Accelerated melting.

[Ice Cool Kim]

The period upto 2007 had a lot of people rightly concerned about "accelerating melting" in the Arctic.

I thought it would be interesting to see how well that model fitted now.

I used NOAA ice extent, re-sampled to 10 day intervals (after suitable anti-alias filtering) and plotted the anomalies from the mean for each 10 day point, based on the whole record.

Accelerating melting would produce a parabola, so I use non-linear least squares to fit such a model using three different periods for the fitting: the whole record ; only 1998-2007 and 1998 onwards.

the 1998-2007 fit produced a strong parabola that was quite a close fit to the data over that period. It would be quite reasonable and accurate to describe that period as "accelerating melting".

The whole record also returns a down turned parabola but of much shallower curvature and the fit is really not that good. Any number of other models could equally well fit the data.

The 1998 onwards fit however, was so gentle it was almost linear, ie virtually no acceleration but a constant decline. The alarming period around 2007 is seen as a notable deviation from this fit, which again is less good overall.

It would seem clear from this that there has been a significant improvement since 2007. Though there is still a steady decline since 1997, the earlier strong acceleration in the melting has ended.

That must evidence either that a negative feedback is in operation since 2007 that has overpowered the positive feedback that was characterising the 1997-2007 period or that whatever was the cause of the melting has eased off.  Possibly a combination of the two.

Looks like a ray of hope for those of us who are not too keen to see run-away climate change.

Others will be disappointed of course.

[werther]

Present this to Tamino....

[Ice Cool Kim]

Present this to Tamino....

Is that an order, a request or an unhelpful suggestion?

Why would I want to do that? I have no interest in his opinion on anything and don't wish pollute any discussion here since my estimation of him would likely cause animosity which Neven would like to avoid. I'm inclined to agree with that approach.


Same thing for CT ice  area.

[anonymous]

Since it's about melting, which reason allows to ignore volume or do you suggest a constant thickness since 1979?

[werther]

Coincidence… Tamino ’s got a comparable graph on his blog…
I combined the ICK graph with it on CAD:

There’s something wrong … the SIE minima don’t fit.
Maybe others can pinpoint which graph is right.
I don’t mind someone being cautious on the Forum… but be sure you can produce good quotes, illustrations and their sources…

[Ice Cool Kim]

which reason allows to ignore volume

What allows me to look at area/extent is what allows climate scientists to look at area/extent. It's data rather than model output.

If they've been launching satellites to measure ice area, it's not for nothing.

[Ice Cool Kim]

There’s something wrong … the SIE minima don’t fit.

I don't see any significant difference , what's your point?

[anonymous]

which reason allows to ignore volume

What allows me to look at area/extent is what allows climate scientists to look at area/extent. It's data rather than model output.

If they've been launching satellites to measure ice area, it's not for nothing.

Sure, it is just scientists then don't call it melting...

[Ice Cool Kim]

There’s something wrong … the SIE minima don’t fit.
Maybe others can pinpoint which graph is right.
I don’t mind someone being cautious on the Forum… but be sure you can produce good quotes, illustrations and their sources…

Don't do "illustrations" , I'm a scientist not an artist.

My graph has the source of the data included.

The small difference in the detail is explained if you read the text that I posted with the headline graph. I used proper data processing techniques. Mr. T used monthly averages, which though commonly used do not properly subsample the data.

Problem 1: "months" are 29.29.30 or 31 days long in no logical order.
Problem 2: averaging is good for removing normal/gaussian noise. It is not a valid means for removing periodic signals. Sub-sampling data with periodic or pseudo periodic component without proper anti-alias filtering WILL lead to spurious periodic signals of much longer period appearing in the data.  That is DSP 101.


If you were expecting the two graphs to superpose identically it is because you had not read either my description or Tammy's.

I don't see any indication from your overlay, that doing similar function fit either dataset would produce a significant difference.

[Neven]

The period upto 2007 had a lot of people rightly concerned about "accelerating melting" in the Arctic.

A lot of people were and still are worried about the Arctic becoming ice-free in September, and then becoming ice-free earlier and earlier in the year, because there's a risk that this process could involve some nasty consequences. A lot of people don't like risks, even if this doesn't concern them directly.

Whether this process comes about through "accelerated melting" or just a plain old linear decrease, is irrelevant for all practical purposes.

I think it'd be great if you can disappoint some of us by showing that the Arctic is a lot more stable than it currently seems, that it will not become free of sea ice in the foreseeable future, and even if it does, there is zero risk involved for humans and their societies.

If you can't disappoint us on this score, then please disappoint others by explaining to them that business-as-usual is no longer an option and that all of us, everyone, need to come up with solutions and implement them as much as we can in our personal lives.

[Ice Cool Kim]

Sure, it is just scientists then don't call it melting...

Call it sublimation if you like. When ice "disappears" between March and September, most people refer to this as the melting season.

Whatever the ice is doing it is doing it more slowly than before rather than accelerating.

Is there anything about that rather good news that disturbs you ?

[werther]

I’ve heard the Dutch Under-secretary of State today drowning in statistical and numerical nonsense over the death of a Russian refugee-applier (RIP Aleksandr Dolmatov) while imprisoned to be repatriated.
I’m not interested getting into likewise nitty-gritty over data interpretation.
You have it your way and be happy with that.

[Ice Cool Kim]

Whether this process comes about through "accelerated melting" or just a plain old linear decrease, is irrelevant for all practical purposes.

I realise your attitude is based on concern but this is not irrelevant.

In physical terms there is a very important difference between a +ve feedback (run away process, "tipping point" , etc)  and a linear decline.

It is also very important when an accelerating process is reduced to a linear one by a later, flatter section.

It's a lot like temperatures. In the 90s temps were going up and getting steeper. That was worrying. After 2000 they flattened off, that is less worrying.

Hopefully most people by this stage will have stopped drawing straight lines or extrapolating exponentials way outside the range of the known data.

We could all wish for less pollution in our food, water and air we breathe, but we're not going to get there while we allow ourselves to be distracted by someone saying "look up there".

Anyhow, without getting too philosophical, perhaps we should restrict this to what we can glean from the satellite data in question , otherwise we're in for a long night

[Ice Cool Kim]

You have it your way and be happy with that.

I'd be happy to look at it "your way" if you actually had a relevant point to make. I think I have explained the minor difference that you highlighted and suggest it is immaterial to the point of this post. If you have a concrete point to make, I'm interested.

Bad news about the asylum seeker, seriously. Lots of people died in Waco too, for no good reason and the USAF/RAF probably bombed another wedding party in Afghanistan. Sadly the world is more concerned with the possibility of what may or may not happen in 50 years time than with what is happening and killing people today.

Oops, there I go getting philosophical again.

[Neven]

You have it your way and be happy with that.

Bad news about the asylum seeker, seriously. Lots of people died in Waco too, for no good reason and the USAF/RAF probably bombed another wedding party in Afghanistan. Sadly the world is more concerned with the possibility of what may or may not happen in 50 years time than with what is happening and killing people today.

We can't help it if folks in general can't handle more than one problem at a time. I'm sure most people here can and aren't saying: "Forget everything else, AGW, AGW, AGW, AGW, AGW, AGW, AGW."

But whether the Arctic becomes ice-free this decade, or in 2030-2040 because the melting keeps de-accelerating, is pretty much irrelevant, as long as the thing that causes it to melt, and causes the bombings of Afghan wedding parties, and causes a financial meltdown, and causes pollution in our food, water and air we breathe, continues.

My hope is that when people acknowledge that there truly are systemic problems, they also see the systemicness of it and demand systemic changes, not only demand them, but do everything they can to implement them in their personal lives.

But it all starts with the acknowledgment that there are problems and that we can't point our fingers to no one but us. And that's what is making AGW (of which the Arctic is the front line) such an interesting philosophical happening.

You see, when we acknowledge that the bombing of Aghan wedding parties (bloody terrorists, all of them) is a problem, we can go: "Ah, it's the armies, the generals, the military-industrial complex. Not us."

Or the credit crisis: "It's those greedy bankers. Not us."

Or air and water pollution: "It's those stinking factories. Move 'em to China!"

But with AGW: "It's those bloody cars we... It's all that stuff that we... It's those long-haul holiday flights that we... Bloody hell, the enemy is us! But how do we change this? We're 100% dependent on the system to continue, and the system needs to grow infinitely or the ultra-rich won't get hyper-mega-rich! Hold on a minute, the system. Who devised the system? God? Moses coming down the mountain with 'Atlas Shrugged' above his head?"

And that's when it gets interesting. 

---

So, Arctic sea ice extent/area loss isn't accelerating? Oo wee, 2030 here we come!

I'm still getting ready to watch the 2013 melting season if you don't mind. 

[Laurent]

Mister Ice cool kim,

You may not be aware of it but there used to be some very long discussions about the relevancy of measuring ice extent (or ice area) on Neven's blog. Personally i have quit this discussions some years ago when I found the piomas volume.
Why because when the ice is melting the ice spread on the ocean and if you measure the surface it does increase but the volume is going down (there is some other reasons but i am not arguing on sea ice extent, sorry).
The volume trend is here :
https://14adebb0-a-62cb3a1a-s-sites.googlegroups.com/site/arctischepinguin/home/piomas/grf/piomas-trnd6.png?attachauth=ANoY7cqLlo0TTxRTkXbO4tyKW4NmkXKTHvEHJjfVHIUCWrWzHoAqHffLZkN7wQtdUM7qKrZKKRYYX9LTU3bPSdQrMBv_qae_hIcXwUe-Rl9J5Db_st48teM7qjGN0dLnDO6HhD2dXr4EprRNt5WGWuWi4DcvedL3r-ciKxhIGMF6R0Ej-GBdUaqrFyB4bYX6RRk3wo0fdNXosippdji7pKk73cIxMTHw33tIf_BlYcRBrnKOZup5zDsgPjOt4LDD80XDVqe8gHdO&attredirects=0

[ChrisReynolds]

  

[Ice Cool Kim]

I'm still getting ready to watch the 2013 melting season if you don't mind. 

I don't mind at all, I'll be watching closely too, though I doubt one more dot will tell us much more, it may give another clue.

I also fully expect the usual jamboree in September when we forget the other 364 days of data and pretend the way the wind blows that week determined the future of the ecosystem.

It will be interesting to see whether the melting season continues to shorten this year or not. will this translate into any recovery in ice?

In the meantime I have pulled a couple of interesting results out the existing data by using all of it and applying standard signal processing.  There is probably more to be had.

[Neven]

Great. Everybody happy. More or less. 

[Ice Cool Kim]

Personally i have quit this discussions some years ago when I found the piomas volume.

I've seen too much of all this modelling to have any confidence in it being done properly and with honest uncertainty assessments. Even the data we get have been tweaked and tuned with bias adjustments. But at least there's some audit trail.

Volume is a better measure of heat capacity and rate of change is the power in or out of the system. However, in the case of the Arctic, the area is where the feedbacks operate. So it is directly relevant.

PIOMAS is looking better but it's still got pretty huge %age errors.

I'm not ready to give up data for models just yet.

[Vergent]

Kim,

Sophisticated argument. The definition of an anomaly changes with time. The anomaly in 2012 is against an average that includes 2007 & 8. the anomaly in 2007 does not. When you look at the actual minimums, the acceleration is obvious.

V

[Ice Cool Kim]

The definition does not change but the data does.

If you prefer this plot separates the recent period and uses the 'climatology' of 2007-2013 as the basis for the anomaly of those years.

This better removes the annual cycles but I decided not to use it because it would be a bone of contention for those not happy with what it shows.

When you look at the actual minimums, the acceleration is obvious.

But why would you chose to effectively choose just one day a year when you have 365. What makes that day more "relevant" than any other?


That is known as selection bias.  I could invite you to only look at the actual maxima, would that be more accurate?

[crandles]

That must evidence either that a negative feedback is in operation since 2007 that has overpowered the positive feedback that was characterising the 1997-2007 period or that whatever was the cause of the melting has eased off.  Possibly a combination of the two.

Looks like a ray of hope for those of us who are not too keen to see run-away climate change.

Others will be disappointed of course.

I think that conclusion is a direct result of choosing a quadratic fit. If you used an exponential fit you would not arrive at this conclusion. The quadratic fits have an upslope which we know does not fit the data so the conclusion that extrapolating quadratic fits does not perform well is a trivial result.

It looks to me like you picked 1998 to make extrapolating the quadratics look particularly bad. You can of course choose dates to make extrapolating exponentials look bad.

Question is do you get some extrapolations on either side of reality?

Exponential fits

so I am not just using one type of fit to reach my conclusion. Here is same for Gompertz:

Gompertz fits

Seems like there isn't much drift in the prediction whether using exponential or gompertz. Maybe Gompertz is a bit hopeful and exponential a little aggressive, but this is rather different than the conclusion you reached from an inappropriate curve type and picking 1998 to make it look bad.

[Ice Cool Kim]

Here's the previous anomaly plot with 6 and 12 month filter applied.

I'm sure all can read that so I won't prejudice with any interpretation of my own.

[Ice Cool Kim]

@crandles

No 'extrapolation' intended in terms of projection or making things look bad. I could have cropped them at the limits of the data fitted but I wanted it to be as easy a possible to see the curvature to emphasise the magnitude of the acceleration.

None of them have any predictive value.

I chose 1998-2007 specifically because that gave a very decent fit to the quadratic ie. the idea of "accelerated melting" that was the subject. My aim was to say, yes, there was something that could accurately be called accelerated melting and indicating a positive feedback.

The other two were to show this model does not really fit that well outside that range and the same description is not applicable.

I did make the point that any number of other functions would fit as well out side of that period. Though you would need a reason to suggest a particular model to fit. Not just pulling functions out of a bag.

[Peter Ellis]

Averaging (or filtering or whatever the heck you want to do) across seasons doesn't really make a lot of sense.  If you look at the anomaly plots or even the raw data, the pattern is obvious.  Up to about 2007, both summer and winter area were declining, although the summer decline was more significant that the winter decline.  Since 2007, the summer decline has continued to accelerate, while the winter decline has slowed noticeably.

The reasons for this are fairly obvious to anyone who thinks about the physics of the situation (hint: it gets really cold in winter.  Ice is gonna form.)

[crandles]

None of them have any predictive value.

I did make the point that any number of other functions would fit as well out side of that period.

Yes you did point out other functions would fit.

However, despite being aware of this, not only did you reach the conclusion that I quoted you said it "must evidence ....". I think Wipneus's graphs show that conclusion does not follow if you used different more appropriate functions like exponential or gompertz.

Therefore it seems to me your conclusion did rely on the function being appropriate.

 

[Ice Cool Kim]

so what's your point? That freezing does count in the winter, but we must worry about melting in the summer?

Why doesn't filtering "make sense"?

Taking anomalies is a means of removing the annual cycle too in order to see the longer term pattern behind the dominant annual variations.

If you want to tell climatologists to stop using anomalies you have your work cut out.

I really don't see what your objection is about.

[Vergent]

The definition does not change but the data does.

If you prefer this plot separates the recent period and uses the 'climatology' of 2007-2013 as the basis for the anomaly of those years.

This better removes the annual cycles but I decided not to use it because it would be a bone of contention for those not happy with what it shows.

When you look at the actual minimums, the acceleration is obvious.

But why would you chose to effectively choose just one day a year when you have 365. What makes that day more "relevant" than any other?


That is known as selection bias.  I could invite you to only look at the actual maxima, would that be more accurate?

Kim,

You are obscuring September minimums among anomalies that happen other times of the year. There is nothing remotely accurate about that. It is adding a noislike signal to the data. You are going to great lengths to obscure the trend.



The decline in older ice as well as area shows acceleration. Loss of thickness can be inferred.



Accelerating loss of thickness can be seen in the observational data from submarines, ICESAT, Icebridge, and CRYOSAT. On the one hand you are obscuring data, on the other you are ignoring data. You are doing so to the end of making an "everything is honky-dory" argument.

If it looks like a duck, walks like a duck, and quacks like a duck, we have at least to consider the possibility that we have a small aquatic bird of the family anatidae on our hands. I personally have acquired a case of  Anatidaephobia.

V

[jdallen]

My turn to weigh in here.

"It would seem clear from this that there has been a significant improvement since 2007. Though there is still a steady decline since 1997, the earlier strong acceleration in the melting has ended.

That must evidence either that a negative feedback is in operation since 2007 that has overpowered the positive feedback that was characterizing the 1997-2007 period or that whatever was the cause of the melting has eased off.  Possibly a combination of the two."


Kim, I'm afraid I'm going to have to quibble with you somewhat on a number accounts here.

First, the correct conclusion in the first statement, should not be that the "earlier strong acceleration in melting has ended".  The only conclusion I can derive from this graph, is that for the period described, it appears to show the "strong accelerating decrease in extent" has ended; and even that can be questioned.

Secondly, the analysis evaluates "extent" as a monolithic, unified object, which it is not.  Year over year, even when extents were more similar in size, there would be differences in what *areas* were covered by the ice extent.  Certainly since 2007, there have been wide variations in what areas of the arctic were and were not covered at different junctures of the season. That's something even my relatively unpracticed layman's eye can determine anecdotally.

Thirdly, extent by itself is a poor measure of total melt on two separate accounts.  First, in as much as it treats all ice as equal - namely that it doesn't consider the identified different physical properties of first year ice versus older ice.  Second, in that it ignores ice volume.

I agree that direct evaluation of observed data trumps models; but for it to be really successful, that data needs to be complete, and the system as a whole considered both for composition and movement of energy.  The fact that also (Fourthly...) the dynamics of the system are inconsistent - insolation, cloud cover, oceanic currents, jet stream flow, ad nauseum - makes assertions based on just one dimension of the system very difficult to support.

I'm reminded some years back when I spent entirely TOO much time attempting to put together genetic algorithms to model stock market behavior in an effort to follow short term swings in stock movement (think minute over minute).  Not. Dooable.  At least, not with my limited time, expertise and the hardware processing cycles I had at the time.  As various people have said... the models worked well, except for when they didn't... which typically meant some input was missed.  I expect arctic climate is far more complex than the stock market, over all.

Your examination of data, and noting that the extent is behaving as it is, is worthwhile.  The next matter is to interpret what the behavior means, and why we see it.

[Ice Cool Kim]

 crandles

However, despite being aware of this, not only did you reach the conclusion that I quoted you said it "must evidence ....". I think Wipneus's graphs show that conclusion does not follow if you used different more appropriate functions like exponential or gompertz.

Therefore it seems to me your conclusion did rely on the function being appropriate.

I don't know whether you noticed but the plots you linked were PIOMAS not ice extent. Is that what you are referring to as Wipneus's graphs ?

Like I said you need to have a reason to fit a function, why is exponential  or gumpitz "more appropriate " ? Please explain.

"must evidence ...." may be contestable if you want to go into. I think it's reasonable but not proven beyond doubt.

Of course my conclusion depended on the function fit. That was the whole point of the the thread: to examine whether the idea of accelerated melting was applicable.

I showed that, yes it was, within a  limited period there was a pretty strong acceleration. I agreed that viewed in around 2007 things did not look good and it was not unreasonable to suggest that it looked like a positive feedback might be in action.

That was why I picked that particular function . It was not an arbitrary choice , neither was it intended to make anything look bad or to ridicule.

I then showed that the acceleration ended in 2007 and that a different model would be required.
I did not say what that should be I simple tested whether the same accelerated melting model applied and showed that it did not.

The length of the dataset is too short to go much beyond such trivial functions which are probably not really representative of the system.

N. Atl and N. Pacific went through strong warming at the end of the century. N. Atl especially has a strong cyclic component.  The record is too short to enable us to see how the Arctic responds to those influences.

All we can tell for the moment is that the accelerated melting has stopped. I'm not a fortune teller so I won't publish a horoscope of where it will go next. Though the current behaviour looks to be more complex than a quadratic or exponential.

[Ice Cool Kim]

Virgent:

You are obscuring September minimums among anomalies that happen other times of the year. There is nothing remotely accurate about that. It is adding a noislike signal to the data. You are going to great lengths to obscure the trend.

No, filtering removes noise it doesn't add it. I am "going to great lenghts" to find the trend , not obscure it.

You seem to have fallen for the media hype on this and for some reason believe that that one day in September is more important that rest of data for the whole year and see it as a foul that I am trying to hide it.

I am not , I am treating all points equally without prejudice. That data point is making its duely weight contribution to the result.

[Ice Cool Kim]

jdallen:

Thirdly, extent by itself is a poor measure of total melt on two separate accounts.  First, in as much as it treats all ice as equal - namely that it doesn't consider the identified different physical properties of first year ice versus older ice.  Second, in that it ignores ice volume.

Sounds like the same point twice actually but yes I agree it is not volume.

As I said above area is important in its own right since the exposed area radiated/absorbs/evaporates etc and there is a signficant change in albedo both in SW and IR. Forcings are calculated in W/m2 , area matters especially from a feedback point of view.

That is to some extent what the question of acceleration is about: +ve f/b.

You will notice that my second post was that same treatment for ice area. Very similar picture.

Volume is an important factor too since it represents the joule counter.  However,  models still seem to experimental to be drawing any conclusions except assessing remaining uncertainty in the model itself.

Cryosat2 has had a very short life so far and all I've seem is 7 dots per year for two years. Not even enough to determine a single min/max properly.

Simulated volume estimates from earlier non volume measuring missions also have large uncertainties.

I think extent/area has its own value beyond being a crude volume indicator and I'm more interested in data that  model output.

Thanks for your thoughtful comments.

[crandles]

I then showed that the acceleration ended in 2007

Sorry, but no you didn't. You picked a date and function such that the noise about the trend allowed you to find an absurdly fast rate of melting acceleration. That that absurdly fast rate of acceleration did not continue is no surprise and is no reason at all to conclude "the acceleration ended in 2007".

You are saying your quadratic fits are not predictive at all but drawing a conclusion that relies on them being a reasonable predictive fit.

That is the way it appears to me. But maybe it is more of an honest mistake than that.

(Yes I was aware graphs linked were PIOMAS. I prefer volume for understanding what is happening but don't mind you choosing extent. Would similar graphs of exponential and gompertz fits of extent show a different story?)

[Vergent]

If you look across the top of the graph, connecting the maximums makes a noise-like signal. This corresponds to variations in seasonal sea ice at low latitudes. This signal is irrelevant to the issue of  the accelerating loss of perranial ice. In your graphs this irrelevant noise is mixed in with and indistinguishable from the signal, the sea ice minimums.

Yes, the September minimum is a uniquely important number, when you are talking about anual losses. If you haven't noticed, it is now closer to zero, than to the historical norms.

I see you are continuing to ignore even observational thickness data.


V

My vote: It's a duck.

[Ice Cool Kim]

You are saying your quadratic fits are not predictive at all but drawing a conclusion that relies on them being a reasonable predictive fit.

the fit is only valid for the period over which the fit was done. I think those periods are clearly indicated.

The quadratic model will neither hind-cast nor forecast and I don't understand how you see my conclusion depending up on prediction. I am only trying to characterise the data that is there, within the fitted periods.

the noise about the trend allowed you to find an absurdly fast rate of melting acceleration.

Nothing to do with the noise. The fit was pretty central. I could fit a similar quadratic to the smoothed version above.

Why 'absurdly' fast. Wasn't the big worry, tipping points (ie +ve f/b) etc.? This is exactly what that would look like, and it will be 'absurdly' fast if you like when it happens, hence the colloquial "tipping point" metaphor.  Once past the tipping point , it's freefall.

Now it was only ten years so maybe it was some other cause that just ended up being close to quadratic and the hypothesised positive feedback due to more exposed water turned out to be wrong.

I was just trying to go along with the story which seemed to fit that part of the data rather well.

Maybe the last six year is just blip and it's about to plunge again. Like I said, I'm not projecting just trying to characterise the data that is there.

The 1997 -2007 period was also special since it had a much smaller swing that both earlier and later periods and that variation was very constant. As we know post 2007 has much larger annual swing due to the open water.

These two observations also lead me to think there is a notable change in behaviour.

That that absurdly fast rate of acceleration did not continue is no surprise and is no reason at all to conclude "the acceleration ended in 2007".

Well if was +ve f/b and it was an acceleration it would be surprising. If I am wrong to conclude it stopped, it may be because it never started. Maybe its just random 'stochastic' variation.

Though I don't see many random parabolae of ten years duration in climate data.

You're right to challenge the ideas I've put forward and your comments are welcome but I don't such a rapid fall as absurd, more like predicted.  The surprise if that is the case is that the massive loss of 2007 provided enough negative feedback to check fall.

This does not seem to be incorporated in the ice models , yet none of the models can reproduce what happened so that may mean they are missing something.

I will have a guess at that. I think they need to account for surface reflection at low angles of incidence on open water.

I asked the CICE group whether they included it and got an evasive reply the first time and ignored the second. So I guess it will be included soon.

[Ice Cool Kim]

If you look across the top of the graph, connecting the maximums makes a noise-like signal. This corresponds to variations in seasonal sea ice at low latitudes. This signal is irrelevant to the issue of  the accelerating loss of perranial ice. In your graphs this irrelevant noise is mixed in with and indistinguishable from the signal, the sea ice minimums.

So one day per year you wish to focalise on is "the signal" the other 364 days are "noise": selection bias.

I never said I was looking at "the issue of  the accelerating loss of perennial ice" , that's your belated attempt to subtly re-frame the discussion.

I'm looking at total ice area/extent.

[jdallen]

As I said above area is important in its own right since the exposed area radiated/absorbs/evaporates etc and there is a signficant change in albedo both in SW and IR. Forcings are calculated in W/m2 , area matters especially from a feedback point of view.

<snippage>

Thanks for your thoughtful comments.

Yes, on the first part, I think you do have a valid argument there; the thing in itself may be what is providing the "brake" you reference on the disappearing extent. What may be at question in my mind is the seasonal increase in albedo simply being overwhelmed by other factors. 

As to the differences in the ice - The less dense nature of first year ice and its lower melting point than multi-year ice is what I was referencing, in addition to volume.  So, even with equivalent extent, the quality of the negative forcing from albedo may be significantly reduced by melt ponds appearing earlier, et. al.  The lower melting point translates into a lower energy threshold in other parts of the system (atmosphere, ocean enthalpy) before the ice can start a phase change.

And you are welcome; you obviously are putting a lot of thought into your work, and certainly have a better grasp of DSP than I do.

[icebgone]

April 17 IJIS extent and Cryosphere Today area are only 188,000 km2 apart.  I am somewhat surprised that they are this close.  As melting accelerates their difference grows reflecting what each measures.  But as ice thins and volume shrinks do we really need either measurement?  As shipping across the Arctic gets closer and closer to the North Pole the churning of their propellers will hasten ice melting and permanent removal.  If we get a strong dipole and lots of sun could this be the year when their measurements grow closer together instead of farther apart?

[Vergent]

If you look across the top of the graph, connecting the maximums makes a noise-like signal. This corresponds to variations in seasonal sea ice at low latitudes. This signal is irrelevant to the issue of  the accelerating loss of perranial ice. In your graphs this irrelevant noise is mixed in with and indistinguishable from the signal, the sea ice minimums.

So one day per year you wish to focalise on is "the signal" the other 364 days are "noise": selection bias.

I never said I was looking at "the issue of  the accelerating loss of perennial ice" , that's your belated attempt to subtly re-frame the discussion.

I'm looking at total ice area/extent.

If you are talking about "accelerated melting", anomaly data from months with no melting is noise. You are also talking about albedo, which only is an issue in the melting months when there is insolation. From your graph, how can you tell if a negative anomaly has a corresponding feedback?



Nope, no acceleration here.

The period upto 2007 had a lot of people rightly concerned about "accelerating melting" in the Arctic.

People are concerned about the accelerating melting because they know that ice is three dimensional. The mass balance loss can only be calculated with September area and thickness. That is all the information you need to answer the question. Anything else is BS. We use PIOMAS, which has been verified by Cryosat, because Cryosat can't measure thickness in September because of the melt ponds. If you want to pretend that the thickness hasn't changed, you are free to do so. We are free to judge the merit of your argument accordingly.

People are concerned because it is accelerating, We know that. You are not concerned. Perhaps you are in denial.

Neven,

I am in favor of moving this thread to "off topic" The only thing that could give this any validity whatsoever would be on a planet where the sea ice thickness was unchanged(maybe you could make a sub-category for denialist arguments, I am not advocating censorship, but lets separate real issues from belligerent sophomoric debate).

V

[Bob Wallace]

Kim - earlier in the discussion you stated -

It's a lot like temperatures. In the 90s temps were going up and getting steeper. That was worrying. After 2000 they flattened off, that is less worrying.

Perhaps you aren't looking at the correct data. 

"There's also a tendency for some people just to concentrate on air temperatures when there are other, more useful, indicators that can perhaps give us a better idea how rapidly the world is warming. Oceans for instance -- due to their immense size and heat storing capability (called 'thermal mass') -- tend to give a much more 'steady' indication of the warming that is happening. Here records show that the Earth has been warming at a steady rate before and since 1998 and there's no signs of it slowing any time soon."

[Neven]

Neven,

I am in favor of moving this thread to "off topic" The only thing that could give this any validity whatsoever would be on a planet where the sea ice thickness was unchanged(maybe you could make a sub-category for denialist arguments, I am not advocating censorship, but lets separate real issues from belligerent sophomoric debate).

Well, I think everyone has had his say, so shall we leave it at that?

[Ice Cool Kim]

Bob

Perhaps you aren't looking at the correct data.

Well I'm looking at the right data set if I want to see how sea ice area is changing. OHC would be another subject entirely.

[Ice Cool Kim]

I am not advocating censorship, but ...

which is of course exactly what you are advocating. Neither is the name calling and referring to what I've presented as BS very becoming.

But don't worry, I don't take offence easily. I have a very thick skin and I understand the whole debate is very emotional for some.

Though it probably does indicate that you realise how tenuous your criticisms have been up until now.

You are also talking about albedo, which only is an issue in the melting months when there is insolation.

No, albedo is an "issue" all year round. Albedo is reflectivity , what is not reflected is absorbed. Absorptivity is the complement of emissivity. I explained it's important above but you either ignored it or did not understand.

You clearly don't have a very good grasp on the science or data processing, so it may be better to ease up on the name calling and calls for censorship and do some more reading. There a lot of knowledgeable contributors here and elsewhere to learn from.


I will quietly ignore further comments of a similar ilk , so please don't be offended if I don't reply.

 

[Ice Cool Kim]

icebgone:

If we get a strong dipole and lots of sun could this be the year when their measurements grow closer together instead of farther apart?

Could you expand on that relationship a little? Maybe the ratio of the two would be worth plotting.
What does it show and is this recognised?

[crandles]

the fit is only valid for the period over which the fit was done. I think those periods are clearly indicated.

The quadratic model will neither hind-cast nor forecast and I don't understand how you see my conclusion depending up on prediction. I am only trying to characterise the data that is there, within the fitted periods.

Nothing to do with the noise. The fit was pretty central.

Why 'absurdly' fast.

Like I said, I'm not projecting just trying to characterise the data that is there.

Just characterising the data but cherry-picking 2007 and choosing a short period and choosing the function.

Your conclusion was "I then showed that the acceleration ended in 2007". How can you do that without some measure of what the rate of decline was in order to show whether it is increasing or decreasing?

Your measure of rate of decline was way too high by cherry-picking the 2007 end date and using a short period like 10 years. (Also not keen on quadratic if the data doesn't look like quadratic see dont-estimate-acceleration-by-fitting-a-quadratic)

1998 was above the trend (your green line in first graph will do for that) and 2007 is way below the trend. So clearly just characterising this limited data gives you a faster rate of decline than the general trend.

If I am wrong to conclude it stopped, it may be because it never started.

Close.

You are wrong to conclude it stopped because it never was as high as your comparison measure.

To reach the conclusion that the acceleration has slowed you cannot just choose 2007 because it is below the trend then have a method that automatically allows you to reach your conclusion.

You either need to show 2007 is not just noise about the trend or at least do a little more work to show that your method still works if you choose other periods like 1979 to 2005 through 1979 to 2009 preferably with functions that look like the data.

In this case it is easier than either of those methods: Just look at your green line and you see that the rate of decline is continuing to increase so the acceleration is continuing at least so far.

[Ice Cool Kim]

jdallen:

Yes, on the first part, I think you do have a valid argument there; the thing in itself may be what is providing the "brake" you reference on the disappearing extent. What may be at question in my mind is the seasonal increase in albedo simply being overwhelmed by other factors.

As to the differences in the ice - The less dense nature of first year ice and its lower melting point than multi-year ice is what I was referencing, in addition to volume.  So, even with equivalent extent, the quality of the negative forcing from albedo may be significantly reduced by melt ponds appearing earlier, et. al.  The lower melting point translates into a lower energy threshold in other parts of the system (atmosphere, ocean enthalpy) before the ice can start a phase change.

Yes, the whole situation is very complicated once you want to model what is doing what. There is some serious work being done on modelling and it's in the right ballpark. There are obviously still some factors that are not right. They are not yet what can be called accurate.

Melt ponds have been included but tend to produce too much melting rather than too little without. CICE has recently added a calculation for a "lid" of ice forming on top of ponds that gets it a bit nearer.

I have suggested to Daniella Flocco of the CICE team that if they took account of low incidence reflection they would be nearer still.  She did not reply so may be they will 'independently' arrive at the same conclusion in their next paper.

[Ice Cool Kim]

@crandles. Thanks for the link. It makes some valid points. eg the first line.  dont-estimate-acceleration-by-fitting-a-quadratic… if your data do not look like a quadratic!

The segment I "cherry-picked" did look like a quadratic and the fit was pretty good. That was why I chose to look at that function which corresponds to accelerating change. One that corresponded to the acceleration many were claiming. My whole point was I agree. 

Fitting the same model to the other periods that I "cherry-picked" , like "cherry-picking" the full dataset, showed that this model did not fit well. My conclusion was that this did not fit the 'accelerating' description that well. This is in agreement with: … if your data do not look like a quadratic!

In this case it is easier than either of those methods: Just look at your green line and you see that the rate of decline is continuing to increase so the acceleration is continuing at least so far.

But the problem is that when I take the whole period the fit is very poor and as the article you point me too rightly emphasises:  dont-estimate-acceleration-by-fitting-a-quadratic… if your data do not look like a quadratic!

This was pretty much my whole point here.  It's a bit crazy you seem to be arguing with me yet we're saying the same thing.

From the same:

Fig. 2. This sea level curve is the integral of the curve in Fig. 1 and thus contains the same information, but when viewed in this way it is hard to judge by eye whether sea-level rise has accelerated. The better way to answer this question is by looking at the rate curve,

I agree looking a rate of change plot if we are interested in the rate of change would be a very good idea.  I'll be back on that ...

[Ice Cool Kim]

Crandles:

How can you do that without some measure of what the rate of decline was in order to show whether it is increasing or decreasing?

It is indeed clearer to see when looking directly at rate of change rather than trying to infer it from the time series. This point is also made in the RC article you pointed me to.

We can clearly see the "acceleration" period, now a linearly increasing, negative rate of change behind the short term oscillations.

That period also seems remarkable for regularity of those oscillations. These are slightly less than 2 years in length and may account for the 1y anti-correlation that I think you remarked on elsewhere.

Are you still having trouble seeing that the downward trend ended with the huge melt in 2007?

You either need to show 2007 is not just noise about the trend or at least do a little more work to show that your method still works if you choose other periods like 1979 to 2005 through 1979 to 2009 preferably with functions that look like the data.

You are asking correct and valid questions. I think this plot shows that 2007 was not noise about the data.

[Jim Williams]

Neven,

I am in favor of moving this thread to "off topic" The only thing that could give this any validity whatsoever would be on a planet where the sea ice thickness was unchanged(maybe you could make a sub-category for denialist arguments, I am not advocating censorship, but lets separate real issues from belligerent sophomoric debate).

Well, I think everyone has had his say, so shall we leave it at that?

I sincerely hope so.  It's gotten boring and I'm having to skip posts.