« ASI 2014 update 2: here comes the Sun | Main | Jakobshavn calves a(nother) big one »


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


2014 is less than 2010 at the end of May according to the text document. 2014 is 21.9 vs 22.2 in 2010. http://psc.apl.washington.edu/wordpress/wp-content/uploads/schweiger/ice_volume/PIOMAS.2sst.monthly.Current.v2.1.txt

All of the end of May values from 2010 to 2014 appear to be effectively the same given the margin of error. Now it all depends on the weather and what's going on under the ice.


I used the file that can be downloaded here. I don't know which file is the correct one.


D has linked the Average for month the end of month figures are
Day 151 31 May:
2014 20.288 4th lowest
2013 20.498
2012 19.591 2nd lowest
2011 19.483 lowest
2010 20.229 3rd lowest
2009 22.431
2008 22.878
2007 21.891


Thanks, Crandles for pointing out the differences. The weather stayed cooler for another week or so this May, I think. Stormy Junes are recovery years, like last year, but now the high pressure is setting up. I guess Neven's right that we're a bit behind the record years so far because of the extra week of storminess in May.

Helen Wiggins (ARCUS)

Hello Arctic Sea Ice Blog community. I wanted to pop in and encourage 'citizen science' submissions to the Arctic Sea Ice Outlook this year! All info is at: http://www.arcus.org/sipn/sea-ice-outlook


Hi Helen

The models fail because they appear incomplete in repeating reality, just comparing their prediction results, may be useful, but not as good as exploring the nature of sea ice in order to find out what they miss. As a better way, I would like to see their actual daily (weekly or monthly) predictions, including and especially ice surface and air temperatures(up to 10 meters ASL). There is enough talent here participating on this blog to help out resolve the major errors in their replication simulations. When a model predicts right or wrong all we know is their names or where they are from. We can't help without more details.

Chris Reynolds

Thanks Helen,

You've spurred me on to update a prediction I made using April PIOMAS data, to using May data. An email is on its way.

For those reading, my prediction is:

NSIDC Extent: 4.06 +/-0.57, M km2.
The range being: Upper limit 4.62M km2, lower limit 3.48M km2

So I'm betting on meeting 2011 or pushing it into fourth place.

My main prediction remains the one after CT Area 20th July data is out, that will be a more confident prediction.

Rob Dekker

Chris, Helen,

I just submitted my ARCUS projection for Sept 2014 to Helen :
Extent : 4.6 +/- 0.445 M km^2
Area : 3.0 +/- 0.371 M km^2

This projection is based almost exclusively on NH land snow cover data (Rutgers) from April and May.

Chris Reynolds

Thanks Rob,

That's three of us I know of. My method is based on May PIOMAS volume, but really needs the Greenland ridging and associated Arctic high to develop.


Will you be putting forum/blog votes in this year?


Larry Hamilton did that last year. I'd have to ask him if he plans on doing it again this year.

John Christensen

Area (CT) and extent (ROOS) predictions from my perspective are based on current ice conditions (thickness, concentration), which are average OK, weather forecast, which is positive for ice conservation due to weak, broad Arctic lows in the coming week, 80N DMI temp, which is low also due to weather, and the NAO forecast, which is negative and could lead to more Atlantic moisture moving North in the coming two weeks:

- Area minimum: 3.4 +/- 0.5 M KM^2
- Extent minimum: 5.2 +/- 0.7 M KM^2

Rob Dekker

Hi Chris,

I wonder if we can combine the two methods and obtain an even better Sept prediction (smaller standard deviation) using only May and earlier data.

Here are my thoughts :

I get good correlation between NH snow cover in April and May (a metric for the amount of ENERGY absorbed by the Arctic regions), and September extent, but that implicitly assumes unchanged thickness of ice.

Specifically, my statistical model implicitly assumes that most of the ice that will melt out during the melting season will be FYI, and that that FYI thickness does not change much year-to-year, and that MYI will mostly stay in place.

Now that makes perfect theoretical sense (I hope) except that the real Arctic does not work like that. There will be years when very little MYI melts out and years when MYI stays nicely in the area which will not melt that year.

Keep that thought in mind, and then look at what my method would have predicted for the last couple of years (again using only land snow data available in May) :

2007: predict 4.69, final 4.30, delta -0.39
2008: predict 5.23, final 4.73, delta -0.50
2009: predict 5.76, final 5.39, delta -0.37
2010: predict 4.17, final 4.93, delta 0.76
2011: predict 4.77, final 4.63, delta -0.14
2012: predict 4.40, final 3.63, delta -0.77
2013: predict 5.78, final 5.35, delta -0.43
2014: predict 4.60, final ???, delta ???

Notice a couple of important things :

(1) Snow cover in April and May alone does explain a 1.38 million km^2 difference between 2012 and 2013 Sept extent. That's 80 % of the 1.72 million km^2 that eventually separated these two years. And that is using data only up till May !

If 80 % of the variability between years is explained at the end of May, then maybe summer weather has less of an influence than we all have attached to it since the 2013 melting season....

(2) that for the past 7 years (with 2014 to be determined), my method has consistently overestimated the amount Sept ice left over, EXCEPT for 2010 !

It is almost as if in 2010, the ice was thicker than in other years, or, the thick ice was located in an area that eventually melted out.

Now, you (Chris) have frequently reported about the great loss of MYI during the 2010 melting season, and the excessive volume loss during that year (possibly even shifting the Arctic into a new 'thinner' MYI state) during that year is also apparent in the Wipneus PIOMAS graphs :

So, I'm wondering.
What was the 2010 ice thickness distribution like during 2010 ?
What caused 2010 to be a year when a lot of heat was in the system, and large volumes of ice melted out that was thicker than other years ?
And I'm wondering how that thick ice in the melting margin in 2010 was related to 2009, an exceptionally cold summer year, comparable to the summer of 2013.

And thus, I wonder how the ice volume spatial distribution charts from PIOMAS could possibly tell us if 2014 is going to follow 2010 with lots of thick ice in the margin, or 2011 for a more average ice thickness distribution...

Any thoughts..?

John Christensen


Great analysis and very interesting that late spring NH snow cover can explain Sept minimum to that degree!

Looking at the data for 2010, my take is that this year saw a significant deconsolidation of the Arctic ice pack in the sense that a lot of MYI was lost and significant fracturing took place, allowing the remaining, broken ice to spread better, bumping up the ice area and more significantly the ice extent.
2010 also had significant positive NAO during summer months, bringing warm moisture to the Arctic, assisting in fracturing the ice.

2007 was the opposite situation, where the MYI held the central pack well together and consolidated, but with significant melting taking place from open waters at the edge of the pack.

For 2013, your analysis predicted a very high Sept minimum, but even assisted by a consistent negative NAO during the summer months and ice-preserving weather conditions overall, the Sept minimum fell below the prediction, probably simply due to the very thin, fragile ice cover.


If we look at mass buoys closer to the Arctic Ocean open shore, we find that underside melting has occurred for a good while now:


all have had lower than 0 C average surface temperatures, but their surface ice temperatures were equal, hence the hypothesis (more refined and explained) found here:


looks good .

The central Arctic Buoy ,


did not show any sign of melting despite similar surface temperatures, the average ice temperatures were mostly always colder for the same latest days.

Again the hypothesis is confirmed.

Thick sea ice is way more complex to understand because it exists with 2 different interfaces. I find this complexity simplified
by knowing refraction and physical phase change physics of water.

By the way its nice to see A-team back, I think a map showing the average temperature differences of sea ice vs surface air would help pinpoint where sea ice underside melts every day. I don't have the software to do it automatically :(.

Chris Reynolds


The 2010 volume loss was in large part due a mass export of MYI from the central Arctic into Beaufort, then Chukchi, and the ESS. This volume export put the MYI into a position where a large amount of volume melted out.

Using week 20 images (late May) from the Drift Age Model shows the export into the Siberian Sector that year.
The PIOMAS thickness distribution for April shows the bulk of thick ice that moved towards Alaska, the start of what led to that tongue of old ice being transported towards Russia.

There also seems to have been a reduced thickening over the late winter, here are Feb to April temperature anomalies.

However I started with the volume export of MYI because by August a persistent mass of ice was present around the Chukchi Sea which stubbornly refused to melt out meaning (IIRC) a lot of people got caught out expecting a lower minimum.
In my opinion the reason for that persistent ice (which was very low concentration later in August) was the presence of MYI which proved surprisingly resistant to melt.

The model I use is the relationship between PIOMAS volume for May (originally April) in the Arctic Ocean (Beaufort round to Bering, Greenland Sea, Central Arctic, and Canadian Arctic Archipelago), and NSIDC Extent, both using monthly averages. For my submission to SIPN I took the residuals between observations for the 2007 to 2012 period, assuming that a new climatic regime during that period was what we should expect in the future and that 2013 was an anomaly. I used the maximum and minimum residual to establish the range of the prediction - really rather simple.

I've been messing around with the residuals since late April, trying to establish relationships between them and summer weather, as the summer is the time of greatest volume loss. But I've had no success.

However as the example of 2010 shows, factors that no simple numeric model can consider come into the equation.

Actually I've just found a better animation of the 2010 export from a blog post this year:

There has been a similar export this year:

Although I said that I thought this export would reduce melt in Beaufort and Chukchi this year, I'm not very convinced about that now due to a mass movement to the Atlantic in May, and the thinner more dispersed nature of the MYI exported.

I'm not very convinced about my prediction based on May data. I still see the June spread of melt ponds as critical to determining the melt season and have doubts that prediction before then can be much more skillful than simply the trend and its uncertainty bounds. My main prediction remains the one in late June. I had been planning on 20 June, but may delay if the June cliff is still continuing on that date. The season has started more slowly than I had expected.

Rob Dekker


Thank you so much for these great animations of ice thickness distribution. Especially the last two (for 2010 and 2014) are illuminating.

The 2010 animation confirms your assessment that there was a large tongue of very thick (5 meter) ice across the Chukchi blocking melting into the Central Arctic Basin.

Of course, 5 meter thick ice requires 5x the energy (or time) to melt out that 1 meter FYI, so it seems reasonable that this tongue caused the slowdown in extent melt in 2010 while still recording record volume losses.

That is consistent with the 'anomaly' in 2010 in my "spring snow cover - Sept ice extent" correlation method.

So I think your animations show that 2010 was indeed special, and maybe the PIOMAS drop in volume, while keeping extent high, was indeed caused by that tongue of thick MYI crossing the Chukchi.

Did I interpret that right ?


Back then we called it The Arm, Rob (not knowing what it was exactly, and wondering whether it would survive the end of the melting season).

Rob Dekker

Neven, wow. That's interesting.
It is totally consistent with Chris' animations of a MYI tongue in 2010.
Did anyone attempt to estimate how large (in km^2) that arm was back then ?

As for 2014 showing a similar tongue, I'm still on the fence.

Chris' animations show a tongue of MYI, but it seems confined to the Beaufort.
IIRC Cryosphere II showed a tongue of MYI from the CAB to the East Siberian shore (although I can't find that image right now; it is late).
And I wonder how good PIOMAS is in simulating MYI drift. Chris' 2010 PIOMAS picture
does not seem to show any arm or tongue to speak of..

Interesting stuff !


I believe this is the CryoSat-2 image you refer to, Rob (from the NSIDC's monthly analysis for May):

Jim Hunt

@Rob - I've gathered together lots of "thickness" related information at:


The "arm" visible in the CryoSat-2 image above seems to be in a rather different place on ASCAT for example, and absent in the PIOMAS gridded data. All very confusing if you're trying to forecast what will happen to the sea ice over the next three months or so!


Confusing! In past years where things have gone off the rails for us is that of figuring out how long a system would stick around. See 2012, 2013 as examples. In general though we did have a pretty good handle of what was happening at the time and usually 2-3 days out at least. This year I have noticed a lot of difficulty figuring out what is going on presently. Very few times have the models even agreed 2-3 days out, and even what is going on at the time seems to be in disagreement at times.
On top of that For example we look at MODIS and see water and extent will say full of ice. If my understanding is right there have been a couple of newish satellites giving us info now. Could it be that there are still some hicups between the raw data being collected and the models giving us the info? And another thing that is different, there seem to be very few researchers up there giving first hand accounts so that we can compare between the two.
In the end it would not surprise me in the least that if we are not surprised at what happens in Sept. we will be surprised next years results. I am convinced there is melting going on the we can not see or misinterpreting because this is definitely a confusing year.

Chris Reynolds


Yes you interpret correctly. However the PIOMAS image was for April and as the DAM shows export continued around the Beaufort Gyre through the summer. PIOMAS doesn't explicitly model MYI/FYI so I use thickness as a proxy for it. Ice over 2m thick being generally MYI because FYI generally only thickens thermodynamically to around 2m thick.


If PIOMAS is correct, and I think it is then there is a lot of melt going on right now. The anomalies of volume show that the unusually large spring melt is indeed underway.

I think the unusually large spring volume losses in PIOMAS are correct because they start from 2010 and since 2010 NSIDC Extent losses in June have jumped - the PIOMAS spring volume loss happens over May and June.

This is why although I now think my prediction of Extent based on May PIOMAS volume in the Arctic Ocean is going to be too low I'm not writing it off altogether.

Rob Dekker

ARCUS sipn June report (based on May data) is out :

And yes, you seem to be on the low end of predictions, but considering the large standard deviations from all of these contributors, indeed we can't write it off.

Chris Reynolds


From my most recent post of 20 June, here's the predictions and uncertainty bounds for the SIPN June outlook submissions.

I may be at the low end, but I'm in good company; UK Met Office, US Navy, and Blanchard-Wrigglesworth are in the same ballpark.

The recent shift in the atmosphere looks promising, I still suspect we've lost any chance of 2014 competing for the top five. The next two weeks and PIOMAS data at the end of that may change that opinion.

Bill Fothergill

@ Chris & Rob

Regarding making the "top five", the NSIDC SIE data on Charctic for the solstice date makes interesting reading.

On June 21, 2012 and 2010 were virtually identical at ~ 10.3 million sq kms. As you know, they are currently in 1st (3.63) and 5th (4.93) places for the Sept average - quite an astonishing divergence.

June 21 2014 is virtually equal to the 2008 figure for the same date at ~ 11 million sq kms. 2008 currently is in 4th place for the Sept average on 4.73 million sq kms. I've no strong view on where 2014 will end up - although I've thought for some time it would likely end up somewhere between 2008 and 2011.

On the SIPN table, it was intriguing to see that the WUWT offering of 6.1 million sq kms neatly bisects the 2003 and 2004 September averages. Looks like Joe Bastardi's oft repeated remark of levels getting back to around 2005 or 2006 have been ditched as being too conservative!

The wuwt number is actually fractionally above the 2013 average for August.


At WUWT being wrong is a vocation!

Jim Hunt

I did try to shed some light in WattsLand, but Anthony wasn't interested for some reason:


Chris Reynolds


For what it's worth.

My prediction bounds are 3.48 to 4.62 million kmsq. The top five lowest years are:
2012 3.580
2007 4.280
2011 4.568
2008 4.695
2010 4.872

Note that SIPN predictions are in terms of average extent for September (that's what those figure are.

I'd consider my prediction as a success with anything from 2008 and lower extent, i.e. an entry into the top 5 knocking 2010 out of the top five lowest years. 2010 is slightly outside my stated bounds, but at 0.07 million out, I don't think that would be worth worrying about.

Actually, now I find, to my great surprise, that if I tally the years from 2000 to 2014 for 20 June NSIDC Extent in order I get...

2010 10.133 5
2012 10.276 1
2011 10.378 3
2006 10.679
2014 10.886 - just in the top 5!
2008 10.948 4
2007 10.950 2
2005 10.966
2001 11.170
2009 11.171
2013 11.181
2004 11.306
2003 11.349
2002 11.376
2000 11.484
The numbers in that list are the positions in the list of top 5 lowest monthly minimum NSIDC Extent.

Doesn't guarantee my prediction won't fail, but I'm pleased to find that at present we're in the top 5! :)

If the current weather holds then holding that position shouldn't be impossible.

Bill Fothergill

Hi Chris,

We seem to be using different figures, albeit only slightly different. Your daily figures are , I think, from the unsmoothed data at...


For daily figures, I was referring to NSIDC's 5-day running average that is plotted on their Charctic tool, available at...


I'm not sure where you obtained or derived your monthly numbers from, but I was using...


I thought that was the "official" source for the numbers that appear in NSIDC's News & Analysis, and hence the "official" monthly average used for SIPN purposes. If I'm looking at the wrong dataset, I'd obviously be happy to be corrected.

Working on slightly longer averages, the NSIDC SIE figures so far this year suggest that 2014 is already very likely to be amongst the top four (maybe 5) annual averages. If the remainder of the year exactly matches 2013, it will end up 5th; anything slightly lower than the mid Jun - end Dec numbers from last year would see 2014 finishing 4th behind 2012, 2011 and 2007, with 2010 dropping to 5th.

I wait with baited breath to see how your Solstice Ice Prediction Notion holds up!

Cheers billf

Chris Reynolds

Yep, I get my data from here:

My average is simply calculated from that, although there are monthly figures here:

Using those monthly figures the picture changes.

year extent
2012 3.63
2007 4.3
2011 4.63
2008 4.73
2010 4.93

I would have to be in the top three to strictly be successful.

However all of my calculations have been done using the data in the first link of my reply (G02135 daily). SIPN may judge on the list above, I'll be sticking to the data I use to decide success or failure.


Very nicely put & while as a Canadian I cringed at D's labeling of "Canada The Good" as a "Corrupt Petrol State", there is more than a grain of truth to it.
The Tar Sands has stained Canada's once bright image & it's the responsibility of Canadian voters to end the madness. Fighting a corrupt machine backed by Big Energy Bucks won't be easy, but as the recent Ontario election has shown, it can be done.
Sorry for drifting so far OT.

The comments to this entry are closed.