Ice free predictions and their uncertainty

[crandles]

Eh, that graph only shows that there's a strong trend in both measurements, which is close-enough-to-linear not to worry about any bendiness.  You'd get an equally good correlation with anything that has a more-or-less linear trend over the same time period.

@Peter Ellis, sorry rather than just repeating myself graphically, I should explain:

If the data show an acceleration in extent decline and an acceleration in emissions then one way to get a more straight line might well be to graph extent against cumulative emission, as has been done.

But do we expect this to continue?

If all the models (assuming BAU increases in emissions) show acceleration in extent decline but then at some point changing to a declining rate as zero ice is approached. Then does it make sense to assume continued linear relationship between ice extent and cumulative emissions?

Or a different exercise: imagine a hypothetical world where we suddenly reduce emissions to zero. We would expect world to continue heating up (thermal inertia of oceans) and ice extent to continue to decline but cumulative emissions would stop dead. This would appear to break the linear relationship being suggested.

Similarly you can imagine a world where human emissions are no longer tied to GHG levels because of natural emissions from soils/permafrost/hydrates etc. Is the ice going to care whether the GHG in the atmosphere are natural or human ff burning sourced?

For these reasons, I am not convinced the relationship is reliably linear.

Just seeing a straight line is not enough to assume it will continue to be straight, it is sensible to also apply some thought to whether it makes sense to assume it will continue to be linear.

Having said this, if you want to know how much sea ice is destroyed by burning a certain quantity of ff to create CO₂ then the answer looks pretty reliable for the immediate past and quite possible the near term future. If it is useful, I have no problem with the relationship being looked at, discussed and used where appropriate.

[jdallen]

Eh, that graph only shows that there's a strong trend in both measurements, which is close-enough-to-linear not to worry about any bendiness.  You'd get an equally good correlation with anything that has a more-or-less linear trend over the same time period.

@Peter Ellis, sorry rather than just repeating myself graphically, I should explain:
<snip>
For these reasons, I am not convinced the relationship is reliably linear.

Just seeing a straight line is not enough to assume it will continue to be straight, it is sensible to also apply some thought to whether it makes sense to assume it will continue to be linear.
<snip>

Here's my metaphor for what's happening, and how things will or will not trend...

https://farm1.staticflickr.com/339/19361016472_83bb01214d_o.gif

I think one way to look at the potential for an ice free arctic is rather than see it as a specific end-state, to instead view it as a threshold - a point at which the energy available to the system is high enough that cyclical events can push it past where summer extent can exist.

So I think we are in fact talking about two different things.  The first, which I think *is* mostly linear, is the rate at which net enthalpy is increasing in the arctic.  The second, which is *not* linear, are the variable and cyclical forces which cause seasonal states to flirt with my aforementioned threshold.

Again, pretty clearly to me, Peter's graph represents increasing potential for that limit - conditions which will preserve seasonal ice - to be passed.  It may be events will cause our observations to proceed straight down that trend without much deviation.  If an event occurs where we *do* break out of that linear trend, I don't think that will mean we've made a final transition to a consistently ice-free summer.  Rather, I think it will mean for that specific event, all of the limbs on our pendulum lined up such that we jumped over it.

So many feedbacks - that's the other problem - the "limbs" of our "pendulum" are changing over time as well.  But I say - focus on the increase in enthalpy, view it as an indicator of the potential of a melt out any given year.

[Peter Ellis]

Just seeing a straight line is not enough to assume it will continue to be straight, it is sensible to also apply some thought to whether it makes sense to assume it will continue to be linear.

Well yes, but my point was rather more fundamental than that. Once you know that ice is declining linearly (and over the last 60 years, a linear fit is "good enough") and that CO2 is also increasing linearly (and over the last 60 years, a linear fit is "good enough") then this graph tells you literally nothing more.

Having said this, if you want to know how much sea ice is destroyed by burning a certain quantity of ff to create CO₂ then the answer looks pretty reliable for the immediate past and quite possible the near term future.

No.  No it doesn't.  The fact that you get a linear correlation between two things that are changing nearly-linearly is not "pretty reliable", it's just how numbers work. Both have a nearly-linear relationship to time, therefore they have a nearly-linear relationship to each other. You could get an equally convincing graph by correlating the summer ice minimum to the height of a slow-growing tree planted in 1950, or world beef production in tonnes per year, or the average life expectancy in Portugal.  You CANNOT infer causation like this, and you certainly can't use it to make predictions without some underlying model for how one variable relates to the other.

[dnem]

I tend to agree with jdallen's view of the big picture.  We will either "trend" our way down to zero or several major drivers will align and we'll get there before the trend predicts.  I don't think that the alignment of multiple drivers is predictable.

What I don't have is a good handle of the hysteresis in the system.  Is one ice free summer enough of an "attractor" to switch the arctic to a new equilibrium (always ice free in the summer)?

[jdallen]

What I don't have is a good handle of the hysteresis in the system.  Is one ice free summer enough of an "attractor" to switch the arctic to a new equilibrium (always ice free in the summer)?

Maybe.  An "ice-free" summer will change two primary "legs"of our pendulum:
- total summer solar uptake by the Arctic ocean
- total available moisture during the fall refreeze

Those two by themselves may be enough to knock the system into a completely different equilibrium; at the least, it would take significant heat loss out of the Arctic to shift back to the old balance.

I do think the key really is total enthalpy in the ocean itself.  The last few years of dropping summer extent I think is an accelerating feedback to that, along with the massive inputs of heat coming in mostly on the Atlantic side via currents.

[be cause]

The dmi 80'N anomaly is rising again and the 850 mb anomalies are going redder with every run , it looks like the rest of the year will be as anomalous as last Christmas . With the Fram pump primed and loaded with our thickest ice , the situation to our North is going critical . Unless something changes quickly Arctic ice is fast heading out of the atlas and into the history book ..

[Jim Williams]

I do think the key really is total enthalpy in the ocean itself.  The last few years of dropping summer extent I think is an accelerating feedback to that, along with the massive inputs of heat coming in mostly on the Atlantic side via currents.

The problem is that you have to take the total ocean, not just the Arctic portion of the ocean.  What really counts here is how much further north the "Atlantic" and the "Pacific" have to get before the Arctic becomes a bay or sound, rather than an ice capped desert.  Hard to guess, but I think the evidence is for December 27th 2015 -- and it is just a matter of time before the residual ice is gone.

[magnamentis]

The first ice free September will be predictable about 2 months before it happens.

Really? Nobody in early September 2016 would have predicted the absurd low appearing in November globally;


Taking that into account, there's really no saying when it will happen. It could be July 2017 with a little (bad) luck.

What people don't seem to realize is that it could just as well be November 2017 as July.

predicting would have been to bold a statement but considering the possibility was indeed posted if one just would read back some posts in september and october this year.

[jdallen]

I do think the key really is total enthalpy in the ocean itself.  The last few years of dropping summer extent I think is an accelerating feedback to that, along with the massive inputs of heat coming in mostly on the Atlantic side via currents.

The problem is that you have to take the total ocean, not just the Arctic portion of the ocean.  What really counts here is how much further north the "Atlantic" and the "Pacific" have to get before the Arctic becomes a bay or sound, rather than an ice capped desert.  Hard to guess, but I think the evidence is for December 27th 2015 -- and it is just a matter of time before the residual ice is gone.

I'll quibble with you here - not all ocean.  The Arctic is mostly isolated behind continents and continental shelves - there isn't as much opportunity for heat exchange via physical transfer as there is in a more open system - such as the boundaries between the Antarctic ocean and the Indian, South Atlantic and South Pacific.  As such, it forms its own mostly isolated heat sink/source.  At a qualitative level you can see that, if you compare ice coverage down to 70N on the Pacific side of the basin as compared to the Atlantic.  The Atlantic side is far more open, and is open right up to Svalbard and FJL.  On the Pacific side, screened from inflow, we have full coverage save for portions of the Chukchi.

Now, this is not to say increased heat in other places isn't factor - it is - but that's not my point.  My point is the primary reservoir of heat which will determine whether we have ice in the summer is the Arctic itself, not the adjacent oceans.

[Jim Williams]

I do think the key really is total enthalpy in the ocean itself.  The last few years of dropping summer extent I think is an accelerating feedback to that, along with the massive inputs of heat coming in mostly on the Atlantic side via currents.

The problem is that you have to take the total ocean, not just the Arctic portion of the ocean.  What really counts here is how much further north the "Atlantic" and the "Pacific" have to get before the Arctic becomes a bay or sound, rather than an ice capped desert.  Hard to guess, but I think the evidence is for December 27th 2015 -- and it is just a matter of time before the residual ice is gone.

I'll quibble with you here - not all ocean.  The Arctic is mostly isolated behind continents and continental shelves - there isn't as much opportunity for heat exchange via physical transfer as there is in a more open system - such as the boundaries between the Antarctic ocean and the Indian, South Atlantic and South Pacific.  As such, it forms its own mostly isolated heat sink/source.  At a qualitative level you can see that, if you compare ice coverage down to 70N on the Pacific side of the basin as compared to the Atlantic.  The Atlantic side is far more open, and is open right up to Svalbard and FJL.  On the Pacific side, screened from inflow, we have full coverage save for portions of the Chukchi.

Now, this is not to say increased heat in other places isn't factor - it is - but that's not my point.  My point is the primary reservoir of heat which will determine whether we have ice in the summer is the Arctic itself, not the adjacent oceans.

If Siberia really does become the climatic North Pole; which it seems to be trying very hard to do, I think it will have a whole lot to say about your analysis.  The arctic is not where are nearly closed as you think.

[RoxTheGeologist]

[
Well yes, but my point was rather more fundamental than that. Once you know that ice is declining linearly (and over the last 60 years, a linear fit is "good enough") and that CO2 is also increasing linearly (and over the last 60 years, a linear fit is "good enough") then this graph tells you literally nothing more.

This is true; I think that there was a study preformed by the cigarette companies a long time ago, with a similar argument about the correlation between lung cancer and cigarettes, and their analogy was that statistics show that people who eat blancmange have more road accidents.

I don't think we will ever come up with a complete model to show the real complexities of our global system, how exactly CO2 emissions cause or do not cause ice loss. We have to use surrogates models where we can. We could be debating the intricacies for the next 50 years, saying we can't be sure, and in the mean time the ice caps will disappear.

As a scientist I understand your argument; As someone who wants to change a mindset it is an excellent graph. Greenhouse gases have been emitted at this rate, and look how it has effected ice: Wip's graph is a great example of this. The systems that control the ice in the hemispheres are very different, but adding the total ice extent together produced a viral image. The earth is getting hotter and the ice caps are disappearing. We smoked cigarettes and it gave us lung cancer. We emitted CO2 and it caused the Arctic ice cap to shrink.

[andy_t_roo]

The line drawn looks to be very good but does it really extend to first and last 15 years well?

I don't think you can rule out the following sort of relationship

One thing to remember is that co2 is the forcing agent, and as such the amount of it in the atmosphere relates to the rate of change, as the equilibrium point gets further away. If we stopped emitting today, that graph would go straight down for a while.

[crandles]

Having said this, if you want to know how much sea ice is destroyed by burning a certain quantity of ff to create CO₂ then the answer looks pretty reliable for the immediate past and quite possible the near term future.

No.  No it doesn't.  The fact that you get a linear correlation between two things that are changing nearly-linearly is not "pretty reliable", it's just how numbers work. Both have a nearly-linear relationship to time, therefore they have a nearly-linear relationship to each other. You could get an equally convincing graph by correlating the summer ice minimum to the height of a slow-growing tree planted in 1950, or world beef production in tonnes per year, or the average life expectancy in Portugal.  You CANNOT infer causation like this, and you certainly can't use it to make predictions without some underlying model for how one variable relates to the other.

Agreed you cannot infer causation from the graph alone. I was trying to add something to say that because I don't think the relationship is reliably linear, that doesn't mean I think the paper is rubbish and shouldn't have been published. Maybe I worded it sloppily.

The graph alone doesn't allow you to infer causation, but we don't have the graph alone we have a whole host of IPCC reports and lots of scientific papers that indicates emissions increase greenhouse gas levels, greenhouse gases warm the world and especially Arctic areas in winter and a warmer Arctic winter is going to mean thinner ice that is easier to melt out. Also that the natural effects have been pretty small. Given such available extra evidence, then I don't see a lot wrong with saying

if you want to know how much sea ice is destroyed by burning a certain quantity of ff to create CO₂ then the answer looks pretty reliable for the immediate past and quite possible the near term future