« 2011 Slush Puppie Animation | Main | Poll Results 2 »

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

Anu

-- Ned Ward | August 24, 2011 at 11:17

2. I would similarly caution people against placing too much confidence in models of volume. Volume is very difficult to measure. Eventually, we'll have pretty good Arctic-wide volume measurements from Cryosat (there are probably a bunch of presentations on this being prepared for AGU), and maybe longer-term changes in volume can be inferred by comparing Cryosat observations to the much sparser and less accurate IceSAT GLAS data from 2003 onwards. But in the meantime I personally would be cautious about making arguments based on models of volume.
===================================================

I agree - I have little confidence in the "PIPS 3.0" ice thickness model, although I have *more* confidence in that one than the old PIPS 2.0 model. They now base their work on HYCOM (HYbrid Coordinate Ocean Model - the work of a large consortium of institutions doing data assimilative modelling of the ocean) and the CICE (Community Ice CodE) model of Los Alamos National Laboratory.

http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticictnnowcast.gif
But 2 and 3 meter ice thickness over most of the Arctic in late August ? I'm not buying it.

Anu

I have more confidence in the PIOMAS model, due to it's validation against IceSat, it's many years of incorporating buoy data and airplane measurements, and it's constant attempt to measure full Arctic ice volume over the years and analyze the trends (PIPS 3.0 strikes me as false-positive overestimation of ice thickness for nuclear sub operations...).
http://psc.apl.washington.edu/wordpress/wp-content/uploads/schweiger/ice_volume/BPIOMASIceVolumeAnomalyCurrentV2_CY.png

I don't care about "longer-term changes in volume" once volume hits 0 (you don't need a decades long medical history of a patient once he's dead). Being "cautious" in ones expectations is no guarantee of being right. If PIOMAS is close to correct (and CryoSat-2 will soon enough confirm or deny this), then "melting away suddenly" can very well happen in the next three years.

Jon Torrance

Paul,

You've designed a bet that has a higher expected value for you based on your stated position than it does for Nick based on his stated position. I wouldn't blame him, if this bet isn't to have a zone of results in which neither of you wins or loses, for insisting on a threshold enough lower than 4.60 to equalise that.

Jon Torrance

To be fair, Nick also proposed a somewhat lopsided bet in his favour if we assume your description of his position is accurate.

D

I realize his opinion is based on trends. I was trying to point out that the trend is decreasing too fast to make it the strict arbiter of reasonableness.

Maybe the fast-melt advocates just had a really lucky decade, but with each passing year it's more likely that they are right and the mainstream consensus is wrong.

Paul Klemencic

Jon, I don't expect Nick to make my bet, because I believe that in his heart and head, he also believes the expected outcomes are the melt are skewed; this is my way of getting him to realize that.

Paul Klemencic

should read "... expected outcomes of the melt are skewed;..."

Lucia (The Blackboard)

Paul--
In your bet offer to Nick, did you mean to bet IJIS would break below 4.6 or 4.25? My reading is Paul put the 50-50% odds at 4.25 not 4.6. So, if I were a money betting sort, I might be tempted to take up your offer. But I only bet quatloos.

My best estimate for the upcoming minimum in the 7-day average is currently very near 4.6.

I noticed people are discussing statistical methods and physics methods as if they must be diametrically opposed. They need not be. It's true that statistics can't for see absolute breaks -- that's a problem. On the other hand, unless we have a very clear understanding of a lot of complex behavior, it's very difficult to anticipate why a break should happen and where it will happen.


I have been posting the odds I get for the 7-day IJIS minimum based on past behavior. It's purely statistical. Moreover, I haven't yet looked at including reported volume estimates or SST etc. in my predictor.

Clearly these must matter somewhat. As I haven't looked I don't know if the current volume are exceptional given the other factors we are observing. That is: the high sst's or low volumes might already be reflected in low areas and fairly large difference between extent and area. That is to say: We would always expect low volume when the area is as low as it is. This means to some extent, once we recognize the area is low, the volume information helps us predict only to the extent that the volume is low given the area.

If the predictive information in 'volume' is already mostly contained in 'area', then the additional information about volume may not help anyone predict any better than merely knowing the area is low.
Or maybe it will turn out that knowing whether the current volume is low given the area will help us predict something. I don't know because I haven't looked. My current educated guesses are based on statistical fits that use (CT area, extent, time) and nothing else.

Paul Klemencic

Lucia, I don't understand your point. He offered an even odds bet at a value that was very close to my 50/50 expectation: for my point of view, the expected outcome from a series of bets like that, would be a net winnings of close to zero. In other words, no reason for me to bet. OTOH, the expectation of Ned and you is that 4.60 is the 50/50 point; so if he shared that POV, then he was in position in his bet to me at even odds at 4.30 to have a very positive expected outcome from a series of bets like that. So, in essence, he was offering me a sucker bet.

I came back to him with a bet proposition that shared the expected value that could be derived from our different points of view. He should take 2:1 odds on a bet at 4.60, if he believed the odds were only 50% that the value would come in below 4.60. I would be willing to offer those 2:1 odds, because I expected the outcome to be less than 4.60 over 90% of the time. Both of us would have positive expected values from a series of bets under those term, based on our DIFFERENT points of view (assessment of the odds).

Lucia, if you haven't done a lot of expected value calculations on risk investments, you should learn about EV calculations. Then we could talk about 'tornado diagrams', and especially skewed probability distribution curves used in these calculations. I learned this stuff evaluating risk investments in the oil industry.

The key problems with your analysis is that you understate the chance of a fundamental system change, and you assume a normal distribution of outcomes. Both are bad assumptions.

Chris Biscan

http://www.iup.uni-bremen.de:8084/amsr/arctic_AMSRE_nic.png


The Eastern Ice pack moved 15-25km inward.

in response to:

http://vortex.plymouth.edu/gifs/110824180611.gif


Why?

http://lance-modis.eosdis.nasa.gov/imagery/subsets/?subset=Arctic_r03c04.2011236.aqua

Because there is crippled thin ice there. It is so thin probably .3 to .7 meters on average that it's nearly blue instead of white since it's transparent.
Also look at the wind Direction coming around the SLP:

http://www.weatheroffice.gc.ca/data/analysis/351_100.gif


I wrote multiple times that there is a mini DPA like situation going down over a 3 day period.

Do they stats account for this?

If that ice was 2-3 meters thick like it should be there this time of year historically. That system would barely effect it. It wouldn't move because it would have no where to go.

it wouldn't melt because the SSTS with that kind of ice pack are way way way WAY WAY WAY WAY WAY WAY WAY WAY WAY COLDER.

I don't believe in anything when it comes to this in terms of politics or sides.

But I do practice Meteorology.

and while I have no idea what odds are this or that for a new record extent.

I know the weather can make it happen. I know the weather with the warm waters left over could still given the worse possible weather conditions drop another 1.5 million km2 of ice off the extent.

One ten day DPA would take out 750-1 mil km2 of that.

If this ice had been subjected to current arctic heat and 2007 winds. we would be down around 3.0 mil km2 for a min.

The winds have been garbage all summer for flushing. Heat killed this ice. And now heat in the water will extend the season longer then it should be.


Chris Biscan

http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticictnnowcast.gif

How can they publish this rubbish.

Polarstern just debunked the entire Eastern half of that map.

Neven

If this ice had been subjected to current arctic heat and 2007 winds. we would be down around 3.0 mil km2 for a min.

The winds have been garbage all summer for flushing. Heat killed this ice. And now heat in the water will extend the season longer then it should be.

I totally agree with you here, Chris.

Regardless of the question whether it would have been a minimum extent of 3.0 or 3.5 million square km, it's important to keep in mind that this year's area and extent has been trailing 2007 so closely, despite having less favourable conditions than then.

If the ice thickness continues to become a more dominant aspect of extent decrease than weather patterns, I don't think it's going to take very long for a melting season to beat 2007 with average weather conditions.

And if in the coming few melting seasons we do get 2007 conditions from start to finish, a 3.0 million square km minimum extent isn't an exaggerated or alarmist expectation. That would be entirely normal.

Neven

I have updated the Fram Strait animation. Not much to see as of yet, but transport could be picking up again in a day or two.

Simon
this year's area and extent has been trailing 2007 so closely, despite having less favourable conditions than then

Wasn't that how a poster here rephrased the no-tipping-points-in-the-models issue: that the first three years' recovery is all you get, because the trend catches up.

Nick Barnes

I only make and take even-odds bets. Let us split the difference between your expected value (4.25) and mine (4.6): 4.425. If the IJIS daily extent minimum is below 4.425 million square kilometres, I will pay you fifty euros. If the IJIS daily extent minimum is above 4.425 million square kilometres, you pay me fifty euros. Result to be decided on 2011-10-01. Do you accept this bet?

Lucia (The Blackboard)

Ned--
Lucia, I don't understand your point.
I'm asking a question not making a point. I'm a bit confused partly because we have Ned-Nick and Paul names in the conversation. But I think it's just down to you talking to Nick. Also, other than quatloos, I pretty much don't bet, so the odds things have me confused.

Anyway-- indulge me here, because I'm trying to get an idea what the bet is. here Nick quoted Paul and offerred a bet.

The Paul bit read:

I suggest by looking at the special causes, that breaking 4.6 million is over 90% certain, and the odds of taking out the 4.25 million low extent record set in 2007 is greater than 50%, and there is a significant change (over 20%) that we end up with less than 4.1 million (if the right weather combo hits, which looks to be at least a one-in-five chance).
To me, this reads as if you think the odds of going below 4.25 is greater than 50%. You think the odds it will be below 4.6 is greater than 90%.

Then, Nick oferred you this even money bet:


Paul, I'll bet you 50 euros, at even odds, that the IJIS extent minimum will be above 4.3 million square kilometres.

When I read that it seemed to me he added a buffer of 0.05 millions-- giving you better odds. Plut, you already think 4.25 gives you better than 50-50 odds. So, based on what you wrote, I thought should see this as better than a 50-50 bet. I couldn't guess how much better- it seemed to me he was offering you a bet you would feel confident you win. I didn't know how confident, but I wasn't expecting you to describe that as a bet you could only win 50% of the time.

Then, in the later comment, it appeared you did not think Nick's offer gave you better than 50%50% odds. So, I was a bit confused. (It's not the first time-- especially with betting.)

But that's why I was asking.

Now, on to the other bit:
I would be willing to offer those 2:1 odds, because I expected the outcome to be less than 4.60 over 90% of the time. Both of us would have positive expected values from a series of bets under those term, based on our DIFFERENT points of view (assessment of the odds).
I'm trying to get a handle of what you are saying:

You think with this bet, you would win 9 times for ever 1 you lost. If you win, you get '1' and if you lose you pay '2' (right? I actually don't do betting, so I'm not sure that's what you are offering.)

But if I understand that right, if you bet $1, based on your expectation of what will happen, you expect to gain:

[9*(+1)+ 1(-2)]/10 =7/10.

On the other hand, if someone (possibly Nick) thinks the odds of being over or under 4.6 are 50%50, and they re right from their point of view, they get:

[1*2 + 1(-1)]/2 = 1/2
Over the long runs, that would be good. This is less good from their POV than you get from your POV.

Did I at least calculate what you think winnings would be on average?

Anyway, I don't know what odds or estimated rates of return you or Nick would expect to get on bets. But I asked about the 4.25 vs. 4.6 because it seemed to me based on what you wrote Nick had offered you a bet that had better than even odds from your point of view. Presumably, I thought that because I misunderstood what you meant in the earlier comment.

Lucia (The Blackboard)

Lucia, if you haven't done a lot of expected value calculations on risk investments, you should learn about EV calculations.
I know about EV calculations.

The key problems with your analysis is that you understate the chance of a fundamental system change, and you assume a normal distribution of outcomes. Both are bad assumptions.
I've agreed that I not only "understate" the chance of fundamental change, I utterly ignore this. I am entirely aware that this can make my prognostications incorrect.

I doubt a single person posting estimates based on past performance is unaware that they are ignoring the chance of a fundamental change. As for me: I'm just saying that I am estimating it based on the past, and this is what I get. Other people who think they can somehow forsee a change are free to do so. This doesn't bother me at all.

Twemoran

Guys - Why not accept each others positions as stated instead of trying to define others through betting propositions

I lived in Las Vegas for long enough to realize that betting behaviors are seldom driven by rationality

In the heat over the differences of opinion, we may have lost touch with JAXA having had a pronounced upward revision today - could this indicate that the sensors were not able to distinguish recently awash ice, or was it an error more due to the extensive cloud cover

Chris Biscan

Via the prelim map. Today looks like a 40-60km2 extent loss day.


This will pick up over the next few days.

the 12z Euro is out and will flush tons of ice out of the Russian side.


http://vortex.plymouth.edu/gifs/110824192330.gif


that is about how the surface winds will look.


As the Eastern side ice gets shunted towards the North Atlantic Side.

The ice along the Canadian Basin gets pushed West towards.


This will help be an exodus for the ice coming down creating divergence in the middle of the pack...which will allow the Eastern side to move faster.

We will see. but there will be quite the bit of movement/compaction.

We will see.

on top of that, the cold is sporadic and not enough to refreeze.

Patrice Pustavrh

Nick, I am with you on taking even odd bets. Cause I can easily imagine myself betting 10 EUR on extent going below 3.9 mio sq km with return to 100 EUR (I didn't calculate the probabilities), but I would find hard to convince myself on bet of 100 EUR at or below 4.6 mio sq m2, even if I'd might have greater chance of winning). But, I think this betting stuff has but us somehow away from what we are looking for: The dynamics of the system and learning for our future world. I personally do think that discussion about arguments (and I find many of yours valid, but on the other hand many of much more alarming ones too (including Paul, Lodger, and many others I cannot mention in a line or two) very very compelling and in the need for stronger study - as Neven does and he is not alarmist by no means at all, but a really great and in my opinion one of the greatest science journalist trying to give us most accurate news in his blog - he really puts his effort where his mouth is, always willing to learn new things and not being scared to post about themes which might be uncomfortable to his well expressed view about GW, but challenges them in true search for truth as much as human can do and I hope I am doing my job here on equal basis too). And yes, I wish some bettings would be taken on at least another post on this blog or even at some other blog. We are not gamblers, men, we like to study.

Chris Biscan

Jaxa revised upwards because it goes on a 2 day average. And the loss the 2nd day was much lower. So by the final, the final loss number averaged with the day before was lower then the first 2 day averaged prelim.


yesterdays loss was negligible. This will hold todays down, so it could be as low as 30km2 today when it is probably 50-60km2 for a single day.

Neven

Patrice, thanks for the nice words (my ego is jumping with joy), but I can assure you that I am very alarmed and worried about what we are observing in the Arctic. I can sleep, I have a great life, but I worry nonetheless.

But with a blog comes responsibility. And the Arctic is a fascinating subject, even without possible consequences for human society and ecosystems or implications for the AGW debate.

I wouldn't be doing this if it wasn't stimulating and fun.

BTW, you guys can bet all you like, but I'd rather have you put in a clause that voids the bet (when extent ends up between x and y). Winning/losing is so boring.

Bob Wallace

Was reading up on NP-38, the Russian floating ice station and saw this...

"The first scientific drifting ice station in the world, called North Pole-1 (NP-1), was established on May 21, 1937 by Otto Yulyevich Schmidt. The ice floe holding the research station drifted 1,770 miles (2,850 kilometers) before a Russian ice breaker picked up the four explorers manning the station nine months later.

Russia did set up drifting stations on ice floes frequently since 1954. One of the most important discoveries made during such a drift was the identification of Lomonosov Ridge, which crosses the Arctic Ocean.

Based on European and U.S. satellite imagery, a team of Russian scientists preselected over two dozen ice fields as potential hosts for NP-38. After closer inspections, most ice floes did not meet the basic requirements.

The researchers look for the "perfect" junk of ice that is about 1.9 by 2.5 miles (3 by 4 kilometers) large, of oval shape and as thick as possible.

Finding an ice floe that is 20 or 30 yards (meters) thick was not a big problem in the past, but now most fields inspected have a thickness of less than two yards (meters)."

Volume. It's about volume, not how much of the surface is covered with ice when the freezing season is finished. Unless something unforeseen turns volume around it really is simply an office betting pool as we guess the year it first all becomes open water.

(And when the "death spiral" graph is next updated I have a feeling that the fitted curves are going to predict even sooner....)

Twemoran

I've been keeping an eye on L Hamilton's excellent graphics, and also one by D Kelly O'Day posted on Arctic io's site - this chart shows daily extent change though the month compared to each year back to 2002 and seems to indicate, at least to me, a fundamental change taking place in 2007
This year has experienced weather patterns that should have been conducive to higher than recent numbers across the board - late opening of Nares with minimal transport - little transport out of Fram - And yet the thin ice and high temperatures see us in a close race with 2007
What will happen when we see a summer with weather patterns like 2007 interacting with ice conditions such as we've experienced this year

Patrice Pustavrh

Neven, I am alarmed too, because even half the ice melt we are observing should be warning sign to us already. But, what I admire here is desire to look at the facts first and then raise the alarm as opposing to sensational levels voicing alarms all over the place (althouhg, sensational news can be helpful for short periods). Sometimes, and maybe it is just my feeling, a decent, sound realization of some trend is just more convincing than sensational claims of records (and it is not good for the headlines). And of course, I find your blog at just exact pace of what and even how should be voiced up. My favorite life motto is go slow to advance fast. We don't need to rush, but deniers do. However, one thing should be done to public: Gather the deniers predictions and compare them to predictions of the guys who really know about the AGW. And show them to public. And yes, if necessary, call the deniers the real liars, if necessary. Gee, sometimes they really makes me sick and angry with their cheap lies, but I repeat to myself: Stay calm, you will get all the help from planet Earth itself. And maybe, just maybe, I will find someone more agree with me.

Paul Klemencic

Lucia, thanks for the opening. To be clear about this, I am not criticizing you as a person, I am talking about your model, and this in turn is influenced by your mental model and prior experience, or lack thereof. With that said, you could learn a lot from this exercise if you choose to, and you aren't close-minded.

Lets start with the statement " I'm just saying that I am estimating it based on the past, and this is what I get." The problem is that this statement is false. You are NOT just estimating it based on the past. You are estimating using the variables you have selected, and using statistical analysis based on the assumption that the minimum extents reported each year conform to a normal distribution.

Do we have enough data points to show the minimum extents are distributed normally?
No, we do not. And what data we do have, suggest a skewed distribution with a bias toward lower minimums.

So why are you using statistics that assume a normal distribution?
Because its what you have done in the past; its standard operating procedure (SOP) for most statisticians. Unfortunately in the case of Arctic ice melt, its a bad assumption, and deserves to be questioned.

More in the next comment...

Patrice Pustavrh

Sorry for the OT, but it really crossed my mind. I remember the song, which states:
You can fool people for some time, but you cannot fool all the people for all the time. But, giving ecology, we are here: We can put some people to live in space for some time, but we cannot put all the people to live in space for all the time.

Wayne Kernochan

My apologies for sounding overheated, but I thought we had covered this before. What is this talk of "volume is not well estimated" and "greater extent drop requiring a fundamental system change"?

The model is of increasing amounts of heat being applied to a fairly thin, very large cube of ice, both from above (the sun/global warming) and below (warm currents from the south). The thickness of the ice in the cube varies, more or less according to a normal curve. As the seasonal heat increases, the effect on volume shows up first. Area and extent only react late in the game, and then near the end they drop much faster than volume. All three accelerate as they go.

PIOMAS' key difference from other models is that it assumes greater underwater melting. PIOMAS' approach has by now been entirely vindicated by observation, as it predicted much less ice thickness than other models by now, and that decreased ice thickness has been observed.

The point of this is that, given the rate of change of PIOMAS volume at minimum from year to year and the observations of very thin ice, we are now around the point where ice thickness enters the "meat" of the normal curve. At this point, more and more of the ice achieves "negative thickness" -- it melts out, and area begins to drop much faster. What happens then depends on the "width" of the normal curve. In this case, it's clear that almost all of the ice is between 0.1 and 2 meters thick, with an average of 1 (as of 2010). Again, following the trend of acceleration of volume decrease (which is exactly what you would expect if the ice is melting both below and above from year to year), we might expect a 10% drop in minimum area next year from 2010, a 50% drop from 2010 by, say, 2013, and a 90% drop from 2010 by, say, 2015.

The way that drop would be manifested is in increases in thin ice being melted "from the bottom" up by warmer water, plus extensions in the melting season because the water is warm when the air is not. These seem to be the kinds of phenomena this site is reporting.

The point of this is that additional melting over past seasons at this point in time is not not not not not not not not not a fundamental system change. On the contrary, to expect that you would get the same amount of melting as in the average past season from this year forward would require that volume suddenly stop decreasing -- and boy, would that be a fundamental system change.

Moreover, the idea that volume is not well measured has far less relevance than you might think. It follows from the above analysis that up until the final stages, volume, however crudely measured, is a better indicator of what's happening to the ice amount than area and extent. And in the end stages that we are entering now, average thickness is becoming a far better predictor of minimum area and extent than past seasons' area and extent data.

The net of this is that (again, all else being equal) extrapolation of past seasons' melt at this point is not the most likely thing; on the contrary, the most likely thing is greater melt than the extrapolation predicts (if, of course, you don't factor in ice thickness). Volume and fundamental system change arguments cut against, not for, such a model.

Wayne Kernochan

oops, that should be 2014 and 2016, not 2013 and 2015. My bad. - w

Steve Bloom

Lucia isn't a statistician, Paul. Tamino OTOH is (actually a time series analyst, even better for these purposes).

Xandra

Maybe a little OT but the great site Skeptical Science http://skepticalscience.com/search.php?Search=arctic&x=0&y=0 is doing a great job to show the real science and has a lots of great articles about Arctic.

Paul Klemencic

Last comment from me on betting:

Lucia, here are correct EVs for the three bet scenarios, with each bettor's assessment of probability.

First the summary, then the detail:

Nick Proposal #1:
Nick EV = 38.5 euro expected profit for 50 euro risk (88.5% chance of winning based on his probability assessment)

Paul EV = 4.0 euro for 50 euro risk (54% chance of winning based on his prob. assessment)

* Please note why I laughed this bet off as sucker bet, in spite of your rejoinder of how this was such a good deal for me, and encouraging me to take this bet.

Nick Proposal #2:
Nick EV = 25.8 euro expected profit for 50 euro risk (75.8% of winning)
Paul EV = 15.0 euro for 50 euro risk (65% chance of winning)

Paul Proposal #1:
Nick EV = 25.0 euro expected profit for 50 euro risk (50% chance of winning)
Paul EV = 35.0 euro expected profit for 100 euro risk (90% chance of winning)

For Paul to negotiate a bet under either of the terms proposed by Nick, is not a good deal, and leaves a lot of "money on the table". My opening offer in the negotiations sought to persuade Nick to move to a more equal risk/reward scenario.

Clearly you need to stick to betting quatloos, and stay away from using real money, or participating in negotiations involving risk investments in the real world.

Details:
I used normal distribution for Nick, with a median of 4.60 and with a standard deviation of 0.25 million sq km.

For Paul's probability assessment I used the following skewed distribution, with the fraction depicting the chance of a probable outcome exceeding that minimal SIE result:
4.10 0.80
4.15 0.66
4.20 0.56
4.25 0.50
4.30 0.46
4.35 0.42
4.40 0.38
4.45 0.32
4.50 0.24
4.55 0.17
4.60 0.10

Nick Barnes

Paul, you should assess whether or not to accept a bet based entirely on your own expectations, not on those of the other party to the bet. If my expectations differ from yours, you think my expectations are incorrect. Why would you base your bets on incorrect expectations? It makes no sense at all.

You're saying that you believe P(minimum extent less than 4.425) is something like 0.36. But you won't take an even-odds bet? That's leaving money on the table.

Nick Barnes

Added to which, where did you get a sigma of 0.25?

crandles

Re: "Why would you base your bets on incorrect expectations? It makes no sense at all.

You're saying that you believe P(minimum extent less than 4.425) is something like 0.36. But you won't take an even-odds bet? That's leaving money on the table."

It isn't a take or leave decision with no alternative. There is an alternative; to try an negotiate a fairer bet where each person gets a more equal expected value. (Though it can be debated whether an equal expected value is the metric of fairness.)


Anu

Posted by: Wayne Kernochan | August 24, 2011 at 22:38
==========================

I agree. The single most fundamental question about Arctic sea ice is - "How accurate is the PIOMAS model?"

And the answer will come from CryoSat-2, once it is calibrated, validated, and ready for prime time. Which I truly hope is the case by this September minimum, because that will be a crucial test of:
http://psc.apl.washington.edu/wordpress/wp-content/uploads/schweiger/ice_volume/BPIOMASIceVolumeAnomalyCurrentV2_CY.png
(although the published results will probably be next year sometime)

If CryoSat-2 verifies a minimum volume of 4000 to 5000 km^3 this September, then throw away those "extent trends". We're looking at an ice free summer within 8 years.

Before CryoSat-2 data is in, arguing/speculating/betting is rather premature.

Ned Ward

Lucia writes:
***********************
"Ned--
Lucia, I don't understand your point.
I'm asking a question not making a point. I'm a bit confused partly because we have Ned-Nick and Paul names in the conversation. But I think it's just down to you talking to Nick."
***********************

You're right that you're confused. The sentence you quote was by Paul K, not by me. Nick, Paul, and I are all completely different people. If you read the "Posted by:" line underneath a given comment, you can determine who wrote it. I have had no part of this discussion of betting.

Timothy Chase

Chris Biscan wrote:

http://www7320.nrlssc.navy.mil/hycomARC/navo/arcticictnnowcast.gif

How can they publish this rubbish.

Polarstern just debunked the entire Eastern half of that map.

Strict military discipline.

Ned Ward

D writes:

But lets look only at extent and area- there is now a 30 year trend that shows the arctic will be seasonally ice free in 20-30 years.

I'd agree with that, more or less, though the number of years depends on what kind of model you use. The basic point is that Arctic sea ice extent has been declining for at least 30 years, and that a continuation of that trend would lead to a seasonally ice-free Arctic ocean well before 2050.

Since just a few years ago some scientists were predicting that the ice would last 100+ years, this is a pretty extreme change.

I agree it's an extreme change, though I haven't tried to keep track of exactly who predicted what when.

While predicting the exact year is impossible, I think it is quite reasonable to publicize the fact that the arctic is melting way faster than predicted in earlier climate models.

I agree completely (well, as long as "the arctic" is shorthand for sea ice ...).

At this point, all the statistics are on the side of the "alarmists".

I don't think statistics take sides, but if you're willing to just say "the situation is alarming" I'd agree with that.

But let's suppose that you now want to reject the statistical argument since it clearly doesn't show what certain groups would want it to.
Do you have any scientific reasons why the 30 year trend is likely to reverse itself in the near future?

Huh? None of that has anything to do with anything I've said here. You seem to be arguing with an imaginary person, not with me.

I think the 30-year trend is likely to continue until there is relatively little ice left in the Arctic ocean at the date of the annual minimum (though some remnant land-buttressed ice will probably persist north of Greenland and in small patches within the Canadian Arctic Archipelago).

At that point, the Arctic ocean will be characterized by a sheet of thin, first-year ice that forms every winter and mostly disappears every summer ... rather like Hudson Bay today, I suppose.

Ned Ward

Nick Barnes writes: D, assuming your comments are aimed at Ned, he is very clear that he bases his opinions on the trends. His criticism is of those who make speculative leaps beyond the trends.

Yes, exactly. Thank you.

Ned Ward

I'm going to remain agnostic about the accuracy of modeled trends in volume until Cryosat starts generating publishable results.

I trust ICESat rather more than I trust the PIPS or PIOMAS models ... and I'm not super confident about the usefulness of ICESat data. I've actually worked with ICESat GLAS data some in my day job, though not for estimating sea ice freeboard. For an experimental system it was interesting, but subject to frustratingly large and difficult-to-characterize errors, even before it died in 2009.

Most new satellite-derived products take a few years (at least) to work out all the issues and start producing reliable information.

Paul Klemencic

Nick, if two rational parties kept at this long enough, they would agree on a 50/50 bet at the minimum SIE point where one party assessment of their chance of winning was the same as the other party's chance of winning.

If we use the two probability curves in my comment, that point is about 4.47, where both parties would have a 70% chance of winning based on their internal assessments. We were getting close, and if you still want to bet, that is the point where I would agree to a 50/50 bet.

Regarding the 0.25 standard deviation; I believe Lucia calculated 0.30 about a week ago, but this is later in the season, and I don't have an updated analysis from Lucia. Two deviations out at 0.3 puts the upper bound at 5.20 million sq km, and we are already below that, so I used a standard deviation a bit smaller and more consistent with the current extent of 5.12 million sq km.

If you don't think that the bet is a fair bet for both parties, then this likely means that you also feel the distribution is skewed, since you won't take a bet that according to the normal distribution probability has a 70% likelihood of being a winner. I suspect you also feel the chances of seeing a final minimum extent lower than 4.47 million are greater than 30%. This means you also have a feeling the distribution is skewed toward a lower minimum.

Anyway, if you want the bet, the 4.47 million mark for a 50/50 bet is on the table.

Lucia (The Blackboard)

Paul,

Thanks for clarifying. I think our computations agree but it appears you describe things based on the "risk" by a party, and I expressed it relative to the "1" in the "2:1" which I called the bet. (I'm totally unfamiliar with betting terminology, so that's what it occurred to me to scale off off.

I only examined 'Paul Proposal 1: and wrote

But if I understand that right, if you bet $1, based on your expectation of what will happen, you expect to gain:

[9*(+1)+ 1(-2)]/10 =7/10.

On the other hand, if someone (possibly Nick) thinks the odds of being over or under 4.6 are 50%50, and they re right from their point of view, they get:

[1*2 + 1(-1)]/2 = 1/2
Over the long runs, that would be good. This is less good from their POV than you get from your POV.

You describe this as:

Paul Proposal #1:

Nick EV = 25.0 euro expected profit for 50 euro risk (50% chance of winning)
Paul EV = 35.0 euro expected profit for 100 euro risk (90% chance of winning)

So, this appears to agree with what I got- but I expressed Paul's 0.7 in terms of the $1 he receives if he wins instead of the $2 he risks if he loses. This 0.7 to 2 scales to 35 Euro profit on average for a 50 euro risk. So that agrees with what you get.

I expressed Nicks in terms of the $1 he risks if he loses (which seems to be conventional) and my 1/2 to 1` scale to 25 Euros profit to 50 risked.


I'll try to remember the convention for the language though I'll probably forget it because I don't generally bet. But it does seem we agree. Thank you for also confirming the first bet Nick offerred did correspond to a profit for you based on your odds. How large I couldn't estimate because I don't think there was sufficient information in comment to figure out that you thought your odds of winning were 54%.-- though maybe that information is there. I don't even know where Nick's second proposed bet is. :)

Clearly you need to stick to betting quatloos, and stay away from using real money, or participating in negotiations involving risk investments in the real world.

Well, since as far as I can tell, I got the exact same numerical values you did, but merely used different language to describe the outcome (using the $1 in the $2:$1 as a reference while use risk to discuss the outcome) I don't see why I couldn't identify which bets are winners and losers assuming I correctly identify the probability distributions. But I don't particularly want to bet money unless it's other people's. So, no worries.

Lucia (The Blackboard)

Paul--
In response to this

Don't worry. I'm not taking this personally. Based on what you write I suspect you are making assumptions about what I am thinking or claiming about my model and some of those assumptions don't match what I think I am doing.

You are NOT just estimating it based on the past. You are estimating using the variables you have selected, and using statistical analysis based on the assumption that the minimum extents reported each year conform to a normal distribution.
I agree I am estimating based on the variables I selected. I am entirely aware of this. In a comment earlier on a thread, I mentioned that I had not yet added in things like volume. I am also aware that to the extent that explanatory variable do not capture the phenomenology, the model will fail.

In my more recent response to you, I was responding to what I thought you were saying was that if there were breaks in the physical process my model will fail. For example: In fluid mechanics we often see a drag crisis resulting in lowered friction factors for external flows as a boundary layer becomes turbulent. These "breaks" will not be discovered by looking at data only in the laminar region or only in the turbulent region. This sort of break is what I though you were referring too. The problem is separate from my model possibly not including even those variables that might have been useful in the past, so I was focusing on the "past" issue and not the "missing variables" issue. Both matter.

I am totally aware of both issues and readily admit that people need to consider this when judging how much credence to put in any uncertainty intervals. Nevertheless, I mention the ones I get using the procedure I described at my blog.

Do we have enough data points to show the minimum extents are distributed normally? No, we do not.
Actually, we have enough to know that if you just take the minimum extents themselves not distributed normally. I never assume the extents themselves are normally distributed. I can't imagine anyone else doing statistical fits assumes this either.

What I do do, is assume is that residuals are distributed normally around candidate fits I try. I've looked at various histograms of residuals, and for some models they are more or less normal (by the eyeball method) in other cases they look skewed. No doubt about it.

As you may or may not know, I draw from what I consider to candidate of models, compute AIC coefficients and weight. This makes the issue of non-normality a bit more complicated than merely worrying about skew because technically, even if the individual candidate models I draw from have normal residuals, the pdf for the weigthed model is not normal. For example: if two candidates models were judged equally likely, and both were normally distributed, the resulting pdf would be bimodal.

Currently, the method is picking out 4 models as somewhat probable. I haven't computed the skewness for the residuals, but for the model with the best AIC coefficient, the median is nearly equal to the mean, and the residuals look fairly normal. For 2 models, the median exceeds the mean; for 2 the median is lower than the mean. So, though three of the distributions have skewed residuals the skewness is not consistent and partly cancels on weighting.

In principle, I could try to come up with a pdf for the weighted model. I may eventually, but for now, I don't. Right now, I compute the standard error (which is the same regardless of distribution) and post the 2 sigma errors.

Why do I do it this way? Budgeting time. I am not going to the trouble to deal with the skew or the multi-model nature until after I look at a few more regressors that I think may matter. I'm doing this in this order because I'm more interested in looking at explanatory variables first. I also think that, as we will know the outcome within a few months, the issue of trying to nail down some sort of precise mythological uncertainty intervals for the upcoming NH sea ice minimum is not all that important. (Also, I notice that the models with better AIC coefficients tend to have less skewed residuals. I suspect there is a reason for this.)

In the meantime: My model is what it is. Someone can bet based on it, or not.

And what data we do have, suggest a skewed distribution with a bias toward lower minimums.
Do you mean for residuals? Or something else? Details appreciated because as I mentioned, the skew on my residuals depends on the individual candidate model I'm looking at.
Because its what you have done in the past; its standard operating procedure (SOP) for most statisticians.
It is admittedly SOP for most statisticians and even many engineers to use normal unless you have pretty good reason to think something is not normal.

The pretty good reasons to not use normal could be:
a) whatever you have assumed is normally distributed doesn't look normally distributed and
b) we understand the random variable sufficiently to chose a different distribution at the outset. (So: log normal, bi-modal or whatever makes sense.)

I don't have any physical reason to know why the residuals to any of the candidate models should not be normal. However, as I've said: some appear skewed so (a) applies. But what I see is the residuals from some candidate models are skewed high and some skewed low.

FWIW, I'm a mechanical engineer. My area is fluid mechanics. Lots of stuff in fluid mechanics is not gaussian. Often, the kurtosis is high. Also, for my weighted model, I'm more sort of more about kurtosis than skew. But that's just because I see that some of my candidate models are skewed left and some right, and I think that may more or less cancel. But kurtosis must be an issue. :)


Frivolousz21

-40Km2 loss on Jaxa. Not bad since yesterday was likely near even or a small loss.


The 00z GFS is out. I only have a good arctic view through 72 hours so far.


Day 1: Right now winds in the ESB have turned out of the East or West pending on which side of 180E you are. The SLP still slides SE over the next 24 hours. Winds are pretty strong around the SLP. between 12-20 KTS over a 200 mile wide swath. These will spread out to a 300-350 mile Swatch and weaken some by 24 hours. However they start to connect with more systems to bring the flow from the Bearing bending to that Islands where the ice is barely touching then back right down the East side of the pack. This will offer some compaction by tomorrow. We should see some of this on the UB map tomorrow. Further South between 125E and 135E IS being pushed towards Russia with a decently strong wind. This ice is diverging and melting. Today that will continue. Early Modis Images show this. They also show the ice in the ESB changing Direction and heading south and still melting. Since the temps are cold there and water temps are relatively cold. The ice is still melting, actually pretty fast. There is likely a lot of bottom melt. I recently read that water temps there indicate bottom melt along the shallow water shelf showing ice melt through mid to late October this year. with 1-2C temps below the surface layer.

The Fram and that part of the ice pack will continue to melt and head south and compact slightly.


Ice in the Beaufort and what is left in the NW Territories is under a warm SE Wind. Which means the pack will be pushed away from land some more and melt. SSTS are 2-4C there.


Day 2: the connection from ESB to the SZ lengthens and gets stronger. By now about 300 miles of the Eastern ice pack is will be in the mean flow here moving towards the Atlantic Ocean. A bit south of there the winds veer toward Russia not the Fram. During this 3 day period this will have a negligible effect on the ice movement except pushing it into warmer waters.

the Fram connection stops by 48 hrs and winds become variable there.

The ice in the Beaufort and NW Territories loses it's flow as well but still has warm temps and SSTs for plenty of late season melt.

The winds north of Greenland now start to blow out of the East pushing the ice towards the Beaufort.

Day 3: http://vortex.plymouth.edu/gifs/110825051915.gif

Those winds are sliding West very slowly by the end of day three. As you can see the winds are booming.

by day 4 the SLP slides West and starts to elongate. The boundary will create 20-30KT+ winds over a 1500 mile area from ESB to the Barrents. the swath will be 400 miles wide over the entire Eastern side of the ice pack.

After what we just saw and give the state of the ice. This set up would severely crush the ice with rapid compaction and melt from the warm waters by the land being churned and smoked into the ice.

I really don't know how else to say it. The rest of the arctic will have negligible winds. The coldest temps are right over this area but given the conditions it won't mean jack for refreeze.

To rub salt in the wound. Temp anomalies of 12-16C are slowly leaked into the Beaufort during days 4-8. While temps of -7 to -11C sit over most of the central and Eastern arctic..but with the howling winds it won't mean jack. And models are to cold because they have climo involved and they are always to warm that far out.

Frivolousz21

00Z GFS IS OUT AND IS BRUTAL FOR ICE COMPACTION FOR THE REST OF THE WEEK. WITH NORMAL COMPACTION STARTING NOW THROUGH DAY 3.

THEN THE BURNERS COME ON WITH 15-30KT+ WINDS FROM DAY 3 TO 8 FROM EAST SIBERIAN SEA TO THE BARRENTS/NORTH ATLANTIC NON STOP. COVERING 400-500 MILES WIDE. GOING TO BE SOME MASSIVE COMPACTION TAKING PLACE. PROBABLY HISTORIC.

AS FAR AS THE EXTENT. WE WILL SEE BUT THE AREA FOR SURE SET A NEW RECORD LIKELY IN THE 2.6-2.8 RANGE. THE EXTENT COULD DROP A FEW 100KM2 LOSS DAYS IN THERE.

Nick Barnes

Paul, our esteemed host has asked for us to have a neutral zone, for which neither of us win.

http://neven1.typepad.com/blog/2011/08/sie-2011-update-17-unfulfilled-potential.html?cid=6a0133f03a1e37970b015390f724c4970b#comment-6a0133f03a1e37970b015390f724c4970b

Put such a zone symmetrically around your 4.47 number and you have a bet.

Nick Barnes

(by the way, I'm travelling from now until Tuesday; I might be able to get online, but if I can't then I'll have to re-assess this once I'm back).

Frivolousz21

The Euro follows suit.


with an ugly day 6-7: sliding the HP west to the NW Territories with an SLP sitting over the Arctic Basin between the Beaufort and ESB

Frivolousz21

UB has now vaulted by 2007 on it's ice extent charts.

NEVEN,

If you get time take a close look at Jaxa and UB extent charts they are identical, except when there is low concentration ice.

Neven

FrivolousZ, thanks for the excellent and thorough update. Be careful with that CapsLock button though. :-)

If you get time take a close look at Jaxa and UB extent charts they are identical, except when there is low concentration ice.

Could you elaborate on that further? What is it exactly they do differently? I check both of them every day of course on the Daily Graphs page.

Espen

North East Greenland:

From what I can see from the pre-modis images of today, true lots of clouds, the whole land fast piece of ice of Joekelbugt is breaken up, that is the largest remaining piece of shoreice in the whole arctic sea, only some small pieces of shoreice left in the north of Greenland and the CA.
http://rapidfire.sci.gsfc.nasa.gov/cgi-bin/imagery/single.cgi?image=crefl2_143.A2011237060500-2011237061000.2km.jpg
Regards Espen

Espen

Shore ice / fast ice:

Even the Russians are confirming the disappearance of shore ice:

http://www.aari.ru/odata/_d0015.php?lang=1&mod=0&yy=2011

crandles

My two models predict:

Gompertz adjusted: 4.39 +/- 1 standard error of 0.18

Fall from 22/8 predicted by weighted average of extent and area 4.52 +/- 1 standard error of 0.175

Weighting average of these giving slight preference to latter model gives 4.46.

4.47 is indistinguishable from this so it looks a fair bet that I would not be interested in. I think the distribution might be slightly skewed but I doubt it is much. I think there is a low possibility of a system change because volume is at record low but that has been the case in many years and (extent-area) is higher than average but certainly not a record high.

I certainly believe the uncertainty is higher than .175 as the errors have been growing.

Newcrusader

Lots of technical jargon here, physics, math- not my real strong points- more into paleo climates, climates etc. I have recently become interested in the Cryosphere.

From my observations over at the NSIDC and the AMSR-E- Just my cheap 2 cents worth.

There is still a 50/50 chance we could melt back to the '07 event- a solid 2nd place finish is now at 75%-80% possibility.

Due to the fact that at my location in eastern Connecticut, being under the threat of a major hurricane Sunday- I will miss reading this stuff- as my power could be down for days.

Thanks all for this most interesting site- and the very intelligent posts.

Neven

Thanks, Newcrusader. Hopefully Irene weakens in strength fast. Good luck.

FrankD

Neven, any chance of a CAPIE update?

As I've mentioned before, I've been tracking IJIS Area against IJIS Extent. This usually shows much the same results, but I'm interested in the nuance of differences. My IA/IE was dropping sharply when I last mentioned it (around the 13th, I think, but there's been a lot of posts since then).

Since then, it bottomed at 75.5%, spiked up around the 15th (as per your CAPIE graph above) to 79.5%, but is now sinking again and is sitting around 76.5% - equal to the 2007 ratio at its bottom - as of a couple of days ago (the area graph is always a bit behind).

If my interpretation of this is right, the rise around the 15th - lot of extent loss, little area loss - indicates compaction of the ice (which gels with increased concentration seen on the CT map at this time), but the drop in the last few days - area falling faster than extent - indicates both spreading and melting in situ. I'd be interested to see if CAPIE tracks the same.

Sorry, I realise that's rather dull hindcasting, so in keeping with the recent zeitgeist, people should feel free to insert any of the following superlatives randomly into the above:
Flambéd!
Scalding!
Nuclear!
TOAST!
PHWOAR!

8^P

Neven

Hi, Frank. CAPIE was going up already, but after that IJIS century break it currently sits at 61.09% (from the record low of 57.39%), on a par with 2007.

2007 starts going up quite fast now and well into September, which is a sign of the heavy compaction it experienced at the prolonged end of the melting season.

So, if we'd want to see more records, CAPIE will have to start going up. That record is already under the belt. :-B

Otherwise we get the same thing as last year, an artificially high minimum extent.

I'll post a new CAPIE map on Saturday.

William Crump

Cryosphere today currently shows about 3.0 million km2 for total northern hemisphere ice area with 2.5 million km2 of the ice in the Arctic Basin. This region makes up more than 80% of the ice which remains at the September minimum.

http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/recent365.anom.region.1.html

2011 ice area in the Arctic Basin is not significantly different from 2010.

Are there any direct measurements of volume for this region which would support the claims of an "ice free" Arctic by 2013?

Generally, this region starts to increase in area before the minimum is reached as melting in other regions outstrips the growth of ice in this region.

The NSIDC MAISE chart for the Central Arctic Basin reported greater area figures for 2011 than 2007 until August 14th. The MAISE graph shows that 2007 had an upward spike of 60,000 km2 after August 16th that 2011 has not matched. As of August 23rd the region was about 60,000 km2 below 2007, however the region has been increasing in area since August 21st.

The rate of decline in these graphs does not support an "ice free by 2013" claim,.

ftp://sidads.colorado.edu/DATASETS/NOAA/G02186/plots/r11_Central_Arctic_ts.png

Neven

Are there any direct measurements of volume for this region which would support the claims of an "ice free" Arctic by 2013?

They measured 0.9 meters in most of the ice around the North Pole. In 2001 they measured 2 meters.

I agree the Arctic Basin is surrounded by a nice buffer zone. It isn't irrational to think however that ice thickness in this area has been decreasing as well, while at the same time the buffer zone on average disappears a bit sooner as melting seasons come and go. This means warm water and air get to nibble at the thinning central ice pack a bit sooner ever year, and winds get to push it around much more easily, which is something we have witnessed last year, and to a lesser extent this year as well.

Ice-free in 2013? Perhaps not (it's the lower bound of Maslowski's prediction of 2016, plus or minus 3 years). But unless some negative feedback or strong natural variability kicks in, an ice-free summer around 2030 wouldn't be an extreme forecast. And this, of course, would be 70 years sooner than the IPCC currently has it. And that would be extreme.

Lucia (The Blackboard)

crandles
I certainly believe the uncertainty is higher than .175 as the errors have been growing.
Are you providing confidence intervals or predictions intervals?
My confidence intervals are larger than 0.175, and my predictions intervals are, of course, wider still.

Part of the reason my intervals are wider is that I use my weighted method, which accounts for the fact that even the model with the highest AICc coefficient might not be the best of the models I've considered, but even without that, I get wider confidence intervals. (That said: I think we aren't drawing from the same candidate models. )

maltose

CT Area now beats 2010...

2010.6959 -1.6625719 3.0774183 minimum
2011.6438 -1.9939168 3.0650997

Seke Rob

2011 Has been ahead of all previous years, including 2007 on many metrics.(For Spencer fans), on first 235 days of this year, we have a negative anomaly [See CT YTD anomaly chart 1979-2011] of -1.188 million km square as at the 23rd, where 2007 is at -1.118 million KmSq. Latter is slowly gaining, but not today: 2011 increased the differential to -70KKmSq average.

Breaking out the fitted JAXA curve from the chart above, extent is also below all previous years in anomaly for first 236 days, 1 day ahead of CT, base 2003-2010) [See JAXA Anom Chart], again 2007 gaining, since it's still 209KKmSq lower than 2007.

If the JAXA Area chart, no numbers available, has a tail (pun) to tell, I'd not be surprised if their extent will align soon, very soon. Panic at 21 Jump Street, the block on the opposite side :D

It will be skimming the edge, so the data spins, but holding on dropping to below 5 million extent on the 26th. That would then have been in 17 days for a 1M decline, ex aequo with the 17 days of 2007 and 2008, where it took 2010 24 days to get from 6M to 5M.

Bob Wallace

"They measured 0.9 meters in most of the ice around the North Pole. In 2001 they measured 2 meters."

If we take those two numbers as accurate average ice thickness then a simple-minded projection would suggest a 0 meter thickness in less than 10 years.

But the decline will almost certainly not be straight-lined. The warming season is starting a little earlier. The season-start extent is a little smaller (meaning less albedo and more exposed water).

Air temperature is continuing to rise. If we have continued breakdown of the circulation patterns which in the past have held in Arctic cold area then more cold is going to get transported out of the area.

I would assume soot will continue to increase, especially with more high altitude forest fires and now tundra fires beginning to appear.

And less first of year volume means that it takes less heat to melt ice thus leaving more heat to warm water which will fight fall freezing, holding down following year volume.

All of these factors work to amplify each other.

I'm starting to look at 2030 as an unlikely late date for the first total melt-out. Do I see it happening in the 2013-2016 frame? No, but if it happens I won't be overly surprised.

Stevemosher.wordpress.com

But unless some negative feedback or strong natural variability kicks in, an ice-free summer around 2030 wouldn't be an extreme forecast.

###

That's no forecast whatsoever neven.
By ice free do you mean no ice whatsoever.
By ice free summer do you mean ice free for the entire summer, or just the end of summer.

Plus by 2030, The best we can say is that temperatures will be maybe .4C higher than today. The range of estimates from a GCM might be from 0C to .8C. And the increase wont be monotonic ( as the last 10 years should have shown us).

Making decadel predictions about the climate is tough. Making them about the arctic where so much depends on non climatic factors seems a bit unwise. Although with everybody making all manner of short term predictions somebody will turn out correct. here is mine

"But unless some positive feedback or strong natural variability kicks in, an ice-free summer around 2030 is virtually impossible."

I Ballantinegray1

Hi Bob!
I think that the 'Perfect storm' synoptics will be the thing to leave us with a 'seasonal' pack (sub 1 million sq km.) After 07' we were told that such unique synoptics came around every 10 to 20 years (the two before 07' having the 10yr gap)so we may well see a seasonal pack before 2020. That said we can see what low ice conditions do to synoptics across the pole over late Autumn/early winter (and the bother they bring us here!!!) so maybe the rapid thinning of the pack and more Northerly storm tracks fetching higher swells into the basin adds another type of destructive synoptic to the mix?
My concerns are based around how easily 'average' summers come close to challenging the exceptional 07' melt year!

crandles

Lucia,

The .175 is a single (not 2) standard error (SE as reported by Excel). It is definitely too low as I tuned the weighting of my weighted average of extent and area (.74 weighting for area vs 1 weighting for extent (so extent is better than area by 22 Aug)) to minimize this standard error. This is bound to be overtuning to some degree.

I also found that predicting the drop using this weighted average performed better than using (extent-area) as the predictor.

A credible error is certainly going to be wider than this standard error which assumes the model used is perfect.

Patrice Pustavrh

Steve Mosher, if you are regarding whole summer (july, aug, september) there is really small chance that this will happen by 2030. But most predictions goes around minimum or at least september "virtualy" ice free state and "summer ice free" generaly refers to this semantics.
So, when someone says to me about ice free Arctic, I always do think that some years minimum extent will fall below 1 mio km2, and most of people discussing this topic do think somehow the same. Yes, you are right in sense that we must be more careful and use excact numbers in predictions, but, just remember, for the sake of debate always putting in numbers is somehow uncomfortable. But, I'll do some caveat here: First summer ice free Arctic will mean something different for ice albedo feedback than Steve's definition - if we are losing most of the ice till end of July, albedo feedback will be much stronger than if we lose it in September.
But, anyway, the first really important point is to see end of melt season ice free Arctic.

Neven

Steven, ice-free for me means below 1 million square km. There will always be nooks and crannies where ice will manage to make it through the melting season, or some extremely thick ice due to compaction in the preceding year.

Initially, I'm looking at one day or a couple of days towards the end of the melting season. Of course, in the period leading up to that sea ice extent/area will also be smaller than it is now.

Are you sure that if AGW continues unabated temps will only go up 0.4 C? Isn't the warming supposed to go faster in the northern latitudes? You know about Arctic amplification, right?

I'm basing what I say on what I have seen so far. Of course, things could switch, but they haven't as of yet, and I wonder how they will. It will have to involve some serious declining of ocean heat flux.

edit: What Patrice said. We discussed this many times before here, Steven. I think most people agree on what is meant by 'ice-free summer'. It's shorthand for 'below 1 million square km at the end of the melting season'. Like Patrice says, that's the important first step.

crandles

Area of 3.065 m km^2 is only .165 from record. A look at the graph page with look of areas per Uni Bremen suggests to me further declines in seas of:
Beauford 20k
East Siberian 60k
Laptev 10k
Kara 10k
Barents 10k
Greenland 30k though depends how much comes in
Hudson 10k
archipelago expext transport in of as much as melts

This is almost as much as needed for a record without considering the Arctic basin which is much more difficult to estimate how much will go.

Anyway, a record looks virtually assured with the 95k decline seen with last data.

The 150k area declines in those seas should probably be expected to reduce extent by at least 300k. That gets us under extent of 4.8 m km^2 before considering arctic basin extent losses.

Does that sound reasonable or miles off? Does this regional approach get us anywhere or are we better off with statistical estimation methods?

Ian Allen

Neven,
I for one don't believe in those nooks and crannies, especially on the American side without the vast freshwater estuaries and so much freakish subsea permafrost. Even now there are only a select few places where the earth is so cold in its bones that ice lingers eg. Foxe Basin.
But the coldness partly stems from the ice joining the world's two biggest islands/continents. I think more people should look at those charts of insolation at the edge of the atmosphere, where the N pole beats the equator in summer. Once the sea ever warms up and hence fogs up in winter, mildness will spread.
On the air conditioning thing, humidity is our friend.
I lived on the Costa Rican Carribean, where it hit the newspapers when they recorded 32C at Manzanillo (sea-level). The deep layer of wet air which would envelop an actual arctic ocean would do something to condition our air, and help stop the low, low humidities necessary for extreme ground temps, such as we got in UK and Ireland recently. If the world's winds wetten, that will definitely help trim the excesses at both ends of the scale.

Espen

The just updated Bremen map shows ice evaporating all around the clock, check it out, amazing stuff, and Healy is at 80 North parallel with not much ice around her:

http://www.iup.uni-bremen.de:8084/amsr/arctic_AMSRE_nic.png

http://mgds.ldeo.columbia.edu/healy/reports/aloftcon/2011/20110825-1701.jpeg

Regards Espen

Neven

Thanks, Espen. That's quite a bit of ice disappearing in the Beaufort Region and in the Bite that the warm waters are taking out of the ice pack in the Laptev Region.

I still find it difficult to translate that into a potential extent decrease number for tomorrow (and we have to await the revised sea ice concentration map), but if it stays like this, I would guess more than 50K.

BTW, we have a couple of small ice packs that have broken away from the main pack, although they are close to melting out. Not the huge chunks we speculated about last year, but still...

Bob Wallace

I've been checking the Healy pictures and trying to locate their position on the Bremen maps. Do wish there was a way to tag their shots with the observed ice concentration for that position. There's a lot of open water around those hunks of floating ice and they cruise for some time with almost no ice in the water around them.

Anyone know where one might see similar pictures from the Polarstern? Up around 90 are they shoving their way through a solid ice sheet or also just pushing floes out of their way? How much of the remaining ice is what we used to think of as "the frozen north" and how much is chunks floating in the punch bowl?

Espen

Another figure to add to the statistic guys
only +/- 42% of the Polar sea ice is now above the 80 North parallel.
Regards Espen

Daniel Bailey

Apologies, Neven, for this one will go OT a bit. But when one sees known disinformationists plying their craft one must stand up for science.

SM's word-games with the definition of "ice free" having already been dealt with, let's examine the rest of his crafted agenda:

    Plus by 2030, The best we can say is that temperatures will be maybe .4C higher than today.

What is intended is for the casual reader to take away the message that "there's no problem". Well the real problem is that SM uses a global temperature and passes it off as also applying to the Arctic, with no regard to Arctic Amplification.

Let's look at the actual Artic temperature rises over just the past 20 years:

[Source]

Well over 1 degree C; given that, it is reasonable to expect another full degree C over the next 20 years in just the Arctic.

So under the existing temperature rise, ice thickness at the North Pole has dwindled from 2 meters to less than 1 (in just the past 10 years)...and SM says this:

    "But unless some positive feedback or strong natural variability kicks in, an ice-free summer around 2030 is virtually impossible."

Hmmm, last 10 years, more than 1 meter thickness lost, less than 1 meter thickness left...and over the next 20 years, SM says an ice-free state is "impossible"?

Can you say "Faster-than-linear?"

Regular readers of climate blogs will recognize what is being purveyed here by the likes of SM by its heady aroma...

Frivolousz21

UB prelim out with some strong compaction all over the place.

Looks like another 50-70km2 loss extent day coming.


12z GFS is out and it would be much less condusive for rapid ice loss next week with a different wind pattern. Will see how the 12z Euro plays out.

Espen

I am sure if we had a reliable figure of Sea ice concentration (not extend/area), we would be well below 2007.
Regards Espen

Philiponfire

Bob Wallace,

If you take 80 N 165 E ( on picture is a W but that I think is a typo) from the picture and apply that to the map you will be right where the ship is give or take.

This being the case my mind went F*&^!
As in real world there is very little ice where the map colours indicate 60 or 70%.
Well that ice isn't worth a damn it is thin most of it cannot be more than a meter thick.

Espen

Frivolousz21: I would not call it compaction but evaporation, because it is simply disappearing.
Regards Espen

Philiponfire

Espen I may be misinterpreting what I am seeing but I see vast areas of compaction and stacking taking place in the laptev sea.

http://lance-modis.eosdis.nasa.gov/imagery/subsets/?subset=Arctic_r04c04.2011237.terra.250m.jpg

Paul Klemencic

My take on the preliminary Bremen map:

First off, the entire eastern side slid toward Eurasia fairly significantly since yesterday, but all along the edge, ice is melting out. In the last four days, the ice edge has tried to slide over 40 km toward the Laptev, the Kara, and toward Franz Josef Land and Svalbard, as evidenced by the movement of portions or protuberances that were able to move that far. But many places in between these, mostly melted out below 15%, and the edge didn't move as far.

The ice pack moving toward Svalbard and the Fram strait didn't get very far at all. The movement of ice already in the Fram, shows the pack moving through the Fram, but on the edge facing Svalbard, the ice is melting as fast, or even faster than it moves in. The map blink comparison shows the Fram actually opening up, even as ice moves in! The same thing is happening north of Svalbard. Weak areas of ice below 70% (green areas) are also showing up and deepening into the pack north of both Svalbard and Franz Josef Land.

Summary of the east side below Severnaya Zemlya: the ice is moving into these regions fairly fast, but melting out even faster in many places.

The hole in the quadrangle between 105E and 135E is opening up, and within 72 hours, there should be open seas with less that 15% ice all the way from the Laptev to the 85N parallel. This is unprecedented in August! Only in 2007 did open seas get to 85N, and that was at the minimum ice extent day in late September.

The ice pack left in the E. Siberian region receded significantly, although some of this may fill back in the final Bremen map today. The remaining concentrated ice in the E. Siberian (lavender and red) is now surrounded by a patchwork of yellow and green lower concentration ice. (Much of this ice will likely melt out by the end of the season.)

The Chukchi doesn't have complete satellite image in the preliminary, but the Healy is at 80N and 165W, and the photos today from the Healy show a lot of open water. Yesterday's map also showed a hole of open water at 80N and 180E, with weak ice intruding into the pack north along 180E. It appears another area of open sea is about to push into the central Arctic Basin at this location.

The Beaufort region saw a big pullback of the ice pack boundary, coupled with large areas of weaker ice pack (green = less than 70%) showing up, and in particular the area next to the Canadian Archipelago.

In summary, the ice reports from all regions are terrible. Sizable extent losses today, followed by new record open sea events in the next several days.

Espen

And what we now see in the Laptev sea, we will soon have an ice free area above 85 North parallel, well we are getting closer to the point?

Regards Espen

Bob Wallace

Phillip, that's what I do but I want it more automatic/objective. I don't always trust my eye.

Like you I look at the map and see it saying that there's >50% ice, then I look at the actual images and realize that ">50% ice" does not look like what I expect when I think about North Pole ice.

We could use a marker prior to "summer melt-out". An identified point at which there is no "massive ice sheet" but only floating remains. At that point Arctic ice has died and all we're watching is to see how long it takes for the corpse to rot away....

Neven

Most excellent update, Paul. I think I'm going to have to thank you every day if this keeps up. I've gotten watching and comparing the UB sea ice concentration maps in my daily routine now as well, and your descriptions are in my opinion 100% accurate.

Anu

Note the unusual melt up to 85°N in the Laptev Sea (from 135°E to past 120°E) in the preliminary Bremen ice concentration map:
http://iup.physik.uni-bremen.de:8084/amsr/arctic_AMSRE_visual.png

Now look how this is exactly over the seabed trench (and surrounding ridges):
http://www.ngdc.noaa.gov/mgg/image/ibca_gebco_comp_cover.jpg
I have another map that shows that the deepest part of this trench is more than 5000 meters below sea level. For comparison, much of the Laptev Sea near Russia is 10, 25 or 50 meters deep.


I wonder if the unusual seabed feature is affecting the thermohaline inversion (warm, deep water currents getting deflected towards vertical flow), or if this is causing extra-severe surface waves (breaking up the thin ice, exposing more surface to water).

Whatever the process, I suspect it's more than a coincidence that this high-latitude melt is located over the deepest trench in the Arctic.

Xandra

Posted by: Daniel Bailey, August 25, 2011 at 19:29

”But when one sees known disinformationists plying their craft one must stand up for science”

Very good post. Thanks Daniel!

Bob Wallace

Paul, off the coast of Greenland, between 75 and 80 degrees, where does the extra purple come from when you flip between 8/23 and today's map?

It isn't coming in from up north, I would think. The long stretch of open water is moving south a little, but not that much. The amount/concentration of ice along the coast seems to hold about steady.

Neven

Anu, I believe the ice retreated quite far over there very early in the season, as air temps were very high and there was lots of insolation as well. I could be wrong, but this seems to be in large part albedo-induced warmth.

Espen

I'm sure we all on this blog sees "a turning point at the moment" and before it gets on the front page of the world's media. It is with fear and sadness on my part!
Regards Espen

Paul Klemencic

Daniel Bailey: thanks to posting a rebuttal to Steve Mosher; he can twist information into incredible pretzels.

His comment here could be nominated for the Golden Horseshoe Award in 2011, but would lose to Joe Bastardi for a wonderfully constructed misconstrued proclamation. It took Tamino an entire post to partly correct one paragraph.

And what is a Golden Horseshoe Award? Here is a clue:
--------------
In Dashiell Hammett’s story The Golden Horseshoe much of the action takes place in a bar of that name in Tijuana. At one point the narrator, an operative for the Continental Detective Agency, kills some time by studying the decorations:

I was reading a sign high on the wall behind the bar:

ONLY GENUINE PRE-WAR AMERICAN AND BRITISH WHISKEYS SERVED HERE

I was trying to count how many lies could be found in those nine words, and had reached four, with promise of more …

Note: Credit for the Golden Horseshoe comparison goes to Eli Rabett at the Rabett Run site, who often links and directs people like me to Neven's site.

Stevemosher.wordpress.com

"Steve Mosher, if you are regarding whole summer (july, aug, september) there is really small chance that this will happen by 2030. But most predictions goes around minimum or at least september "virtualy" ice free state and "summer ice free" generaly refers to this semantics.
So, when someone says to me about ice free Arctic, I always do think that some years minimum extent will fall below 1 mio km2, and most of people discussing this topic do think somehow the same. Yes, you are right in sense that we must be more careful and use excact numbers in predictions, but, just remember, for the sake of debate always putting in numbers is somehow uncomfortable. But, I'll do some caveat here: First summer ice free Arctic will mean something different for ice albedo feedback than Steve's definition - if we are losing most of the ice till end of July, albedo feedback will be much stronger than if we lose it in September.
But, anyway, the first really important point is to see end of melt season ice free Arctic."

I appreciate the clarification. As one of the few people ( R Gates is the other ) who devotes some time to correcting the misperceptions of skeptics, my job is made easier if I don't have to 'explain' what people mean when they say 'ice free'. I also think it's important to avoid making scientific 'icons' of things. Yes, it's meaningful and makes good copy, but you also may create unintended consequences.

Paul Klemencic

Bob Wallace wrote: Paul, off the coast of Greenland, between 75 and 80 degrees, where does the extra purple come from when you flip between 8/23 and today's map? ... It isn't coming in from up north, I would think.

Bob, I generally ignore colors of red, lavender, ,and purple, because they all involve concentrations of ice 80-90% plus, and melt pond refreeze, compaction, etc. swing these colors dramatically day-to-day.

I also generally ignore yellow (up t0 80%) for the same reasons, unless associated with nearby green (less than 70%) areas. Once green areas show up, they generally persist, even if disappearing for a day or two due to compaction, they usually reappear a few days later. So the green areas show the weak ice pack areas that are emerging.

The area you point to, concentrations above 70-80% can swing dramatically day to day, and generally don't mean very much.

Stevemosher.wordpress.com

Actually Daniel you missed the point.

the point of crafting a prediction of that said
"VIRTUALLY, impossible, ICE FREE, SUMMER."

Was to draw attention to the way in which ambiguity can be used both ways. Hence, I caution against using ambiguous statements like " virtually certain we will have ice free summers" The point of the exercise was to demonstrate how ambiguity and unstated assumptions can lead to all sorts of trouble.
So, for example. If, in 2030 we had 10ksqkm of ice left in september, a weasel could say.. 10K is not ice free. or what if we had 0K ice, A weasel could say.."I meant the WHOLE summer, from june 21 on." The point is this. I'm not making a prediction. I'm making a point about predictions. If you want to be scientific about things you need to make predictions that are testable. So, that means using numbers. When I see no numbers, I know rhetoric is in action. When a skeptic says "the next 10 years will cool" I want numbers, otherwise he is just engaging in rhetoric.

Further, you should be aware that the best science we have cannot rule out cooling over short periods. It's best to take notice of that when making definitive statements. Or we end up in the situation where we are now.

Let me put it another way. Our knowledge of how polar ice will evolve in a warming regime is less certain that our knowledge that it will warm. It makes no epistemic sense to focus one's arguments and ones credibility on the issues that are less well known than on those that are more well known.

Seke Rob

MASIE [See Chart with detail legend] dropped a century+ from the 23rd to the 24th... exact -103609.90 KmSq, with the biggest shakeup in CA -30862,54 (thought it curious for the Russian map to have the main gate solidly closed... lest I'm daltonic too). Essentially all areas that had sea ice before saw a decline.

ttyl

Neven

So, for example. If, in 2030 we had 10ksqkm of ice left in september, a weasel could say.. 10K is not ice free. or what if we had 0K ice, A weasel could say.."I meant the WHOLE summer, from june 21 on."

Thanks for letting us know what the weasels will say, REGARDLESS of what 'we' say or do.

Our knowledge of how polar ice will evolve in a warming regime is less certain that our knowledge that it will warm. It makes no epistemic sense to focus one's arguments and ones credibility on the issues that are less well known than on those that are more well known.

The uncertainty makes it worse, not better. If you find uncertainty in this context reassuring, I'm happy for you.

---

Look, Steven Mosher, even though you're not my best friend, I don't in principle mind you commenting here. I'd rather not have you comment, because people are going to jump on you based on your reputation and this will spoil the good and friendly atmosphere here. I won't let that happen.

It could be a coincidence that after almost a year and a half of relative calm just at this very moment the first lukewarmers and below start to make their appearance here, but I'm making one thing very clear right now: I'm not going to let this blog and community be poisoned by discussions on semantics, PR tactics and credibility by people who cannot handle what is seemingly happening in the Arctic right now.

I will block and ban immediately, as there is a whole parallel universe out there for people who don't like this particular echo chamber. I like it and I want it to stay that way. I'm not going to moderate or berate. I don't have the time, patience or philosophy for it.

Let the ice do the talking.

Paul Klemencic

More info for Bob Wallace (and taking a break from semantics lectures):

Bob, I know you are relatively new to the game of observing the ice pack through the melt season, so I thought I might explain what we should be seeing from the pack, if this was a normal season.

Normally ice in the pack is pushed toward what arctic ice scientists call the 'switchyard' region of ice pack north of Ellesmere Island in the Canadian Archipelago. Most of the ice is forced to change directions, and some moves toward the Beaufort, and some moves toward the Fram and the Greenland Sea.

At this point in the season, the ice pack should be moving down into the Beaufort region (the region consisting of the Beaufort Sea and up to the 80N parallel), and covering the mouth of the Parry Channel (the most direct route of the Northwest Channel). Typically ice extent in the Beaufort region should be increasing as high concentration (over 90%) ice pack is pushed into the Beaufort.

So what is happening this year? The ice pack not only isn't covering the mouth of the Parry Channel, its receding with ice extent in the Beaufort region declining. The recent storm damaged the ice pack resulting in the overnight 'flash melt' of over 100k sq km of ice extent in the Beaufort and Chukchi regions (the other regions gained extent on that day). Not only did the storm take out a lot of extent, but it apparently weakened the ice pack left in the Beaufort and Chukchi. As a result, the ice edge is moving away from the Parry Channel, and weak ice in the Beaufort extends all the way north to at least the 78N parallel, and seems to be moving toward above 80N.

If this area of ice melts out, then the mouth of the Parry Channel won't be closed with ice until October! The direct route of the NW passage could be open for another six weeks! Unprecedented.

continued in next comment...

Newcrusader

On the rise of global temperatures since the pre industrial era. We have risen 0.8 degrees C- we have reached the warmest part of the Holocene. In 10 years- that rise will be around 1 degrees. This is globally by the way- the arctic has warmed more.

I read climate progress edited by Joe Romm, and also John Cooks Skeptical science- they both feel we should see an 'ice free arctic' under the perimeters spoken here by Neven by 2020.

C02 now at 391ppm (now near the lows of the year) will likely reach 397ppm in mid Spring 2012.

By 2020- C02 at 410ppm? or perhaps more. We are now the highest in 15 million years- remember ice did not begin to form in the arctic in winter till 12 million years ago.

C02 will breach 450ppm in 2030-2035? This point when reached, according to Dr. Hansen; we will reach A major tipping point- we will not be able to stop a 3 degrees rise or more in global temperatures by mid century.

Hansen, Mann among others have really refined their paleo climate data with ice and CO2 levels. The melting we are seeing now is very close to what has been found in ancient climates of the past.

Remember also in the Eemian interglacial 125,000 yrs ago, when global temperatures where perhaps .0.2 degrees warmer then today- sea levels where 4-6 meters higher (this extra warming due mostly to a change in the earths tilt).

What we are seeing today fits almost perfectly into the past with warmth and ice in the arctic- we are fools to let C02 rise much higher then today.

Paul Klemencic

Ok, then, the Beaufort should be getting ice moving in (the Beaufort Gyre), but isn't. And the ice in Beaufort is melting out at high rates, and one day in the last week, an extraordinary rate.

What about the rest of the pack? The ice on the other side of the switchyard should be moving into the Fram Strait, and choking it solid with ice. This year that isn't happening either. The ice is moving slower than normal toward the Fram and Svalbard, but most melts out before it reaches the Fram or stacks up against Svalbard. The normal 'choke point' (my terminology) at the mouth of the Fram is gone. There is a huge amount of open sea (with relatively warmer waters) that the ice pack could 'flush' into. The potential exists for a massive and unprecedented ice flush on this side of the pack.

Now look at the shape of the ice pack, and compare to the pack at this point in 2007. The pack in 2007 was still firmly buttressed against Siberia on one side, and ice was already pushing into the Beaufort and toward the mouth of the Parry Channel. However in 2007, the E. Siberian region ice was mostly gone; whereas this year there is still a large extent of ice (about 580k sq km more than '07) in the E. Siberian. However, most of that ice is weak, and over 80% of the 580k difference should melt out this year.

This leaves the ice in the central Arctic basin. Which way will it move? There is plenty of room to take ice toward the Fram and Svalbard; and there is (or soon will be) plenty of room to take ice in the Beaufort. Open sea areas are eating into the central Arctic Basin from the eastside from Kara to Svalbard, most worrying, from the Laptev (already at 85N), and now from the Chukchi.

In short, we now see a weakened Arctic ice pack, precariously balanced as though on top of a ladder. The remaining concentrated ice in the E. Siberian is too far away to fill in any central basin gaps, and the central Arctic Basin ice pack can collapse to either the eastside, or toward the Beaufort.

Not a happy situation for planet Earth.

Lucia (The Blackboard)

Crandles--
I wasn't suggesting yours were wrong. I'm trying to figure out how to relate mine to yours. I know we aren't doing the fits the same way, so I don't expect exact matches, but we are also doing similar stuff, so I was curious about which features were included in yours.

The .175 is a single (not 2) standard error (SE as reported by Excel).
The 'SE as reported by Excel' is part of the answer explains why mine for similar sorts of single regressions are larger.

A credible error is certainly going to be wider than this standard error which assumes the model used is perfect.
Yes. Mine are larger because I take into account to factors I know how to quantify: 1) The uncertainty in the estimate of the fit parameters (i.e. 'm' and c ) in an regression of the form y=mx+c . This estimate at least assumed that functional form is ok.

2) the a portion in the uncertainty associated with identifying the type of model y=f(x).

Whether or not you were dealing with issue (1) is what I wondered about so I could see how your values happened to relate to mine.

I was pretty sure you weren't trying to do (2). I try to capture a portion of the error under (2) by looking at many candidate models weighting by AICc.

The comments to this entry are closed.