What will the Arctic resemble in 2050?

[ChrisReynolds]

Andrew,

The off-hand dismissal of models and of modellers is something that angers me as much as the drivel the denialists come up with - in general it's just as thoughtless. I've found modelling studies to be of invaluable help in figuring out what's going on.

However I don't think they're showing a volume loss that is as aggressive as that seen in PIOMAS, NAME, or the real world. I think we can put a sea ice free September well after 2020, because of growth of ice after the insolation set period in which current minima are set. i.e. roughly halfway through September the sun sets and ice grows.

From my understanding we are in a Rapid Ice Loss Event, so a greater warming trend is to be expected. At present the greatest +ve anomalies are in autumn (Oct Nov) due to energy being released, energy that was gained in the summer. More open water will increase this. Furthermore in a 1980s paper (forget which one) it was calculated that over winter the MYI emitted as much heat as the FYI, i.e. thicker ice emitted far less than thinner. Then the pack was about 9/10 MYI, which puts heat flux from thinner ice into context, this is important because the pack is now much  thinner overall. However while Oct/Nov show continuing increase in temperature, the winter temperatures (DJF) seem to have levelled since 2007, is this due to increased IR acting as a negative feedback?

I'll stop rambling and sum up - while I'm confident we're in a RILE, and that we face a rapid transition, I'm not sure about what follows. And I think there is a great deal of warming to come. We've had about 3degC warming so far, I think you can easily double that by 2030 beyond which period I really haven't a clue. At present the ice pegs summer temperatures at zero, with more summer open water temperatures will go up.

I've used NCEP/NCAR temperature for the comments about temperature trends.

[AndrewP]

I would guess that when parameterized with the actual SAT change that has occurred, many of the models, especially those in the subset chosen by Zhang, would accurately model the decline in volume. Given that the subset chosen by Zhang is capable of modelling the decline in SIA, they can't be too far off on volume either. The data must be out there.

So again the primary question is how much does arctic SAT increase in the future? I don't see any reason the climate models would not have a reasonable handle on this question in the long-run. If we're going to call arctic SAT projections into radical doubt, why not call the global projections into radical doubt as well, as the deniers do? And again, I think my point stands about the arctic SAT increase rate overshooting the global rate the last 15 years (5-10X the rate instead of the theoretical and long-term ratio of 2-3X). I don't see how you can predict arctic SAT increase of 3C in light of these facts. Looking at both the long-term ratio and the modeled ratio would suggest we are due for a pause in or much slower rate of arctic warming the next 10 years.

Jim does bring up an interesting question about ocean heat transport. How has ocean heat transport compared to model predictions? If it has been higher, should we expect it to continue to be higher, or should we expect it to revert to the modeled trend?

Again, too little attention has been paid to whether the disparity with models is due to a flaw in the models, or internal variability which we can expect to "swing the other way" in the near future.

I agree the rapid decline has raised uncertainty. One cannot rule out an ice-free day before 2030 or even 2020. But I don't think you can rule out the possibility that this will not occur until the 2040s or even later. My guess is the 2030s, which is a little earlier than modeling would suggest, but also assumes there is some sort of stabilization in the near future. An ice-free month would, as you say, likely occur after that.

[Bob Wallace]

Let's revisit that CMIP5 graph after this year's melt.  One more year of melting as has been happening and I think we can toss some sauce on that spaghetti.  It's cooked.

[AndrewP]

Oh and to respond to your point that the rapid loss of sea ice is causing the rapid arctic warming: I think you have the cause and effect mostly reversed. People often seem to overestimate the albedo feedback. The forcing from a 2 or 3 million SIA anomaly is not that large globally, and the forcing regionally is limited by atmospheric heat transport.

An example calculation:

March-September 2012 had a SIA anomaly of 2 million sq km. Over the same period solar insolation in the arctic averages 200W/m2. The difference between the albedo of water and ice is .8. The surface area of the earth is 510 million sq km. Thus, 200W/m2 * 2 million sq km / 510 million sq km * .8 =  a global RF of only .6W/m2. Now because it is regionally concentrated the global RF would probably actually be less than half that because of a lower efficacy. The paper below finds a global RF of only .3W/m2 for a summer ice free state. By comparison, the RF for doubling CO2 is over 12X that (3.7W/m2). Thus a seasonally ice free state warms global temperatures, assuming a sensitivity of 3C/CO2 doubling, of less than .3C. They find the RF due to ice loss from 1979-2007 was a mere .1W/m2, causing less than .1C of warming at equilibrium. Obviously the warming would be concentrated in the arctic since this is where the RF is concentrated. So perhaps .3-.5C of the 2C of warming in the arctic since 1980 is due to albedo. The rest is driven by AGW and other sources of internal variability unrelated to albedo. The flow of causation is about 80% temperature to ice and 20% in the reverse direction as a feedback.

[Csnavywx]

And again, I think my point stands about the arctic SAT increase rate overshooting the global rate the last 15 years (5-10X the rate instead of the theoretical and long-term ratio of 2-3X).

Careful here. I think we're wandering into chicken/egg territory. The theoretical rate was at least partially based on slower degradation of ice cover. Now that this portion of the prediction was under-forecast, it undermines the original assumption, at least until the albedo-feedback mechanism weakens.

Very little guidance (other than the few already listed) suggested a rapid non-linear response like the one we've experienced so far. There could be another mechanism to slow it down (other than increased emitted radiation in the autumn/winter from open water), but I haven't found one or have seen a decent argument for it.

Edit:

On the topic of albedo forcing, radiative forcing and atmospheric heat transport:

1) Surplus heating from albedo differentials is, at its fundamental level, still a local diabatic process. Thus, this extra heat originates and generally affects its immediate surroundings most strongly.

2) Given #1, taking the regional forcing from local (or even regional, as is the case here) albedo changes and extrapolating it to a global average is rather arbitrary and not very useful. Again, much of that heat remains in the Arctic.

3) Your radiative forcing calculation, though I understand it to be just a back of the envelope-type calculation, is too rough to be really useful here. The region above the Arctic Circle represents only about 6% of the Earth's surface (which might be a bit generous in this discussion since we're talking about sea ice here). Using a more conservative RF of .07W/m2 for ice melt gives around .8-.9C of warming (assuming 1W/m2 = .75C increase via "fast feedback" equilibrium temperature). Using your figure of .1W gives well over 1C.

4) Given that the recent heating of the Arctic reduces dT/dx (temp gradient), this also generally reduces jet speeds (and wind speeds in general), weakening Ferrell/Polar cell interaction and advective heat and moisture transport, especially during summer and "shoulder months". Stronger differential heating along the ice edge will prove a local exception to that rule.

[AndrewP]

And again, I think my point stands about the arctic SAT increase rate overshooting the global rate the last 15 years (5-10X the rate instead of the theoretical and long-term ratio of 2-3X).

Careful here. I think we're wandering into chicken/egg territory. The theoretical rate was at least partially based on slower degradation of ice cover. Now that this portion of the prediction was under-forecast, it undermines the original assumption, at least until the albedo-feedback mechanism weakens.

Very little guidance (other than the few already listed) suggested a rapid non-linear response like the one we've experienced so far. There could be another mechanism to slow it down (other than increased emitted radiation in the autumn/winter from open water), but I haven't found one or have seen a decent argument for it.

As I suggest above, SAT is mostly causative of low sea ice not the other way around. This has been repeatedly demonstrated by studies which fully recover SIA in just a few years, even when all sea ice is artificially removed. SIA resides in close equilibrium with global temperature and thus CO2. You cannot permanently warm the globe or the arctic by temporarily removing sea ice. You can only do so temporarily.  Arctic SIA and SAT are determined by CO2.

[Csnavywx]

Please see my edit. Looks like we hit the button almost at the same time. 

[AndrewP]

I am not sure I follow you in #3. You seem to be assuming that all the heat is trapped in the arctic and there is no change in atmospheric heat transport. This is a wildly incorrect assumption. You can apply a candle to only one side of a pot of water, but the entire pot will warm, and the side you warm will only warm slightly more than the side that has no candle.

If what you say were even remotely close to true the arctic would have a massive runaway feedback that was never stable. Think about it, a small decline in SIA would have a regional warming effect of >1C.

By contrast models and history show SIA and arctic SAT to stay in close equilibrium with global T and CO2.

[Csnavywx]

And again, I think my point stands about the arctic SAT increase rate overshooting the global rate the last 15 years (5-10X the rate instead of the theoretical and long-term ratio of 2-3X).

Careful here. I think we're wandering into chicken/egg territory. The theoretical rate was at least partially based on slower degradation of ice cover. Now that this portion of the prediction was under-forecast, it undermines the original assumption, at least until the albedo-feedback mechanism weakens.

Very little guidance (other than the few already listed) suggested a rapid non-linear response like the one we've experienced so far. There could be another mechanism to slow it down (other than increased emitted radiation in the autumn/winter from open water), but I haven't found one or have seen a decent argument for it.

As I suggest above, SAT is mostly causative of low sea ice not the other way around. This has been repeatedly demonstrated by studies which fully recover SIA in just a few years, even when all sea ice is artificially removed. SIA resides in close equilibrium with global temperature and thus CO2. You cannot permanently warm the globe or the arctic by temporarily removing sea ice. You can only do so temporarily.

Underline: By the very same models in which carry the same weaknesses that generally do not allow for the non-linear response that has been seen. This is likely most strongly related to parameterizations (or a lack of decent ones in a lot of cases) and model resolution. Trying to correct that problem takes a lot of precious computational power, which is expensive. Often times, a balance has to be made in that case, and the price paid is often a penalty to accuracy.

Bold: This assumption does not/has not fit observation.

Italics: Net heat transport is INTO the Arctic via advective processes. Surplus heat generated in the Arctic via albedo changes generally stays there. In fact, this surplus heat weakens the advective process, emphasizing albedo and its related feedbacks as more important with time.


Just to note, I'm not arguing about CO2 being the primary initial driver. What I'm arguing is that while CO2 started this circus, albedo changes have moved to the front seat of the bus, right along side the driver. I would put the feedback forcing at this point at a higher emphasis than you have (closer to 60/40, rather than your 80/20). It's not so cut and dry though, since other changes are now taking place (atmospheric and oceanic responses) that will become important "passengers" too. I guess our main disagreement is on the actual strength of the albedo-feedback response on sea ice. Would that be a fair assumption in your opinon?

[Csnavywx]

I am not sure I follow you in #3. You seem to be assuming that all the heat is trapped in the arctic and there is no change in atmospheric heat transport. This is a wildly incorrect assumption. You can apply a candle to only one side of a pot of water, but the entire pot will warm, and the side you warm will only warm slightly more than the side that has no candle.

If what you say were even remotely close to true the arctic would have a massive runaway feedback that was never stable. Think about it, a small decline in SIA would have a regional warming effect of >1C.

By contrast models and history show SIA and arctic SAT to stay in close equilibrium with global T and CO2.

Of course not all of that remains in the Arctic, just most of it. It doesn't runaway for two reasons:

1) Advective processes (that run on dT/dx) weaken as it warms. This is due to weakened wind speeds. This would result in a reduction of heat transport from lower latitudes.

2) Significant increases in outgoing longwave radiation occur over newly open water, which also dampens the response.


My actual argument was that closer to 1C of the temperature change there was due to albedo. This has come as a result of a significant loss of SIA/SIV, not a small decrease.

[Carex]

Andrew, Csnavy and Chris, beautiful.  This is the reason we lurkers are here.  Some red meat, some disagreement.  A little education from those who know the work much better than I do or ever will.  So some questions from the ignorant to further our education.  Andrew, why in your minimal effects of albedo argument do you distribute the albedo effect globally?  It seems to me that the effects would be almost entirely local and seasonal.  Local albedo increases local near surface water temperatures which increases ice melt.  Most of the heating would be radiated out in fall before the water is transported out of the arctic.  The global temperature change seems insignificant to the question.  This problem is mediated some what by using, probably, the maximum albedo difference between water and snow. 

In a melting situation much of energy is going into phase change.  If you put a candle next to a pot full of snow the snow close the candle is going to melt long before the snow on the far side of the pot.

Also, the comparison to the global effect of doubling CO2 does not seem relevant to what is being observed on the ground.  CO2 is globally emitted throughout much of the globe and is circulated and active as a Greenhouse gas relatively quickly.  And while doubling CO2will result in an increase radiative forcing of 3.7W/m2 or about 1.0C warming, there has only been a 17% increase in CO2 during the period of comparison or about 0.2C.  Your figure of 0.1C from increased albedo 1980 to 2007 seems quite significant.  Although you obviously understand the physics of weather and climate much better than I, you seem to be comparing apples and oranges.

This also cuts to the core of distrust and perhaps misunderstanding of models by wee lay folk.  Because models must work on a global scale many of the local details must be eliminated.  Because effective models should work backwards and forwards, the fit to geologic changes does not seem to be easy to reconcile with the rapid rates of change we are living in. Over the long term I trust the models and modelers and I trust that the will balance out. But over the next ten or 20 years it seems more likely that regional forcing that is calculated over global areas can have stronger regional effects than the models may detect.  Mr. Zang's results sound encouraging.  Were the parameterized models run out beyond current conditions?  If so, how do the parameterized and non-parameterized runs compare.

I, lowly citizen, lay my discussion on your table to be  flayed.

[AndrewP]

My actual argument was that closer to 1C of the temperature change there was due to albedo. This has come as a result of a significant loss of SIA/SIV, not a small decrease.

OK so we agree that much of the temperature increase is spread globally by advection. If the forcing 1979-2007 was .1W/m2, as the paper I posted calculates, then the forcing through 2013 would be what .12W/m2? Now if this forcing was concentrated over 6% of the earth with no advection, regional temperatures would rise 1.5C. With advection, it's going to be way less than that. Well under the 1C you suggest. Perhaps the in the vicinity of .2-.5C. I'm not really basing this on anything other than what seems reasonable so if I'm overestimating the effect of advection let me know.

Over the same period the arctic warmed 2C, so albedo change explains a mere 10-30% of the increase.

Then we have to factor in the fact that climate models, even if they didn't model 100% of the decline in sea ice, modeled a large portion of it (say 60-80%).

So now returning to the original question which was "How much of the underestimation of warming of the arctic is explained by model underestimation of sea ice loss?"

(20-40%) * (10-30%) = 2%-12%


And come to think of it, aren't any of the models parameterized with observational ice extents? If so, it renders the whole question pointless. In that case, none of the underestimation of arctic warming is due to underestimation of SIA loss.

[AndrewP]

Andrew, Csnavy and Chris, beautiful.  This is the reason we lurkers are here.  Some red meat, some disagreement.  A little education from those who know the work much better than I do or ever will.  So some questions from the ignorant to further our education.  Andrew, why in your minimal effects of albedo argument do you distribute the albedo effect globally?  It seems to me that the effects would be almost entirely local and seasonal.  Local albedo increases local near surface water temperatures which increases ice melt.  Most of the heating would be radiated out in fall before the water is transported out of the arctic.  The global temperature change seems insignificant to the question.  This problem is mediated some what by using, probably, the maximum albedo difference between water and snow. 

In a melting situation much of energy is going into phase change.  If you put a candle next to a pot full of snow the snow close the candle is going to melt long before the snow on the far side of the pot.

Also, the comparison to the global effect of doubling CO2 does not seem relevant to what is being observed on the ground.  CO2 is globally emitted throughout much of the globe and is circulated and active as a Greenhouse gas relatively quickly.  And while doubling CO2will result in an increase radiative forcing of 3.7W/m2 or about 1.0C warming, there has only been a 17% increase in CO2 during the period of comparison or about 0.2C.  Your figure of 0.1C from increased albedo 1980 to 2007 seems quite significant.  Although you obviously understand the physics of weather and climate much better than I, you seem to be comparing apples and oranges.

This also cuts to the core of distrust and perhaps misunderstanding of models by wee lay folk.  Because models must work on a global scale many of the local details must be eliminated.  Because effective models should work backwards and forwards, the fit to geologic changes does not seem to be easy to reconcile with the rapid rates of change we are living in. Over the long term I trust the models and modelers and I trust that the will balance out. But over the next ten or 20 years it seems more likely that regional forcing that is calculated over global areas can have stronger regional effects than the models may detect.  Mr. Zang's results sound encouraging.  Were the parameterized models run out beyond current conditions?  If so, how do the parameterized and non-parameterized runs compare.

I, lowly citizen, lay my discussion on your table to be  flayed.

I've identified several points and questions here which I have attempted to respond to:

1. Shouldn't the radiative forcing of a local albedo change be distributed almost entirely locally?

I could probably have been clearer in my presentation, but the end result is the same. Advection would distribute much of the heating from a local RF globally. I basically calculated the global RF and then worked backwards to how much of this might reasonably be expected to concentrate at the source. It would probably make more sense to calculate the local RF and then estimate how much of this heat would reasonably be expected to distribute globally. The local RF, when an area that used to have ice is now open water, is about 160W/m2!!! depending on solar insolation at the location (Solar insolation averages ~200W/m2 in the arctic summer and 200*.8=160).

Now we've thrown the system way out of equilibrium. Outside of the summer months, the arctic relies upon an incredible amount of thermal advection from outside the arctic and also on heat accumulated in summer months in order to maintain temperatures above absolute zero (-273C). The region that has become ice-free in August and September, but used to have ice, has now had a fire taken to it which burns at 160W/m2 during those months. The "new" energy will go into increasing OHC, thinning surrounding ice, and into the atmosphere via evaporation, conduction, and LW radiation. A lot of it will indeed go into the atmosphere (via the local increase in evaporation, conduction, and LW radiation) and be transported globally during August and September. But the portion that goes into increasing OHC and thinning surrounding ice must be dissipated October-April. Fortunately, by the time April has rolled around all or nearly all has dissipated and the ice is as thick or nearly as thick as at this time the year before. Thus, over the course of a year the energy is distributed globally and there is little or no permanent accumulation of heat. It isn't distributed globally immediately, and so during August, September and October there is an increase in temperature in the immediate environment. But even during this time huge amounts of heat are being dissipated via an increase in evaporation, conduction, and LW radiation to the atmosphere. Hopefully this explains why there is not anywhere close to a 120C+ increase in local temperature from the 160W/m2 local RF.

2. A .1W/m2 RF from SIA 1979-2007 seems significant compared to a the RF from a 17% increase in CO2 concentration in the same period.

Two problems. First, you've got the RF of a 17% increase in CO2 wrong. A 17% increase in CO2 provides more than 17% of the RF of a single doubling (3.7W/m2), because it's logarithmic (the first 1% provides more RF than the last 1% on the way to a doubling). The equation for determining the RF of a change in CO2 concentration is:

5.35*ln(CO2/CO2').. in this case 5.35*ln(1.17) = .84W/m2

The Planck response (~1.1C/3.7W/m2) yields .25C (slightly more than the .2C you got)

the second problem is the bigger one. You've made an apples to oranges comparison. You've compared the planck response of a 17% increase in CO2 to the equilibrium response (including feedbacks) to a .1W/m2  RF from SIA. Including feedbacks, the response to a 17% increase in CO2 would be .68C. That is nearly 10X the .075C (to be more precise) expected from a .1W/m2 RF from SIA loss.

Or we could just compare the RFs directly. A 17% increase in CO2 is an RF of .84W/m2. The RF from SIA during the same period was .1W/m2. Of course most of this RF was actually a feedback to the CO2 RF.


3. Were the parameterized models selected by Zhang run into the future?

The models weren't actually parameterized. They simply got the relationship between arctic SAT and SIA correct. Which means they've got the ice physics reasonably correct, similar to PIOMAS.

If they were parameterized, as PIOMAS is, they would have a similar accuracy at hindcasting as PIOMAS does. Definitely for SIA. And as I suggested in response to Chris, I would assume for SIV as well. (How else are they getting the relationship between SIA and SAT correct if they can't do SIV pretty well too?)

These models, which have hindcasting skills similar to those of PIOMAS, when run into the future predict an ice free arctic between 2037 and 2065. Earlier than the rest of the models, but still not the 2010s or 2020s a lot of people expect.

[ChrisReynolds]

Andrew,

Sea ice loss is driving Arctic Amplification in large part. This is seen in the lower tropo warming, and has been confirmed in research e.g. Simmonds - "The central role of diminishing sea ice in recent Arctic temperature amplification."

Actually I'm suggesting Arctic warming of 6degC as it's already warmed by 3degc.

Consider the following graphs:
http://farm9.staticflickr.com/8103/8481914926_3f4b4511eb_o.jpg
and
http://farm9.staticflickr.com/8226/8468995762_a2009cf3f3_o.jpg

I was of a similar opinion to you:
http://dosbat.blogspot.co.uk/2011/07/arctic-sea-ice-free-this-decade.html
But I have since then changed my opinion:
http://dosbat.blogspot.co.uk/2012/09/a-sea-ice-free-arctic-by-2020-and-maybe.html

I'll address the rest of what you say when I have more time after work - got to dash.

[Wipneus]

Newer models accurately simulate the decline in extent and volume.

If a model can accurately simulate the recent declines, then there should be reasonable confidence in its predictive power.

A lot of the CMIP5 models are actually pretty decent. One must remember that the models do not include internal variability. They only represent the forced portion of the trend. They don't have actual weather observations parameterized as PIOMAS does. As I've said above, much of the recent decline is likely not forced. Arctic warming has been much faster the last 15 years than it is likely to be the next 15 years.

Models have their own internal variability, they are part of the equation  and you would hope that they are consistent with observations. It is one reason why multiple runs with the same model are provided.

I picked the 1980's because arctic amplification is not that apparent and  volume decline is not that fast yet. Look at the Stroeve presentation to convince yourself that the same conclusions can be drawn from the 1850-1980 era.
Without ANY reference to observations or PIOMAS the models differ in opinion about ice volume by a factor of 20.  You cannot even BEGIN to contemplate modelling the decline when the more steady state before it is already that badly represented (in the sense that there is very little agreement).

As Zhang 2010 points out, models also considerably underestimate surface air temperature increase over the recent period as well. In addition, the subset of models chosen are able to accurately simulate the rapid loss of sea ice in 2007.

"... the extreme event of sea ice cover loss in summer 2007 also fall into the range of changes simulated by the subset."

Hmm, Zhang talks about area here. My argument is restricted for the moment about volume. Further the models used here are all a release older than the CMIP5 models. I would expect the CMIP5 models  to perform at least as good as those used in the article.
Thanks for the link anyway, had not red it before.

PS. Zhang includes observations to 2008. If you fill in 2009-2012, I find that 2011 and 2012 are as extreme as, and even extremer than 2007. Without extreme surface temperatures comparable to 2007.  This seems far from settled.

[ChrisReynolds]

Andrew,

OK, sorry but this will be long, so I'm splitting it....

Before addressing the points you raise in more detail, I should explain my preceding post: I was rather vague when referencing this graph.
http://farm9.staticflickr.com/8226/8468995762_a2009cf3f3_o.jpg
My point in presenting this is rather simple. I had been sceptical of talk of an imminent sea ice free state, preferring to accept the GCM's projecting around 2030. This was mainly because I was expecting strong negative feedbacks to kick in: The heat loss due to anomalous open water in autumn causing autumn warming to be the greatest during the year. The heat loss due to thinner ice. Increased IR emission due to warmer atmosphere. The reduced insolation around the pole due to lower angle of incidence.

In essence I thought that "the main problem is that people seem to be confusing the effects of thinner ice cover with the suggested intial symptoms of a rapid transition to a seasonally sea-ice free state."

However what the above graph shows is that short term, simple, extrapolation brings melt season volume losses equal to winter maximum volume by 2012. This implies that after that point there will be zero volume during the late summer. Now the problem is this: These negative feedbacks are already at play, yet within 5 years volume in the summer should reach zero. Therefore is it reasonable to suppose that these negative feedbacks have enough power to overcome the volume loss? Furthermore if something is going to appear to stop the volume loss it has only 5 years to do it!

[ChrisReynolds]

Andrew,

Now onto your points:

"If we're going to call arctic SAT projections into radical doubt, why not call the global projections into radical doubt as well, as the deniers do?"

I do call the global projections into doubt. I think the early loss of the Arctic has global consequences that have not been factored into projections. The primary issue is in the occurence of extreme events, the Arctic is being implicated in flooding and drought events, also in the lack of boreal winter warming in the northern hemisphere. The difference between me and a denier is I don't throw the models out for now reason, and I keep using model based studies for their diagnostic utility. I have to agree with Csnavywx that trying to work out RFs globally from Arctic processes is not really useful. This doesn't tell us about the myriad complex effects on atmosphere and within the Arctic Ocean itself.

"...albedo feedback..."

Actually albedo feedback doesn't lead to a great degree of direct warming. It happens in the ocean, which has a high specific heat capacity and vertical mixing so warms less than land would with the removal of snow. Furthermore it happens in the summer when sea ice pegs temperatures it its vicinity down to zero, 'spare' energy goes into melting ice not sensible warming (sensible = that which can be sensed). Indeed there's a paper examining the 2007 ice loss which finds that while ice albedo feedback had a role increased downwelling IR due to increased atmospheric moisture had (IIRC) at least as big a role in melting ice.

The sensible warming that the albedo feedback causes is delayed. The ice/ocean system acts as a capacitor during the summer, storing energy, and a large part of this is latent heat involved in melting ice. This is because before the ocean can freeze its surface must lose energy to the atmosphere in order to cool so as to freeze, then the freezing process itself releases heat - effectively the heat used to transition the ice from solid to liquid during the summer.

"You cannot permanently warm the globe or the arctic by temporarily removing sea ice. You can only do so temporarily.  Arctic SIA and SAT are determined by CO2."

I'm sorry but I think you're very wrong here. I've already directed you to Simmonds in my post above - that paper shows that the autumn/winter warming is directly due to reduced sea ice. And as I outlined above the period when warming is causing sea ice loss is not actually warming because the energy from that notional warming (from the sun) goes into melting ice. Francis & Hunter "Changes in the fabric of the Arctic’s greenhouse blanket" examines outgoing infra-red over the Arctic and finds that water vapour and clouds are the prime mover behind this change, not CO2. This is not to say that CO2 has no role, but in the Arctic it is the force majeur behind the scene driving a mass of feedbacks. If I can refer to my blog again, both Johanessen and Notz & Mortzke show that behind the changes CO2 is the prime mover:
http://dosbat.blogspot.co.uk/2012/05/co2-causes-arctic-sea-ice-loss.html
The reason for this is that CO2 is driving global warming and increased heat flux from ocean and atmosphere have a strong role in changes in the Arctic. Add to this the integrating nature of sea ice volume which 'remembers' impacts upon it, and is declining in an environment now inimical to it, and to that add the relationship between volume and area/extent (open water formation efficiency). And a qualitative sense of what's going on in the Arctic emerges.

The reason why Zhang's linkage between SIA and SAT works is that he restricts it to the summer JJA, other periods in the year seem to suggest recent acceleration away from the relationship. The increasingly large areas of open water will warm, due to ice albedo feedback, and this will be anomalous, creating warm anomalies. These warm anomalies appear in a regional average, and because the area of open water is increasing the anomalously warmer areas increase in area, making a greater impact on the regional average. Therefore the temperature anomalies are IMO both cause and effect, which makes relying on Zhang's linkage of SIA and SAT rather unsatisfactory for me.

It's worth noting here that I agree with Wipneus about the effect of including more recent years.

[AndrewP]

Models have their own internal variability, they are part of the equation  and you would hope that they are consistent with observations. It is one reason why multiple runs with the same model are provided.

I picked the 1980's because arctic amplification is not that apparent and  volume decline is not that fast yet. Look at the Stroeve presentation to convince yourself that the same conclusions can be drawn from the 1850-1980 era.
Without ANY reference to observations or PIOMAS the models differ in opinion about ice volume by a factor of 20.  You cannot even BEGIN to contemplate modelling the decline when the more steady state before it is already that badly represented (in the sense that there is very little agreement).

Yes Chris corrected me on that. Although with so many lines it's hard to see that some of them do show a precipitous drop at periods, similar to what we have seen at present. I agree as a whole, the models are not very good. But choosing a subset which when prescribed with observed weather parameters performs similarly to PIOMAS (which we all agree is fairly good), as Zhang does, should yield a much more accurate picture.

As Zhang 2010 points out, models also considerably underestimate surface air temperature increase over the recent period as well. In addition, the subset of models chosen are able to accurately simulate the rapid loss of sea ice in 2007.

"... the extreme event of sea ice cover loss in summer 2007 also fall into the range of changes simulated by the subset."

Hmm, Zhang talks about area here. My argument is restricted for the moment about volume. Further the models used here are all a release older than the CMIP5 models. I would expect the CMIP5 models  to perform at least as good as those used in the article.
Thanks for the link anyway, had not red it before.

PS. Zhang includes observations to 2008. If you fill in 2009-2012, I find that 2011 and 2012 are as extreme as, and even extremer than 2007. Without extreme surface temperatures comparable to 2007.  This seems far from settled.

I would guess that if they are doing fairly well on area, then they probably do fairly well on volume too. It's also possible that although 2011 and 2012 were not as warm as 2007, the models would correctly show that it was still warm enough for continued thinning and area losses. Including more recent years is just as likely to validate the subset of models chosen by Zhang as invalidate them.

An updated version of that paper, preferably that included a discussion of volume, would be very interesting. Especially if it used CMIP5 models.

[AndrewP]

Andrew,

OK, sorry but this will be long, so I'm splitting it....

Before addressing the points you raise in more detail, I should explain my preceding post: I was rather vague when referencing this graph.
http://farm9.staticflickr.com/8226/8468995762_a2009cf3f3_o.jpg
My point in presenting this is rather simple. I had been sceptical of talk of an imminent sea ice free state, preferring to accept the GCM's projecting around 2030. This was mainly because I was expecting strong negative feedbacks to kick in: The heat loss due to anomalous open water in autumn causing autumn warming to be the greatest during the year. The heat loss due to thinner ice. Increased IR emission due to warmer atmosphere. The reduced insolation around the pole due to lower angle of incidence.

In essence I thought that "the main problem is that people seem to be confusing the effects of thinner ice cover with the suggested intial symptoms of a rapid transition to a seasonally sea-ice free state."

However what the above graph shows is that short term, simple, extrapolation brings melt season volume losses equal to winter maximum volume by 2012. This implies that after that point there will be zero volume during the late summer. Now the problem is this: These negative feedbacks are already at play, yet within 5 years volume in the summer should reach zero. Therefore is it reasonable to suppose that these negative feedbacks have enough power to overcome the volume loss? Furthermore if something is going to appear to stop the volume loss it has only 5 years to do it!

Yes I understand where the predictions of imminent ice-free conditions come from. And the trend-lines do introduce some serious doubt into conclusions that rely upon modelling.

As I've suggested, I believe the recent rapid rise in arctic temperatures is due not only to albedo amplification, but also internal variability in weather and ocean currents that are unlikely to continue. You've disagreed and blame it almost entirely on albedo effects, which are likely to continue and get worse. I'll get to those posts below.

I do have one objection to what is above and one supporting observation for the strength of negative feedbacks. The negative feedbacks are roughly directly proportional to the minimum SIA, not volume (although thinning does provide some negative feedback via increased heat loss through thinner ice in the fall). The primary negative feedback is heat loss by water in the re-freeze in the fall. Prior to 2012, minimum SIA had only decreased 2-2.5 million from the 80s and early 90s norm. That means the strength of the negative feedbacks had only reached half of their full potential. We had witnessed some increase in volume generation during the re-freeze, but not a lot. The negative feedbacks were probably being largely canceled by warming winter temperatures and OHC storage from summer.

The strength of the negative feedback can be measured by the volume of ice generated from minimum to maximum, as seen in the chart you posted:

http://farm9.staticflickr.com/8103/8481914926_3f4b4511eb_o.jpg

We see some evidence of the negative feedbacks 2007-2011 which generated 2-3 million km3 more ice in the re-freeze process than the historical norm.


However, 2012 had nearly .7+ million sq km less SIA than any other year. So it is a perfect test-case for the strength of negative feedbacks. Thus far we have generated 1 million cubic km more ice than last year, and possibly still growing. If PIOMAS is right, the negative feedbacks have activated and are keeping us in a holding pattern where we will start 2013 with no less ice than 2012 (or 2011). If we're starting every spring off from the same volume and extent, then the positive feedbacks can't get any stronger than they were in summer 2011 and 2012.

It will take further external forcing to finish off the last of the ice.

The ice loss thus far has taken place in an environment of little global warming. The positive feedbacks, and in my opinion (with which I think you disagree) unfavorable weather and ocean patterns some of which may be partially reversible, were able to run the show. If winter 11-12 and especially 12-13 are any example, the positive feedbacks (and bad weather/currents) may have taken us as far as we're going to get.

[Jim Hunt]

Firstly may I congratulate all the participants on the quality of the current debate.

Secondly we appear to have a professional "weather forecaster" taking part. Is there by any chance a professional "climate modeller" in the house also?

[ChrisReynolds]

Andrew,

This season and the winter volume increase.

But we start this season with unprecedented levels of first year ice.
http://dosbat.blogspot.co.uk/2013/02/ascat-piomas-and-dam.html

And first year ice gains almost 1/3 more energy over the melt season than old ice.
http://dosbat.blogspot.co.uk/2012/09/sea-ice-albedo-changes-and-seasonal.html

In PIOMAS the thinning of the ice in 2010 has lead to a more aggressive spring melt over the last three years.
http://dosbat.blogspot.co.uk/2013/03/what-caused-volume-loss-in-2010-part-2.html

Therefore this year's starting volume is less significant than the overall composition of the pack. There will be an aggressive spring melt this year, as with the previous years, and this will be followed by enhanced energy absorption due to predominant FYI. So the slightly higher volume will be more than made up for by losses throughout the season.

Negative feedbacks.

Yes the ice growth/thickness feedback is active, yes it has been apparent in the post 2007 increase in annual range for area/extent and volume. However the volume losses continue without abatement.
http://farm9.staticflickr.com/8103/8481914926_3f4b4511eb_o.jpg

Yes the negative feedback of growth/thickness and heat loss in autumn due to anomalous open water has not yet reached its full potential. That they will increase in strength is not the issue, there will be a tail of remnant late summer ice after this RILE is spent. But the volume losses are increasing and since 2001 only 2008 has shown an increase in summer volume. Against the volume loss, maintaining that there will not be virtually sea ice free conditions until 2040 just doesn't strike me as realistic.

Albedo feedback.

I won't restate what I said in my earlier post. But I did not say that albedo feedback is driving the process, it's importance is increasing, but the process of the removal of old ice is many decades old. Area/Extent is a sideshow, the prime mover is volume.

Weather that may change?

I have mentioned the issue of the summer dipole anomaly. This is not weather it is a response to, and augmentor of sea ice loss. This is demonstrated by the fact that the causative atmospheric pattern appears as a regular summer feature after the 2007 crash.
http://dosbat.blogspot.co.uk/2011/09/summer-daze.html
Dipole anomalies are associated with sea ice loss events.

Ocean heat flux plays an important role in the background forcing driving the loss of sea ice, and is observed to be higher than for hundreds of years (I forget the ref for that). Bitz et al 2006, "The influence of sea ice on ocean heat uptake in response to increasing CO2", gives useful insight into what may be happening now:
http://www.atmos.washington.edu/~bitz/Bitzetal_OHU.pdf
While Ocean heat transport declines approaching 2XCO2, brine rejection from formation of new ice in the autumn in a low sea ice regime causes increased mixing and draws in warm water from the Atlantic. In the model this is at 2XCO2, but the current low levels of MYI and predominance of FYI is a result of processes which would apply regardless of CO2 level or warming, but are strictly dependent on the levels of sea ice and the amounts of open water and remnant ice surviving the summer.

So here are but two reasons why we might suspect that atmospheric and ocean heat transport are not going to decline. In any case, I am bored of the idea that the current ice recession is an accident of weather, it just doesn't tally with the evidence.

[AndrewP]

Sorry Chris I didn't get around to responding to the role of albedo in arctic amplification. If I understand the Simmonds paper correctly I don't think it is fully supporting your assertion.

My contention has been that arctic SAT will not increase as quickly the next 15 years as the last 15 because in the last 15 arctic temperatures increased 5-10X as fast as the rest of the globe and climate models say the ratio in the long-run will be 2-3X. A number of papers, and the leaked IPCC report, suggest short-term internal variability may have had a role in recent sea ice losses. However, you say that it increased like this because of unexpected sea ice loss and albedo loss, not "weather."

However, why then has Nov-April arctic SAT increased over 1C over the last 15 years while the globe has increased 1/10th of that over the same period? You can't blame that on albedo.. maybe some of it you can blame on heat loss through thinner ice. But a lot of that occurs in October. I don't think that can come close to explaining the huge 10X amplification of .1C globally to >1C in the arctic winter. Only weather can IMO.

Simmonds looks at a long-term period and determines that albedo loss explains a large portion of the amplification @ 2-3X global rates. Clearly albedo makes for a significant feedback. But I don't think that the Simmonds paper is a specific endorsement of the idea that all, or even the majority, of the arctic amplification the last 15 years can be traced to albedo. Obviously some of it has. But it doesn't convince me that a large portion was not also due to weather. Especially considering that warming was actually fastest from Dec-May, in winter, when albedo changes have been much more modest.

Even ignoring that, I don't know that Simmonds specifically supports the idea that a decline in SIA of 2 million sq km is responsible for 1C+ of arctic warming, when the globe has only warmed .1C.  I think one would need a paper which artificially reduces SIA by the observed amount and measures how much warming is produced. Or something along those lines. Do you know of any papers like that?

Can a decline in SIA of 2 mil sq km cause 1C+ of arctic warming when the earth has only warmed .1C?

I can only read the abstract from Simmonds, but so far it doesn't convince me of this specific claim.

[AndrewP]

Now back to your more recent post Chris and the topic of negative feedbacks, this years volume, and the FYI proportion.

You point out that this year has much less MYI than last year and that FYI absorbs 1/3 more energy than MYI. But how significant is this when 75%+ of the ice was FYI already in previous years at the start of melt season? And also I am guessing that part of the reason FYI has a lower albedo is that it is thinner. As the graphic you posted shows, FYI reached an "open water" state that MYI did not. And the "pond evolution" phase probably had a lot more leads opening up in the first year ice. This year the FYI is thicker than in previous years (deduced from the fact that we have more FYI but same volume as last two years).

So what's worse, 75% FYI where the FYI is 1.5m thick, or 80% FYI where the FYI is 1.7m thick? You're probably right, the latter is worse, but how much worse?

Can I also presume from the above that if in spring 2014 we enter the year with volume equal to 2011-2013 and a FYI proportion equal to 2011-2012, that we have reached an equilibrium which can be broken by further external warming? Because right now the only argument that we have not reached an equilibrium with 2011-2012, is that 2013 has a higher FYI fraction. Were that not true in 2014, would you agree a temporary equilibrium has been reached? That would also mark 4 years in a row with unchanged max volume.


I think we both agree that max spring volume 2007-2012 provided evidence of negative feedbacks in the fall/winter, but positive feedbacks and AGW in summer continued to drive the minimum volume lower. My argument is that the negative feedbacks are proportional to SIA. 2007-2011 all had similar minimum SIA. 2012 is a new test case with dramatically lower SIA. I believe the volume generation this fall/winter will stick out like a sore thumb on your graph.

http://farm9.staticflickr.com/8103/8481914926_3f4b4511eb_o.jpg

The annual range could approach 20 million cubic km this year. Well above any previous year.


That could be the start of something new. Winter volume generation is proportional to the SIA minimum. The lowest minimum thus far was 2.2 million, and the average of the last 3 years was ~2.8 million. That gives us substantial buffer room.

This is why simply extrapolating the volume trends is a bad methodology. The SIA trend is more indicative of the balance between positive feedbacks, negative feedbacks, and the externally forced environment. Thus far volume has decreased faster than area. This trend cannot continue. There is a substantial SIA buffer remaining.

This is probably the most significant point in my opinion. Ultimately, what determines whether we are gaining or losing ice is the AREA of ice that exists, not the volume. The AREA of ice must exist in equilibrium with the global climate and whatever positive and negative feedbacks exist. The AREA of ice is what determines the energy flows into and out of the arctic. The AREA is what determines whether there is or is not an energy equilibrium. If the climate warms, we lose area until a new equilibrium is reached. The current climate dictates an area of 2.2-3.2 million sq km of ice. We will need a warmer climate to dictate an area of <1 million.


In the end, extrapolating SIA should have more predictive power than extrapolating SIV.

[crandles]

Can I also presume from the above that if in spring 2014 we enter the year with volume equal to 2011-2013 and a FYI proportion equal to 2011-2012, that we have reached an equilibrium which can be broken by further external warming?

I agree that if that happens it would be a strong point in your favour.

Winter volume generation is proportional to the SIA minimum. The lowest minimum thus far was 2.2 million, and the average of the last 3 years was ~2.8 million. That gives us substantial buffer room.

This is why simply extrapolating the volume trends is a bad methodology. The SIA trend is more indicative of the balance between positive feedbacks, negative feedbacks, and the externally forced environment. Thus far volume has decreased faster than area. This trend cannot continue. There is a substantial SIA buffer remaining.

I disagree with the characterization of 'proportional to the SIA minimum'.

I think proportional to how far below a winter maximum equilibrium volume is better because thicknesses in winter approach a thermal equilibrium thickness. So if you start off with less ice volume and area at minimum there is a tendency to quickly catch up to the volume of previous years. Once it has caught up there is no more ice thickness reason for it to continue to grow faster than in previous years. If it does continue to grow faster then I would tend to blame less snow cover and that means lower albedo for a faster start to next seasons melt.

>"Because right now the only argument that we have not reached an equilibrium with 2011-2012"

So actually 3 reasons:
1. Reduced MYI area/volume so faster take up of heat through albedo difference
2. Snow cover might be low due to late formation of ice after low minimum area causing lower albedo and faster start to melt season.
3. GHG levels keep increasing.

I accept that 3 is likely to be a fairly slow factor so if factors 1 and 2 are shown not to be important by 2013 not reaching a new low and volume returning to 2011-2013 maximum volume levels then it will look like ice free will be some way off. Even then that could be caused by random weather or some tendency to bounce after a new low in 2012 such that a 2014 or 2015 new low could still be possible. But if it happens it would support your views.

[ChrisReynolds]

Andrew,

"why then has Nov-April arctic SAT increased over 1C over the last 15 years while the globe has increased 1/10th of that over the same period? You can't blame that on albedo.. maybe some of it you can blame on heat loss through thinner ice."

You seem to have forgotten that, as I explained, summer energy gains (sensible and latent heat) are being released in the autumn causing the autumn warming. The ice/ocean system is acting like a capacitor storing heat in the summer and releasing in the autumn. Then during the winter thinner ice is allowing more of this gained heat to escape. But in tandem with these processes the volume is going down year on year, which shows that the whole system energy is increasing.

As the summer has more open water, there will be more warming away from the ice edge, increasing summer temperatures, and more release of heat to the atmosphere in autumn increasing autumn temperatures.


I don't know where you get your thicknesses from. But this year is down on thickness from 2011 and 2012 has only met this year's thickness because 2012 thinned in the winter.
http://psc.apl.washington.edu/wordpress/wp-content/uploads/schweiger/ice_volume/Bpiomas_plot_daily_heff.2sst.png


The amount of FYI this year is substantially greater than in the previous two years, and those two years show much higher FYI than any previous year.
http://dosbat.blogspot.co.uk/2013/02/ascat-piomas-and-dam.html

This is backed up by analysis of PIOMAS gridded data. The following graphic is a table from a spreadsheet: September volume broken down into contributions from 25cm thickness bands each row expressed as percentage of total volume for that March. (sorry - forgot to include years)
http://farm9.staticflickr.com/8237/8582706478_d1b73f6619_o.jpg
From this it can be seen that not only is the thickest ice virtually gone, but also the peak below the thickest ice has dropped to the order of 1m thick. That was the state up to just before this freeze season. PIOMAS gridded is updated per year early the following year.


This year's gain from daily min to March 1st was of the order of that following 2007.
http://farm9.staticflickr.com/8251/8582706458_0202832116_o.jpg
However scatter plot against daily min for the preceding year does not suggest a strongly non linear relationship between the volume at min and the volume gain in the following autumn/winter. No need to do a time series as you can appreciate volume generation this winter - it's large but not exceptionally large for the post 2007 era.


I don't agree that an equilibrium has been reached. After 2007 there was a rebound in thickness - going back to the PIOMAS thickness graph.
http://psc.apl.washington.edu/wordpress/wp-content/uploads/schweiger/ice_volume/Bpiomas_plot_daily_heff.2sst.png
This was also reflected in 2008 being a gain in volume. Here is the most recent (Mar 1) date in PIOMAS daily since 2000, year, volume, change from previous year:

2000   25.26   -0.027
2001   25.786   0.526
2003   25.25   -0.536
2004   23.777   -1.473
2005   23.715   -0.062
2006   23.001   -0.714
2007   21.958   -1.043
2008   22.703   0.745
2009   22.734   0.031
2010   21.193   -1.541
2011   19.856   -1.337
2012   19.671   -0.185
2013   19.746   0.075

This is not a rebound like 2008, 2012's loss was not a deviation below equilibrium like 2007.

However 2012 was not a major volume event, the last such event was 2010, and after 2010 there has been no rebound. Therefore we have not seen a dip below equilibrium, with following years hunting back up to the equilibrium (albeit these are pseudo-equilibria against a shifting sweet spot).


Area and extent are a sideline, their trend is epiphenomenal of underlying changes in volume. Area and extent are easy to measure, this alone accounts for their popularity and the supposed significance attached to them. Open Water Formation Efficiency is the relationship between area/extent and volume. Holland et al, 2006, "Future abrupt reductions in the summer Arctic sea ice" define OWFE as 'percentage of open water formed per cm of ice melt'. The following is a graphic of OWFE as a function of March thickness.
http://farm9.staticflickr.com/8237/8466170424_22569325ba_o.jpg
Essentially as the ice has thinned so the ease with which open water is formed has increased. Area/extent changes are not driving the process of sea ice loss, volume loss is driving the process.


"The annual range could approach 20 million cubic km this year. Well above any previous year."

Given that last year was 3.261k km^3, and 1/3/13 was 19.746k km^3, a gain of 16.485k km^3, with a typical 1k km^3 to go to the maximum in April making about 17.5k km^3. I seriously doubt if the gain will be as much as 20k km^3, we'd need to gain 2.5k km^3, 2.5 times the typical gain from March 1st to April peak.

[ChrisReynolds]

Andrew,

Before I pack up for the afternoon, I just knocked this graph together.
http://farm9.staticflickr.com/8514/8581832089_f33e65cb58_o.jpg

It shows what you're claiming in words, but using the data. As area goes down the following seasons ice growth goes up. But this is no surprise, it's because of what I keep calling the growth/thickness feedback, and is due to heat flux through ice being greater for thinner ice. I've extended the best fit curve and coincidentally it meets the volume axis at about current winter volume levels.

However it really doesn't tell us anything about the pace of change in area/volume.

I've also previously calculated OWFE for the gridded PIOMAS data.
http://farm9.staticflickr.com/8366/8465073749_657a618e8c_o.jpg

You'll note that it becomes spread out in more recent years, this is something seen in models during RILEs. You'll also note that it follows a similar curve to the OWFE graph posted in my previous reply. But bear in mind it can cease before reaching the OWFE = 1 point, because open water can reach 100% for a seasonal melt of more than 1m.

[anonymous]

... I think we both agree that max spring volume 2007-2012 provided evidence of negative feedbacks in the fall/winter, but positive feedbacks and AGW in summer continued to drive the minimum volume lower. My argument is that the negative feedbacks are proportional to SIA. 2007-2011 all had similar minimum SIA. 2012 is a new test case with dramatically lower SIA. ...

AndrewP, this appears to be the foundation of your point. What makes you sure a few years with similar spring volume indicate a negative feedback at work? How do you distinguish it from usual deviation?

[Ice Cool Kim]

Since our previous conversation along similar lines I've got CICE up and running quite nicely on my 6 core desktop. The supplied data is from 1997 however!

Hey Cool ! How about setting up a R-Pi cluster??

Are you part of the team or is the code available?

[Jim Hunt]

Hi Kim,

Currently I'm the only member! (Of DASIM not CICE). For further information please see:

http://econnexus.org/projects/the-distributed-arctic-sea-ice-model/

and/or

https://forum.arctic-sea-ice.net/index.php/topic,108.0.html

which would seem to be the best place to continue this discussion. You'll see that RasPi clusters do get a mention!  I've since acquired the source for Wavewatch III and CESM 1.0 too, but I haven't tried to run them on my Pi yet!

At the moment I'm anxiously waiting for a pukka climate modeller to pass by, and pass comment. You're not one by any chance are you?

If you're keen to experiment with CICE 4 yourself, it can be freely downloaded via:

http://oceans11.lanl.gov/trac/CICE/wiki/SourceCode

[Ice Cool Kim]

Jim, thanks for that link. Source available via svn , brilliant! All science should be conducted this way.

No, I'm not working as a climate modeller although I do have some professional experience in 3D integrals in the atmosphere and interactions of radiation with cloud and precipitation, etc.

[Nightvid Cole]

The albedo discussion here has been completely ignoring the all-important contribution from snow.

The ice effect is never really that huge because the magnitude of the area anomalies peaks in September or October, when the Arctic isn't getting much insolation any more. Additionally, the open water areas in the Arctic Ocean tend to be very cloudy, further reducing the albedo effect from melting sea ice.

Snow, on the other hand, has its biggest deficit in May and especially June. And anomalously early melting of snow on N America / Eurasia doesn't appear to lead to high local cloud cover. (If anything, it has the opposite effect!) Thus the albedo effect of snow, unlike ice, is given an enormous boost by its proximity in time to the Summer Solstice, and the feedback is not reduced by cloud effects.

Right now, the snow over the Arctic sea ice isn't gone until the end of June, particularly near the Pole and on the CAA side of the Arctic. So the ice is to some degree 'protected' from the highest insolation of the Summer Solstice around June 21st.

When the snow atop the Arctic sea ice is ALL gone before the middle of June and the ice heavily ponded (a few years from now ), the ocean under the ice will warm up much more than it has been even in recent years, for three reasons. One, more insolation at the Solstice than later in the season. Two, higher solar radiation incidence angles near the Solstice mean a larger percentage of that insolation gets through the ice to be absorbed by the water underneath. And three, a longer snow-free period means more thinning of the ice, in turn making the ice not only more penetrable by the light, but penetrable by a larger increase than before, due to the exponential nature of the Beer-Lambert Law. Result: Near-complete melt-out by September due to catastrophic amounts of 'bottom-melt'.

The timing of snow cover retreat has advanced 3-4 weeks at very high Northern latitudes over the past few decades, and this trend is accelerating. Once it advances by an additional two weeks or so, the ice is toast.

[Ice Cool Kim]

The ice effect is never really that huge because the magnitude of the area anomalies peaks in September or October, when the Arctic isn't getting much insolation any more. Additionally, the open water areas in the Arctic Ocean tend to be very cloudy, further reducing the albedo effect from melting sea ice.

What you are saying is that the albedo can not be producing much of a positive feedback. That does not mean it is not having some significant effect.

Albedo (or reflectivity) is the complement of absorptivity, which is the same as emissivity.

So the lower albedo also implies greater emissivity.

The coeffs for short wave (incoming only) and longwave IR (in and out) are different but the same relationships apply.

Emissivity of open water is much higher than ice and/or snow (though ice is about 30% in long IR). So less ice provides a negative feedback effect by open water emitting more IR. As you point cloud cover will impede incoming SW and IR , it will not stop the surface emitting.

Open water also evaporates massively more than ice sublimates. Another negative feedback process.The increased emission and evaporative losses will continue through the long winter "night".

These two effects probably account for why we have not seen the run away melting a lot of simplistic commentators were expecting.

Since the 2007 low there has been much larger swings in the annual cycle. The years immediately following showed some clear recovery, even of the one-day-a-year metric (min) which is not representative of overall state of the Arctic climate. That is not compatible with a dominant positive feedback.

Dr Giles of UCL CPOM group recently commented on an apparent negative feedback being visible but did not forward a mechanism.

I think the above probably provides that mechanism.

[Peter Ellis]

Since the 2007 low there has been much larger swings in the annual cycle. The years immediately following showed some clear recovery...

I don't think that claim's justified.  The "recovery" after 2007 was no more than a reversion to the long-term mean rate of decline.
http://nsidc.org/images/arcticseaicenews/20091005_Figure3.png

[crandles]

Since the 2007 low there has been much larger swings in the annual cycle. The years immediately following showed some clear recovery, even of the one-day-a-year metric (min) which is not representative of overall state of the Arctic climate. That is not compatible with a dominant positive feedback.

Certainly much larger annual swings:


You could have negative one and two year auto-correlation but a dominant positive feedback on a timescale of 3+ years. The more obvious one is noise happening to overwhelm the dominant positive feedback signal over short periods (less than 3 years).

[Peter Ellis]

Yes, I should have edited out the bit about the swings - it's blindingly obvious that the annual cycle is getting wider, as summer ice is being lost faster than winter ice. 

My point was rather that there's no good evidence for "recovery" of the summer minimum at any point.  Simply returning to the (accelerating) downward trend does not constitute recovery. One- and two-year negative autocorrelation between successive summer minima is possible, but you certainly can't establish it and/or reliably distinguish it from regression to the mean, using just the six years' data since 2007!

[ChrisReynolds]

Ice Cool Kim,

A major factor of differences in albedo between FYI and MYI is that being ridged and 'crumpled' MYI doesn't hold melt ponds very well. Whereas FYI being generally flatter holds melt ponds. So most of the difference in albedo occurs in the melt season, with freeze season emissivities being very similar, which is why the IR images don't show the MYI/FYI differences.

[Ice Cool Kim]

Chris: "So most of the difference in albedo occurs in the melt season, with freeze season emissivities being very similar, which is why the IR images don't show the MYI/FYI differences."

My point was about the increased emissivity of increased area of open water vs ice rather than MYI/FYI differences, though melt pools are probably a useful refinement.


Peter Ellis:  The "recovery" after 2007 was no more than a reversion to the long-term mean rate of decline. .... My point was rather that there's no good evidence for "recovery" of the summer minimum at any point.  Simply returning to the (accelerating) downward trend does not constitute recovery.

Peter, I really don't think it meaningful to fit a linear model to variations in ice. Climate science abounds with this kind of inappropriate and simplistic analysis. It is really not helpful to developing understanding of what is driving it.

There is also a rather blatant contradiction in suggesting a linear rate of change and then talking of returning to an accelerating downward trend.

I would say that up to 2007 there was clearly an accelerating melting. That acceleration stopped in 2007 with the short lived recovery.

"... it's blindingly obvious that the annual cycle is getting wider, as summer ice is being lost faster than winter ice.  "

Indeed it is blindingly obvious , so we should take some effort to see how this affects the system.
How the system responds to this increased amplitude gives us a chance to increase our understanding of the mechanisms and feedback that are operating. That's what I was attempting to look at.

Crandles: "You could have negative one and two year auto-correlation but a dominant positive feedback on a timescale of 3+ years. The more obvious one is noise happening to overwhelm the dominant positive feedback signal over short periods (less than 3 years)."

All this needs a more in depth analysis. It is difficult to see how the accelerating melting that was happening before 2007 could have stopped since then (that's 5 years ago now) in the presence of a dominant positive feedback .

The characteristic of a +ve f/b is a run away change. This is not compatible with an accelerations that stops or a 'linear' rate of melting.

That kind of hypothesis was a possible explanation up to 2007 but really does not fit what has happened since.

Thanks for all the good comments.

[Ice Cool Kim]

Crandles: Certainly much larger annual swings:

What is very noticeable in the SIE graph you posted above is that the period with the steepest rate of melting , circa 1997-2007, stands out from the rest of the record before and after as being a time of much lesser annual variation.

Don't know if there is any linkage but it's a notable feature.

[crandles]

All this needs a more in depth analysis. It is difficult to see how the accelerating melting that was happening before 2007 could have stopped since then (that's 5 years ago now) in the presence of a dominant positive feedback .

The characteristic of a +ve f/b is a run away change. This is not compatible with an accelerations that stops or a 'linear' rate of melting.

That kind of hypothesis was a possible explanation up to 2007 but really does not fit what has happened since.

There are a few papers that have noted the one year negative auto-correlation.

Peter, Yes I agree things may have changed around 2007 (certainly a change in seasonal cycle including amplitude) and 6 years since is not enough. Still if there was such 1 year lag auto-correlation before the change, is there any reason or evidence that the change has caused that to cease?

I really don't think it meaningful to fit a linear model to variations in ice.

It doesn't matter whether you fit a linear model or an accelerating change. 2008 and 2009 could be noise or 1 year negative auto-correlation plus noise.

While I see evidence there is a change in behaviour around 2007, I don't see evidence there was a both a change in behaviour around 2007 and a further change in behaviour after that. I don't think there is enough evidence to support that especially when 1 year lag negative auto-correlation was statistically significant.

[Ice Cool Kim]

There are a few papers that have noted the one year negative auto-correlation.

That's interesting, yesterday I was looking at power spectrum and found a significant 2.01 year signal.  Not sure how that could be caused but maybe there's a link.

Maybe a neg. f/b with an annual delay would cause that kind of oscillation.  There are also clear sidebands that suggest the 2 yr is modulated by a 12.7 yr oscillation. That can be found in N. Atl SST and it would not be unreasonable to suppose exchange of waters with N.Atl would modulate part of the Arctic ice coverage.


If the most important of these oscillatory variations can be identified and subtracted we will get a clearer idea of longer terms patterns.

[crandles]

Maybe a neg. f/b with an annual delay would cause that kind of oscillation.

For 1 year negative auto-correlation, the sort of explanation I imagine is snow cover. A particularly low year means it is late before the ice can support snow meaning low snow cover which means less insulation and causes thicker ice. Because some of the thickened ice does not melt it can support snow earlier so thick snow and thin ice which melts out....

[Ice Cool Kim]

Just found the quote from Dr Katherine Giles commenting on the CS2 data:

The smaller relative decline in winter volume highlights an interesting "negative feedback".

I take that to imply it shows a feedback that is not included in the model.

Another possible cause that I did not mention before is specular reflection. With the very low angle of incidence in much of the region surface reflection will be significant for much of part of the year when there is light at all.

All these discussions of albedo seem to take overall , straight down values. I had a quick look at CICE and could not see mention of surface reflection being accounted for. However, I do not have access to 'trunk' so that may have gone in more recently. I suspect not though.

This will not affect outgoing IR but it will make a notable change to both long and shortwave incoming on both open water and melt pools.

[Richard Rathbone]

Am I right in thinking that you need to account for ice transport too?

Increased velocity of ice is compensating for decreasing thickness, and the negative feedback in Fram Strait export won't actually kick in until there is an ice free season with no ice to export?

[crandles]

Just found the quote from Dr Katherine Giles commenting on the CS2 data:

The smaller relative decline in winter volume highlights an interesting "negative feedback".

I take that to imply it shows a feedback that is not included in the model.

I don't believe it isn't included. I think a literature review would quickly find that the dominant feedback processes are albedo insulation and the winter insulation effect. So it is well known and important. If that wasn't included what would be included?

'Interesting' doesn't seem to imply omitted to me.

[Ice Cool Kim]

CPOM use CICE , I'm trying to get access to have a closer look.

'Interesting' doesn't seem to imply omitted to me.

Yes, 'Interesting' could mean a lot of things, so I'm guessing what she meant.

It could be linked to 1 year anti-correl we've already discussed or something longer. If it was something that was already recognised and incorporated into the model, I'm not sure she would be calling it interesting.

Also she was commenting on  a feature that was shown in CS2 but is not apparent in PIOMAS for the same period.

the public version of CICE is hopelessly out of date now. When I've done my homework I may contact her to ask what she meant.

Unfortunately this group seems to publish everything in JGR.


fuller quote:

The smaller relative decline in winter volume highlights an interesting "negative feedback".
Seymour Laxon Seymour Laxon pioneered the techniques that made Cryosat observations possible

"Thin ice grows more quickly than thick ice in the winter. Ice acts as an insulator - the thinner the ice, the more heat can be lost to the atmosphere and the faster the water beneath the ice can freeze," Dr Giles told BBC News.

So, no surprises.

[kevin_s]

The characteristic of a +ve f/b is a run away change.

A feedback only causes run-away change if the gain factor is >1. There are several feedback mechanisms in climate/weather--if they all ran away to infinity we wouldn't be around to talk about it, right?

[Ice Cool Kim]

A feedback only causes run-away change if the gain factor is >1.

That's pretty much the definition of +ve f/b isn't it?
gain<1 is implies negative f/b.

There are several feedback mechanisms in climate/weather--if they all ran away to infinity we wouldn't be around to talk about it, right?

Indeed. Something a lot of people seem to forgot (or rather not think about in the first place).

Climate has suffered far more extreme perturbations in the past than the small additional forcing attributable to GHG etc., including several thousand ppm of CO2.

Whatever mix of mechanisms exist with both +ve and -ve f/b , if the -ve ones did not dominate, as you say, we would not be here.

That is why I was dubious about crandle's suggestion of a dominant +ve feedback above.

The bistable behaviour of glaciation demonstrates it well . A positive f/b causes it to swing rapidly form one state to the other.  A dominant -ve f/b ensures it stays within certain limits.
 

[crandles]

A feedback only causes run-away change if the gain factor is >1.

That's pretty much the definition of +ve f/b isn't it?
gain<1 is implies negative f/b.

No, you can have a positive feedback that doesn't run-away, at least in std climate language.

So gain <0 implies negative feedback.

Gain between 0 and 1 implies positive feedback that stabilises at a higher level. Eg push temp up 1C and with gain of 0.67, the response is a further 0.67C rise and the response to that is 4/9C and response to that is 6/27C leading to an eventual temp rise of 3C. 3C is greater than the forcing of 1C so the feedback is positive but it doesn't run away.

With climate heat emitted varies with 4th power of temperature and this will usually dominate in the end. So ultimately there is a dominant negative feedback. This still allows Venus and Pluto like temperatures so within the ranges we normally consider it is quite weak and only when pushed a long way does this negative feedback begin to gain strength and ultimately dominate.

I intended to say a +ve albedo feedback and a -'ve ice insulation feedback (but muffed it up brain getting ahead of fingers). The +'ve albedo dominates in summer. -ve insulation dominates in winter. If the +'ve albedo feedback wins out over the range we are currently thinking about then sea ice disappears and the albedo feedback considerably weakens itself through running out of further areas that are easy to convert to low albedo.

[gfwellman]

The +'ve albedo dominates in summer. -ve insulation dominates in winter. If the +'ve albedo feedback wins out over the range we are currently thinking about then sea ice disappears and the albedo feedback considerably weakens itself through running out of further areas that are easy to convert to low albedo.

Yep.  This is why a seasonally ice free arctic is very likely a stable phenomenon and would have a wider "swing" between minimum (i.e. zero) and maximum extent than present, at least until the forcings are cranked way up.  (And indeed why the present has a wider swing than the baseline.)

[kevin_s]

Climate has suffered far more extreme perturbations in the past than the small additional forcing attributable to GHG etc., including several thousand ppm of CO2.

Whatever mix of mechanisms exist with both +ve and -ve f/b , if the -ve ones did not dominate, as you say, we would not be here.

No, if feedbacks worked as you describe, negative feedbacks would be just as destabilizing as positive ones.

It's two sides of a coin; given an initial forcing in one direction (e.g. warming) a feedback may be positive and amplify the response (e.g. warmer air holds more water vapor which enhances greenhouse effect and further warms the air). If an opposite forcing happens (so, cooling in this example) the same feedback mechanism amplifies that response (cooler air, less water vapor, further cooling). Run away in either direction and you've got an uninhabitable rock.

A climate feedback mechanism is therefore only positive or negative by virtue of the context. It has no intrinsic directionality. Therefore talking about a long-term dominance of negative feedbacks is meaningless (or worse).