Quantifying Uncertainties in Climate Science

Another date for your diaries - the Royal Met Soc's meeting on uncertainty in climate science.

Climate models produce different projections of future climate change under identical pathways of future greenhouse gases. This meeting will highlight recent studies that have attempted to quantify those uncertainties using different approaches.

Programme: 

Time	No.	Presenting author	Title
16:40		Prof Reto Knutti, (ETH Zürich)	Projection uncertainties: The multi model perspective.
17:10		Dr Paul Williams, University of Reading.	Climate models: The importance of being stochastic.
14:10		Dr Jonty Rougier, University of Bristol	Background and philosophy
14:40		Dr David Sexton, UK Met Office	UK climate projections.
15:10		Dr Tamsin Edwards, University of Bristol	Palaeo-constraints on climate sensitivity.
16:10		Dr Lindsay Lee, University of Leeds	Constraining aerosol models.

Details here.

Comments

[sam]

Tamsin

I hope that you and Richard will answer my question about the use of Mann and Jones 2003 and Mann Bradley Hughes 1999 by modellers. It may be a silly question which only requires "No" as an answer.
Thanks

[Richard Drake]

Tamsin:

I note down about twenty times more things to read or think about than I get round to following up.

Which reminds me, I wanted to pick your brains on Evernote, if you're still using it. I'm still not sure I can make it next Wednesday - and there probably wouldn't be time then anyway - but something else for us both to note down! I'm particularly interested in how you find as a pretty technical user.

[HaroldW]

Tamsin -
Thanks for the response. It might make a good topic for your blog - a list of parametric and structural uncertainty areas, possibly arranged in (your opinion of) decreasing order of their effects upon GCM accuracy. [You might choose a more specific criterion.] The idea being to stimulate thought into how to work at reducing the more important factors.

[Hilary Ostrov]

So while we are all waiting for Richard Betts to find some time in his busy life to respond to one or more of the questions that have been put to him (assuming he didn't miss them twixt his 5.600+ tweets to date) - and/or address the responses to his questions to others - as a Bridgeplayer, I thought it might be useful to review the bidding, so to speak.

OK,OK, I know, I know ... this doesn't even begin to approximate the relevance to this thread of Richard D's interest in Tamsin's opinion of Evernote. But humour me, folks!

In response (for want of a better word) to my comment of Dec 6, 2012 at 11:04 PM, which, after presenting my evidence, I had concluded with:

So anytime you'd like to apologize for misleading Richard B. and quite possibly others - and/or attempting to boost your own credibility by diminishing mine - will be just fine with me.

Then perhaps we can all "get over it" ... at least until the next time;-)

Richard Drake wrote [Dec 7, 2012 at 9:22 AM]:

Hilary: your original comment wasn't to me, I mentioned you only in passing and you shouldn't have bothered with me further.

Is this an apology I see before me?! Not bloomin' likely!

But, just for the record, lets take a look at this "mention only in passing" that I gather I am supposed to have ignored in accordance with Richard's Rules of Order™

Richard Betts seems to have made Nic Lewis priority so far. I personally think that's a good call. Unlike Hilary Ostrov I didn't interpret one of Richard's terse answers as dismissive of Nic; quite the reverse, I thought it showed a busy man was really trying to help [Richard Drake, Dec 5, 2012 at 8:41 PM]

If I might quote someone or other, "Wow. Just wow". But moving right along ...

Some might wonder which subsection(s) of Richard's Rules of Order™ determine which of his pronouncements are worthy of further mention and which are not. But I couldn't possibly comment. Although it is quite apparent that - in the rules of his game, whatever it might be - Richard's reframing and interpretation will trump anyone's actual words any time he chooses.

Interesting bit of "revisionism" twixt Dec. 5 and Dec. 7 on his part, though! Who would guess that - during the course of this mere 'mention in passing' - he had taken the liberty of misrepresenting my words, wrongly summarizing them in his interpetation that I had, in effect, suggested another poster was being "dismissive". But, according to Richard's Rules of Order™, I'm supposed to have ignored this, evidently, because my "original comment wasn't to" him.

Amazing. Simply amazing.

Others' MMV, but I find it quite astounding that one who feels free to ride his hobby-horse into whatever thread his little heart desires - and takes extreme umbrage whenever he believes that another poster might have misrepresented his words - should fail to recognize that others at least have the right to request that he refrain from misrepresenting theirs.

[Richard D. had continued:]

Read Richard [B] carefully and learn that less would be have been more for him too.

Considering that we are discussing what turned out to be an unnecessary 201 word paragraph on a bureaucratic procedure, ironically, less on his part would certainly have been more, wouldn't it?! But, frankly, because I come from the school of "if it's worth doing, it's worth doing well and if I'm going to respond to a post, I should read it, rather than skim it", I'd rather not learn from him, thank you very much! Consequently...

I did read Richard B. carefully. Obviously far more so than you seem capable of doing. But here, take another look at that paragraph [Richard Betts Dec 5, 2012 at 1:15 PM] you've chosen to ignore:

Sorry, in my haste I genuinely thought I was addressing the most important point in advising Nic on how to get his voice heard (following Don's comments about "silencing" earlier). [emphasis added to assist Richard D. in his "reading" -hro]

Is there something about "the most important point" that you do not understand, Richard? The "most important point" from his reading was explaining a bureaucratic procedure because someone else had mentioned "silencing". Ooops ... did I say "reading"?! Sorry ... skimming as he seems to prefer because it's only a blogpost, so accuracy doesn't really matter, regardless of how much thought might have been put into the post to which he was responding.

Which raises another point. I went back through the thread looking for "Don's comments about 'silencing' earlier." because I had no recollection of Don making any mention of "silencing" in the context of any Reviewer Comments he might or might not have made! Here's what I found:

Professor Phil Jones tried to get me "silenced", before saner minds prevailed

and

I myself have had climate "scientists" discuss how to "silence" me and "make it difficult" for me at my place of work, simply because I asked questions of them.

Nope. Nothing about Reviewer Comments in either of these. I did, however, find the following in a comment addressed to Don:

You talk about people trying to "silence" you - well, I've asked you several times if you would consider being an Expert Reviewer for the IPCC 5th Assessment Report, particularly my chapter on terrestrial ecosystems

Now, some might ask, is this guy fan of Poor Phil, or it that he just doesn't get it?! But I couldn't possibly comment. And I'll bet you'll never guess who wrote it.

Bottom line: "Don's comments about silencing" had absolutely nothing to do with IPCC Expert Reviewer Comments reaching the right parties, i.e. that which supposedly led Richard B to make his "most important point".

So if you - or anyone - want to accept such justifications for such choice of posting behaviours, by all means be my guest. But I am no more obliged to accept Richard B's choices without comment - than I am to accept yours. And I don't.

I retain a high regard for your contributions in other areas.

I honestly wish I could reciprocate, Richard. However, over the past few months, I'm sad to say that you've offered fewer and fewer reasons for me to do so.

And, that's a decline you cannot hide! Although I must say that - while very disappointing - it's been quite fascinating watching the two Richards' attempts to waltz away from their very own words in this particular thread.

[mct]

Once in a while we get a thread where the "mainstream" choose to engage, and people try to derail it.

Why?

[Peter Lang]

mct,

I agree. What a pity this excellent discussion is being disrupted.

[Richard Betts]

Dec 7, 2012 at 12:49 PM | sam

Hi Sam

If your question means "do climate modellers use palaeoclimate reconstructions to test or develop the models", the answer is, not to my knowledge. Sometimes the models and reconstructions are compared with each other as alternative ways of figuring out what might have happened in the past, but in my experience the reconstructions are generally regarded as being too uncertain to usefully constrain the models. Also, the proxies that are used in the reconstructions are often more representative of individual locations, and assumptions need to be made to scale these up to large areas such as the northern hemisphere average, whereas the models have the opposite problem - they are more reliable over large scale averages and at small scales the uncertainties get larger. Hence it is quite tricky to reliably compare models and proxies in a like-with-like way.

Moreover, the papers you cite are now very old, especially MBH99.

[sam]

Hello Richard

Thank you for your reply.

Hegerl may not be a modeller but she uses Mann and Jones 2003 about which I was curious, in this study: Constraints on climate sensitivity from temperature reconstructions of the past seven centuries. The paper can be found here:
http://www.geos.ed.ac.uk/homes/ghegerl/Hegerl_et_al_sensitivity.pdf
Mann and Jones 2003 is a "hide the decline" paper and there are difficulties about the other reconstructions used in the paper if my memory of McIntyre's deconstructions are accurate.

[Richard Betts]

Thanks Sam - I stand corrected.

Maybe Tamsin will talk about updates to this kind of work at the RMS tomorrow? As I said, the papers you mention are quite old now.

[Richard Betts]

@Nic Lewis

Has the paper you submitted to J. Clim been published yet? I can't find it online, although I see it was only submitted in July so I guess it's a bit soon. Has it been accepted? If so, can I get a copy please? Thanks!

[sam]

Hello Richard

How is "climate" defined - physically or mathematically - in "climate sensitivity?

[Richard Betts]

Dec 11, 2012 at 1:28 PM | sam

"Climate sensitivity" is defined as the change in global mean temperature for a given radiative forcing (conventionally the forcing associated with a doubling of CO2 relative to pre-industrial), so in this context "climate" is merely global mean temperature. However of course climate is much, much more than that - the whole radiative forcing / climate sensitivity conceptual model is very simple (some might say too simple!)

[sam]

Thanks, Richard

[John Shade]

Quote from Dr Betts (Dec 11, 2:05PM)
'the whole radiative forcing / climate sensitivity conceptual model is very simple (some might say too simple!)'

I hope I am not reading too much into this by hoping it means that this peculiar model is undergoing fundamental review amongst relevant experts.

[sam]

Richard

Will you be my mentor for a little longer, please? What does radiative forcing mean? What is meant by change in global mean temperature? In practice what does this mean - where and how measured, for example? Thanks

[Ruth Dixon]

The RMS meeting was very interesting and it was great to meet Tamsin Edwards, Jonty Rougier, Josh, and others.

Tamsin's talk was essentially an update of her post from April 2012: http://allmodelsarewrong.com/a-sensitive-subject/ . She described how she (with her co-authors) are using two paleoclimate eras (the mid-Holocene, 6000 years ago, and the Last Glacial Maximum 21,000 years ago) to constrain climate sensitivity. The result (the climate sensitivity itself) is still to come.

This method has the advantage of testing models that have been created to simulate present-day climates against completely independent data, that is, reconstructions of past climates. Otherwise you risk 'double-counting' (as David Sexton explained in his talk) - both optimising the model to match the present climate and also validating the model against the same observations. His talk explained how attempts are made to get round that using only contemporary data but there is always an element of circularity (e.g. testing one model against others that also use the same data). But of course the details of paleoclimates are less well known than the present climate, given the sparse distribution of proxies, etc.

(I hope others will correct me if I've misrepresented anything!)

In answer to John Shade (Dec 12, 2012 at 1:03 PM) - I did not get the impression that there is a major re-think of the validity of the concept of climate sensitivity. Though as Tamsin pointed out in her talk, it is not a well-defined concept. For instance, should one consider timescales of 100s or 1000s of years for 'equilibrium' to be reached following a change in forcing (as in 'equilibrium climate sensitivity')?

Altogether an interesting afternoon, and I'm sorry I couldn't come to the pub afterwards.

You can read Tweets from the meeting using #QUCS.

[Ruth Dixon]

The meeting's Tweets have been collected here:

http://storify.com/ramblemuse/rmets-quantifying-uncertainties-in-climate-science

The popularity of the meeting can be judged by the fact that the mulled wine was only enough for about 1/4 of the audience (it ran out long before I got out of the lecture theatre). :-(

[John Shade]

Thank you for the report from the meeting, Ruth Dixon. I wish I had been able to get to it.

Thank you also for addressing my query about the computer/conceptual forcing model. I note what you say about the sensitivity concept, but my main puzzle is over how these models are set up to estimate the effects of increasing CO2 levels (and other changes in atmospheric composition such as aerosol amounts). Bear with me while I try to spell that out in a way which I hope all our readers will find accessible (if not necessarily convincing!). I would very much appreciate corrections and other criticisms since I am very aware that my view is not the result of deep study but more of impressions gained and fragments of info taken in over recent years.

The method used by the climate model programmers is ‘top of the atmosphere forcing’ and I think they use it because it makes modeling tractable within current computer capabilities. The problem I have is with the back-to-front nature of this device, a device which does not attempt to match how the system actually works but assumes that by working backwards from a presumed effect (the net 'forcing' introduced at the top), the models will settle down to a reasonably realistic state that will by some act of faith represent a real atmosphere's behaviour, esp. down near the surface where we are.

The real atmosphere works differently, The radiation budget at the ‘top’ of the real atmosphere is a result, an effect, of a spherical planet having very intense incoming solar heating per unit area at the tropical and subtropical surfaces, plus atmospheric and ocean flows transferring heat polewards. The curvature of the earth means relatively modest incoming heat per unit area from the sun as the poles are approached. This results in more heat in than out at low latitudes and more heat out than in at high ones These 'top of the atmosphere' differences in radiation budgets do not drive anything, they are not forces, not 'causes' so much as 'effects'. .

The programmers turn this around, inserting an instantaneous change in this budget as a cause, a driver of, inter alia, temperature changes below it. As I understand it, they install the change (instantaneously at the top of their model atmosphere) and wait and watch (the models do seem to need to be watched because they can go on excursions to produce unacceptable results) while their atmospheres re-adjust ti the disturbance. This is what I mean by the back-to-front nature of this device, and it is not clear to me that it should work. Especially for CO2 changes, which generally begin at the surface in relatively very high concentrations that vary continuously (not instantaneously) in both time and space, and which take a while (at least weeks) to be dispersed and mixed-in, a period in which their contribution is ignored by the GCM modelers.

Now to model that more directly would, I presume, be out of our current grasp on the global scale. But I wonder (as I think does Rhoda who posts comments on this blog) if such a detailed model could be done on a local or regional scale in such a way that testable hypotheses could be produced that observations could confirm or refute. The effect of CO2 is so small compared to other factors in the system (esp. water in all its phases and the changing of these phases, and surface heat driving convective and advective flows) that this might be too much for both our observational and our computational abilities at this stage, but given the colossal sums of money at stake thanks to the political success of the campaigns by people alarmed by CO2 , it seems to me as a somewhat naïve observer that some considerable effort in this area would be highly desirable. Campaigners, though, tell us that the science is settled, and they have been widely believed.

[Richard Betts]

Dec 13, 2012 at 8:17 AM | sam

Radiative forcing is basically the imbalance between energy input and output to/from the Earth, this energy being in the form of radiation received from the sun (some of which is reflected and some absorbed) and emitted by the Earth.

Global mean temperature usually refers to the average temperature of the Earth's surface. It can't be measured directly, only estimated from measurements at numerous points across the surface.

[Richard Betts]

Dec 13, 2012 at 11:16 AM | John Shade

Actually your 4th paragraph is what the models do. Dividing the Earth's surface into several thousand grid squares, at each square they simulate the input of radiation from the sun according to the position on the Earth surface, angle of tilt of the axis and position in it's orbit. This heats the Earth's surface, which exchanges energy with the atmosphere through latent and sensible heat fluxes and the emission of longwave radiation, and the interaction of these with the atmosphere is then also simulated. As you say, transfer of energy from equator to poles is very important, and this is what the models do. All this drives the atmospheric circulation within the model (which is what makes weather forecasting possible - the models are simulating the flow of a fluid on a rotating sphere, influenced by differences in temperature across the surface.) The TOA radiation budget is, as you say, merely a diagnostic not an input.

[John Shade]

Dec 13, 2012 at 1:25 PM | Richard Betts
Thank you for your quick response, which is reasurringly in line with my understanding of the models.
Could you also summarise for us here how the models actually model CO2 changes?
Thanks in advance!

[Nic Lewis]

Richard Betts: "Has the paper you submitted to J. Climate been published yet?"

Not yet, but I am emailing you a copy.

BTW, in my comment posted on Dec 5, 2012 at 3:57 PM in response to one of yours, I used too low an uncertainty range for my observational estimate of climate sensitity. I had misread the aerosol forcing uncertainty as relating to a 5-95% range but that particular figure related to +/- 1 standard deviation. Correcting this mistake, and addressing other forcing etc. uncertainties in a somewhat more realistic manner, increases the upper 95% bound on the climate sensitivity estimate from 2.6 C to circa 2.9 C.

[Richard Betts]

@Nic Lewis

Thanks for sending the paper, and for reminding me about Graeme Stephens - I'd not got round to asking him about his statement about the best estimate of the indirect aerosol effect being zero, but following your email I did ask him and he has responded as follows (and was happy to be quoted):

I may have said that and as I don’t follow blogs I cant say how out of context that is – I think all evidence (but not yet agreed upon by all) is that the indirect forcing is much smaller in reality than is applied in models and there are many reasons for this expectation, among the most important being that there is clear evidence that Twomey effects (so-called 1st indirect effect), and that which is mostly dominant in models, is practically invisible in the real world because of other compensating effects that occur – but we don’t fully understand all these compensating effects at all yet it appears these effects offset any Twomey effect- I think my comment on zero is based more on sound science arguments – we know aerosol can affect clouds, we know in the real world these effects are a net result of various processes not yet understood, some positive and some definitely negative, and as a science exercise we ought to consider a climate projection enterprise without AIE completely – the worrying thing for me is that climate model sensitivity is grossly defined by how small or large an AIE is in models and if this is indeed the case then there isnt much scientific stock to be placed on climate sensitivity at all because its based on processes we don’t understand and are artificially represented in models and I would argue do damage to our ability to understand and diagnose the internal processes that also shape climate sensitvity- as an exercise it would be far better to perform projections with the unknown and not yet understood AIE turned off completely and thus may well be much closer to reality (for reasons mentioned) than projections that are based on inidirect forcings of 1, 2 or more Wm-2 . I am sure this view isnt accepted by all

So yes, I can confirm you are correct about his view.

BTW for any readers who are not aware, Graeme is a Lead Author on the IPCC AR5 WG1 chapter on "Anthropogenic and Natural Radiative Forcing" (Chapter 8).

[John Shade]

I've downloaded the 2nd order draft of that chapter, and I hope it will clarify the peculiar virtual world that turns effects into causes, adds fluxes ad hoc in search of verisimilitude, and whose LA now seems in anxious search of further tweaks for that same admirable end. Since these models have long been used in support of catastrophic 'projections', those who would distance themselves from catastrophism might need to distance themselves from the models as well. Or are they versatile enough to produce more politically acceptable, i.e. more temperate, output as the tide of public/political opinion turns?