Who do we listen to? Which science?

[ccgwebmaster]

Having recently apparently managed to give the impression to some people that I am anti-science, and with the general note that bodies such as the IPCC are somewhat vilified even amongst portions of the scientific community, I thought some debate around what we take as relevant science might be illuminating.

A view expressed by another person in another thread was that we should consume the conclusions of bodies of "experts" with respect to climate change (with the implication that personal speculation and progression along the spectrum of expertise was irrelevant and unhelpful).

I wanted to - in the specific context of arctic sea ice (but with the note that the same principles apply to a great number of other earth system predictions with respect to anticipated effects and timescales thereof from climate change) raise this as something worth a closer examination.

Perspective 1

My impression is that the mainstream scientific community has consensus views little shifted from 2007 - predicting total Arctic sea ice loss later this century (certainly not within the next few years, and - on average - not even without the next few decades). This is a recent paper appearing to state that very conclusion:

http://www.pnas.org/content/early/2013/07/10/1219716110

IPCC AR4 doesn't really appear to predict total ice loss this century:

http://www.ipcc.ch/publications_and_data/ar4/wg2/en/ch15s15-3-3.html
http://www.ipcc.ch/publications_and_data/ar4/wg2/en/figure-15-3.html

Although Neven is picking holes in it in this example, it contains the necessary quotes to illustrate that AR5 is likely to be similarly conservative:

http://neven1.typepad.com/blog/2012/12/the-real-ar5-bombshell.html

The Hadley Centre (Met Office) weighed in - they expect later century ice loss also, as predicted by their models - this article contains their views as to why projections of near future ice loss ought to be discarded and ignored:

http://www.guardian.co.uk/environment/2009/feb/11/climate-change-misleading-claims

All of the above, I argue is mainstream science. It is mainstream because it reflects the typical view held by the public, by policy makers and by the scientific community at large (noting that most scientists are not experts in any arbitrarily selected field, in this case sea ice, as science is a vast entity).

It is a matter of public record that the models used to attempt to predict these things are still incomplete and still advancing quite fast (indicating a long way to go, if I might be cynical).

Perspective 2

Observational evidence, which I submit is still science appears to indicate that the models are not accurately predicting the state of the Arctic (notwithstanding that these models usually take base runs starting well over a century ago historically and are actually only off by a few decades - something fairly pointed out to me by someone defending the modelling community - my counter point was that those decades really matter once you're inside them)

We have measurements by satellites (which I would hope most people using this forum are familiar with as I don't want to have to compile a list). We also have corroboration from ground measurements - sometimes that actually suggest that the satellites appear to be painting a somewhat more rosy picture than what's going on at ground level.

I can't remember where I came across this (might have been this forum):
http://home.cc.umanitoba.ca/~dbarber/David_Barber/Sea_Ice_Research_files/Barber_etal_GRL%2709.pdf

Between PIOMAS and Cryosat 2 it seems reasonable to state that ice volume has crashed and is continuing to decline precipitously with very little time left for any negative feedbacks to put a floor under ice loss before we start to see ice free conditions in the Arctic for part of the year. It is very hard to see how the ice can persist perennially for decades to come given that the pressure on the ice is increasing over time as greenhouse gases continue to be released and the warming from already released gases continues to build (noting lag in the climate system in this respect).

There are multiple experts (and I use that word strongly) in the field of sea ice who have lent their name to near future predictions for total ice loss. This is also science - in the sense that it is presumably based upon extensive expertise and knowledge with the science, even if it is not precisely mainstream.

Professor Peter Wadhams - http://www.guardian.co.uk/environment/2012/sep/17/arctic-collapse-sea-ice
Professor Carlos Duarte - http://theconversation.com/final-frontiers-the-arctic-12911
Dr Wieslaw Maslowski - http://www.bbc.co.uk/news/science-environment-13002706

I feel it is worth noting that if the climate models in Perspective 1 are failing to accurately predict the loss of the sea ice, they are also failing to accurately predict knock on consequences of losing the sea ice. That is to say that there are at least two sources of possible errors from the models:

1. The models may well have failed to capture key details that cause things to happen faster than expected as apparently with the sea ice
2. The models may capture some processes accurately but nonetheless fail to predict timescale as the forces driving these processes do not occur within the models until much later

It seems in little doubt to me that the experts in the field of sea ice being quoted above do not subscribe to the consensus mainstream views in perspective 1.

The Big Question

Who do we listen to?

I choose to go with what observations appear to be saying, and I tend to take a more pessimistic view. I do believe a pessimistic view is justifiable in terms of the surprises the earth system has sent our way so far, and am unclear what process could meaningful delay (by more than a matter of years) total ice loss during the Arctic melt season.

Bodies such as the IPCC may say numerous events are very unlikely, but also acknowledge the uncertainties and limitations of their methods. That is ignoring the statistical near certainty that if you have enough very unlikely events to pick from - at least some of them will happen.

I've chosen sea ice as the terms of reference for this debate about which part of science we listen to - and want to be clear that I do not listen so much to mainstream science as to what the experts in the specific field in question are saying (this is perhaps where I offended some people in another thread).

I think this is a relevant debate not only to examine more closely the portions of the scientific spectrum we base our views on - but also to explain why I (and possibly others, who I won't claim to speak for) take an outlook sometimes generously described as pessimistic (and sometimes insulting described as much worse things).

It's also relevant to help identify when alarmist and apparently extreme statements are justifiable.

Perspective 1 or Perspective 2? Is there any middle ground?

[birthmark]

I'm always one for observation over models, particularly when the models aren't performing very well.  Such is the case in the Arctic.

However, to answer your title questions: We should listen to all the credible scientists and all the credible science.

[OldLeatherneck]

To be fair to the scientists, we have to understand that they are trying to model a system that is currently in chaos.  While each of the many uncontrolled variables may be understood to a great degree, I'm beginning to question whether there is a good grasp of the interaction between these variables. 

I'm familiar with the basic concepts of control system modeling, such as are used in the development of guided weapons.  In those cases, the controlled variables are very well understood and the distribution curves in the model are reasonably accurate.  Also the feedbacks and interaction between these variables are equally well represented in the models.  Even the uncontrolled variables are generally understood with the random distribution curves reasonably accurate.

When it comes to global climate models there is only one variable that humans can control and that is the level of emissions, GHGs and sulfates.  Beyond that, every other variable is either a positive or negative feedback and I'm not sure the interactions between the feedbacks are well understood or whether the random distribution curves adequately represent what is likely to happen in the various emissions scenarios.

Disclaimer: I've never developed a model nor written the code for any models.  My understanding is at a user level in systems engineering, systems testing and strategic planning.  Therefore, everything I said above is subject to further clarification by those who are more familiar with the development of Global Climate Models.

[ccgwebmaster]

Material diverted from another thread: https://forum.arctic-sea-ice.net/index.php/topic,290.msg10309.html#msg10309

I think you are too soft on the IPCC. Where do they say the models they use for their reports are lagging the real world and have important feedbacks missing or inadequately represented? 

Perhaps - my criticisms of them are somewhat different. I remember reading IPCC material identifying important feedbacks that were omitted as they were insufficiently understood, but struggle to find it again - a testament to either the inferiority of my search skills or the vastness of the material they've tossed out there.

I'd be surprised if statements to those effects couldn't be found in the vast body of literature they've generated - but you're probably looking at a sentence or two amongst thousands of pages.

My view of the IPCC is that it's turned into an obsolete monster. It's years behind the curve (and I'd agree with McPherson when he describe how and why - the time taken to not only author the original papers but to collate and process them through the IPCC framework) in a time when climate change is proceeding increasingly rapidly on a timescale faster than the IPCC can hope to keep up.

They promote dangerous views on climate change, given that the media and politicians are apparently incapable of engaging the topic in sufficient detail to read the caveats or to understand the not inconsiderable cumulative risks hidden away in what they say. The IPCC talks primarily about what they consider likely, and as you say have substantial elements missing from the modelling (and real world observations suggest their timescales for changes are wildly optimistic, but again consider just how out of date their reports are now).

Again, as an example, see Kevin Schaefer's and John Mitchell's comments  found here:
http://www.brusselsblog.co.uk/do-you-belive-the-european-commission-on-climate-change/

The response from John Mitchell at that link about modelling reminds me of a similar conversation I saw in another place with respect to sea ice modelling - except there the suggestion offered was that there must be important processes we don't know exist (or understand properly) at work, and that a few decades out wasn't too bad for model runs of ~150 years...

I think that is "We don't understand it so it doesn't exist".

Pretty much. The IPCC are guilty of that - they list the things they don't understand (but do know about - obviously nobody can list the things nobody knows about) - and then proceed to ignore them and pretend they don't exist for modelling purposes and in stating conclusions.

The real problem here is that the IPCC are a political animal seeking a mass consensus. Truth is not necessarily approached through mass consensus, just as I'm a little skeptical of the idea of reducing uncertainty by averaging together model run ensembles. There is only one truth (the real observed world) and there will be specific models that do better than others (but will be diluted by averaging in a whole lot of other results).

The IPCC are also not able to categorically rule out numerous potentially catastrophic events - using what might be tagged "weasel words" on Wikipedia - highly unlikely. The other way of reading that is within the realm of possibility.

[ccgwebmaster]

To be fair to the scientists, we have to understand that they are trying to model a system that is currently in chaos.  While each of the many uncontrolled variables may be understood to a great degree, I'm beginning to question whether there is a good grasp of the interaction between these variables. 

I think it is entirely unreasonable to expect to model the planet accurately under such circumstances. I'd argue the models concentrate primarily on physical and chemical processes - but there are plenty of responses and annoying little details that can have significant impacts that I'm unclear how any human is going to anticipate and model accurately.

In trying to model natural carbon feedbacks for example (which to the best of my knowledge they do not, yet) how can anyone anticipate something like pine beetles?

In trying to model agricultural production the same problem exists (on an even vaster scale) with respect to agricultural pests.

Except in the broadest and most qualitative of senses - can we actually get anything useful out of models? Anything useful enough to gamble our species on?

When it comes to global climate models there is only one variable that humans can control and that is the level of emissions, GHGs and sulfates.  Beyond that, every other variable is either a positive or negative feedback and I'm not sure the interactions between the feedbacks are well understood or whether the random distribution curves adequately represent what is likely to happen in the various emissions scenarios.

I'd agree more strongly if you said only one major variable - other non emission factors do count, for example land use changes.

Disclaimer: I've never developed a model nor written the code for any models.  My understanding is at a user level in systems engineering, systems testing and strategic planning.  Therefore, everything I said above is subject to further clarification by those who are more familiar with the development of Global Climate Models.

I've also never developed a climate model, nor ever worked upon anything of comparable complexity. I consider myself a software engineer (albeit mostly self taught, I have no credentials) and the first complete program I ever wrote was just about the simplest ecosystem simulation imaginable. I'd read a few chapters of a VB3 book (conditional branching, loops and variables, mostly) and got bored enough to want to write an actual program - so I put in some very basic rules:

• a user defined starting amount of red squares (grazers) moved to the adjacent square with the most "food" once per cycle, and consumed the food there
• grazers consumed a unit of food per turn in energy cost
• food (grass) was randomly distributed by a user defined amount over a user defined grid each "cycle"
• grazers reproduced if during a given turn they exceeded a user defined food storage value (asexually spawning the offspring adjacently to their position)

After I got over the initial surprise of the program actually working as intended, I had hours of fun playing with it and watching red squares dance across the screen. What struck me then and still strikes me now though - is that despite the incredible simplicity of the rules and code, I saw outcomes that had never been designed for. The grazers tended to move in herds (presumably by virtue of adjacently spawning) chasing the highest concentration of food. If food ran out they tended to sit around starving as food accumulated further away, until food appeared nearby and they found a route to a high food area again. I saw the population oscillate in booms and busts as the numbers expanded above carrying capacity and then crashed back below it (carrying capacity was strictly fixed, of course - no erosion thereof). With virtually every parameter user definable, I could explore how things changed if I altered the rate of food growth, the spawning threshold of the grazers, etc and try to find stable solutions (in a grid no larger than 200x200).

A little later I tried adding blue squares (hunters), but they were too efficient in such a small grid and I never managed to find a long term stable set of parameters where they didn't either starve to death chasing their prey or exterminate their prey and then starve to death (leaving grass growing to infinity - a rather more boring world).

Over the last 20 years I've gained a lot more skill in programming technologies (and also better understand the surprise of the teacher who then ran the department when he saw what I'd done in the context within which it was done) and I suspect the models attempting to simulate the earth system are probably some of the most complex and sophisticated endeavours the human race has ever attempted. Technologically, they're probably shining jewels of excellence.

Unfortunately, I think they're also of rather limited value and rather dangerous as the temptation appears to be to peer in at them, watching the dancing red squares - oblivious to what's happening outside the window (and to the damning warnings from paleoclimate). Why the obsession with how fast we will see what effects - and by implication how close to the line of damnation we can get away with treading - when the paleoclimatic record already suggests our behaviour will lead to a very significantly changed planet (of which we have precisely one)?

Most damning of all, when the models start to agree with some of the implications of paleoclimate and start to tell us we're in big trouble - along with observations outside the window - why do we continue to think the answer lies in more modelling, as opposed to diverting the same investment in funding and intellectual resources into solutions?

Why does so much mainstream climate science focus so strongly on the models and not on historical lessons from the past and direct observations?

[GeoffBeacon]

CCG

Thanks.

[Glenn Tamblyn]

"Why does so much mainstream climate science focus so strongly on the models and not on historical lessons from the past and direct observations?"

One part of this might be derived from the principle that Correlation is not Causation. Historical data says stuff happened, but doesn't necessarily provide the causal mechanism for why it happened. Hence the resort to modelling (defined in the broadest sense - a single eqn is a model) to attempt to derive the underlying causal mechanism for the observed behavior.

Then the models may be inadequate, at least at their current level of development so they aren't adequately agreeing with observations. But their is a lot invested in the idea of the models, that they are the pathway to ultimately deriving the desired understanding of the causation.

Then you get a confusion of purposes. Discussion of the models can be all about debate amongst the scientists about how to advance the models. This is ultimately the scientists day job, is not really for public consumption. At the same time their is the need to communicate what is known to the wider world. The official organs of communication - IPCC etc - are tasked with being conservative, measured, cautious in their statements.

Add the essentially conservative nature of science itself and you get a situation where a claim for something won't be made if there isn't solid support for it - both observations AND theory of the causation. So within scientific circles the conversations most likely would be of the form 'the models are only moderately successful at showing X'. This would simply be taken as simply providing a status report on development, grist for the mill etc, with the participants understanding that this is work in progress.

But then expect those scientists  to comment on the current state of play outside the scientific community and they are in trouble.

If they say that the models aren't reproducing the observations very well, that will be taken by the general public (eagerly assisted by the deniers) as 'the scientists have no idea what is going on' which is a major misrepresentation of the situation.

If they go out on a limb and make statements about what is going to happen that are not supported BY THEIR OWN WORK, then they are breaking part of the central code of conduct of science.

So instead they give a 'our models are saying X' response. Non-committal. Trying to straddle a barbed-wire fence. Not wanting to let misrepresentations flourish, but at the same time resistant to making unsupported claims - that can ruin your career.

[LRC1962]

Came across this piece. http://www.marketwatch.com/story/aer-scientists-uncover-link-between-declining-arctic-sea-ice-and-colder-winters-2013-09-12
"This research shows for the first time how both sea ice melt and rapid snow advance are related to more severe winter weather," Judah Cohen, Ph.D. principal scientist and director of seasonal forecasting at AER, who is lead author of the study. "The results suggest that record sea ice melt last autumn contributed to the individual extreme weather events of last winter. Using a statistical rather than dynamical model, we established the relationship between declining sea ice, more extensive snow cover in northern Eurasia, and a weakened polar vortex, which allows cold air normally confined to the northern latitudes to spread southward."
Now I am very confused. I am sure I had been reading that from climate change researchers for years. If I am right this is what makes the public even more confused and sceptical. My view is: follow the money. Even if it is from a well established university where the money comes from is very important. It used to be said that if it was government funded and at arms length it could be trusted. Based on how some governments are now working, that is no longer the case.
There is a lot of interference from both government and business sectors to get scientist to spin there research in favour of what their respective positions are. This goes to the point of firings, public vilification of their credibility, and closing of research departments. In Canada with respect to Arctic research our present government is notorious for doing that. Business is also known for doing the same thing. The problem then becomes even if we are able to find out where the money the researchers get their money from, many times now it is not the same ones as to where the money originated from and that is the difficult thing.
My preference is to listen to the researchers who are going out into the field and say that based of what is going on now and what we thought we would find then we can easily see this happening by this time.
I do understand that scientist need to work with models, but many times because things are changing so fast and the models take so long to develop and prove they are out of date by the time they publish.
An example of a scientist who ran into this kind of difficulty was Galileo.
Maybe what we should do is find those scientist who have been vilified and sidelined by big business interests and governments and see what they have to say. They wont have the models but they will have knowledge that has been hard won and probably more accurate then many today. 2 scientist that I can think of right now are Michael Mann and Hansen.

[wili]

LRC, are you saying you don't trust the Cohen study for some reason? It is pretty obvious that more open sea water would make for more evaporation would make for a snowier winter somewhere in the Northern Hemisphere.

On the relation of models to reality (map vs territory), there is some nice discussion going on at realclimate right now: http://www.realclimate.org/index.php/archives/2013/09/on-mismatches-between-models-and-observations/

The conclusion actually mentions the mismatch between the modeled and the actual rate of Arctic sea ice loss, and blames the discrepancy on "model structure" but claims that the latest modeling program shows improvement (it would be rather embarrassing if it were worse than the wildly erroneous ones it replaced.)

[ChrisReynolds]

LRC,

I don't understand the problem you have with the Cohen study, which is actually just the latest in a long series of papers on the influence of Eurasian snow-line advance during autumn on the winter atmosphere. The recent Kosaka/Xie paper showing a role for the ENSO in cooler NH winters and the wider lack of GW this century notably predicts a warming during winter over much of the same region Cohen cites as being due to snowline advance. That can be taken as additional evidence that Cohen is correct. FWIW I'm working on a blog post on that matter right now.

[LRC1962]

Sorry reread it again. seems my brain was miss firing on all cylinders when trying to get what it was saying. Got it now and do agree with it.

[Steven]

Re: Judah Cohen's work.  Two of his papers were cited in the new IPCC report, more precisely in Chapter 2 (links added):

Much interest has focussed on the period since 1998 and an apparent flattening (‘hiatus’) in trends, most marked in NH winter (Cohen et al., 2012)

Cohen et al., 2009 reported an increase in the number of Arctic sudden stratospheric warmings during the last two decades. However, interannual variability in the Arctic Polar Vortex is large, uncertainties in reanalysis products are high (Tegtmeier et al., 2008), and trends depend strongly on the time period analysed (Langematz and Kunze, 2008).

The 2012 paper by Cohen et al. was also cited in the influential new paper by Kosaka and Xie, and in a blog post about it by Tamino.  Maybe this will help to bring more attention to the physical mechanism proposed by Cohen et al.: the causal connections between Arctic sea ice melt, advances of Siberian snow cover, sudden stratospheric warming, etc.  I think his theory makes a lot of sense.