Thursday, 21 May 2015

The Slow Transition: The BCE Region

The BCE region is the region covered by the Beaufort Sea, Chukchi Sea, and the East Siberian Sea. For there to be a virtually ice free Arctic any September soon, the BCE region must melt out early in the season. This post is the penultimate post in my argument that we face a slow transition of the Arctic sea ice, after which I will wrap up the argument with a summary, before moving on to watch this season.

Coincidentally, as others including Neven have noted, Beaufort is very interesting right now, this winter's export of multi year ice makes me suspect it will stall, but an early start in Beaufort is promising for a more exciting season than the last two years. Here's hoping...

The BCE region is shown in the modification of a graphic from Cryopshere Today shown below. The BCE region is filled in red.

April PIOMAS thickness calculated on a per grid box basis for the BCE region shows a continuing decline in April thickness. Behaviour from 2004 onwards my evolve into a slower decline rate of April thickness, but I will get onto that.

Gice is the sub grid thickness distribution used in the PIOMAS model, described more fully here. It uses 12 thickness bands to allow the model to cope with a smaller scale than the size of a grid box, which is typically several hundred square km in area. The following graphic shows thickness distributions for December from 1979 to 2014. Individual lines are not important, the colour scale is from dark blue for earlier years to pale blue for the latest years and shows qualitatively what has been happening to PIOMAS thickness distribution in the BCE region.

Growth of new ice after September shows as the peak around 1.46m thick, this has declined in volume somewhat and has spread out, but the major source of loss of December volume is from the thickest ice. Moving through the winter to April, the month of peak volume for the whole PIOMAS model domain, it can be seen that while ice below 3.3m thick has remained fairly level (actually increasing slightly), ice above 3.3m thick has been the source of the overall volume decline. 3.3m is the demarcation between the 2.61m and 4.23m thickness bands in the above graphic.

I have previously posted on the Fast Transition, the rapid transition of the peripheral seas of the Arctic Ocean to a virtually seasonal ice cover. The BCE region shows the same behaviour, and the same uptick in extent, and area, over the last two cool Arctic summers. But sticking to the BCE region, is there a connection between the loss of thicker ice and the decline in September extent?

The following graphic has the vertical axes deliberately offset to bring volume and extent in line. It shows April PIOMAS gice volume from ice above and below 3.3m thickness plotted together with September extent, all for the BCE region.

Extent in September is seen to track the decline in April ice volume over 3.3m thick, as I have said before for other regions, this is because a decline in thick ice increases open water formation efficiency. So leaving aside the last two years, which were a result of weather so tell us nothing of the underlying process in question, for the 2007 to 2012 period the BCE region has seen extremely low extent. The general decline of extent seems to be related to the decline in the thicker ice, and a causal relationship between the two indices is to be expected due to increased open water formation efficiency. For a given seasonal thinning thinner ice is more likely to melt out to give open ocean.

Turning back to thickness. Using the data behind the first graphic of this post I have a series of April thickness for the BCE Region. I have taken the thickness for the regional seas of the BCE region, Beaufort Sea, Chukchi Sea and East Siberian Sea, and have plotted them as thickness differences from the overall thickness for the whole BCE Region.

The effect seen in the above graphic is one of decreasing variance as time proceeds, as the ice approaches 2m thick in April the plots of the three regions converge. That 2m thick is the typical sort of thickness expected for growth of ice from open water is not a coincidence. This was found by Dromicosuchus in the comments of my original Slow Transition post.

I have taken the data behind the above graph and have recalculated the data for each sea as absolute values (i.e. all positive, no sign +/- used) and have calculated the average for the three seas. This is presented as the following graph, which basically shows the decline in variation of thickness within the BCE region.

Obviously this decline cannot carry on to zero, since that would imply a pack of uniform thickness throughout the entire BCE region. So what does this mean?

The clue is in the thickness distribution plot posted earlier. But for this part I have re-drawn for April gice and annotated.

As argued in my previous post, The Mono-modal Spike, the relative lack of change at the thin ice end of the distribution is marked compared to the decline in thicker ice. This is because of the thermodynamic thickening of ice over winter, a process described previously. That post was done using grid box effective thickness, and the same process is seen in the sub grid thickness distribution, gice. Note that the wall is drawn as a fuzzy line, it is fixed only by winter temperatures, and as the winter warms that will dictate thinner ice. However the winter is not warming fast enough for claims of a virtually ice free September this decade to be at all plausible.

What years like 2007 to 2012 mean is that most of the ice melts out. This prevents the BCE region being a transit for the exported multi year ice from the Central Arctic, and turns it into a killing ground for that ice. As 2013 and 2014 show, this does not happen every year, it just needs to happen most years. And even 2013 and 2014 did not bring extent up to the levels before 2003, because thicker ice volume was so much less. 2003 may be a significant year because of the dominance of the Arctic Dipole from around that time, i.e. here and the references therein. So for as long as the BCE region remains in the state seen since 2007 it can be expected to revert to its new default state; virtually seasonal.

I should justify the description of virtually sea ice free for 2007 to 2012. For the BCE region, the 2007 to 2012 period has an average September extent 14% of that for the 1980s. For comparison with the whole Northern Hemisphere extent, 14% of the 1980s average would be 0.979M km^2, which fits the commonly accepted definition of below 1M km^2 for which the Arctic be considered virtually ice free in September. Therefore BCE can be considered virtually ice free in September for the period 2007 to 2012.

So the BCE region is already at the stage where, after 2007, most years have been what can be described as a virtually seasonal pack. The future is likely to see that condition entrench and continue. However that does not mean they are out of the equation of the transition of the entire Arctic Ocean to a virtually seasonal ice pack (<1M km^2 extent for September). This is because as long as there is ice there to melt during the summer that ice must melt out to enable the ice edge to make significant inroads into the Central Arctic. The melt season is fixed by the solar cycle, melt has from March until September, and unless we see a crash in April thickness, there will be similar volumes of ice to act as a barrier to large ice edge encroachment into the Central Arctic region.

The issue is that 'the wall' described above suggests that ice growth over winter will continue to keep winter volume in the BCE region up, countering the past losses from thicker ice. In such a situation the melt season will still have a lot of ice to melt and the post 2007 behaviour of summer extent would become a new normal, not a prelude to a collapse of the Arctic Ocean's ice cover.

The obvious question then becomes: What is going on in the Central Arctic and what are the prospects for a collapse of summer ice cover in the Central Arctic? That will be the subject of my next post.


Nightvid said...

"For there to be a virtually ice free Arctic any September soon, the BCE region must melt out early in the season."

Why? The biggest factors affecting ice melt are LOCAL. If conditions are anomalously conducive to melt in the central Arctic, but less conducive than normal in those peripheral areas, it is entirely conceivable to have the extent seem relatively normal up to maybe mid-July, and then suddenly have the whole thing just melt out.

The snow melt always seems to portend the ice melt, but precedes it by about 2 - 2.5 months, locally. If a given area still has snow cover on July 1, this pushes the end of the range of melt dates to September 15, at which point the melt season has ended. Thus it may not melt at all that year.

The reason that the pack gives the illusion of melting from the outside edges inward, rather than from the top and bottom surfaces, is that the ice is simply following the snow, with a delay of 2 - 2.5 months.

If the snow melted in the central Arctic first, then the ice would as well.

Chris Reynolds said...


You are criticising that statement on the basis of a hypothetical event. Let me know when it happens.

Meanwhile, every year on record, the melt starts on the Siberian and Alaskan coasts, not from northern Greenland and the Canadian Arctic Archipelago.

Kevin O'Neill said...

Compare any good polar stereographic map with the August 1st Cryosphere Today historical sea ice maps for 2007 and 2012.

The conclusion seems pretty obvious to me and matches what I would logically expect: the melt by and large follows latitude bands. I.e., we see open water first at 60N, then 65N, then 70N, then 75N .....

Northern Ellesmere Island and northern Greenland are hundreds of miles further north than the Bering Straight and the Alaskan/Siberian coasts that adjoin it. Alert is 508 miles from the north pole; Prudhoe Bay is 1200 miles from the pole.

Just by looking at the lines of latitude we would expect vast stretches of the Beaufort, Chukchi, and East Siberian Seas to melt first. And that is in fact what we see.

Landmasses and geography also play a role. The Canadian Archipelago and the ice bridge that often exists near Severnaya Zemlya are two good examples -but in general we see melt marching north.

This is similar to what we see on land. The boundary for lakes that freeze over completely each winter is moving north. The range of many plants and animals that do not like cold winter temperatures is moving north. Some of them - the pine beetle - with devastating consequences.

An examination of the historical data by latitude band should show that all relevant metrics start decreasing (increasing) at lower latitudes first, but should show a steady progression northwards.

IIRC, it was Judah Cohen who made the point that eventually the snow and cold will give way to rain. And then the ice goes away.

Kevin O'Neill said...

To elaborate a bit on precipitation: "Overall, precipitation in the High Arctic is less than half that of the low and middle Arctic and one-quarter that of the subarctic. For these reasons, much of Canada's High Arctic is classified by soil and botanists as polar desert."

Warmer air brings with it more water-vapour capacity. Models should be able to handle the physics, but we know how poorly they've done in the Arctic to date. I rarely see the transition from desert to increased precipitation and the transition from snowfall to rainfall as a positive feedback, but it will have an effect.

Kevin O'Neill said...

Nightvid writes:"The snow melt always seems to portend the ice melt, but precedes it by about 2 - 2.5 months, locally. If a given area still has snow cover on July 1, this pushes the end of the range of melt dates to September 15, at which point the melt season has ended. Thus it may not melt at all that year."

There is a lot of ambiguity in your statement, and on its face it is unsupportable. We have ice melting out now - are you claiming that the snow melt began 2 1/2 months ago in those areas? The beginning of March?

When you say 'snow melt' do you mean the onset of snow melt, the end of snow melt (all snow melted), or the period of rapid snow melt? Snow begins by melting a few hours per day. Oftentimes this is followed by periods of cold weather that can last for days - even weeks - before it begins again. Two different analysts looking at the same data can come with totally different dates for the 'onset' of snow melt.

A case in point would be Ice Buoy 2012L. The 'official' onset of snow melt is listed as 06/09/2013. Yet looking at the accompanying chart, it seems obvious to me that the snow began melting on or about 05/27/2013, lasted for three or four days, paused for more than a week, then resumed. The snow was completely gone by 6/18/2013.

Regardless which date we use, the buoy appeared to be bobbing in open water on 9/22/2013 which is 3.5 months after the onset of snow melt and 3 full months after the snow was completely melted. More important, a different buoy at a higher latitude might see similar snow melt onset dates, but might not see the snow ever completely melt out - much less the ice itself.

Nightvid said...

Chris, I know what all previous years have done. However, we are discussing the unprecedented, so you've got to question your assumptions.

Nightvid said...

Kevin, you are referencing data from a buoy that started the season at a very thin spot in the snow and thick part of the ice.

Try looking at JAXA RGB data instead, which gives a broader view rather than just a single point measurement.

Kevin O'Neill said...

Nightvid, you failed to answer my questions on amibguity. I've looked at lots of buoy data - I see no 2 to 2.5 month correlation between snow melt and eventual ice melt.

I'll ask again, we are already seeing areas of open water, did the snow start melting in these areas 2.5 months ago?

Chris Reynolds said...


I do question my assumptions regularly, this is the eighth step in my argument that we face a slow transition not a rapid crash. The ninth will be published later today.

Do you have a correlation study to back up your snow argument?

To Kevin's citing Beaufort this year, I add Laptev last year to counter your 2 month lag on snow cover claim.


I'm in agreement.

Kevin O'Neill said...

Chris, have you looked at recent buoy data and compared it to the PIOMAS data as a sanity check? I'm looking specifically at IMB Buoy 2014F which was deployed in the Beafort last September (77.5101,-146.2074).

2014F had a thickness on 10/1/2014 of approximately 1.54m. Thru winter and spring it has moved 100 miles south and currently sits at 75.7015, -151.8498. What I find interesting is that it never even reached 1.7m in thickness; topped out at 1.67m.

From this it would appear that MYI is becoming increasingly misleading. Ice simply can't grow beyond 1.75m or more in many of the areas outside the CAB anymore. It could sit there for 3 or 4 years and probably never reach 2m thickness.

What value is PIOMAS putting on ice in this area for thickness? I know that it assimilates data - but does the model actually believe what it sees :)

Chris Reynolds said...


With a grid box size of around 500km^2, and a sub grid distribution of 12 bands from 0 to 25m thick. With just one data point there are many possibilities...

Cryosat however is in line with PIOMAS.
(Sorry, but I still considered Cryosat to be an unknown quantity).


And (not sure if this will work) Wipneus's AMSR2 thickness plots seem to be to agree.;topic=1112.0;attach=16547;image

With those animated gifs you have to try to figure out the average thickness. I would say about 2m.

I put the thickness in Beaufort from those three at 2m +/- 0.25m nominal - I suspect you hate that word.

If I am correct then your IMB buoy is at the low end of thickness. That said I would have expcted more thickening.

Anonymous said...

There seems to be several buoys in the Beaufort Sea that surpassed 2m this winter. In fact Kevin, the buoy you picked looks like an outlier ( on the low side)


Kevin O'Neill said...

2015E 1m
2015D 2m
2015B 1.7m
2015A max 1.96m
2014i 1.32m on 10/1/2014
1.97m max thickness
2014F 1.54 on 10/1/2014
1.67 max thickness
2013F 1.1m on 10/1/2014
2.06 max thickness

Average max thickness =1.72m

I wouldn't consider 1.67 much of an outlier.

Nightvid said...


Those open water areas this early in the season are from advection of ice, not melt.


Snow melt and melt pond formation are tightly linked, so the recent study about melt ponds in May supports the conclusion if water-logged snow is being counted by the algorithm as melt ponds, as I argued in the ASIF.

Chris Reynolds said...


I suspect that even wet snow would play a similar role to actual water as the albedo would plummet in both cases. Whether the insolation is absorbed early in the season into water or wet snow might not make much of a practical difference. Be it the snow/ice, or melt pond/ice thermodynamic system, a drop in albedo still means more energy input from the sun. And such an event happening in May should prime either 'system' for peak insolation around late June.

Chris Reynolds said...

Nightvid, Kevin,

I haven't been able to find an account of why the snow first melts in Siberia in the papers I have or on the 'net.

I have weekly extent of snow from Rutgers

Would it be useful if I tried a correlation analysis, with say the ESS and Laptev extent?