Friday, 28 September 2012

Sea Ice Albedo Changes and the Seasonal Cycle.

I'm currently ploughing through my stack of Arctic related papers. It seems to me that in order to fully appreciate the spectacle I think the coming years will deliver, knowing the scientific literature is a pre-requisite. So over the autumn and winter I'll be making more of an effort to post commentaries on what I've been reading. One aspect of the fallout of this year is that there will be a series of papers by various researchers who've been following this year. I can't wait!

However along the way I'll be distracted from my campaign of re-reading and catching up with my 'to be read' folder by research I've not been aware of.

Like these papers...

Thanks to Kevin O'Neill for bringing my attention to these two papers. The first is Rosel & Kaleschke, 2012, "Exceptional melt pond occurrence in the years 2007 and 2011 on the Arctic sea ice revealed from MODIS satellite data." The second is Perovich & Polashenski, 2012, "Albedo evolution of seasonal Arctic sea ice." Rosel is pronounced Roesel - there should be an umlaut over the 'o', but I'm too lazy/busy to go through the post cutting and pasting the appropriate character - call it anglo-saxon linguistic arrogance if you like.

As the title implies, Rosel and Kaleschke find exceptional melt pond areas in 2007 and 2011. This might partly explain the CT Area anomalies I refer to in preference to the other extent measures. The reason I prefer CT Area is that it seems to hold information lost in the calculation of extent.

That's from a previous post, so the August cyclone isn't of relevance here. What I need to draw attention to is that in 2012, 2011, and 2007 early June anomalies fall off a cliff. What these years have in common is that they were all area records. Also all the post 2007 years show drops around this time of year.

The significance of Rosel & Kaleschke is in their figure 3b, shown below.

The horizontal axis are dates - 18 May, 1 June, 15 June, etc. This shows the area of melt ponds for 2007 (red) and 2011 (magenta) against the average for 2000 to 2011 (black with one sigma bounds in dashed lines). So the area of melt in 2007 and 2011 rose rapidly in early June, exactly when the CT area anomalies drop off a cliff. More than that, the difference between the recent average and peak area of melt ponds is around 0.5M km^2, which is a substantial fraction of the drop in anomalies in 2007 and 2011 shown above.

The overall path of the CT area anomalies through the summer isn't due to melt ponds, the end of season state of the ice shows this. However one reason why the professionals tend to prefer extent to area is that area is known to be prone to error from counting melt ponds as open ocean. Is there something about the processing behind CT area that amplifies this effect in early June? I don't know enough about their processing technique to say so, but it's a possibility that has to be borne in mind. I've interpreted the rapid drops in CT area anomalies as rapid retreat of ice as the melt edge enters to basin: In 2007 due to fast early melt as a result of weather conditions (Arctic Dipole). In 2011 due to fast retreat due to first year ice prevalance and previous thinning. However part of this drop in anomalies could be due to melt ponds. Against that however, It is worth bearing in mind that ice extent series, such as JAXA also show a deviation from previous periods in recent years from early June. JAXA extent anomalies also show a drop in June, but not as precipitous as CT area.

2011 didn't have the open skies that 2007 had, in general the weather wasn't as conducive to ice melt. Rosel & Kaleschke note a role of weather, but add that in 2011 the location of open skies in July ties in with the melt ponds, which were concentrated in the Beaufort Sea and Canadian Arctic Archipelago (CAA). This feeds in with another obsession of mine - the role of the atmosphere in hastening the loss of ice, but I'll get on to that later in this post. However all of this doesn't mean that first year ice is 'off the hook' as a likely candidate factor.

In Perovitch & Poshenski the authors find that there is a significant difference in albedo, and seasonal evolution of albedo changes between seasonal or first year ice (FYI) and perennial or multi-year ice (MYI). They use insolation data from 1998, the year of the SHEBA experiment, and apply the difference they have found between FYI and MYI albedo. Over a square meter covered by MYI they find that by September cumulative solar heat input amounts to around 900MJ/M^2, however for a square meter of FYI the gain over the same period, with the same solar input is almost 1200MJ/M^2. That's almost 1/3 more energy gain. Which is what I call a feedback!

Figure 3 of Perovitch & Poshenski shows the annotated time evolution of albedo for seasonal ice (FYI) and MYI.

Above I noted that all years since 2007 show drops in CT area anomalies in early June, I don't think this is weather, it happens in such a short time window at a specific period of the seasonal cycle that interannual variation in weather argues against a weather driven process. The common factor seems to be the ice itself. The transition to thin FYI for large swathes of the Arctic post 2007, combined with the implications of the above graph, a large early deviation in albedo between FYI and MYI, and a possible role for melt ponds due to decreased albedo of FYI during early June seems to me to make a better explanation.

Going back to Rosel & Kaleschke, they provide a graph of the fraction of ice area covered by melt ponds.

If 2007 was due to extreme weather, what happened in 2010 and 2011. As I've discussed previously PIOMAS suggests massive losses of MYI in 2010 starting with a failure of thickening during the winter of 2009/10 and then massive losses off the Canadian Arctic Archipelago. Melt pond area doesn't show such an increase for area, but area has been dropping significantly overall, so I suspect that the best way to consider melt pond changes is using the relative measure of fraction, which takes into account overall decline in sea ice area. Is the increase in 2010 and 2011 due to larger amounts of FYI following the volume loss attributable to MYI in early 2010? I don't know, but I think it could have a role.

Recently I've posted about the late summer Arctic Dipole (AD) and have asked how much of a role this has played in the post 2007 behaviour, a question prompted by the anomalous low index AD following that year. There must be a role for it. However as other studies (Comiso 2011 and Polyavkov 2011) show, after 2007 there was a substantial loss of MYI. Due to my work on PIOMAS gridded data I'm convinced that the transition to a mainly FYI pack can now be considered largely complete. Even for those who aren't persuaded by this, the consistently lower sea ice areas post 2007 automatically imply about 1M km^2 more FYI at the end of each season w.r.t pre-2007 years.

So whilst the early summer AD must have an impact; equally, as Perovitch & Poshenski show, a transition to a largely FYI pack must also have an impact. As for their relative importance, I think that can only be solved by modelling, and the best model for that remains PIOMAS.

Here's a graph of CT area loss from 1 June to minimum of each year in the record.

There are at least two reasons for the jump seen in the years after 2007, changes in ice albedo, and the late summer AD. What other factors are at play? And are other factors as important as those two?

Rosel & Kaleschke, 2012, "Exceptional melt pond occurrence in the years 2007 and 2011 on the Arctic sea ice revealed from MODIS satellite data."
Perovich & Polashenski, 2012, "Albedo evolution of seasonal Arctic sea ice."


Anonymous said...


something for your reading. It seems very new.

no time to digest. Only let it be said that I think that the many people who are giving disdain to the modeling effort, may have overlooked that those people have made real progress. anyhow, have pleasure reading it.

Chris Reynolds said...

Thanks Anon,

That paper is a reworking of the earlier "A sea ice free summer Arctic within 30 years", this time based on CMIP 5. Which I blogged on here, towards the end of that post.

You'll find the original paper that precedes this one here.

This latest one is a reworking of the same principle - that the forced change of seasons should resemble the forced change of AGW in terms of sea ice response.

I could take figure 1 of Overland & Wang 2012 as meaning I've called it too early in recently stating I think we'll see a virtually sea ice free state this decade. However I still think that volume and implied thickness from PIOMAS, and what anecdotal information there is about thickness indicate a we're in a RILE, and that it will end with a virtually sea ice free state within 8 years.

Kevin O'Neill said...

Chris, to continue with melt ponds there's Incorporation of a physically based melt pond scheme into the sea ice component of a climate model,Flocco, D., D. L. Feltham, and A. K. Turner (2010), J. Geophys. Res.

I was actually looking for Impact of melt ponds on Arctic sea ice simulations from 1990 to 2007 D. Schroeder, D. Flocco, and D. L. Feltham - but haven't found a non-paywalled copy yet.

Chris Reynolds said...

Thanks Kevin,

The Flocco paper on their model is interesting, in particular figure 2 (IIRC) which shows substantially increased seasonal cycle and lower summer extent when the model's included.

Neven said...

Hi, Chris. Nice post, as always.

I wanted to do something on melt ponds, but am suffering from temporary blogger burn-out. I'll share my short thoughts here:

This quote from a Climate Central article from last week got my interest:

For example, one new study shows that the melt ponds that form on top of sea ice floes in June and July can dramatically accelerate sea ice melt. These ponds, which form as snow and ice melt under the Arctic sun, can dramatically increase the amount of solar radiation the ice absorbs. This warms the surface and eventually allows more heat to ocean waters below, in effect melting sea ice from the top and bottom.

The study found that the melt rate beneath pond-covered ice is up to three times greater than that of bare ice.

“In a crude sense, pond-covered ice is more akin to open water. So, because ponded ice reflects less of the solar radiation, there is more heat available to melt the surface of the ice,” said Daniel Feltham, a researcher at the Center for Polar Observation and Modeling at the University College London, and co-author of the study, in an email.

The study, published in the Journal of Geophysical Research, showed that computer models incorporating melt pond information tend to perform better in simulating historical conditions than models that don’t take such ponds into account.

The study also points to an unexpected relationship between snow cover and sea ice loss. The more snow on top of the ice at the beginning of the sea ice melt season, the greater potential there is for melt ponds to form, and thereby speed up sea ice loss, the study found.

How about looking at CAPIE instead of CT area? However crude it is, it does potentially say something about melt ponding during the May-mid-August period.

I've uploaded last week's CAPIE graph here.
See the huge drop from June 1st onwards?

The problem, however, is that not all of the drop can be ascribed to melt ponding. First of all, the ice pack was also diverging a lot in several places. Second, the switch from AMSR-E to WindSat by IJIS made comparing 2012 CAPIE to previous years a bit of a Golden Delicious-Elstar comparison. Apple to apple, but not entirely the same apples.

I always thought that CAPIE is a useful metric if you want to know towards the end of the melting season whether the ice pack is diverging or converging. But it could prove useful to guestimate the development of the melting season based on the first few weeks. I'll be looking at it more intently at the start of melting season 2013.

I wrote a piece about melt ponding in February: New Data: Melt Ponds in Arctic Sea Ice, comparing 8-day composites of melt pond cover fraction from KlimaCampus (based on MODIS data). I've asked if they already had a look at this year, but unfortunately, no budget.

My conclusion for now: melt ponding more important than I thought. Lack of data. CAPIE too crude and influenced by other factors as well. Still, if CAPIE is going down hard at the start of the season, expect fireworks.

Neven said...


Chris Reynolds said...


Blogger burn out - It's been a long hard season, my suggestion is have a break, open new threads for comments for a while if needs be. Perhaps with a short post with questions to discuss?

Thanks for the heads up regarding CAPIE. As I no longer keep up to date spreadsheets of extent I've not been following CAPIE. You're correct that the combination of the area & extent methods makes interpretation dauntingly difficult.

The problem I have with the CAPIE June 2012 crash is that there was a similar crash in 2008, which in terms of all measures (area/extent/vol) was a bounce back from 2007. Whereas all post 2007 years have shown a CT area crash in early June. 2012 being the biggest of the lot. Despite my uncertainty as to what's causing this early june area crash (rapid loss of thin ice and/or melt ponds skewing the area reporting) I still think it's epiphenomenal of critically thin ice, and expect to see it happen again.

As for the albedo issue. What staggered me was the net gain in heat between a mainly MYI pack and mainly FYI. Almost a 1/3 gain in energy from the former to the latter!

Anonymous said...

Hi Chris,

Just reading Eli's post The Ice Melts at Midnight, the CO2 concentration's seasonality is is much more dramatic in the Arctic than I'd realized. From just a couple of locations, it perhaps falls short of support for an enhanced spring melt, but it's certainly not inconsistent with it. Any thoughts on whether the effect, if true generally at those latitudes, would account for any significant fraction of the change in sea ice's seasonal patterns?


Chris Reynolds said...

Hi Simon,

I don't think the seasonal change in CO2 is a major impactor on the seasonal changes in sea ice. The change in albedo described in the above post acts upon a solar flux of over about 100W/m^2, whereas CO2's forcing increase is of the order of around a watt/m^2.

In a paper by Francis & Hunter "Changes in the fabric of the Arctic’s greenhouse blanket", 2007, they find that most of the increase of downwelling infra-red is actually due to water vapour (in the Arctic). Because the Arctic has typically been so cold, its atmosphere has been very dry. The warming that has occurred, together with increased influx of warmer moist air from outside the Arctic, has lead to a far greater increase in humidity than outside the Arctic. And water vapour is a very powerful greenhouse gas.

I've previously posted on CO2 and sea ice loss, here. As Eli points out, the 0.25W/m^2 change in RF due to the seasonal CO2 change could be significant. However I see other factors, ice albedo, water vapour, heat loss due to increased open water, as more massive players than this small CO2 contribution. The long term effect of this CO2 increase however is a continual 'bias', and amongst the various factors aiding and acting against sea ice survival, I see the CO2 'bias' as a continual player acting against sea ice survival. Hence the agreement between sea ice loss and CO2. You could think of it like glacier mass balance, the glacier may have a mass of 1 million tonnes in the summer and 10 million just before spring, but even a small loss of 0.01 million tons a year will ultimately lead to the glacier's demise: Despite that annual loss being dwarfed by the other statistics of the glacier, including its seasonal cycle.

Hope this helps.

Chris Reynolds said...

Actually, 1 million vs 10 million doesn't make sense, that would merely imply a snowpack of 9 million tons.