When will the Arctic Go Ice Free?

[Glen Koehler]

      And since I am pontificating on statistics, here are some take away messages from the recent graphical posts by Oren, binntho, Archimid and El Cid (and thanks to all).

RE binntho's Extent and Area straight line trend
https://forum.arctic-sea-ice.net/index.php/topic,2348.msg227604.html#msg227604
     While it certainly looks like a significant downward trend, you can't say the slope is different from zero without doing the stats.  It probably is, but your use of the visual assessment method is no more valid than it is for the folks arguing that the process has stalled because it looks that way in the last 10-13 years (again I am shameless, the same applies to me too, my sinful nature was noted in previous post.  We are all fallen creatures.)

RE Oren's CAB volume trend and thickness graph
https://forum.arctic-sea-ice.net/index.php/topic,2348.msg227570.html#msg227570
    That's almost the chart I was hoping for, but it would be even better with a straight line regression trendline, tested for difference from zero, and then extended out 20 years to 2040.  FWIW, if you squint and draw a straight line through the CAB volume trend for Day 243, aka end of melt season, the slope of that line will indicate about 4 million km3 decline from 2000 to 2019, i.e. 19 years.  If that trend continues, then take another 4M km3 over next 19 years and it reaches zero in ca. 2038.  That's only a few years later than the Wipneus straight line projection of sea ice volume trend for the entire Arctic.

    The key characteristic about Oren's chart is that it is limited to ice volume in the CAB.  Thus, it presumably removes possible inflation of losses by peripheral seas that are melting out sooner than the CAB.  What started this phase of the discussion was the notion that future loss rate would decline because the CAB would be more resistant to melting.  I think the Oren chart refutes that. 

     I was surprised how strongly negative the CAB end-of-melt-season (i.e. annual minimum, day 243 data) volume is.  The CAB may look like it's been hanging on, but apparently that is the deceptive Extent curve at work.  The CAB has been rotting out from the inside.  As for the future, the presence of ice in the peripheral seas late into the summer might have reduced past losses in CAB.  Their presence has kept Arctic Ocean albedo high and almost certainly reduced pack rotation and transport out through the Farm Strait (and thanks to Tor for insight on importance of  export losses).  With less protection from ice in those peripheral seas as they melt out earlier in the year, the rate of CAB losses could markedly increase in the future. 

   Archimd's graph shows that CAB volume losses appear to already be increasing https://forum.arctic-sea-ice.net/index.php/topic,2348.msg227455.html#msg227455.

     In addition,  the wider amplitude of the fluctuations in El Cid's graph
https://forum.arctic-sea-ice.net/index.php/topic,2348.msg227470.html#msg227470 gives me a bit of the willies because one of the predictors for a nonlinear chaotic system reaching a tipping point is higher variability.  I may be misapplying that concept because max to min amplitude is not the same as variability between years, but I allow myself my own superstitions.

    But 2038 as the projected zero year for CAB sea ice volume is over a century earlier than binntho's trend extension showing Extent not reaching zero until 2187. How can that be?   Extent is not declining as fast as volume because the remaining volume is being contained in thinner and thinner ice, and thus the Extent does not decline as much as it would if thickness remained constant.  But as the thinnest ice contributing to Extent reaches zero thickness, it stops contributing to the Extent number.  In the end, the Extent curve and the Volume curves have to meet because zero volume provides zero ice for Extent.

    Which brings me back to Oren's thickness graph.  Total conjecture, but my guess is that once average thickness gets below 1 meter we will start to see the end-of-melt-season Extent curve start catching up with its parent Volume curve.  Ice melting comments elsewhere on ASIF point to the much lower melt resistance of thin vs thick ice.  Regardless of my conjecture, the Extent curve HAS to catch up to the Volume curve eventually.

    Stay tuned.  I think there are wild times ahead for ASI in the very near future because it is on the edge of the precipice.  It will be entertaining for those of us who like to watch numerical systems evolve.  Too bad it isn't just a horse race or some other innocuous event, but is instead the loss of a crucial component for meteorological and climatic stability on the only planet in the universe known to host self-aware, so-called "intelligent" life (actually any life, but I think we will soon see that microbes are just about as common as water).  As my brother, a conservative who bought into the climate hoax BS for a while, but who is too smart to stay ignorant, said when he came to see the big picture: "This story does not end well".

   

[binntho]

Thanks Glen for some very thoughtful posts.

[Klondike Kat]

Glen,
Yes, Oren's graphs are nice and the 19-year trendline does have volume going to zero fairly quickly.  However, his day 243 trend appears to change trajectory at the half-way point (2010).  The overall trend is ~-0.25 million km3 / year, but the R2 is only about 0.75.  Conversely, the trend for the first dozen years is much higher (-.45), with an R2 of ~0.9.  Typically, the R2 will increase with increasing data, if no deviation from the observed trend occurs.  The trendline since 2010 is effectively zero.  The Chow test would indicate if there is indeed a structural break in the trend.

Instead of refuting the premise that future loss rate would decline because the CAB is more resistant to melting, I feel that his graph support that premise.

[binntho]

Glen,
Yes, Oren's graphs are nice and the 19-year trendline does have volume going to zero fairly quickly.  However, his day 243 trend appears to change trajectory at the half-way point (2010).  The overall trend is ~-0.25 million km3 / year, but the R2 is only about 0.75.  Conversely, the trend for the first dozen years is much higher (-.45), with an R2 of ~0.9.  Typically, the R2 will increase with increasing data, if no deviation from the observed trend occurs.  The trendline since 2010 is effectively zero.  The Chow test would indicate if there is indeed a structural break in the trend.

Instead of refuting the premise that future loss rate would decline because the CAB is more resistant to melting, I feel that his graph support that premise.

I'm not sure if you are allowed to make such assumptions based on data from within a line arbitrarily drawn preciesely in order to seperate the periphery (that tends to melt out, at least shortly) from the center (that tends not to melt much as long as it is protected by the periphery).

This feels like a circular argument - selecting the piece of ocean that hasn't yet started to melt out properly (because it is still being protected by the periphery), seeing that it doesn' melt out properly, and then claiming that it is somehow "resistent" to melt.

I'd think that the changes in Oren's CAB graphs are a mixture of the outlier of 2012, which brought a lot of action into the CAB, and the bounce-back of the following years, which moved most of the action back out of the CAB.

And what physical properties of CAB ice do you think could make it more "resistent" to melt than the other bits of ice floating on the surface of the Arctic Ocean?

[Klondike Kat]

Glen,
Yes, Oren's graphs are nice and the 19-year trendline does have volume going to zero fairly quickly.  However, his day 243 trend appears to change trajectory at the half-way point (2010).  The overall trend is ~-0.25 million km3 / year, but the R2 is only about 0.75.  Conversely, the trend for the first dozen years is much higher (-.45), with an R2 of ~0.9.  Typically, the R2 will increase with increasing data, if no deviation from the observed trend occurs.  The trendline since 2010 is effectively zero.  The Chow test would indicate if there is indeed a structural break in the trend.

Instead of refuting the premise that future loss rate would decline because the CAB is more resistant to melting, I feel that his graph support that premise.

I'm not sure if you are allowed to make such assumptions based on data from within a line arbitrarily drawn preciesely in order to seperate the periphery (that tends to melt out, at least shortly) from the center (that tends not to melt much as long as it is protected by the periphery).

This feels like a circular argument - selecting the piece of ocean that hasn't yet started to melt out properly (because it is still being protected by the periphery), seeing that it doesn' melt out properly, and then claiming that it is somehow "resistent" to melt.

I'd think that the changes in Oren's CAB graphs are a mixture of the outlier of 2012, which brought a lot of action into the CAB, and the bounce-back of the following years, which moved most of the action back out of the CAB.

And what physical properties of CAB ice do you think could make it more "resistent" to melt than the other bits of ice floating on the surface of the Arctic Ocean?

Not sure what you mean by "hasn't started to melt out yet."  According to this graph, the CAB has lost half its volume at minimum between 2000 and 2012.  How much has to disappear before it can be said to start melting?

[binntho]

Not sure what you mean by "hasn't started to melt out yet."  According to this graph, the CAB has lost half its volume at minimum between 2000 and 2012.  How much has to disappear before it can be said to start melting?

I was thinking about extent when I said "melt out", the point being that volume would start going down when extent melt started making proper inroads into the CAB.

The half between 2000 and 2012 is presumably the multi year ice that is exported but never imported again (as it was in the past). This eventually results in what Chris Reynolds pointed out, that every year the ice grows back to more or less the same as the year before, and therefore volume at max flattens out for the entire Arctic, and with the CAB being artificially defined as the area with the most ice (whether intentional or not), the effect is felt the strongest there.

CAB volume will therefore probably not see large drops before the surrounding seas start seeing extent disappear earlier in the season, leading to large extent losses within the CAB.

[Shared Humanity]

And just to remind you - my involvement in this discussion started when somebody tried to explain the percieved stall since 2010 (or whenever) based on a totally erroneous posting, and since I pointed out that a) there was no stall, and b) the explanation was erroneous anyway, a lot of people have used a lot of effort to convince me of the mechanisms behind this stall that they claim is there for all to see.

But - my graphs show a steady decline since 2010 or whatever. No stall. As I stated very clearly:

I suspect that the "someone" was me so I would like to clarify what I said. At no point did I suggest a stall in melt. The melt continues apace with the condition of the ice looking progressively worse every year. Peripheral seas no longer freeze over as they have in the past (Barents, Bering) while interior seas melt out earlier and freeze later. The pack overall is more fragmented and mobile with far less MYI and no prospect for the oldest MYI to return. The ice at minimum is generally more disperse due to the higher mobility.

What I have stated and will continue to maintain is that the decline in SIE at minimum has slowed over the past decade. I have previously provided some suggested reasons for this behavior. Since someone (I don't know who) has defined an ice free Arctic or BOE as less than 1 million km2 of extent at minimum, this slow down in the rate of decline at minimum is relevant to the question posed by this thread. When will the Arctic go ice free?

For the record, IMHO...

• We will not go ice free in the coming decade.
• Our 1st BOE will occur some time between 2030 and 2040.
• After the 1st BOE we will see a rebound in SIE at minimum for the same reasons that SIE at minimum has slowed recently. We will not immediately see BOE's occur every fall after the 1st occurs.

[Shared Humanity]

I would also argue that there has been a similar slowdown in the decline of sea ice volume at minimum.

[Shared Humanity]

If we look at the behavior of SIE throughout the year, we can actually reword the question to read "When do we expect the CAB and CAA to be ice free?" as most of the other seas approach or reach ice free by the end of the melt season. In 2012, these 2 seas reached 2.8 million km2, the lowest to date. When can we expect them to fall below 1 million km2?

[binntho]

I suspect that the "someone" was me so I would like to clarify what I said. At no point did I suggest a stall in melt.

I can't remember who said what, I think we've all said what we need to say, and a good deal more than that perhaps.

For the record, IMHO...

• We will not go ice free in the coming decade.
• Our 1st BOE will occur some time between 2030 and 2040.
• After the 1st BOE we will see a rebound in SIE at minimum for the same reasons that SIE at minimum has slowed recently. We will not immediately see BOE's occur every fall after the 1st occurs.

Agree.

[grixm]

I'm rather new to this but personally I am not convinced that melt have slowed down, certainly not stalled completely. The last few years can all be explained by simple noise around a linear downwards trend, as seen in this image which I think has been posted here before. Even though we've been (un?)lucky at the times of minima, the average level keeps dropping and we regularly see different times of year with record low ice, both extent and volume.

I think also extent is very misleading compared to volume. Melt keeps going relentlessly the whole season, but sometimes we may not see this in extent because it's spent making the ice thinner. Then suddenly when an area reaches 0 thickness, large extents can disappear almost at once. I think the CAB can act the same way when a BOE happens, taking everyone by surprise. Funny thing, I heard a military commander describe the fight against ISIS like that too. They could surround one of their occupied cities for weeks or months with seemingly stalled progress. But what was happening was just that ISIS's reserves of soldiers were slowly diminishing, until suddenly they didn't have enough left to maintain the frontlines, and the whole city fell in days.

Exactly when we get an ice-free minima, I'm hesitant to guess, but it becomes more and more likely every year. I think even next year it may be theoretically possible if absolutely everything aligns, even if the chance is just 0.1%. But then in 10 years it may be a 10% chance per year, and in 20 years maybe 50% chance per year. And so on until it becomes inevitable.

[Klondike Kat]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

[binntho]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

Whether he uses anomalies or not makes no difference. Volume does not have a September minima. And if volume is dropping on average through the year, but appears to stall for a few years at minimum, then that's indicative of random noise (unless you can come up with a physical explanation for why general melt continues, but minimum stalls?)

[grixm]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

The whole point was that I think anomalies through the year is a better indicator than minima.

[crandles]

And if volume is dropping on average through the year, but appears to stall for a few years at minimum, then that's indicative of random noise (unless you can come up with a physical explanation for why general melt continues, but minimum stalls?)

Could be random noise. Could also be remaining ice now in harder to melt locations eg where it has been compressed, deep water that allows warm salty water to sink, ....

Different months have pretty much the same shape, though people have suggested 4 parameter gompertz is better fit in Sept than in Apr. Think I agree differences are likely just random noise could be random noise. Or maybe the curve is more apparent in Sept than in April. That perhaps gets me thinking it is less about MYI and more about remaining ice now in harder to melt locations.

[Klondike Kat]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

The whole point was that I think anomalies through the year is a better indicator than minima.

For the general state of the ice in the Arctic, I would agree.  However, when we talk about an ice-free Arctic, we are specifically talking about the ice at minimum.  Therefore, ice at minima is more relevant to this thread.

[Shared Humanity]

However, when we talk about an ice-free Arctic, we are specifically talking about the ice at minimum.  Therefore, ice at minima is more relevant to this thread.

Yes. It would seem obvious.

[grixm]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

The whole point was that I think anomalies through the year is a better indicator than minima.

For the general state of the ice in the Arctic, I would agree.  However, when we talk about an ice-free Arctic, we are specifically talking about the ice at minimum.  Therefore, ice at minima is more relevant to this thread.

What I'm saying is that the general state of the ice is a better indicator for future minima including an ice-free arctic, than past minima is. The actual minimum just comes down to luck because weather is noisy, but the odds of that luck is decided by the overall state of the ice, which can be studied throughout the year, not just at peak melt season.

Maybe past minima could have been useful in analysis if we simply had more of them, but since 2012 and the supposed slowdown started, there have only been 6 or 7 such data points. That's not statistically significant and thus can't really be used to make any sort of conclusion or prediction. It is much more meaningful to use the continuous streams of ice data we have from the rest of those years too.

[crandles]

Updated 4 parameter gompertz PIOMAS trends for each month.

July stands out as unusually still declining.

RMSE and parameters are

Code: [Select]

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE 0.9185809389 0.8863242069 0.8284413011 0.8146835816 0.8125406488 0.9620798129 1.056417406 1.00667826 1.010139747 0.9967588282 0.9508929005 0.9195709234
Parameters
GompertzP1 9.269435488 9.191963973 9.438560823 9.729099738 10.06706469 11.6564007 11.95225403 10.93870146 10.24143293 10.22148939 9.569257407 9.022173629
GompertzP2 0.145297716 0.1414869572 0.1246895738 0.1150470147 0.127664166 0.1449587939 0.1476787046 0.1473753748 0.1645007519 0.1728136333 0.1680035308 0.1684974314
GompertzP3 2004.945362 2005.133057 2005.427536 2005.482432 2005.630612 2006.125992 2006.009189 2005.485213 2005.052459 2004.717624 2004.516176 2004.080202
GompertzP4 16.50378959 19.2461913 21.17258536 22.06952654 20.94481892 16.29242535 9.683299438 5.748983986 4.938447892 6.186087261 9.41611277 13.09319642
P4 sets the floor at which it levels out.
P1 says how much higher the ceiling is.
P3 is the date of point of inflection.
P2 dictates how close to the point of inflection the curve happens; high values cause curve to happen close to point of inflection.

Oct has highest Parameter 2 value and April lowest value. April is highest month but October isn't lowest month. There may be some explanation, but I haven't figured it out, at least not yet.

May has lowest RMSE, July highest. Not sure I want to venture explanation there either.

[Klondike Kat]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

The whole point was that I think anomalies through the year is a better indicator than minima.

For the general state of the ice in the Arctic, I would agree.  However, when we talk about an ice-free Arctic, we are specifically talking about the ice at minimum.  Therefore, ice at minima is more relevant to this thread.

What I'm saying is that the general state of the ice is a better indicator for future minima including an ice-free arctic, than past minima is. The actual minimum just comes down to luck because weather is noisy, but the odds of that luck is decided by the overall state of the ice, which can be studied throughout the year, not just at peak melt season.

Maybe past minima could have been useful in analysis if we simply had more of them, but since 2012 and the supposed slowdown started, there have only been 6 or 7 such data points. That's not statistically significant and thus can't really be used to make any sort of conclusion or prediction. It is much more meaningful to use the continuous streams of ice data we have from the rest of those years too.

Inferring what will happen at minimum by using daily anomalies does not make much sense.  Sure, there are more data points and the fit is better, but without evidence that other times of the year is behaving similarly to September, the fit is meaningless.  That would akin to determining the change is summertime high temperatures in the U.S. based on global average temperature change.  They are not measures of the same thing.  Sure, the slowdown is based on only ten years of data, but that is a better indicator than forty years of daily anomalies, which are not necessarily correlated.  See crandles previous plots for monthly changes.

[crandles]

Inferring what will happen at minimum by using daily anomalies does not make much sense.  Sure, there are more data points and the fit is better, but without evidence that other times of the year is behaving similarly to September, the fit is meaningless.  That would akin to determining the change is summertime high temperatures in the U.S. based on global average temperature change.  They are not measures of the same thing.  Sure, the slowdown is based on only ten years of data, but that is a better indicator than forty years of daily anomalies, which are not necessarily correlated.  See crandles previous plots for monthly changes.

Daily creates lots of datapoints, but there will be a lot of autocorrelation. Really need to know the time needed to break the autocorrelation and make residuals pretty much random. I think, but am not terribly sure that this is about 3 months. It is certainly less than 6 months as we have commented on residuals at max and min not being correlated.

13 years above linear trend followed by 9 years below linear trend, could amount to 88 independent data items, a large proportion of which are indicating the actual trend is steeper than the linear trend.

While one line of 40 datapoints showing a slight curve might be dismissed as possibly random variation giving impression of curve. However, when all 12 months are showing curves in same direction at more or less the same times, dismissing this as random noise seems highly likely to be wrong.

[jdallen]

Except that your graph shows anomalies throughout the year.  While volume has continued dropping at other times of the year, that cannot be said for the Sept. minima.  Check what others posters have presented regarding volume at minimum.

The whole point was that I think anomalies through the year is a better indicator than minima.

Based on my expectations of 2019 extrapolated from conditions, and the current surprising (higher than I expected) numbers, I'm thinking there may be sense to this hypothesis.

[Shared Humanity]

crandles. Thank you for this data dive. Very surprised by the chart.

[Klondike Kat]

Yes, Interesting how they all show the same general trend, although the fall months showing greater flattening than the others.

[crandles]

I need to admit I got the time to decorrelate completely wrong. (Maybe I was thinking about extent where time to decorrelate was well under 6 months.) 

Here are the residuals. Spikes would align even better if month made a difference to placement on x axis.

Anyway for PIOMAS other months don't provide much extra information. Sorry about the error.

[MyACIsDying]

Updated 4 parameter gompertz PIOMAS trends for each month.

July stands out as unusually still declining.

RMSE and parameters are

Code: [Select]

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RMSE 0.9185809389 0.8863242069 0.8284413011 0.8146835816 0.8125406488 0.9620798129 1.056417406 1.00667826 1.010139747 0.9967588282 0.9508929005 0.9195709234
Parameters
GompertzP1 9.269435488 9.191963973 9.438560823 9.729099738 10.06706469 11.6564007 11.95225403 10.93870146 10.24143293 10.22148939 9.569257407 9.022173629
GompertzP2 0.145297716 0.1414869572 0.1246895738 0.1150470147 0.127664166 0.1449587939 0.1476787046 0.1473753748 0.1645007519 0.1728136333 0.1680035308 0.1684974314
GompertzP3 2004.945362 2005.133057 2005.427536 2005.482432 2005.630612 2006.125992 2006.009189 2005.485213 2005.052459 2004.717624 2004.516176 2004.080202
GompertzP4 16.50378959 19.2461913 21.17258536 22.06952654 20.94481892 16.29242535 9.683299438 5.748983986 4.938447892 6.186087261 9.41611277 13.09319642
P4 sets the floor at which it levels out.
P1 says how much higher the ceiling is.
P3 is the date of point of inflection.
P2 dictates how close to the point of inflection the curve happens; high values cause curve to happen close to point of inflection.

Oct has highest Parameter 2 value and April lowest value. April is highest month but October isn't lowest month. There may be some explanation, but I haven't figured it out, at least not yet.

May has lowest RMSE, July highest. Not sure I want to venture explanation there either.

The majority of trendlines starts under the actual and end up above the latest values. Isn't that an indication a Gompertz is not a valid fit for this data?

[crandles]

If it was 4 or more years at each end in different directions, that would be a indicator that something was wrong with the fit, maybe shape but perhaps with this shape curve it would be more likely to be the optimisation of the fit.

With only 1 or 2, well the can't be perfect everywhere or it is badly overfitted. If it is not perfect then 1 or 2 in same direction before switching sides at each end is the most likely situation.

The optimisation of this fit can easily move the ends up or down independently but chose not to change either because it is at the appropriate height. Moving either up or down might make fit better for some points but there is more worsening the fit by doing so.

In short, the answer is no that isn't a reason to say gompertz shape isn't a valid fit for the data.

[crandles]

Tried adding a slope to gompertz fit. Means 5 parameters rather than 4 and the difference is inconsequential. (yes 4 lines are plotted below.)

Optimum slopes added each year -0.086 for Mar and -0.000038 for Sept.

Yes, I don't believe the extrapolations saying Sept would reach 0 in 128000 years time but March would reach 0 in just 247 years. 

So completely unbelievable and differences negligible and extra parameter makes it easy to decide not to use this function for most purposes. Nevertheless, perhaps interesting to know that if a slope is added the effect are neglibible.

[Archimid]

So a Gompertz fit for now, then “no melting since 2012”  or we can call it a hiatus or very slight melt for the next 50 years. Global warming can be safely ignored as it won’t affect the linear fit. We can bet the well being of the world that no non linearities will ocurr.


I don’t buy it.

[dnem]

I too have zero faith in this sort of curve-fitting or in trying to draw conclusions from the short time series of data of questionable quality (failure of areal measures to capture ice quality; reliance on modeling to estimate volume) that we have.  The arctic environment is changing too quickly and too erratically.

[gerontocrat]

I too have zero faith in this sort of curve-fitting or in trying to draw conclusions from the short time series of data of questionable quality (failure of areal measures to capture ice quality; reliance on modeling to estimate volume) that we have.  The arctic environment is changing too quickly and too erratically.

Chartists rule, OK. On this thread at least.

Meanwhile, Bolsonaro, The Aussie Govt, Trump, Russia, Indonesia, etc etc, etc, are doing their best to ensure CO2 ppm and methane ppb continue to increase at an accelerating rate, (if you don't burn stuff, you can do the job by destroying the carbon sinks or letting more methane escape), & the UN etc wring their hands and say how awful it all is.

So maybe Hansen is an optimist to say Global Temps will rise at about 0.4 celsius per decade instead of the consensus of 0.2, and therefore maybe the Arctic will therefore warm by nearly another 1 degree in the next 10 years.

After all, the Masters of the Universe still decree we must have minimum World Real Terms Growth of 3.5% per annum (compound).

It is this sort of stuff that convinces me a lot of summer ice (and maybe of winter volume too) will be gone by 2030, not the charts

[nanning]

Thank you for giving your view gerontocrat. And I fully agree with your reasoning and wording.

[crandles]

So a Gompertz fit for now, then “no melting since 2012”  or we can call it a hiatus or very slight melt for the next 50 years. Global warming can be safely ignored as it won’t affect the linear fit. We can bet the well being of the world that no non linearities will ocurr.


I don’t buy it.

>So a Gompertz fit for now, then
looks like it to me.

>“no melting since 2012”  or we can call it a hiatus or very slight melt for the next 50 years

According to the optimisation of the fits, the trend does appear to be only very slightly down in last few years. I don't believe this either, it has to be steeper but each year I expect new data to be lower and affect the trend accordingly. While this hasn't happened much so far, it is only a few years and maybe just a few more will start to do so and even show deviations from the gompertz shape. Just because it hasn't yet, doesn't mean it won't happen.
 
>Global warming can be safely ignored
Absolutely not. Even if we could trust these 4p gompertz trends for the arctic sea ice (which we can't see above), other effects of global warming will continue. It perhaps suggests we should be more concerned about other effects of global warming rather than arctic sea ice.

[oren]

Somehow each new year spends a high number of days in the "lowest ever for the date" category of either extent, area and/or volume. 2016, 2017, 2018, 2019, each had its own major claim to fame. I don't think the downtrend is finished, I think it's still with us big time. The Sept minimum is not everything, and many indicators show the deteriorating state of the sea ice. Thus BOE chances increase every year until it pops up with complete (but unjustified) surprise.

[Shared Humanity]

So a Gompertz fit for now, then “no melting since 2012”  or we can call it a hiatus or very slight melt for the next 50 years. Global warming can be safely ignored as it won’t affect the linear fit. We can bet the well being of the world that no non linearities will ocurr.


I don’t buy it.

No one here has said that but, of course, you already know this.

[Shared Humanity]

I too have zero faith in this sort of curve-fitting or in trying to draw conclusions from the short time series of data of questionable quality (failure of areal measures to capture ice quality; reliance on modeling to estimate volume) that we have.  The arctic environment is changing too quickly and too erratically.

I could, of course, be mistaken but it is my understanding that SIE is a reliable measure with a longer record of data. Since a BOE will be determined by this metric, exactly how is it that SIE is "data of questionable quality"?

[Shared Humanity]

It perhaps suggests we should be more concerned about other effects of global warming rather than arctic sea ice.

We absolutely should.

[bluesky]

 
>Global warming can be safely ignored
Absolutely not. Even if we could trust these 4p gompertz trends for the arctic sea ice (which we can't see above), other effects of global warming will continue. It perhaps suggests we should be more concerned about other effects of global warming rather than arctic sea ice.
[/quote]

You could be right,  it reminds me a conference given by James Hansen that  I attended in London a few years ago, Hansen said that  If we would  decrease significantly our CO2 emission the Arctic sea ice will re expand (still think it's purely theoretical considering we are still on a BAU RCP 8.5 scenario, and we could possibly see a  BOE within 5  to 30 years ),  while recent estimates  shows that 16  to  20%  of the Amazonian has gone already, and around 5  to 8  % more and the threshold of no return would likely be crossed  breaking the self evaporation / rain recycling effect from the forest itself, than once the forest is gone, it's for ever.  Although, we could have well passed the threshold of no  return in the Arctic too, but the Arctic is probably to complex to have a clear view of the threshold of no return.

[Archimid]

So Arctic Amplification can be ignored because it poses no threat of acceleration. The Arctic will keep warming but ice melt will not increase much more and freezing will stay about the same, for 50 years. The jetstreams will keep shape shifting into new an unusual forms increasing WAA, but that is no biggie. The oceans will keep warming. Albedo will keep decreasing not only over the Arctic Ocean, but over the hemisphere.

But the ice won't melt much faster, even the when the records are coming more frequently. During winter we'll get all the ice we need back even when peripheral oceans are showing memory.

Nah.

I'm convinced that there is a hiatus if we take a 2012 starting point. I'm also convinced that there is an accelerating trend if we take 2013 as a starting point.

This freezing season is terribly important. If we get a 2013 like recovery it means the arctic is warming slowly after the MYI phase change. If we get a 2017 like recovery or worse, we are done.  The arctic already spent cold reserves and we are heading towards collapse.

[crandles]

So Arctic Amplification can be ignored because it poses no threat of acceleration.

Arctic amplification is driven in part by ice retreat causing albedo feedback. So if the ice retreat has/does slow down then so does the arctic amplification.

However,

1. The slow down in ice retreat is not certain, and
2. Ice retreat is not the only driver of arctic amplification. Warm tropics and heat flows to other parts of globe. Warm the arctic and the heat stays in the arctic. Therefore warm everywhere and the arctic warms most.

So some arctic amplification continues even if ice retreat slows down.

My 4p gompertz trends have the ice retreat coming to a dead stop, but I don't believe this while GHGs continue to increase which they clearly will for well over a couple of decades. Just because the trends indicate something, it doesn't mean we have to believe this will continue.

[dnem]

I too have zero faith in this sort of curve-fitting or in trying to draw conclusions from the short time series of data of questionable quality (failure of areal measures to capture ice quality; reliance on modeling to estimate volume) that we have.  The arctic environment is changing too quickly and too erratically.

I could, of course, be mistaken but it is my understanding that SIE is a reliable measure with a longer record of data. Since a BOE will be determined by this metric, exactly how is it that SIE is "data of questionable quality"?

Not the best wording. It is fairly long time series of well-managed data of good quality.  As I said above, I believe (mainly because I keep hearing it here!) that it fails to fully capture a consistent downward trend in ice quality that is making the pack increasingly more vulnerable than the trend in SIE indicates.

I concur with Oren's observations above.

[gerontocrat]

So Arctic Amplification can be ignored because it poses no threat of acceleration.

Arctic amplification is driven in part by ice retreat causing albedo feedback. So if the ice retreat has/does slow down then so does the arctic amplification.

My 4p gompertz trends have the ice retreat coming to a dead stop, but I don't believe this while GHGs continue to increase which they clearly will for well over a couple of decades. Just because the trends indicate something, it doesn't mean we have to believe this will continue.

In 2019, Albedo Warming Potential was pretty much a record.

I attach 2 of Nico Sun's AWP Anomaly graphs. It was the Bering, the Beaufort, Chukchi, ESS, Laptev & Kara that did the damage. The Central Arctic Sea was AWOL this year.

The last graph I attach shows actual Accumulated AWP against the potential maximum from an ice-free Arctic. As you can see, progress has been slow but remorseless and significant.

Arctic temperature increase amplification from AWP change has not stopped.

https://cryospherecomputing.tk/NRTawp

[crandles]

I keep getting thing wrong. The solver function I was using wasn't optimising correctly for 5 parameters.

The revised 5 parameter sloped gompertz fit now have estimate for 0 ice in 264 years for Mar and 38 years for Sep. A little more sensible this time.

RMSE for 4 parameter gompertx fit for March is .82844 reduced by 5 parameter model to .82607

For September RMSE reduced from 1.01014 to 0.98997.

Chart shows differences are larger now but still fairly small.

Looks more sensible now, so maybe 5th parameter is (approaching?) being justified.

[Klondike Kat]

I'm convinced that there is a hiatus if we take a 2012 starting point. I'm also convinced that there is an accelerating trend if we take 2013 as a starting point.

This freezing season is terribly important. If we get a 2013 like recovery it means the arctic is warming slowly after the MYI phase change. If we get a 2017 like recovery or worse, we are done.  The arctic already spent cold reserves and we are heading towards collapse.

Using arbitrary starting points to match ones personal preferences is poor science.  However, using a linear straight line fit because it is simpler, is not much better.  I cannot say for certain whether the gompertz fit, Taminos three-sloped trends, or another polymeric fit is the best, but they all match the data better.  Just because no one has presented a definitive reason for the changes, does not mean they have not occurred.  The best science is to try an devise a theory to match the data, not the other way around.

[grixm]

Statisticians weep when they see people trying to fit 4-parameters functions to series with so few data points that vary so much.

That's another argument for the linear fit instead. To just see what the real experts use. I already posted the official PIOMAS graphs, which includes linear trend lines:

http://psc.apl.uw.edu/wordpress/wp-content/uploads/schweiger/ice_volume/BPIOMASIceVolumeAnomalyCurrentV2.1.png

http://psc.apl.uw.edu/wordpress/wp-content/uploads/schweiger/ice_volume/BPIOMASIceVolumeAprSepCurrent.png

And then there is the graphs in NSIDC's official extent discussions, such as this:

http://nsidc.org/arcticseaicenews/files/2019/09/Figure-3.png

Arctic Data archive System allows you to overlay a linear trend line but no other forms of trend line:

https://i.imgur.com/MbD14S3.png

JAXA includes a graph with a linear trend line in their Satellite Monitoring for Environmental Studies front page:

https://kuroshio.eorc.jaxa.jp/JASMES/climate/data/graph/JASMES_CLIMATE_SIE_197811_000000_5DAVG_PS_9999_LINE_NHM_201.png

And NOAA:

https://arctic.noaa.gov/Portals/7/easygalleryimages/8/412/arc18_seaice_perovich_fig2.png

Meanwhile I don't see any serious attempts from the scientists in these organizations to get fancy with 3 or 4 or god forbid 5 parameter regression curves. Why do you think that is?

[Shared Humanity]

Yes, but when will the Arctic go ice free?

[blumenkraft]

We can't possibly know!

[Klondike Kat]

[quote author=grixm link=topic=2348.msg228684#msg228684 date=1568531536

Meanwhile I don't see any serious attempts from the scientists in these organizations to get fancy with 3 or 4 or god forbid 5 parameter regression curves. Why do you think that is?
[/quote]

Really?

https://www.searcharcticscience.org/files/search/sea-ice-outlook/2011/07/pdf/pan-arctic/hamilton_panarctic_july.pdf

https://www.esrl.noaa.gov/psd/events/2014/arctic-predictions-science/presentations/wed/arctic-wkshp-051414-stroeve.pdf

[Tom_Mazanec]

We can't possibly know!

Until it happens, then we will all know.

[crandles]

It appears there is more independent data in the extent data. Here are monthly 4 parameter gompertz fits for NSIDC monthly extent averages. March April May are better fitted with 3 parameter exponential than 4 parameter gompertz and you may prefer less parameters for some (or all) of the other months.

4 is dangerously high.

but then if Tamino finds this statistically significant and that uses 6 parameters


then how many parameters are justified?

I want to emphasise all models are wrong. We don't know shape of trend; it could be anything between 4 parameter gompertz at one extreme, Tamino's 3 linear pieces, anything in between, or perhaps having a steeper trend now than Tamino shows.

A couple of ways of deciding which to prefer between these include:
What has least parameters while still giving a good fit.
What looks like most of the models as a physical explanation for what is likely to happen.

Apart from above, I do admit grixm makes a good point.