.

When the March extent this year came in at a record low, there was a greater year over year loss than average, which means that the predicted year over year losses for all future months increased. This predicted change dies out rapidly, so there was a substantial change in the prediction for April, but essentially no observable change in the prediction for September.

Colorado Bob

Fires and smoke in northern Alaska
22:40 UTC


The fires are starting to cross the Brooks Range on to the coastal plain. The smoke is now blowing out over the Arctic Sea Ice.

Colorado Bob

The Northeast coast of Greenland today :


21:20 UTC

Note the smoke coming from Alaska and Canada.

Colorado Bob

Fires and smoke in northern Canada

19:10 UTC


Chris Reynolds


Thanks for that detail.

Larry Hamilton,

I was wrong about Prof Wadhams, he is serious about a crash in sea ice this year or within a couple of years.
I must admit myself puzzled as to how he reaches this conclusion.

Jim Hunt

Thanks for the heads up Chris. My consequential musings:

Is Time Running Out for Arctic Sea Ice?

Yes is the answer to that question, according to Prof. Wadhams at least.

Bill Fothergill


In the video clip you linked, Prof. Wadhams was certainly unequivocal in his views. However, just like you said, there was nothing there which was particularly convincing. (Hand-waving and smiling earnestly at the camera doesn't count.)

I take it that you also spotted the bloopers.

Those with some familiarity with Wadhams' involvement with the Arctic Methane Emergency Group would not have been surprised to see the relatively shallow East Siberian Sea getting a prominent mention.

Then, staring at about 10:10, Wadhams twice reversed the locations of the Kara and Laptev Seas. If it had happened just the once, one would be tempted to write it off as someone accidentally saying something in the wrong order. However, to mis-identify them twice was less than impressive - especially for someone with his background.

Even worse was the utterly bone-headed question asked by the interviewer at about 16:55. She had fixated that methane was 27 times more powerful a GHG than CO2, and then asked if a single molecule of methane broke down (oxidised) into 27 molecules of CO2.

I actually had to play that bit several times, as I could scarcely believe the inanity of the question.

Jim Hunt

Here's your starter for ten Bill:

What was the title of Peter Wadhams' PhD thesis?

Here's a little clue for you. I'm seriously considering writing an academic paper on the same topic!

Colorado Bob

The smoke leaving Alaska, and on to the Arctic ocean -

Note the 2 fires burning on the tundra.

06:15 UTC


Colorado Bob

Smoke over the Greenland Sea

04:30 UTC


It appears these “rivers of smoke” are now a part of our atmosphere this summer.

Colorado Bob

Fires and smoke in northern Canada

19:00 UTC


The long hot summer , really takes hold. Note the smoke in the upper right hand corner . that’s Alaskan and NWT smoke.

Too bad “senator snowball” isn’t there throwing snowballs at these fires.

Colorado Bob

Smoke is wild card , it's not in the models, and if the northern bogs, and peat, and tiaga really start to burn, and they are. We are deep trouble.


We're deep trouble, indeed, CB. :-)

Bill Fothergill


The thesis was called ...
"The effect of a sea ice cover on ocean surface waves".

Obviously, my extremely limited brain does not carry such arcana, and I had to look it up.

Presumably I can now confer with myself for the 3 bonus questions?

cheers Bamber

Bill Fothergill

Actually, his thesis paper dovetails very neatly with something that Prof W was talking about on the video.

He stated that an increasing ice-free area would act as a kind of positive feedback (although I'm pretty sure that he didn't use that term) through the mechanism of increased wave action upon the remaining ice. Almost a corollary to his original thesis.

Chris Reynolds


I must admit that after the first few minutes I concluded 'more of the same' and was doing something else whilst only vaguely listening. Sorry to inflict the job of paying attention on you.

Wadhams increasingly reminds me of Lovelock, as Lovelock was. Revenge of Gaia was equally hand wavy, with an interesting core argument, then a massive leap to DOOM! DOOM! DOOM! Recently Lovelock has retracted his prognostications of imminent doom.

Compare 'Revenge of Gaia' with Hansen's 'Storms of my Grandchildren', or Archer's 'Understanding the forecast' and the difference between alarmist* hyperbole and densely reasoned, evidenced, and subtle scientific argument becomes very apparent.

Wadhams is wrong. We'll see exactly how wrong (in million kmsq) by September. By September 2019 we'll see that the 2016 +/-3 was wrong too. Everything points towards the post 2007 regime being a new regime of decline, not the prelude to a fast crash.

*Alarmist - the definition of this is "someone who exaggerates a danger and so causes needless worry or panic." Raising the alarm about the serious issue of climate change is not alarmist.

Bill Fothergill


"Compare 'Revenge of Gaia' with Hansen's 'Storms of my Grandchildren', or Archer's 'Understanding the forecast' and the difference between alarmist* hyperbole and densely reasoned, evidenced, and subtle scientific argument becomes very apparent."

Yep, I've got both "Revenge ..." and "Understanding ...", and I've read "Storms ..." a couple of times. I couldn't agree more with your assessment.

Bill Fothergill


NOAA has just published the June figures for the Nino 3.4 region (+0.97C), and that will push the rolling 3 month figure for April-June up to +0.86C - which will duly be rounded up to +0.9C. This represents the highest 3-month value in 5 years.

So why did I say "Farewell to el Nino"?

Well, it would appear that NOAA have performed a re-analysis on the numbers, and this extends throughout the entire file. By way of example, from Nov 1985 onwards, the climatology figures have gone up by, typically, between +0.02C and +0.04C.

The absolute figures have also moved slightly throughout the entire file - they would obviously have had to do so in order for the climatology values to change.

The immediate effect has been to take us out of a fully declared el Nino event. Prior to the reanalysis, the rolling 3 month figure had crossed the +0.5C threshold at the Oct - Dec period last year, and had remained above the threshold ever since.

However, post the re-analysis, the Jan - Mar period has now dropped to +0.43 - which, of course, rounds down to +0.4C

As we no longer have a minimum of 5 contiguous 3-month periods at, or above, +0.5C, all that red at the bottom of NOAA's Oceanic Nino Index (ONI) page has just disappeared. When the August monthly figures appear, I fully expect that the 5 rolling periods, FMA - JJA, will all match or exceed the threshold, and a somewhat cutailed splurge of red will reappear.

For those of you who stick these figures into a spreadsheet as part of your SIPN calculations, or for any other reason, please note that this is NOT the normal 5-yearly update on climatology. The next one of those is scheduled for early 2016.

When that update is implemented, the climatology applied to every year from 2001 onwards will be updated, but there will be no effect on earlier values.

cheers Bill F


Chris: Wadhams is off in his own little corner of the graph. Apparently he has let his emotions cloud his observations. Earlier this year I anticipated the heat from the "blob" in the north Pacific plus El Nino and warmth in the Barents would cause accelerated melting and a record melt year. But that's not what's happening. Too many clouds and variable winds have led to a modest melt year to date. So be it. We're lucky that the melting has slowed down since 2012.

Wadhams has made the mistake of getting emotionally invested in the Arctic Methane Emergency Group. Their proposed geoengineering schemes are ill considered and frightening to me. But I do not agree with:

"Everything points towards the post 2007 regime being a new regime of decline, not the prelude to a fast crash."

Everything does not point. The Arctic continues to be variable and unpredictable. We don't understand it very well. As Colorado Bob points out, smoke is not in the models. A shift back to more deep convection in the north Atlantic as took place in the early '90's isn't factored into the models AFAIK. There a a large number of variables affecting melt rates over the next decade that are very hard to predict.

So, I'm watching and learning what I can. This year could still surprise us in the Arctic but the surprise is unlikely to be a year like 2012.

Bryant Morganelli

I find myself disagreeing a bit with Bob. The wildfires in Alaska are mostly contained for the moment and it still has a long way to go to catch up to the 2004 total for the entire fire season.


Bryant, you might want to follow what the authorities are saying directly. We just hit July and this season is *already* in the top ten burn years.


Bryant Morganelli

Jdallen, that's not really surprising at all. The fact remains it still has a ways to go to catch up to 2004 and conditions are improving at the moment. Not great, but not as bad as a week ago. Indeed, wildfires become more common as time goes by with warming. This is nothing unexpected to scientists.

Bryant Morganelli

Not to say it's not concerning. It certainly is, but as with sea ice this year and people saying it would be below 2012, let's wait and reserve judgment.

Colorado Bob

Arctic temps warmer than Miami? We have a serious methane problem!

Current Arctic Weather Conditions

According to Arctic News, as of July 2nd, “While the media gives wide coverage to the heat waves that have been hitting populous countries such as India, Pakistan, the U.S., Spain and France recently, less attention is given to heat waves hitting the Arctic.”

Furthermore, “The heat waves that hit Alaska and Russia recently are now followed up by a heat wave in East Siberia… a location well within the Arctic Circle… temperatures as high as 37.1°C (98.78°F) were recorded on July 2, 2015.”

And, even more, “With temperatures as high as the 37.1°C (98.78°F) recorded on July 2, 2015, huge melting can be expected where there still is sea ice in the waters off the coast of Siberia, while the waters where the sea ice is already gone will warm up rapidly. Note that the waters off the coast of Siberia are less than 50 m (164 ft.) deep, so warming can quickly extend all the way down to the seabed, that can contain enormous amounts of methane in the form of free gas and hydrates.”

Also, on July 1, 2015, a temperature of 36°C (96.8°F) was recorded near the Kolyma River that flows into the East Siberian Sea.

The Arctic is hotter than Miami!

Somehow or other, 98°F in the Arctic makes the world seem upside down/sideways. Is it?


Colorado Bob

Bryant Morganelli

I'm sticking with NOAA -

Alaska wildfire season worst on record so far

Fires are raging in Alaska, and there’s no end in sight.

More than 600 fires have burned in excess of 1.8 million acres in the state, according to the Alaska Interagency Coordination Center, making this year the worst wildfire season so far in Alaska’s history. Fires have caused evacuations, highway closures, and rail and flight disruptions. More than 350 structures have been damaged, including about 70 homes.


Colorado Bob

Bryant Morganelli -

Our analysis of 65 years of Alaska wildfire data shows:

The number of large wildfires (larger than 1,000 acres) suddenly increased in the 1990s, and the 2000s saw nearly twice as many large wildfires as the 1950s and 60s.

In the Arctic region, the number of large wildfires increased nearly tenfold in the 2000s compared to the 1950s and 60s. Only three years in the 1950s and 1960s saw large wildfires; there have been 33 large wildfires in the Arctic since 2000.

The area burned in large wildfires each year is increasing. In just two years, 2004 and 2005, wildfires burned a larger area than in the 15 years from 1950-1964 combined. In particular, there has been a dramatic increase in wildfires larger than 10,000 acres but smaller than 50,000 acres.

Alaska’s wildfire season is about 40 percent longer now than it was in the 1950s. The first wildfires start earlier in the year, and the last wildfires are burning longer into the fall. Overall, the wildfire season has increased more than 35 days and is now more than three months long, running from May through early August.

Rising temperatures across Alaska have been concurrent with the rise in the number and size of Alaskan wildfires. Years with the hottest May to July temperatures also tend to be years with the most fires, and the greatest area burned.

According to the National Climate Assessment, the amount of area burned in Alaskan wildfires is projected to double by 2050 and triple by 2100 under continued emissions and further warming.


Colorado Bob

Bryant Morganelli _

App;es to apples :

June of 2004 saw 216 fires in June 2004, scorching of 1,153,258 acres; this June saw 404 fires and 1,875,984 acres consumed, the division reports.


Colorado Bob

Bryant Morganelli _

The Greenland melt page note the melt graph -


Note how we've blown through 2 standard deviations like a volcano.

Bryant Morganelli

The fact that you use Arctic News as a source is enough to make me question your credibility, what with their focus on spikes instead of overall trends. If you look closely at the NOAA CH4 trends, you will see a less alarming trend. And I've kept careful track of the wildfire situation, and the mistake you're making is looking at just the month of June instead of the whole year.

Bryant Morganelli

And the sea ice is on track for well above 2012. It's even above several years prior to 2015, so that goes to show you that your melt projections are off like Wadhams.


@Colorado Bob
I was just about to post that link :)

Prodigious melting this year, and I think with the pack melt, both combined stats will make a truly remarkable year.


"...The fact that you use Arctic News as a source is enough to make me question your credibility..."

Ad Hominem foul!

You need to present arguments which favor your assertion. You want to rule out the source presented? Present fact which refutes what's presented.

"And the sea ice is on track for well above 2012. It's even above several years prior to 2015, so that goes to show you that your melt projections are off like Wadhams."

Please. There are a lot of very smart people who disagree, and I think the assignation of "Off like Wadhams" Has more subjective slant to it than box full of Fox commentators.

If you are so sure, rather than posture here, Bryant, come post your data and analysis on the forums.


"It (the sea ice) is even above several years prior to 2015"

This is such an oversimplified misleading assertion. It reminds me to the typical sentence from that other guy, Cincinnatus.

Bryant Morganelli

Whatever, you two are [snip]

[Come on, keep your cool; N.]


Wow sorry man this is on you


Neven is doing a great job of keeping this blog focused on the science and the data. I, for one, appreciate how he encourages differences of opinion as long as it stays on topic and sticks to the science.

This summer the warmth in the Canadian Arctic is going to test the models and remote sensing measurements. The narrow passages and strong currents in the channels may melt quickly under the intense heat. We'll see if the models can handle the details.
-George aka FishOutofWater

John Christensen

September extent: 5.4 +/- 0.2

Factors in my consideration:
- Still more volume in central Arctic basin compared to last year (Waiting for latest PIOMAS numbers to confirm)
- Current extent now a bit higher than 2013 and 2014

However - the unknown factor seems to be the forecast: AO seems to stay medium-negative leading to higher air pressure, less cloud cover. While this should lead to less oceanic heat and moisture reaching the Arctic, it will increase in situ melting, which is still significant only a couple of weeks since mid-summer..

Robert S

The more I read about and watch the arctic ice dynamics, the more I think that extent as a single measure is not what we need to be focused on. While I recognize the difficulties, I think we're getting close to the point where we ought to begin switching to a sequence of measures - a hierarchy of volume, then area, and then extent (really for use with area and base ice albedo/melt ponds) as a proxy for total system albedo... or possibly a hybrid number which would capture the energy dynamics of the arctic ocean system... which at the end of the day is probably what we are all interested in, in the face of global warming.

The focus on extent seems to me to be an artifact of an earlier day when we had less data and less ability to process data.


What strikes me most, looking at the CT area comparison 2014/2015 for July 5th is how little ice there was in the Hudson Bay/Baffin bay in 2014, compared with how much ice there is in the same place in 2015.

Yet 2015 is slightly below 2014.

Also the completely different pattern of melt ponding with, subjectively viewing, something like 50% more melt ponding in 2015 and some of that close to the pole.

We know that the ice in Hudson and Baffin will melt and can melt very rapidly. So, in essence, some rapid drops could be in the chart again in the near future. Which would put 2015 back in the top 3/4 again.

More monitoring needed.

Chris Reynolds

John, Robert,

Using PIOMAS daily gridded data the error stats on my calculations against the PIOMAS main series of volume is no more than 0.16k km^3 out, with an SD of 0.002k km^3. So I have used that data to do an early calculation of the PIOMAS spring volume anomaly.


So the spring volume loss is muted again this year and doesn't suggest a large loss this year. e.g. see what happened in the above plot for 2012.

But I have pointed out over at the forum that the Bremen plots show a persistent tendency to lower concentration for the last 15 days over the Pacific half of the Arctic Ocean. I'm not changing my opinion for now (4.5 to 5.15M km^2), but I think a large loss this year might happen as decline in compactness/area within the bounds of extent leads to strong drops in extent. Much like Slater's persistence model suggests.


Anyone who thinks we are going to have another rebound year might want to take a look at the Bremen AMSR2 image from today.

It's either massive melt ponding confusing the sensors or it's ice actively vanishing. Either way that ice is not long for the world.

That's a LOT of ice currently vanishing.

Chris Reynolds

Neil T,

Normally I would say wait and see, but I think this is real and could be the start of something exciting.

I do agree, it is very likely that this year will not be another year like 2013 and 2014.


Chris, I was not that surprised because I've been watching the Barrow webcam page and it has a weatherbug temperature include on the top right. It's been hovering between around 43F and 63F for days now with the general daytime temps around 50F to 60F. Clear skies mostly and warm.

So I was wondering when that would turn into melting in the region and it looks like it might have.

Rob Dekker

Just submitted my revised SIPN prediction for the July report :

4.6 M km^2, with a standard deviation of 370 k km^2.

Most of you will know that my method is based on a metric for the "whiteness" of the Arctic in June, as a predictor for the amount of ice that will melt out between June and September. I use Rutgers' snow cover as a variable, as well as NSIDC's (extent - area) as metric for "dark water next to ice" as well as of course, "area" as a metric for whiteness of the Arctic in June.

Reasoning is explained here :
and below in the "short description" of the method as I submitted to SIPN.

Now, Rutgers snow cover in June 2015 just came in and it is VERY low (second lowest only after 2012) :

Also, while in May, sea ice was pretty compact (VERY low extent while area was just below average) June showed a major decrease in area, while "extent" reduced much slowed, and thus ice "concentration" collapsed. You can see that very well in Wipneus' graph :

Result is that at the end of June, snow cover being VERY low, a LOT of heat is being absorbed by land, and major areas of the Arctic sea ice pack show melting ponds.

This means that the Arctic became a LOT darker in June than it was in May, and thus my method, which predicted 4.9 M km^2 in May, now predicts significantly less ice in September : 4.6 M km^2.

Standard deviation on this prediction is about 370 k km^2.

Here is the method explained in more detail, as submitted under the "short description" to SIPN ARCUS for the July report :


*** September 2015 monthly average projection :

Pan Arctic Ice EXTENT : 4.6 million km^2 (with 370 k km^2 standard deviation on the prediction)

*** Short explanation of outlook method :

The basic concept behind my method pertains to estimating albedo-based Arctic amplification during the melting season.

I use the “whiteness” of the Arctic in June as a predictor for how much ice will melt out between June and September.

Specifically, I set up a formula which reflects how “dark” areas near the Arctic in June would create heat that will melt out ice over the months until the September minimum.

As an educated guess, such a formula could take the following form :

Melt_formula = 0.25 * Snow - 1.0 * (Extent - Area) + 0.5 * Area

With factors explained like this :

For (Extent - Area): 1.0 (assuming that ALL solar radiation onto melting ice and into polynia will cause ice to melt later in the season.

For (Area): 0.5 (assuming that half of the heat absorbed in the ocean OUTSIDE of the main pack will cause ice melt (while the other half would cause the ocean to warm up.

For (snow cover): 0.25 (assuming that half the heat from lack of snow cover will be blown North, and half of that will go to ice melt.

Then I set up a regression equation for how much ice will melt out between June and September :

september_extent - june_area = alpha + beta * (Melt_Formula) ;

When I tweek the factors, to obtain the best fit over the 1995-2012 range, the ‘Melt_Formula' that obtains the best correlation (R=0.93) is this one (centered to (extent - area):

Melt_Formula = 0.434*snowcover - 1.0*(extent - area) + 0.65*area

Which is remarkably close to the “educated guess” factors explained above. This suggests that this formula is realistic, and the effect is physically real.

Using this formula, for the period 1992 - 2013, I obtain R=0.93, beta = 0.5588, and a prediction for Sept 2015 ice extent of 4.61 million km^2 with a standard deviation of 370 k km^2.

The “beta” of 0.5588 means that for every km^2 of polynia/melting ponds in June, an extra 0.558 km^2 of sea ice will melt out between June and September.

And the 0.434 factor on snow cover with that beta means that for every 1 km^2 snow cover loss in June, some 0.242 km^2 of sea ice will melt out by September.

The interesting issue is that the standard deviation (at 370 k km^2) is significantly better than the 500 k km^2 or so that would be achieved for a simple linear trend. This means that the June “whiteness” signal is apparent in the September sea ice minimum, and serves well as a predictor.

As for past performance, here are the results for what this method would have predicted for the past couple of years :


2006: predict 5.34, final 5.91, delta 0.57

2007:predict 4.79, final 4.29, delta -0.50

2008: predict 5.01, final 4.72, delta -0.29

2009: predict 5.66, final 5.38, delta -0.28

2010: predict 4.39, final 4.92, delta 0.53

2011: predict 4.60, final 4.61, delta 0.01

2012: predict 3.71, final 3.62, delta -0.09

# -2013: predict 4.89, final 5.35, delta 0.46

# -2014: predict 4.93, final 5.28, delta 0.35

# -2015: predict 4.61, final ????

Note that the years tagged with a # are NOT part of the 1992-2012 regression learning period.

This suggests that the “whiteness” of the Arctic in June, as expressed in the regression formula, using snow cover and (extent-area) as well as June “area” itself, explains a large part of the increase in September ice extent during the 2013 and 2014 season w.r.t. 2012 and other years.


Well done, Rob. My compliments.

Rob Dekker

Thanks Neven, but we'll see in September.

Also, I must say that I am relieved that the Arctic did not immediately continue its nose-dive after 2012, but instead continued a more linear downtrend.

Which of course is good news (or at least not terribly bad news) for the Arctic, although it puts my bet with William Connelley at risk :

But I must say again, as I did back then in 2011, when I stated that the Arctic sea ice decline was running way below the linear decline line, and showed signs of quadratic collapse (in volume), that :

I sure hope that I am wrong. Because if I am right, Arctic sea ice is in much worse shape than the IPCC expects it to be, with potentially disastrous consequences for Arctic wildlife and climate patterns across the entire Northern Hemisphere.

And thus, I'd be happy to pay up next year if I was proven wrong.

Jim Hunt

Rob - I've also got some hard cash at stake on the state of Arctic Sea ice! All in a good cause of course. See:


et seq. On the other side of the wagers are a certain "Bit Chilly" and a name you may be more familiar with, Chris Reynolds.

Rob Dekker

Interesting bet, Jim !

My gut feeling is that you have a good chance of loosing that one.
And I also think that Chris Reynolds is overly confident with his 1000-to-1 bet.

But I did not run the numbers yet.
So let me do that first, and I'll get back to you with my opinion on your bet.

Jim Hunt

Rob - I await your return with your numbers with much interest!

Do you also fancy ensuring $1000 ultimately goes to a good cause, one way or another?

Rob Dekker

One more note about my projection adjustment. About latent heat and time delay from heat sources.

During the final stages of the melting season (late August and early September) the Arctic does not receive enough sunlight any more to melt by insolation directly.

The heat that melts the sea ice at during these final melting stages is thus caused by latent heat. The heat from the land, the air, and the water around the Arctic which has build up during the earlier months.

So late stage melting (August and September) should be affected by how much heat was build up at lower latitudes.

And thus, it is my expectation that my main variable (snow cover in high summer) will affect late stage melting (in August and maybe even September).

With this year's June snow cover being VERY low (second after 2012) this is a good one to test my theory that snow cover in summer affects mostly late stage Arctic melting.

If this year's minimum will end up above 5 M km^2, then my theory (of snow cover's delayed effect) is probably incorrect. Maybe latent heat build-up by snow cover decline in summer does not affect late stage melting much.

But if this year's melting season will show persistent declines in August and maybe even September (and thus end up at 4.6 or below) then we should consider that snow cover early on DOES have a strong effect on latent heat buidup and thus late season melting...

Food for thought.

Jim Hunt

Here's Andrew Slater's latest SPIE "prediction":


More food for thought? How about Prof. Wadhams' latest video?


What might be the unfortunate side effects of a "GAC 2015" or two?



SPIE prediction: How reliable is an algorithm that forecasts a huge drop, then a stall, and then moderate decay, all this during August 2015?

Perhaps somebody could point me to the source where the model details are explained . . . because I must be missing something


OK I got it. Slater Probabilistic Ice Extent. Will be reading it :)

Rob Dekker

Hi Jim,
I owe you an opinion about your bet with 'bitchilly' and with Chris Reynolds.

If I understand correctly, the bet with 'bitchilly' is that you win if Sea Ice Area as recorded by Cryosphere Today dips below 1 million km^2 by 2022. Otherwise you lose.
And if I understand the bet with Chris Reynolds correctly, it uses the same conditions, but the wager is 1000 to 1 in your favor.

Now, I'm assuming CT SIA is from this record :

First, I found the minimum SIA for each year for the entire record (1979-2014), and regressed it against time.

I obtained a beta (SIA minimum decline per year) of 0.069590.
So each year the CT SIA minimum reduces 69 k km^2.
Also, I obtained an 'alpha' (reference at year 2000) of 4.114881 million km^2, a correlation factor R=0.88 and a standard deviation over the decline line of 388 k km^2.

What that means in plain English is that the predictions for various years, including 2022 for CT SIA minimum, using a long term decline line, comes out like this :

2012: predict 3.28, final 2.23, delta -1.05
2013: predict 3.21, final 3.55, delta 0.34
2014: predict 3.14, final 3.48, delta 0.34
2015: predict 3.07, final ????
2022: predict 2.58, final ????

with the long term standard deviation of 338 k km^2 over that prediction, the 1.58 million km^2 difference by 2022 represents some 4.67 sigma. Now 3.2 sigma is already 0.1 % probability, so if we go by the long term trend, and we assume Gaussian distribution, then you made 2 bad bets, which you are likely to loose (even considering Chris' 1000/1 odds.

Second test I did was regress over a shorter period, which reflects how SIA decline accellerated, and using the period since 1992-2012 which excludes the 'rebound' 2013 and 2014 years.

That experiment is kind of a 'best-case' scenario for your bet, and here are the results :

21 years analysed :
beta : -0.113878 (114 k km^2/year decline)
alpha (referenced to year 2000) : 4.235367
Correlation (R): -0.888165

Standard deviation over the predicted decline line is 357 k km^2, and here are some key years :

2012: predict 2.87, final 2.23, delta -0.63
2013: predict 2.75, final 3.55, delta 0.80
2014: predict 2.64, final 3.48, delta 0.84
2015: predict 2.53, final ????
2022: predict 1.73, final ????

The 0.73 million km^2 difference between 1.73 and 1.0 to win your bet is now only 2 sigma, which means that with this modern trend you have something like 2 % probability of winning these bets.
That means that 'bitchilly' still made a good bet (by a factor of 50 in 'bitchilly's favor), and Chris Reynolds made a bad bet (by a factor of 20 in your favor).

Now, all that said, let me add a note.
These regressions are based on the assumption that sea ice decline will be somewhat linear over the 7 years to come, and that the SIA uncertainty distribution is roughly Gaussian.

And that is not guaranteed.

It may be that there are more non-linear effects kicking in once the Arctic SIA cover get closer and closer to 1 million km^2. Which, after all, represents a wide open Arctic ocean, one that the planet has not experienced for a very long time, so we don't have any president for that situation.

It seems to me that you have placed your bets on that situation realizing before 2022.

That is possible, but we don't have any evidence of that being a realistic scenario.

I would like to spend some more time to see if there is any indication of such accellerated melt for particular years that may suggest that accellerated (non-Gaussian) melt occurs when SIA dips exceptionally low ((2012 comes to mind).

But until then, my opinion is that you probably made two bad bets.

Rob Dekker

That last sentence should read as follows :
But until then, my opinion is that you are probably going to loose both bets, but your bet with Chris Reynolds is still in your favor.

That is because I think that the accellerated shorter term (since 1992) SIA trend is a fairer representation of reality than the long term trend (since 1979).

Jim Hunt

Some pertinent comments in a personal communication from Peter Wadhams:

A-Team - "Re. the Sikuliaq cruise, my own work on board will be a close-in wave-ice experiment where we will deploy 10 buoys just outside and just inside the ice edge to measure the microscale scattering of wave energy by individual floes at the edge. Since the ship will be staying around into the freeze-up period we will then measure wave penetration through the new ice (which may be in the form of pancake ice)

Re press officer, I only wish we could do it, but we have had to battle for every berth just to do our basic work. The ship is really crammed. Which of course will raise an interesting question if there does indeed end up being an exceptional retreat this year. We sail on Sept 28, which is 10 days after the normal sea ice minimum, but we may end up having to go a very long way to find ice, and this would certainly be of interest to the media."

Thanks for that analysis Rob, but as some small comfort - "My SIPN prediction is an outlier but it is in fact what is predicted by the 5-year trend in ice volume in September. In 2013 and 2014 the volume in summer came in above this trend, but since then we have started a new El Nino which tends to increase air temperatures more. I am just staking this area out as a marker, and expect to be shot down in flames."

Chris Reynolds


My bet was financially poor for me, being that if I win Jim pays me £1, if Jim wins I pay him £1000. But it wasn't a financial bet, it was a test of confidence. My skewed monetary value (i.e. not £1000:£1000) was purely because I would have felt bed taking money off Jim.


Hi Jim,

After a swim in one of the Great Lakes really good tropical blue fresh water to avoid sharks :), I must say the water languished indeed colder, testimony to winter just past, nicely summed up this way: http://eh2r.blogspot.ca/2015/02/rrr-explained-by-open-sea-water-and.html

I did expect Baffin Bay ice to be stubbornly resilient because of the imprint of a dry winter and therefore a lot of late melting in that region.

Therefore all those guys who strictly judged extent, say 2 weeks ago, who commented how not so strong a melt was occurring were fooled by not considering last winter imprints. However Neven
being quite a good sea ice analysts nailed it right as it happened.
Must keep in mind the staggering difference between last year late August melt blitz, and the same look one month earlier now particularly in Polar sector 90W to 90E clockwise. It will be 2012 record bye bye announcement time soon.

Mean time I would like to point out to contrarians fake skeptic champion Cincinnatus failing his "no ice melting past Wardle Island" , perhaps one of the greatest prediction bomb of this year. I remember reminding him that open water in Bering Strait in March is conducive to such events we witness now. Of which bombastic, bellicose and especially none descriptive predictions, arm chair stances, based on "a gut feeling" are absolutely not to be taken seriously.

Which leaves me with Wadhams:

"In 2013 and 2014 the volume in summer came in above this trend, but since then we have started a new El Nino which tends to increase air temperatures more. I am just staking this area out as a marker, and expect to be shot down in flames."

Good to read from you Dr Wadhams! Please clarify your prediction for this year more! There is a distinction to be made, between those who think that annoying researchers with stupid repetitive prognosis or projections will flood the internet
and therefore bury the gem true science sources to the oblivion of confusion.


Hi, Rob. I enjoyed seeing your hindcast above, and was particularly impressed by the delta=-0.09 for 2012.

I have an idea for correcting for the "data point in the training set" bias you refer to:

Calculate your parameters by deleting the year you are calculating.

(I.e. if you are using 1992-2012 as the baseline, then when calculating 2006, you'd use 1992-2005 union 2007-2012, but not 2006)

My guess is that it would still be fairly darn good and would allow you to more accurately determine of the correct confidence interval, etc.

Rob Dekker

Thanks, Bfraser !
I was surprised myself that the 2012 extent was so nicely in line with my hindcast (using snow cover and open water next to ice from the June records only).

However, it may be that the 2012 hindcast was merely a coincidence of choosing the training period (1992-2012 in my case).
A different training period (especially one including the years before 1992 and after 2012) would obtain a higher trend line, and thus a prediction further away from the actual 2012 minimum.

Problem is that we really don't know exactly where the "trend" line is.

What is more interesting in my opinion is that my method (of including snow cover in the prediction) obtains a better standard deviation than a plain linear trend (or even a Gomperz curve prediction).

That tells me that the albedo effect during summer melting season is quantifiable in snow cover and extent-area in June and this significantly affects the September minimum.

What I'm still looking for is if the summer melting season is significantly affected by other variables, or if going to a smaller timescale (like weeks instead of months as I did so far) would improve our estimate of albedo feedback during summer.

Rob Dekker

Also, I appreciate your suggestion to exclude each year from the training set.
Let me see how that affects the results, by apart from the obvious expectation that it will increase the standard deviation somewhat, it seems to me that it has no physical meaning to do so, and instead would be only be a statistical exercise.


The physical meaning is that it lets you ask the question "Do all years obey the same rules governing melt?"

As I understand it, the plan is to run the "tweaking the variables" process multiple times, each time leaving out a single year from the training set, and then seeing how well the excluded year is forecast by the (tweaked) variables.

That gives you two bits of information.

1) Are all years forecast equally well, or are there individual 'outlier' years that are poorly forecast? A reason for this could be (e.g.) an unusual wind setup that only brought winds into the Arctic through the Bering Strait, over sea rather than over land. This would mean that for that specific year, land snow cover would have less effect than in other years.

2) When you compare the different runs, so you always end up with the same parameters, or is it wildly variable between runs? If you get a different set of parameters for each different training set, it becomes much harder to believe that the overall process is reliable.

The comments to this entry are closed.