What is the cause of the PIOMAS volume loss?

[ChrisReynolds]

Zhang and Rothrock have done a paper in 2005 covering the period 1948 to 1999, pdf here. In a nutshell; during that early period they find that the volume loss is due to loss of thinner mechanically undeformed ice, in other words - thinning of younger ice.

Such a thinning has continued into the PIOMAS record for this century, but most of the volume loss this century has come from the central Arctic and loss of thick, mechanically deformed multi year ice. I can provide graphics if needed, but for now I want to crack on.

So what's been happening to cause volume loss in PIOMAS last century is different from after 1995, when the large rapid drop in volume happened, in short Zhang and Rothrock 2005 is of no use because the mechanism has changed. It's possible that some of the other gridded data from PIOMAS might be able to help, but I'm trying to see if there's a quick and easy way to the answer before I a) ask Dr Zhang, b) try a more advanced approach using the other gridded PIOMAS data.

I can break the year into two discrete periods with their own processes going on, again I can go into why this is valid, but want to crack on with this train of thought. So I have the melt season and the freeze season. Being able to tie the loss to either would be advantageous as I could bring other research to bear in the search for the cause of the volume loss. (I have been playing around with this since before Christmas!)

Using PIOMAS monthly averages calculated from their main season I use April to September as the melt season, September to April as the freeze season, with both seasons stated for the year in which April falls. This gives me a series of numbers for volume gain and volume loss.

Year   Freeze   Melt
1980   15.334   15.925
1981   14.431   17.937
1982   16.165   15.468
1983   16.887   15.195
1984   15.137   15.705
1985   16.244   16.294
1986   16.359   14.863
1987   15.721   16.440
1988   15.842   16.214
1989   15.128   15.347
1990   15.139   16.090
1991   16.930   17.153
1992   16.056   14.565
1993   15.348   17.985
1994   17.292   15.877
1995   14.578   17.208
1996   16.220   13.495
1997   15.416   16.146
1998   16.188   17.795
1999   16.834   17.414
2000   16.117   16.074
2001   16.554   15.367
2002   15.159   16.587
2003   16.401   16.963
2004   15.473   15.716
2005   16.012   16.772
2006   15.830   16.002
2007   14.656   17.236
2008   18.468   17.750
2009   17.717   17.986
2010   16.267   18.658
2011   17.188   17.564
2012   17.520   18.358
2013   18.309   16.637

Taking these as zig zagging through the years the first month used is September 1979, the last September 2013, over that period there's been a volume loss of 11.866k km^3. If I sum the above columns I get: 548.920 and 560.786, subtract those numbers and the result is -11.866, no surprise there, the total loss is as a result of an imbalance between volume gains over autumn/winter and losses over the spring/summer.

The problem is I don't know whether freeze season volume gains are less than they 'should' be, melt season losses greater than they 'should' be, or a combination of those two factors.

I make up two synthetic series of volume loss, one using a melt season that is losing 0.5k km^3 per year more than the freeze season gains, the other with a freeze season that produces 0.5k km^3 less ice than is lost in the melt season. This illustrates the two exclusive possibilities, loss of volume due to freeze season processes, and loss of volume due to melt season processes.

Say I fix the nominal freeze and melt season to be 15k km^3, so without an offset the peak volume stays at the initial value, which I could set to 30k km^3. When I apply the 0.5k km^3 offset to either the melt or freeze season I get the same result, the melt season losses are larger than the freeze season gains, either because I've set the melt season to be 0.5k km^3 larger, or the freeze season to be 0.5k km^3 smaller. The point is that the observation that total melt season losses exceed total freeze season gains, by the amount of volume lost, does not tell us whether the losses have been from the melt season, freeze season or indeed both.

Anyone got any ideas as to how I might seperate out the relative roles of melt and freeze seasons? Or indeed is it likely to be impossible?

In the Zhang Rothrock paper I linked to above they say that losses may be from either melt or freeze season processes, suggesting they've not been able to determine which - or perhaps just didn't have the time to do the extra digging into the far more detailed data they had...

[ChrisReynolds]

Perhaps this should be more of a general thread on the causes of the volume decline and how we might tie it down, consider the above my first amateur fumblings...

Has anyone tried asking Google? I get pages about the volume loss, but the explanations are of this sort: "The $64,000 question is what's causing this decline? It's certainly consistent with what we expect from global warming, but we also need to understand better the natural variation that occurs in the system on perhaps decadal timescales." from a BBC article. It's bugged me for ages that at least with PIOMAS it should be possible to identify the cause(s).

Anyway, at risk of 'doing an ice cool kim' here but here's something really quite amazing. Has anyone else pondered how staggering it is that with the massive volume declines (shown by Max and Min volume) the annual range (Freeze and Melt) have barely changed? Available insolation energy must play a role in the summer melt, but not the case in winter, and there's been proportionately much more open water in recent years, but only a small increase in annual range.

[crandles]

The problem is I don't know whether freeze season volume gains are less than they 'should' be, melt season losses greater than they 'should' be, or a combination of those two factors.

Umm, why should they be different from what they are? Surely they are what they are.

Isn't it a matter of whether, how and why they are changing?

Up to 1990 freeze averages 15.7 and melt 16.0. By changing the years you use there is some uncertainty in these amounts but only by about 0.3

The last 10 years average 16.7 and 17.3 which is well outside the ranges indicated above.

It appears the melt has increased by slightly more than the freeze has increased.

[crandles]

If there was a fixed thermal equilibrium thickness that was always reached in winter giving fixed max volume, then there would be a 1:1 correlation between min volume and freeze volume.

Clearly that is not happening. Also clear is that the equilibrium thicknesses are decreasing over time.

So the freeze volume is increasing because of the lower min thermal range volume (i.e. volume ignoring any ice over 2m thick)

But it is also reducing because of the falls in equilibrium thickness.

And it just so happens that these changes are close to cancelling each other out (until recently when the min volume is getting really low)?

Regarding melt
As you melt ice over 2m thick at maximum, you don't increase the open water formation efficiency. Once most of that is gone and maximum thickness is reducing under 2m then OWFE increases causing albedo feedback and the energy available to melt ice increases causing an upturn recently. Albedo feedback is the major player in summer so I think we can be reasonably sure this represents most of the recent change in melt volume.

[anonymous]

Chris, let me add some more confusion, just for the sake of looking at all things of the multi-causal network you're using to catch insights. Neven blogs from time to time his so called 'crude' plot of average sea ice thickness using piomas volume divided by uuic area.

Above is the whole story. Beside the obvious: 1) some strange thing happens every 15 years and 2) max thickness is no longer in September, but in June since 2010. I find it hard to reason about, without re-engineering piomas or checking its input in detail, however, that option exists. Simply speaking I know of 2 main drivers of seasonal ice loss, export and melting. The latter in summer, the former mostly in winter. It is said, a lof of thick MYI has been lost in latest decades. I can barely see this effect above. There is thinning, but what happened then 2010?

I'm not discussing the errors of the plot above, the methods/results being compared haven't changed during the relevant time span. However, area is closer to reality and since piomas is a model, which is en route to calc the same winter max for the 4th time, I can't exclude probs here, too.

Zhang et al wrote a paper comparing piomas before and after 2006 in depth. One result is export volume must decline since there is simply less ice to export under same area. Can you see that effect somewhere here or elsewhere? Has winter export loss decline actually stopped? (last 4 winter max?)

Look the shift of thickness max by 3 month above. Has it to do with volume now melting comparably faster than area in June to September as before? And what does that mean? A major summer flux change installed itself 2010? But then why does 2013 looks same? It was effing cold.

Honestly, it doesn't makes sense to me any longer. Actually I'm close to reset all I know and start over from the beginning with the buoys. So, if you are going to spend some light, I'm all ears.


 

[Neven]

Anyone got any ideas as to how I might seperate out the relative roles of melt and freeze seasons? Or indeed is it likely to be impossible?

Here's my 2 cents, or 0.002 cents actually. I wouldn't know how to separate out the relative roles of melt and freeze seasons, but if we assume that AGW is mainly behind volume loss, one would think that most of the decline happens during the freezing season, ie less freezing. Simply because the AGW signal wrt temperatures is most pronounced during winter. Then again, like crandles says: albedo feedback.

Another factor could be the Arctic Oscillation. When looking at the AO during winter we see it was highly positive during the 90's:
 


This more or less caused this:



Or, in words, this from Wunderground:

When one looks at the wintertime pattern of the Arctic Oscillation (AO) over the past 100 years, a mostly random pattern of positive and negative AO modes is apparent (see figure below). However, one anomalous period is very striking: a string of seven consecutive years with a positive AO, including two years (1989 and 1990) with the highest AO index ever observed. During this period, strong westerly winds rapidly flushed more than 80% of the oldest, thickest sea ice out of the Arctic Ocean, leaving most of the Arctic covered with ice less than three years old. Younger ice is much thinner, and melts much more readily. Rigor and Wallace (2004) estimate that at least half of the loss of sea ice in the Arctic since 1979 is due to these six years of strange weather with very low surface pressure over the Arctic.

It's a bit simplistic, I know, but what you see, is what you get. 

[jdallen]

Echoing Neven, checking the correlation of loss to the AO may reveal something.  Driving to the heart of it, I think the answer may lie in understanding the fundamentally different routes of heat exchange that Drive melt vs refreeze.

Melt is driven by the direct kinetic transfer of captured heat from seawater to the ice (export is mostly a sideshow, and in winter, is rapidly compensated for by the freezing of leads).

Refreeze is driven by the transfer and re-radiation of heat from seawater to or through the atmosphere, either directly or through ice. These mechanics are modified via very different feed backs.

I'd note also  that in your starting numbers, there appears to be a statistically significant difference between pre and post 2007 volatility.  This implies to me that total enthalpy... Sensible plus potential heat... Has increased past some point of symmetry in the total system.  It seems to me, that this increase and the symmetry point may be measured by examining the total increase in Arctic Ocean water temperature, and the loss of "buffer" inherent in the arctic ice itself.  Hopefully, I will have more time to elaborate on this later, but this is my initial musing.

[anonymous]

Mmmh, can't think of the AO as a cause. It's an index like Dow Jones or NASDAQ. If your Apple portfolio goes south, there is no reason to check NASDAQ, except you imply some psychological or social impact, which can be safely excluded in the Arctic.

[jdallen]

Mmmh, can't think of the AO as a cause. It's an index like Dow Jones or NASDAQ. If your Apple portfolio goes south, there is no reason to check NASDAQ, except you imply some psychological or social impact, which can be safely excluded in the Arctic.

True, but I'm thinking that the correlation my reveal a relationship with factors determining the AO, not that the AO itself is a cause.

Rethinking my previous statement slightly, it might be better to say the symmetry point of the system has changed, rather than some aspect of the system passed some sort of arbitrary boundary.

The key question then becomes not one of ice, but rather then of heat.  Has less heat been lost in refreeze, or more gained in melt?  Secondarily, does the current state of the system reflect a net change in total heat content? If so, what are the expected changes in its mechanical behavior?

[anonymous]

True, but I'm thinking that the correlation my reveal a relationship....

There is a little trap near the word correlation...

[Neven]

Mmmh, can't think of the AO as a cause.

I didn't mean to posit is a cause, but only as evidence that perhaps the volume loss has mostly been taking place during freeze seasons. Now, how to tease that out of the PIOMAS data (the thread subject) is beyond me.

[anonymous]

I agree it would be extremely helpful to know, when and how much ice melted or drifted into the Atlantic, yet I do not understand how the AO or any other index, could provide a single bit of useful information. Or the other way round, how does one get from an dimensionless metric to m³/day.

[jdallen]

True, but I'm thinking that the correlation my reveal a relationship....

There is a little trap near the word correlation...

I hate autocorrect

[ChrisReynolds]

The AO plays a role but doesn't explain all.

Got to dash off for work, but I've woken up and realised I may have much more of the answer, but forgot I had the paper, and/or didn't think it addressed the issue. But over night my subconscious has been working*.
http://psc.apl.washington.edu/zhang/Pubs/Lindsay_Zhang_tipping_point.pdf

The
largest source of variability is in the summer melt,
which shows a consistent trend of increasing melt
over the 56-yr study period and a marked increase in
the melt trend in the last 16 yr (Fig. 7). Winter freezing
rates follow the summer melt rates: when there is
increased summer melt, there is increased winter ice
production. Net advection averaged over the basin
has not changed much over the study period but
there was a spike in the volume export at Fram Strait
just after the 1987 maximum and again in 1995 (Fig.
14). Since then, the volume export has diminished
because the mean thickness crossing Fram Strait is
less, not because the area transport is less.

So the melt has become more aggressive, especially in from 1990s to 2005 (when the acceleration starts), freeze occurs in response to melt. And export doesn't account for the post 1995 acceleration.

I'll reply properly tomorrow later today when I get back from work.

*Why do I have to ask questions publicly before my subconscious gets properly to work?

[jdallen]

... But over night my subconscious has been working*.
http://psc.apl.washington.edu/zhang/Pubs/Lindsay_Zhang_tipping_point.pdf

Excellent paper, and I love it when that happens (subconscious going to work for you while you sleep).  I find it very rewarding when I wake up and it presents me with a programming solution to my current analysis problem.  Joyfully, that happens with some frequency.

*Why do I have to ask questions publicly before my subconscious gets properly to work?

<Snerk> Welcome to my world.

[Sourabh]

Hey Chris,

I am not qualified and/or experienced enough to add much, but I found these videos. I have shared first video before as well. It shows changes in ocean current in Arctic.





It might help you if you haven't seen this video before.

[crandles]

Nice paper, useful for this

These terms also show significant interannual variability. The annual total thermodynamic growth and net export, averaged over the Arctic Ocean (Figs. 7a and 7b), show the average winter growth is 1.30 m yr^-1 and the average summer melt is 0.91 m yr^-1. The net advection is nearly zero in the summer and negative in the winter, averaging 0.41 m yr^-1. This term represents the net export of ice from the basin. The average thinning rate due to both processes over the entire 56 yr period is 0.02 m yr^-1.

...

Since 1988 the largest net change in the surface fluxes compared to the mean values, amounting to the equivalent of about 3 m of ice loss, is that due to the net solar flux (Fig. . This loss is partially compensated by a change in net longwave flux equivalent to about 2 m of ice gain, a large amount because of the heat lost from the warmer open water or thin ice surfaces. The loss of ice during this period due to ocean heat flux is about 2-m ice equivalent. Some of this ocean heat is from the solar heat absorbed by open water. Notably the change in ice thickness due to changes in the turbulent sensible and latent heat terms is relatively small. These fluxes are largely determined by the air temperature (relative to the ocean temperature), so the recent changes in the mean ice thickness are not primarily due to recent changes in the surface air temperature.

The increased net solar flux in the simulations can only arise from changes in the model albedo because the cloud fraction in the model, and hence the estimate of the downwelling solar flux, has no interannual variability. These simulations isolate the ice–albedo feedback from possible real changes in the downwelling solar flux.

Net advection 0.41 versus melt of 0.91 seems like a high proportion.

Fig 8 shows changes in components from 1948 to 2003. So we could estimate from that graph what is happening recently compared to a period with little change in thickness (from graph 3 we could pick something like 1950 to 1963 or I am using a more recent period 1968 to 1985).

Estimates from graph:
Component____1968-1985Est____2003
Net solar_____ -3.2m/yr_________-3.7m/yr
Ocean________-0.4m/yr_________-0.6m/yr
Net Longwave_+1.95m/yr_______+2.25m/yr
Sensible______+1.2m/yr________+1.25m/yr
Latent________+0.4m/yr________+0.45m/yr

Splitting these figures into melt season and freeze season and bringing up to date would be nice.

Don't know if this represents any progress Chris.

[ChrisReynolds]

Crandles,

Yes very useful, another model study shows the 'Tietsche effect' of large amounts of heat being vented that stops the ice/ocean system crashing rapidly to zero ice.

Ocean has gone up, but the ocean is revealed to be a relatively small player. That net solar is so massive is indicated by the way the seasonal cycle of volume loss (time derivative of volume) follows the insolation cycle so strongly.

I'll have a look at the other figures when I give the paper a re-read, it does however make the earlier paper I linked to rather superflouous.

With regards your initial reply, you noted that avg melt increased more than freeze season gains in recent years, but as I outlined with the discussion of the synthetic data approach that doesn't tell us whether it is actually freeze season gains that are failing to catch up on the melt season losses. Of course all that is largely irrelevant now we've got this paper.

Neven, Arcticio, JDAllen,

Download the Lindsay/Zhang paper, and refer to figure 13 and the text below. Those show correlations between modelled ice thickness and the AO and PDO, and as you'll see from the quote below, the authors of this paper (in common with many other scientists see the 1990s +ve AO event as pivotal to what's followed. It's no coincidence that the increasingly steep volume loss trend has started from the mid 1990s.

Sourabh,
thanks for that, if you have questions just ask.


The following is the summary of what the authors (Zhang & Lindsay) think has driven the volume (thickness) decline since the 1950s.

In summary, the thinning is due to

1) fall, winter, and spring air temperatures over the Arctic Ocean gradually increasing over the last 56 yr, leading to reduced thickness of first-year ice at the start of summer (the preconditioning);

2) a temporary shift, starting in 1989, of two principal climate indices causing a flushing of some of the older, thicker, ice out of the basin and an increase in the summer open water extent (the trigger); and

3) the recent increasing amounts of summer open water allows increased absorption of solar radiation, which melts the ice, warms the water, and promotes creation of thinner first-year ice, ice that then often entirely melts by the end of the subsequent summer (the feedback).

FWIW my translation of the above, and a simple summary of what has caused the PIOMAS volume loss.

The increasing temperatures have meant that ice grows less thick in winter. So what was happening in the earlier period (1948 to 1999 - Rothrock & Zhang) was that the thicker deformed ice was increasing while the thickness of undeformed (generally first year ice) was going down.

That's different from what I've shown usng gridded PIOMAS data, that in recent years (since about 1995) almost all the volume loss has been from multi-year ice over 2m thick in the central Arctic.

So what's changed? In the early 1990s the AO (Arctic Oscillation) and PDO (Pacific Decadal Oscillation) shifted such that there was greater export of ice from the Arctic Ocean. This lead to the start of a period of increasing summer open water (see fig 11!!!).

That's when the snowball started to roll downhill...

The PIOMAS volume loss has been highly non linear, with losses increasing as time moves on, since about 1995. When you see such behaviour the thing to start looking for is a strong positive feedback. The feedback here is the ice albedo feedback.

This paper only takes us to 2005, the authors couldn't have foreseen 2007 and 2012.


There's a lot in this paper, anyone up for disecting it?

To start with, figs 7 c & d, cumulative anomaly - I can't figure out how they've done these so don't understand them, can anyone clarify?

[anonymous]

... Those show correlations between modelled ice thickness and the AO and PDO, and as you'll see from the quote below, the authors of this paper (in common with many other scientists see the 1990s +ve AO event as pivotal to what's followed. It's no coincidence that the increasingly steep volume loss trend has started from the mid 1990s.

The AO, NAO, PDO and all the other indexes are arbitrary chosen by humans and in some regard helpful. But the above reads to me like the index of sold T-Shirts strongly dictates summer temperatures and is supported by major economic leaders to estimate soft drink consumption. I'd really like to know when these indexes became an acting feature of the earth?

Beside the fun part, Chris, many thanks digging out that paper. Did you read the recent Meyer paper pointing at major flaws in sea ice area/extent math? How does PIOMAS retrieves daily area? Do you think, after correction the above explanations are still valid?

[crandles]

FWIW my translation of the above, and a simple summary of what has caused the PIOMAS volume loss.

The increasing temperatures have meant that ice grows less thick in winter. So what was happening in the earlier period (1948 to 1999 - Rothrock & Zhang) was that the thicker deformed ice was increasing while the thickness of undeformed (generally first year ice) was going down.

That's different from what I've shown usng gridded PIOMAS data, that in recent years (since about 1995) almost all the volume loss has been from multi-year ice over 2m thick in the central Arctic.

So what's changed? In the early 1990s the AO (Arctic Oscillation) and PDO (Pacific Decadal Oscillation) shifted such that there was greater export of ice from the Arctic Ocean. This lead to the start of a period of increasing summer open water (see fig 11!!!).

That's when the snowball started to roll downhill...

The PIOMAS volume loss has been highly non linear, with losses increasing as time moves on, since about 1995. When you see such behaviour the thing to start looking for is a strong positive feedback. The feedback here is the ice albedo feedback.

This paper only takes us to 2005, the authors couldn't have foreseen 2007 and 2012.


There's a lot in this paper, anyone up for disecting it?

To start with, figs 7 c & d, cumulative anomaly - I can't figure out how they've done these so don't understand them, can anyone clarify?

That is a little different from my translation of what you quoted from the paper, though I need to read the paper more thoroughly.

You said

The increasing temperatures have meant that ice grows less thick in winter.

Apologies if I am being rather nit-picky about this but

'meant' could be seen to imply causation. While the paper might seem to say this with

The winter air temperature over the Arctic Ocean
has gradually warmed over the 56-yr period leading
to a reduced equilibrium ice thickness (Fig. 9).

but it also says

These fluxes [turbulent sensible and latent heat terms] are largely determined by the air temperature (relative to the ocean temperature), so the recent changes in the mean ice thickness are not primarily due to recent changes in the surface air temperature.

So I think we should interpret this as:
 
Reduced thickness causes higher temperatures. (Certainly/especially in fall and early winter.)
Higher temperatures likely has 'some effect' causing reducing thermal equilibrium thickness. (Especially / particularly if the higher temperatures last all the way through the winter. 'Some effect' is not necessarily a large portion of the causation of thinner un-deformed FYI.)

I am suggesting this interpretation because the relative importance of factors causing reduced thermal equilibrium thickness (such as GHG levels, upward heat flux, air temperatures, trends in cloud cover ...) are not discussed.

There is another slight problem with your sentence quoted above: I think it should refer only to first year ice if you are trying to follow the logic of what the paper is saying.

Thicker deformed ice increasing or decreasing?

Fig 3 shows ridged ice increasing to a maximum in 1988. So that is consistent with the paper indicating that in the earlier period ridged ice was increasing (but only up to 1988 rather than the 1999 that you suggest). You have shown loss of deformed ice in recent years and that is entirely consistent with the paper.

figs 7 c & d, cumulative anomaly - figure out how they've done these

The graphs starting and ending at 0 on y axis does look a bit odd but it doesn't seem difficult.
Anomaly = Actual - average over full period

Cumulative anomaly means, err cumulative. Ie for first year data point is the first year anomaly.
For second year add anomaly for year 1 and year 2. For 3rd year sum anomalies for years 1 to 3.... Adding all the differences from average should of course be zero so the data value for the last year should be 0.

Interpreting what those graphs mean is also a bit odd. If there is acceleration in rate of ice loss the shape of the line will be a hump above the x axis and if decelerating rate of ice loss it is a trough below the x axis.

To me, it seems more sensible to set a base period like 1968-1985 where ice volume didn't change or vary much and then plot anomalies from average of that base period and cumulative anomalies then wouldn't end at 0.

[Sourabh]

Dear Chris,

I wanted to ask if salinity influences Arctic sea-ice formation. Based on the video I posted, lots of fresh water is entering Arctic ocean. The following article also suggests that fresh water helps in protecting ice.

http://www.nasa.gov/topics/earth/features/earth20120104.html#.UtaIHvZSbas

http://www.bbc.co.uk/news/science-environment-16657122

Furthermore, as happening in Antarctica, sea-ice extent is increasing due to melting of glaciers that supply lots of fresh water. Can similar phenomenon occur in Arctic?

Can Russian floods and river runoffs have any correlation with  sea-ice losses? Lack of  fresh water would discourage ice formation and/or increase melting. So, may be in 2012, 2007, there wasn't enough fresh water, but remaining years had huge fresh water running into Arctic Basic due to flood in Canada or Russia. Therefore, other years including 2013 were not record melt years.

Does this proposition have any validity?

~Sourabh

[crandles]

Sourabh,

It certainly has an effect as you indicate - it is easier to freeze fresh water than it is to freeze salty water - the salt has to be expelled which takes energy to overcome the increased entropy in separating out the salt into a brinier part that doesn't freeze and fresh water that does freeze.

There is also the effect of sensible heat coming with the river discharges if we are talking about Russian rivers (this effect is obviously much less for Greenland).

If this was a dominant effect, then it would occur mainly in the freeze season. Increased absolute humidity results in higher rainfall and increased river run off in addition to melting permafrost. So there would be increased fresh water after global warming gets going. This would result in an increased volume of fresh water and more freezing during the winter.

What we are seeing in reality is increased melt in summer partially compensated by increased freeze in winter.

The effect is in the opposite direction to what is happening and in the wrong season, so we can rule out this having a dominant role. It certainly does have a role but it is overwhelmed by other changes. I don't see any reason to think this feature isn't well dealt with by models - it certainly isn't a newly discovered feature of the system and models certainly track salinity levels.

Another thing to mention is that the Arctic is stratified by salinity not temperature. So adding a whole lot of fresh water might only have a fairly negligible effect on the depth of the fresh water layer and negligible effect on the salinity of water in contact with the ice.

[Sourabh]

Crandles,

Thanks for the explanation. It was really helpful.

I am sure many models incorporate the effects of the fresh water. However, one of the videos I posted shows significant increase in fresh water content of Arctic Ocean.

My question is that can inflow of fresh water prevent other factors from becoming dominant although fresh water itself is not dominant? I can be completely wrong, but consider this theory:

Before large influx of fresh water, sources that provided heat to melt ice included solar radiations wind, and water mixing (deeper layer, Atlantic, Pacific etc.). But lots of fresh water created an independent or isolated pool of fresh water that melts or freezes without interacting with broader Arctic climate/ocean system. Imagine a very large 'saucer' near Beaufort floating in Arctic Ocean.

In other words, now we have two sub-systems in Arctic that are not interacting much. One sub-system composed of this pool, which is receiving heat only through sun and wind, but not much or at all from water mixing. Another sub-system, may be composed of bottom layer, which is receiving heat from mixing, but not enough from sun and wind. So, fresh water separated Arctic system in two smaller systems over period of years.

Before I think any further, I would like to see you thoughts over this crazy theory.

Btw,

I found this presentation. I  could not entirely understand the findings. So, if you don't mind going through it, I would appreciate that.

http://www.whoi.edu/fileserver.do?id=169345&pt=2&p=180769

[ChrisReynolds]

Crandles,

My reference to temperature reducing thickness was for the early period, up to 1989, although I did give the impression of 1999 by the reference I cited. You are correct that it is in the later period that temperature is not the prime mover of volume loss. However section 5 (Analysis of the recent thinning) subsection 5a is titled "The preconditioning: Warming winter air temperature" and the text of both this paper and Rothrock & Zhang 2005 supports the impression that the authors are of the view that increased air temperatures drove the decline in undeformed ice from 1948 to 1989.

I should point out that I did not 'find' the decline of thick ice in a scientific sense, my wording was poor, this has been a topic discussed in the scientific literature. It was something I found (in an amateur sense) when delving into the PIOMAS gridded data. But you are correct in pointing out that the recent decline of deformed ice is not at odds with the paper.



With regards other factors.

In common with Rothrock & Zhang 2005, this paper has no mention of ocean heat flux, i.e. from Rothrock & Zhang:

What explains the reduction in ice thickness?

By separating the wind
component and the temperature component of the interannual
forcing we find that even though the variance in
volume derives equally from the wind and temperature
responses, only the thermal component VT seems to have
a significant downward trend:
0.07  103 km3 yr
1 or

3% of V per decade. Although from the late 1980s
through the mid-1990s the wind was the dominant factor
in the rapid decline in ice thickness, overall the wind
appears to cause large oscillations but not a multidecadal
downward trend. The volume response to rising temperatures,
on the other hand, seems to be more steadily
downward, accounting for a reduction of over 25% in
volume over 5 decades.

This surprises me because I have always thought of ocean heat content as playing a role in the preconditioning.

There is mention of clouds in Lindsay & Zhang:

The increased net solar flux in the simulations can
only arise from changes in the model albedo because
the cloud fraction in the model, and hence the estimate
of the downwelling solar flux, has no interannual variability.

...The downwelling solar fluxes may not, in fact,
be constant...

With regards the cloud radiative feedback (infra-red) and sunlight effects, I'm not sure off the top of my head what the position is. But there has been an increase in downwelling longwave radiation (DLR).

Francis & Hunter "New Insight Into the Disappearing Arctic Sea Ice" (EOS) find that increased downwelling infra red (DLR) plays a role in the loss of ice since the mid 1970s (which is during the 'recent period'). They find that:

In addition to warming and increased
clouds, more abundant liquid-water-containing
clouds [Zuidema et al., 2005] and increased
atmospheric water vapor content [Wang and
Key, 2005] also appear to be influential.
Trends in satellite-derived observations
(Table 1) likely are caused by a combination
of increased moisture transport from lower
latitudes [Groves and Francis, 2002], as well
as by evaporation from additional ice-free
areas and from an earlier start to the melt
season [Belchansky et al., 2003]. These
changes constitute a positive feedback to
Arctic warming, augmenting the much anticipated
ice-albedo feedback.

The EOS arcticle is based on the Env Reserach Letters paper "Changes in the fabric of the Arctic’s
greenhouse blanket."
http://media.cigionline.org/geoeng/2007%20-%20Francis%20and%20Hunter%20-%20Changes%20in%20Arctic's%20Greenhouse%20blanket.pdf
That paper gives more detail about the magnitude of increase of DLR, many regions have decadal trends increasing by up to 10W/m^2 from 1980 to 2005. So increased DLR directly due to GHG increases over that period are swamped by a feedback due to increased humidity of the atmopshere.

Thanks for the explanation about cumulative anomalies.

[ChrisReynolds]

Sourabh,

Sorry for not getting back to you earlier, I can't post from work. I agree with what Crandles had to say anyway. I would add that Arctic Russia has warmed significantly, which has warmed the rivers pouring into the Arctic.

I'm not sure about your idea with regards freshwater and can't imagine how such a pool of fresh water wouldn't interact with the ice and ocean. As Crandles points out, during the melt season much of the fresh water added is due to unusually large melt of ice. But this is countered in autumn when the ice re-freezes producing dense cold brine which causes mixing in the ocean.

It might interest you to know that I was prompted to start this post by a paper linked to by Sidd:
http://forum.arctic-sea-ice.net/index.php/topic,711.msg18790.html#msg18790
The final lines from the paper are:

However, even in the Nansen
Basin, where the AtlanticWater is the warmest and the upper
halocline is generally absent, the seasonal ice melt appears
independent from the increasing Atlantic heat content. During
the 2000s, freshening of the upper ocean is decreasing
the depth of winter convection and the established stratification
restricts, at least during summer, the upward mixing of
Atlantic heat. Comparison of the hydrographic melt estimate
to the potential ice melt obtained from reanalysed heat fluxes
indicates that the seasonal ice melt rather reflects the variability
in atmospheric heat fluxes than changes in the deep
ocean heat content.

I've long been interested in the heat flux from the ocean as a player in the preconditioning of sea ice for the current losses. But very sceptical about a primary role. That paper showed that in summer the wamer deep waters (Atlantic Water) were not playing a role in recent losses, which made me determined to see if I could figure out whether the volume loss was coming from a certain season.

What the Lindsay & Zhang paper we're discussing (about loss of volume) shows is that during this century the increase in volume loss has mainly been due to process in the melt season, not autumn and winter. Due to mixing by brine rejected from ice formation during the freeze season ocean heat should have a greater impact on ice over autumn and winter.

[Sourabh]

Thanks for the description.  I should start reading more about Arctic to become familiar with more technical terms and climatology.

I was interpreting the first video. I was trying to argue that after 2000s, freshening has further stratified Arctic Ocean, in at least in some basins. So, during winter, mixing and transport of heat due to convection reduced. Therefore, deep ocean water no longer interact or mixes with upper layer, which is what I meant by two-sub systems, upper ocean layer and deeper ocean layer. Or, there could be multiple systems, one for each basin. Before influx of fresh water, deeper ocean interacted with upper layer through mixing in the previous century i.e. before 2000s.

(The link you provided has the paper you were excited about. I read that paper, though not completely. So, I may be missing many things. I will read it again to see what mistakes I am making. )

After 2000s, fresh water has thickened the upper layer. So, during winter mixing occurs but only within the upper layer, not in deep ocean unlike previous century , when mixing with deep ocean occurred. . Is that correct? In other words, fresh water influx is somewhat protecting ice melt. If it wasn't for fresh water, Arctic would be ice-free by now. Or did I get it totally opposite of what is correct?

[ChrisReynolds]

I must confess I've not had the time to fully read that paper, I've read the abstract, the conclusions, and the section that applies to the conclusion I was drawing from it. So I might have missed that bit about winter- but I thought it was mainly in summer that the deeper warmer water was strongly isolated from the ice...

I hope I'm not going over stuff you already know here.

Take ITP73:
http://www.whoi.edu/page.do?pid=125776

Its location is near the Atlantic edge of the ice pack, which is why I chose it. Select the salinity and temperature plots for that ITP (ice tethered profiler) - "Plot of ITP T & S Contours ".

The first plot is the top of the ocean, below is a deeper temperature plot. You can see the layer of warmer water below the surface. Its actually from 200m to 500m down below the surface. with a cap of colder water above it. If you scroll down to the bottom two panels you can see the reason why it's a warm body of water capped by cold, that water is far more salty than the surface, but the surface layer isn't 'fresh' water, its just less salty.

The warm layer is called the Atlantic Water layer because it comes from the Atlantic. But along the ice edge in the Atlantic the ocean bottom drops into the abyssal basins of the Arctic Ocean. You'll see what I mean on this bathymetry map:
http://geology.com/world/arctic-ocean-bathymetry-map.jpg
Atually, IIRC, there's a map of basins in the paper you've been looking at.

The ice edge in the Atlantic is at a generally higher latitude than in other sectors because warm Atlantic water melts the ice. But as it falls off the edge into the Arctic Ocean it falls out of contact with the surface and is insulated from the ice.

Now consider these maps from DMI:
http://brunnur.vedur.is/pub/trausti/Iskort/Jpg/1900/1900_allt.jpg
That's the ice edge in 1900, check out April (left column middle pane), note how although the ice edge in Barents is much further than we see these days, below Spitzbergen theres a notch in the ice, that's due to the West Spitzbergen current.
More maps here:
http://brunnur.vedur.is/pub/trausti/Iskort/Jpg/

EDIT - forgot to explain - the West Spitzbergen current is a major source of Atlantic flow into the Arctic, althugh the recession of ice edge in Barents Sea is also due to the warming of the Atlantic with GW, and with a greater flow of water into that region (I think I'm right in saying but am too rushed to go looking for that paper - sorry).

Hope this ramble makes some sense. There are others more up to date on the ocean, I barely have enough time to keep up to date with the sea ice.

[crandles]

At Tamino's
http://tamino.wordpress.com/2014/01/15/southern-discomfort

KR says

Hank Roberts – the Manabe et al 1991 paper Dumb Scientist pointed to discusses this; according to their modeling increased precipitation from warming freshens the surface waters, increasing the halocline gradient (salty below, fresh above) and thus reducing exchange. The surface waters are then less connected to the warmer deep waters, a smaller thermal mass, and more ice occurs. I’m sure increased meltwater doesn’t help either, but I don’t believe that was part of the Manabe model.

A counter-intuitive result, and certainly not one of the outputs of most climate models.

That all seemed to be more or less agreeing with what I said, until the last sentence which I find surprising but don't trust me on that. Anyway, the Manabe et al 1991 paper is a lot more authoritative than what I said.

Manabe et al 1991

[ChrisReynolds]

From Lindsay & Zhang, there are three periods, .

1948 to 1988 - Preconditioning due to warmer air temperature.
1989 to 1995 - Trigger, the AO and PDO.
1996 to present - Feedback, the authors cite the ice albedo feedback as the primary one, which is likely correct IMO.

Here are the three periods shown over the September average volume calculated from gridded PIOMAS data.



Note that Preconditioning shows a downward trend, but is only the end of a longer period stretching back to 1948, the study only goes back to that date because NCEP/NCAR only goes back to that date, in reality the preconditioning phase probably goes back further.

During the Trigger phase there are seen to be a succession of years of substantial volatility. Note that strictly speaking the paper identifies the first year with low area/extent in summer as 1990, but I'd rather keep the periods disrcete.

It is during the Feedback phase that the serious decline in volume starts.

Here is the September volume broken down into regions (after Cryosphere Today regions), within the Arctic Ocean.



During the Feedback period the Central Arctic has been responsible for 78% of the overall volume loss. Why?

One reason will be loss of multi year ice (MYI), which is concentrated in the Central Arctic. Indeed in the above plot it is apparent that the peripheral seas of the Arctic Ocean are virtually seasonal since 2007. So MYI that is present in those seas will have been exported from the Central Artic. Due to the transition of the peripheral seas to a seasonally sea ice free state, the ice edge and the region of greater melt within it will have impacted the Central Arctic more. But I want to just concentrate on one issue.

The Beaufort Gyre Flywheel system (BGF) was once a stabiliser for the Arctic, circulating and ageing ice from the Central Arctic, i.e. Woods Hole Oceanographic Institute. However that page refers to the BGF as a facet of the 'healthy' Arctic system, considering that the Beaufort Sea is now largely seasonal the BGF is now broken, ice that moves into it doesn not circulate and age, it melts out in the summer.

In "Contribution of melt in the Beaufort Sea to the decline in Arctic multiyear sea ice coverage: 1993–2009." Kwok & Cunningham, GRL 2010 (PDF), the authors examine the issue of what happens to MYI that is transported into the Beaufort Sea. They find that half of the MYI volume loss (as measured by ICESat) is due to transport into the Beaufort Sea, the ice transported there has been subject to increased melt during the summer, culminating in the post 2007 period in which Beaufort ha transitioned to a virtually sea ice free state.

[jdallen]

Excellent post as usual, Chris. I will add, that is an excellent and very scary graphic.

[ChrisReynolds]

I guess it is scary. I spend so much time with this stuff it no longers strikes me like that. Check out 2013 though - virtually all the summer volume gain is from the central Arctic!

[Sourabh]

Chris,

I see that the years that are frequently discussed here are 2007 and 2012. As I joined this forum last year, I did not know what was discussed during 2010 melting season. I find 2010 strange year. It had lower volume than 2007, but higher extent and area coverage than 2007. It is the only year with such anomaly. Was there any reason for it? Was it important?

[wili]

Sourabh, I would do some searching over at the blog on that. There was indeed much discussion of that year over there.

Chris, I concur with jdallen that this is really great work, once again. And that it's scary.

Your write: "Note that Preconditioning shows a downward trend, but is only the end of a longer period stretching back to 1948"

Do you know of any measurements or proxies that give us any idea of what total volumes were before 1979? Is there a gradual slope going up as you go back in the years? I guess I'm wondering what the long-term volume is, and if and how we know it. Presumably the various navies operating under the ice through the cold war had some idea.

[jdallen]

A critical and key takeaway/reason that graphic is so disturbing:

Even after a "recovery" in 2013, the total volume of ice in the central arctic is less than that of ice in peripheral seas in 1978.

Mull that for a bit.

[crandles]

Your write: "Note that Preconditioning shows a downward trend, but is only the end of a longer period stretching back to 1948"

Do you know of any measurements or proxies that give us any idea of what total volumes were before 1979? Is there a gradual slope going up as you go back in the years? I guess I'm wondering what the long-term volume is, and if and how we know it. Presumably the various navies operating under the ice through the cold war had some idea.

See fig 3 of the paper we are discussing. Note this is PIOMAS model not actual measurements but hopefully it has been tuned to give result at least somewhat similar to submarine sonar data.

http://psc.apl.washington.edu/zhang/Pubs/Lindsay_Zhang_tipping_point.pdf


Chris,

I think it is important to point out that the paper is saying it is the decline in thickness of the level ice (which is undeformed ice) which is the preconditioning. (The way you have labelled your graphs make it seem it is total ice volume which is the trigger, but it is mainly the first year ice.

Before 1988, level ice was declining in thickness while ridged ice was increasing at a faster rate per Fig 3.

You graph


shows continued thinning.


What do you make of the language: preconditioning, trigger and feedbacks? Does it imply an episode of rapid decline caused by the trigger and after the period of rapid decline the system might return to a slower rate of decline? 

Or do you think the language of tipping point and feedback period are saying we are past a point of no return?


If we hadn't had the trigger of AO and PDO lining up to cause thin ice starting around 1987, how much harder would it have been to get to the feedbacks stage? How much later might it have been before the feedback stage began? Or are these unreasonable questions?

[ChrisReynolds]

Chris,

I see that the years that are frequently discussed here are 2007 and 2012. As I joined this forum last year, I did not know what was discussed during 2010 melting season. I find 2010 strange year. It had lower volume than 2007, but higher extent and area coverage than 2007. It is the only year with such anomaly. Was there any reason for it? Was it important?

Sourabh,

I've blogged previously on the 2010 volume loss event, the reason area/extent were higher was a large transport of ice from the central arctic into the peripheral seas, this lead to 2010 being as large a volume loss as 2007 and the following years having a changed seasonnal cycle in PIOMAS, with a drop in thicker multi year ice proportion.

More here:
http://dosbat.blogspot.co.uk/2013/03/what-caused-volume-loss-in-2010-part-2.html

Everyone else,

Got to rush to work, will reply later.

[ChrisReynolds]

Do you know of any measurements or proxies that give us any idea of what total volumes were before 1979? Is there a gradual slope going up as you go back in the years? I guess I'm wondering what the long-term volume is, and if and how we know it. Presumably the various navies operating under the ice through the cold war had some idea.

In 1998 (IIRC) Al Gore arranged the release of submarine upward looking sonar data to scientists, previously it had been considered too secret to release. This data is available here:
http://nsidc.org/data/g02194.html
It only goes back to 1979 though. 

There are snippets of earlier data available, if it works I've attached some supplementary materials from a 2009 paper by Kwok. That gives some of an earlier period of sub cruises (1958 to 1976 and 1993 to 1997) compared with the ICESat period (2003 to 2007). Hope it's of use, it shows massive drops of thickness between the earliest and latest periods (50 to 60% typical).

[ChrisReynolds]

A critical and key takeaway/reason that graphic is so disturbing:

Even after a "recovery" in 2013, the total volume of ice in the central arctic is less than that of ice in peripheral seas in 1978.

Mull that for a bit.

Blimey, well spotted, I hadn't noticed.

Beaufort through to Greenland Sea Sept 1978: 4568.84 km^3
Central Arctic Sept 2013: 4394.8 km^3

That's only about 4% greater, with the accuracy of my regions it might be safest to say that:

Even after a "recovery" in 2013, the total volume of ice in the central arctic is roughly equal to that of ice in peripheral seas in 1978.

Although thinking about it, the errors will be in the boundaries, which will be reduced using two small areas, both with siginificant land edges (no error there as the PIOMAS land mask is used).

So at 4% it probably is safe to say bigger.

[ChrisReynolds]

Crandles,

All I was trying to do with the labels was to give some reference to the total volume graphs we're used to seeing, and for good measure the volume breakdowns I do. You're right that the implication that preconditioning only applies to undeformed ice isn't made clear.

I've not got round to considering the issue of how much later the process would be without the 'trigger' stage. Perhaps this is a major reason why the free runnning GCMs that make their own climate haven't kept up with observed area/extent losses, and losses modelled by assimilating models (PIOMAS/NAME)?

I don't see it as a classic bifurcatory tipping point, a series of really cold years at the early stage of 'feedback' might conceivably have put things back on a slow path. Although - perhaps I need to think about this a bit more.

What does intrigue me is that the preconditioning phase process - warmer air temperatures reducing the equilibrium thickness of thermodynamic growth in the winter has never really ended. Temperatures have continued to rise since 1989 and even though much of this warming in autumn/winter is due to anomalous open water and thinner ice, it doesn't mean that it hasn't impacted the thickness of thermodynamice growth.

Here is the thickness distribution for the East Siberian Sea in December, with thickness along the bottom axis, volume in the vertical, and pentads (5 year groups) for average thickness distribution.



The problem with this graph in terms of trying to gauge thermodynamic growth is we can't tell what's deformed or not. But in 78 to 82 and 83 to 87 we see at the right most side of the graph the sort of out of range spike associated with the thickest deformed ice that is seen in the 2007 distribution in the graph you posted. So that's the presence of thick deformed ice in the ESS during the early period 1978 to 1987. Maybe I ought to re-do this for April, because the peaks around 2.5m for that period can't be thermodynamically thickened ice (surely). However by 88 to 92 and 93 to 97 the peak is about 1.5m, 98 to 02 is slightly thicker peak at 1.8m. But 03 to 07, 08 to 12 and post 2010 (including 2013) all have a peak about 1m thick in December - certainly in the last two periods when the ESS has been virtually ice free by the end of the melt season all of this is thermodynamic growth.

Could the thickening by December have dropped so much due to warming and delay of ice fromation as the ocean loses heat? It would be better if we had some indicator of deformed/undeformed in the PIOMAS data.

I will re run that for April or March.

[wili]

Thanks, Chris.

[crandles]

Thanks Chris,

I doubt 3 month old ice would have grown to 2m thick even back in 1979. However back in early years 79-90 the beaufort gyre flywheel would still be bringing MYI to ESS. Since 2003 there is nothing over 3m and since 2008 nothing over 1.6m.

You could use a thickness cut off and say any ice over 2m in April is MYI. However it might need a moving threshold e.g.for 1979 a 2.5m cutoff gradually reducing to 2m in 2013. Then you could sum volume and area of cells under the threshold to plot an estimated thickness for 6 month old ice. Not sure if that is considerably more trouble for you to do or if you just end up with a graph showing the cut-offs used.

[ChrisReynolds]

I doubt it too, the other option would be to use Okhotsk, Bering, and Hudson, seas which have had negligible MYI throughout the period since 1978. I'll look at that this morning.

[ChrisReynolds]

I've attached a copy of peak thickness in March for the seas outside the Arctic Ocean. These are worked out as follows: Take the thickness distribution (see the graphs Crandles and I have posted) and calculate the peak volume for any of the thickness bands for each year. So the following graphs show the peak of the distribution curve.

Okhotsk and Bering are in the Pacific and show less of s strong trend, I suspect that this is due to the influence of the PDO, although I haven't time to analyse that further.

Both Hudson and Baffin show downwards trends. Hudson has the least variability of all the seas shown. The trends are both downwards, for Hudson there's loss of 0.0034m/yr, for Baffin 0.008m per yr in Baffin. While Lindsay & Zhang find large thinning rates in regions of thick ice for 1988 to 2005 are high the thinning over the Siberian sector is low, but not as low as the rates suggested in the following graph for Hudson and Baffin.

[ChrisReynolds]

Polyakov et al 2003, "Long-Term Ice Variability in Arctic Marginal Seas" Section 4 and following page, find that there are conflicting trends between the locations where long term samples have been obtained, and that no clear picture of thinning due to AGW is visible. Rather local variations in climate are dominant. As referenced in that paper, Wadhams has suggested that fast ice should show the impact of thermodynamic thinning for AGW.

Dmitrenko et al 2007, "The long-term and interannual variability of summer fresh water storage over the eastern Siberian shelf: Implication for climatic change", use NCEP/NCAR reanalysis to determine that the thickness of fast ice in sites used by Polyakov - 

for both the Laptev and East Siberian Seas, the correlation coefficient between fast ice thickness and winter SAT NCEP data are -0.54 and -0.60 (statistically significant at the 95% level of confidence). This indicates that interannual variation in sea ice thickness of land-fast ice is thermodynamic in nature.

The Lindsay & Zhang paper's model findings of a decrease of undeformed ice thickness are for the whole Arctic Ocean, it shouldn't be surprising that four sites in certain regions don't bring the thinning out, any more than four weather stations will bring out the signal of AGW.

It's quite possible that this model study will never be verified with observations for the 'preconditioning' phase and what the paper finds to be the dominant mechanism - surface temperature increases driving thinning of underformed sea ice.

[ktonine]

Chris, To determine whether volume loss is primarily due to changes in melt or freeze,  it may help to start with the possiblities: most of the following is beating a dead horse, but I think better if I lay it out line by line

1) Level or Increased melt and decreased freeze
2) Increased melt and increased freeze (but of lower magnitude than melt)
3) Level or Decreased melt and decreased freeze (but of greater magnitude than melt)
4) It's not thermodynamic

I don't believe a reasonable case can be made for #4 - so I'll cross that off the list.

Looking at the data both melt and freeze show an increasing trend with melt averages slightly larger than freeze averages, so we can cross off #s 1 and 3.

With confidence we can then state that Arctic volume losses are a result of increased melt and increased freeze, with melt increases outweighing freeze increases.

The next question is one of timing: is there a relationship between the melt and freeze on an annual (or otherwise) basis?  I.e., does melt track freeze or does freeze track melt?  Looking at the data it's easy to show that freeze tracks melt.  The correlation coefficient for Freeze > Melt as you have the data posted is 0.158, but if we look at the correlation between Melt and the following Freeze season the coefficient is 0.615

Having established a correlation, the causality is obvious.  Melt determines the subsequent Freeze. It would be truly exciting if we could prove that this winter’s freeze determined last summer’s melt – a whole new branch of physics would be opened up – but alas we’re stuck with the mundane -- cause precedes effect

Let’s now consider the nature of the correlation.  If we take the average of the Melt data and break the data set into those Melt seasons above and below average we find there is little correlation between low melt years and the following refreeze (0.181), but a stronger link between above average melt years and the following freeze (0.467).  More important, the average volume change for below average melt seasons and the following freeze is + 0.05, while the change for above average melt seasons and the following freeze is -0.76.

From this I think we can conclude that volume loss is due to above average melt seasons – despite above average freeze in the subsequent winter.  In only 1 of 15 above average melt years did the following winter create a volume gain (Melt 2007, Freeze 2008).  Both Melt and Freeze are increasing during these years (because Tietsche), but without increased melt there would likely be little volume change.

[ChrisReynolds]

Kevin,

Your initial crossing off of 1 and 3 isn't sound, see my comments regarding the use of a synthetic datset in the first post of this thread.

However your reasoning regards the correlation relationship of melt determining the subsequent freeze is sound. It's something that has occurred to me, I've previously found similar correlations using CT Area at minimum vs the subsequent freeze period gain, and what sets the area at minimum is ultimately the volume loss through the melt season. I just wasn't confident to put as much weight on this relationship as you have, perhaps I was being overly cautious.

Anyway breaking the data down into low and high melt seasons really is a clincher, well done! Thanks for that.

[ktonine]

Chris,

You must have misunderstood the reasoning.  We can arrive at the same volume change by several different scenarios.  I examined each of the mathematical possibilities and compared their theoretical properties to that of the actual data.

Scenario #1 stipulates that Melt is increasing and freeze is decreasing.  The data shows that freeze is NOT decreasing so we can disregard this scenario as a possibility. 

Scenario #3 stipulates that both melt and freeze are decreasing, but that freeze is decreasing with a larger magnitude.  Again, this contradicts the data so we can cross it off.

The only scenario that survives comparison to the data is that both melt and freeze are increasing, but that melt is increasing with a greater magnitude.  To me, up to this point I was just stating the obvious. 

Yes, the interesting numbers are those related to the divide between below average and above average melt years. 

[crandles]

The problem is I don't know whether freeze season volume gains are less than they 'should' be, melt season losses greater than they 'should' be, or a combination of those two factors.

What does 'should' be mean? We have PIOMAS record stating what freeze and melt has been.

Given that we can see what they have been in the 80s' and 90s' then we can say they are larger in the late 00s' and 10s'

[ktonine]

Chris R: "Anyone got any ideas as to how I might seperate out the relative roles of melt and freeze seasons? Or indeed is it likely to be impossible? "

crandles: "What does 'should' be mean? We have PIOMAS record stating what freeze and melt has been."

*If* the data showed a steady melt volume over time, but that the freeze was declining, we could attribute the volume change to the lack of freeze.

*If* the data showed a steady freeze volume over time, but that the melt volume was increasing we could attribute the volume change to increased melt.

Neither of these scenarios would lead to definitive statements, but we'd have a pretty strong clue.

Most other scenarios lead to 'chicken or egg' problems - at least theoretically.  Before looking at the data I would have argued for Chris' 2nd sentence - that it would likely be impossible to determine attribution.  After looking at the data I think it more reasonable to believe that volume loss is due to larger melt season losses.  Rarely is an above average melt loss followed by a winter able to 'recover' the loss.

[ChrisReynolds]

The problem is I don't know whether freeze season volume gains are less than they 'should' be, melt season losses greater than they 'should' be, or a combination of those two factors.

What does 'should' be mean? We have PIOMAS record stating what freeze and melt has been.

Given that we can see what they have been in the 80s' and 90s' then we can say they are larger in the late 00s' and 10s'

I've got a sense of deja-vu here - haven't you made this point previously? The 'should' relates to my use of synthetic datasets with the losses taken from freeze or melt. With a trend of loss the melt must always be greater than freeze - even if the losses are actually coming from the freeze season. It turns out there is no way to extract whether the loss is coming from melt or freeze without knowing what the numbers 'should' be given the choices made in calculating the synthetic data.

This point also relates to Kevin's latest reply.

However Kevin's seperating out low and high melt seasons addresses the issue, something that couldn't be done in the synthetic data I was playing around with. BTW I had also played around with sinusoidal synthetic seasonal data - same problem there as with the freeze/melt simplification.

The reasonable assumption all along was that it was melt driving the loss, but I had a suspicion that ocean heat flux might be limiting winter growth and wanted to see if winter could be ruled out.