Tuesday 19 March 2013

What caused the volume loss in 2010: Part 2.

Back in September 2012 I wrote some posts on the 2010 volume loss in PIOMAS, why it was important, and what caused it. I'm now able to add another layer of detail into the explanation of what caused the 2010 volume loss and consider whether it has any reality outside of the PIOMAS model. I'm more confident that the 2010 volume loss was real, not just a blip in a model.


I'm not going to re-hash the earlier posts, but I will write a few words to save busy people the bother of reading those posts.

Basically the 2010 volume loss is important because it was one of a series of five large volume loss events, three of which occurred at a time when the ice pack had much more volume, so were proportionately not such large events. The volume loss event preceding 2010 was the 2007 sea ice crash, this puts the scale of 2010 in what I consider to be its correct context.



This loss of volume in PIOMAS has caused a change to the seasonal cycle, in that a more pronounced and aggressive spring melt (April to June) is occurring as shown in the following graphic (PIOMAS anomalies, baseline 1980 to 1999).


In my post on what caused the 2010 volume loss I argued that warmer temperatures had retarded ice thickening over the Arctic coast of the Canadian Arctic Archipelago (CAA). Not having access to a model, and with certain gridded data only running to 2004, I am unable to further confirm this view point. However the failure to thicken still seems to me to be likely to have a thermodynamic component, and the additional factor I outline in this post doesn't seem to me to explain the loss of thick ice along the coast of the CAA.

I've recently broken down the PIOMAS volume/thickness breakdowns into three regions, see here. This further detail has enabled me to examine what was going on within the sub regions in 2010. I've neglected the Atlantic sector because this didn't play a large role.

The following image is from a spreadsheet, showing years from 1998 to 2012 for the Siberian region. The table of numbers are volumes for different thicknesses of sea ice in June, the actual thicknesses aren't important right now. Clicking on it will enlarge to full screen, but the notable feature is visible anyway: 2007 cleared out the thickest bands of ice from the Siberian region, except for 2010 when the thickest ice made a reappearance.


So where did this ice come from? Applying the regional boundaries to images from the Drift Age Model, source, shows what happened.


Week 20 is in late May, close enough to June not to make any substantive difference. What can be seen in that after 2006 there is a marked drop in old ice in the Siberian (SIB) sector, however in 2010 a large swathe of old ice enters the Siberian sector. This transport of ice won't all have occurred between April and June, the ice just doesn't move that fast. Trying to compare volume for June brings in the complication of ice thinning and volumes cascading down to thinner categories. However the difference between 2009 and 2010 in April for the Siberian sector (SIB) was +1.35k km^3, and for the American sector (AME) was -2.22 k km^3. So the gain in the Siberian sector accounts for over half the drop in the American, which is one reason why I still suspect a role for the warm anomalies over the CAA.
 
The Siberian sector shows marked gain in volume for ice over 2.5m thick, at a sum of 3.93k km^2, post 2007 years show nowhere near that sum total of volume of ice in grid cells over 2.5m thick. While in the American sector ice over 3.25m thick shows a notable deficit, with grid cells from 2.5m thick up to 3.25m thick showing a notable gain in volume, which I suspect is the result of grid cells that would have reported thicker ice falling into lower thickness groups due to less thermodynamic thickening and possibly less ridging as ice was freely exported from the region, rather than being crushed up against the CAA.

What caused this transport was a strong anticyclonic set up over the Arctic, for example this is shown in the following plot of SLP from NCEP/NCAR, note the Dipole formed between the North Pacific low and the Arctic high pressure centres.


This will have led to clockwise winds around this Arctic which drove the clockwise ice movement from Beaufort towards Siberia seen between week 1 and week 20 in the DAM: DAM Week 1, DAM Week 20.

The transport of thick ice shown in the DAM, and implied by PIOMAS volume changes and the presence of a strong SLP dipole, can therefore be accepted as not just an artefact of PIOMAS. This movement of ice really happened and this fact is confirmed by the ASCAT scatterometer, 365/2009, 140/2010.

Then we have the thermodynamics issue. NCEP/NCAR shows average absolute temperature was around -25degC for February to April 2010, this being some 4 degrees warmer than average, that's a factor of 1.16. As the heat flux through the ice is inversely directly proportional to the difference in temperature between the ocean at the base of the ice and the air at the surface, this 4 degC warm anomaly corresponds to a heat flux of 86% of the climatological baseline, based on temperature and neglecting issues such as snow cover. This heat flux is crucial because it is that flux of heat that causes new ice to freeze onto the base of the ice pack at the ice/ocean interface. However, although 2010 was the warmest year for surface temperatures north of 70degN, other recent years were also warm.

Against this broad average however there are the very warm surface temperature anomalies over the Canadian Arctic Archipelago and the nearby Arctic Ocean.

That right next to this large warm region there was a failure in winter thickening of the thick old ice off the CAA seems to me to beg too much of a coincidence. The following graphic shows the deviation from the 1980 to 1999 average thickness for each grid point in April. The dark area off the CAA shows the deficit in thickness in April 2010.



However without further data I can't be sure, so I'm not putting the atmospheric conditions and implications for thickening forward as strong evidence in support of the reality of the 2010 event.

So it seems to me that we have only one strong line of evidence that suggests that the volume loss of 2010 may not have been just an artefact of PIOMAS - the large export of old ice into the Siberian sector. But is there any further corroboration?

The PIOMAS volume loss of 2010 was followed by a change in the seasonal cycle due to a more aggressive spring melt. So is there anything not derived from PIOMAS that stands out about the years following 2010? Cryosphere Today anomalies (my anomaly calculations; baseline 1980 to 1999) show what I think is a result of the volume loss of 2010. Here's the 1980s, the 1990s, and below are the periods 2000 to 2006 and the post 2007 years, since that latter period needs to be considered as a separate phase of behaviour.



2000 to 2006.




Post 2007.

What stands out here is that 2007, 2010, 2011 and 2012 are common in that they all exhibit a massive drop around early June, days 160 to 170. Other years such as 1999 and 2005 also show such a drop, but there is no similar grouping of behaviour since 1979. However beyond this early June drop 2007, 2011 and 2012 are common having a steady drop thereafter. I think the reason for 2010's anomalies not dropping like those years is that the process of thick ice loss had not completed, there was still the substantial tongue of old ice in the Siberian sector that had to be melted out, i.e. Bremen plot from August here ( - was the spur towards Chukchi due to this older ice?). By 2011 and 2012 this process was complete, so area was free to fall unimpeded by thicker old ice. The fall in 2007 was driven by the unusual weather and preconditioning of the ice. So what has caused the CT Area anomalies for 2011 and 2012 to behave so atypically, if not the preceding volume loss of 2010, with the attendant aggressive spring melt in following years that PIOMAS reveals.
 
The resultant state of the pack can be appreciated from the following June thickness plot from PIOMAS gridded data, 2011 and 2012 stand out from the rest. Note the reduced size of the blue region (over 2m thick) and the lack of dark blue off the CAA post 2010.


So the 2010 volume loss, an event as large as 2007, seems to me to be a real event which has had real world repercussions.

As the Arctic ice pack transitions to a seasonally sea ice free state the spring volume loss has important implications. Until the Arctic winter warms substantially conditions will still be cold enough to grow ice substantially thicker than 1m thick, up to somewhat over 2m thick. This means that to get a repeated sea ice free state each year as much ice must be lost in the melt season as grows in the freeze season. As growth of new ice will be vigorous due to the growth/thickness feedback (thin ice grows faster than thick), commensurately increased loss of ice during the melt season will be a pre-requisite for a seasonally sea ice free state, resulting in a greatly increased seasonal cycle of area/extent and volume. The spring volume loss will play an increasingly important part in the attainment of this state, losing volume and preconditioning the ice for losses during the summer which will usher in a largely sea ice free ocean state probably by the end of this decade, possibly within a few years.

14 comments:

Anonymous said...

In my post on what caused the 2010 volume loss I argued that warmer temperatures had retarded ice thickening over the Arctic coast of the Canadian Arctic Archipelago (CAA). Not having access to a model, and with certain gridded data only running to 2004, I am unable to further confirm this view point.

I am not sure if you are aware that dr. Zhang recently added the iceprod.Hyyyy files: growth/melt rate.

Chris Reynolds said...

Thanks for that Wipneus,

I wasn't aware of it, and I'll be having a look. Although I'm not sure how much extra information it provides. Isn't this really the same information as doing an inter-month difference on the thickness data?

The one I had my hopes on was oflux - the ocean heat flux used to melt ice. With that it would be possible to partition between 'ocean' and 'other' heat fluxes. Which could be of use.

Anonymous said...

Just noticed by chance when looking for the snow files.

delta-thickness is sum of melt/freeze and advection. Advection is the 'transport' term (in the math formula's it is the divergence, the upside down delta symbol).

oflux - the ocean heat flux used to melt ice

Hmm, does that term get negative when the ice freezes?

Anyway, if you ever ask dr. Zhang about the oflux files, perhaps ask the advection files as well. They are fundamental for a complete description.

arcticio said...

So, old and thick ice drifting into summer melting zones leads to significant and unrecoverable volume loss. That's bad news for the floes around Lincoln Sea considered to survive until the end.

Chris Reynolds said...

Thanks Wipneus,

I'd forgotten advection, but of course it is a crucial part of the puzzle, having that and oflux would go some way towards a mass balance of each grid cell.

Arcticio,

Such drift of ice is a common and always has been a feature of the pack, it's just that now the drift into Beaufort doesn't age the ice, it sends it to destruction. The transpolar drift will always shift ice up against N Greenland and the CAA, but what happens to this mechanically thickened ice after that IMO ensures low volumes of MYI into the future - however long it is until a seasonally ice free state.

Anonymous said...

The oflux numbers are always non-negative.
So the upward heat flux from the bottom of the ice, through the ice to the atmosphere seems not to be included.

So I still think the iceprod.Hyyyy files are the ones that make up the balance, but make no distinction between bottom and top.

Chris Reynolds said...

I had looked at oflux last year, and now you remind me it fits with the model details I've been reading in the relevant papers. Heat flux through the ice always seems to go from ocean upwards to atmosphere. This makes sense because the ocean, being liquid is always warmer than the ice. Surface ice temperature at a maximum is the triple point. So while oflux is described as:

"Ocean heat flux used to melt ice, unit: (meter of ice)/s"

I suspect it is actually heat flux through the ice, whether ice is growing or melting.

However it's been a long while since I really looked into these other variables, and you're much more adept than I ever was.

Neven said...

Thanks for this, Chris.

I remember The Arm from 2010 well, that basically kept sea ice extent/area higher than it would've been if the melting season had lasted a week or so more. Same in 2011, but didn't stand a chance in 2012. And now there's practically nothing there - except for weather conditions - to prevent a massacre on the Pacific side of the Arctic. But I've said all of that already.

At the risk of sounding stupid, could those warm 2010 air temps you allude to have caused the water temps to stay higher as well, because of the decreased temperature difference between air and water (in leads)?

BTW, what is a DAM?

Neven said...

Following...

crandles said...

The melt thickness required in each cells to reach melt volume reduction in the melt season shows a fairly linear trend but 2010 was unusually high:

Year, Melt thickness in all cells to reach volume reduction from max to min
2003 1.421715
2004 1.36406
2005 1.46563
2006 1.42854
2007 1.56865
2008 1.53596
2009 1.53736
2010 1.768
2011 1.6302
2012 1.7688

Do you think that the deviations from the trend might be predictable by area above certain thicknesses?

ie in 2010 were there an unusually large number of cells below 1.768m in April?

One of Zhang's SEARCH contributions suggested area under 0.5m thickness was their best predictor. I wonder if the thickness used each year has to change in line with the trend in the above typical thickness reduction numbers?

Does that seem a sensible sort of analysis to carry out?

crandles said...

I should have given a better introduction. What I am wondering is did 2010 melt a large volume because of the way the volume was distributed?

area distribution or thickness distribution?

Or is it something else like weather or ...?

Chris Reynolds said...

Neven,

DAM = Drift Age Model. The sea ice age work of Maslanik/Fowler/Tschudi.

I know the feeling, I'm starting to sound like a stuck record.

Crandles,

What is causing this ties in with our discussion with Dungeon Master at the forum. Basically thin ice has a greater spread of loss of thickness over the season than thicker ice (thickness at start of season -April). I'll explain over there. I intend to do a blog post this weekend on the issue of 'why'.

Part of the reason for 2010 was the position of a large influx of MYI into regions that melted out, so the answer to your question: "What I am wondering is did 2010 melt a large volume because of the way the volume was distributed?" is 'Yes'. But there may be a contribution from lack of thinning.

I'll reply again after I've eaten, the DAM image FTP site is down. But looking at ASCAT I'm now wondering if the thickest ice along the CAA was ejected in 2010.

Steve Bloom said...

OT: Chris, I have the (draft but readable) English translation of that Russian blocking event paper. Send your email to me at stevebloom55 -at- deletethisbit.gmail.com and I'll reply with it.

Chris Reynolds said...

Thanks Steve,

Email on the way.