Click images for more details



Recent comments
Recent posts
Currently discussing

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > Evidence, confidence and uncertainties

Dec 6, 2011 at 9:38 PM | Philip

Hi Philip

Thanks for highlighting this very interesting paper!

it seems to me that the GCMs hit problems of realism in exactly the regime where we would want them to be most reliable and robust!

I don't think that is quite what Lovejoy and Schertzer are arguing. They seem to be noting that there are aspects of variability in the data that are not captured by the models in terms of internal variability alone but this does not preclude a better match between models and data when external forcing (which could be either anthropogenic or natural) is included. Indeed they specifically say at lines 112-114:

The aim of this paper is to understand the natural variability so that the important question of whether or not GCM’s with realistic forcings might be able to reproduce the low frequency climate regime is outside our present scope.

So it seems that this paper is not commenting on the realism of the models, more on whether the observed spectrum of variability in the data can be explained in terms of internal variability (as emergent from climate models) alone.

Thanks for the pointer though - it's a fascinating paper!

Dec 9, 2011 at 12:18 AM | Unregistered CommenterRichard Betts

Dec 7, 2011 at 6:14 PM | sam

I confess to being on very uncertain ground when I wonder why adjustments are made anyway.

Hi sam,

The adjustments were made in order to remove a bias in the early 20th-century SST data which would otherwise have exaggerated the 20th Century warming.

The issue is that when sea temperatures were measured in those days, they threw buckets over the side of ships to collect water, pulled the bucket back up to the deck and then put the thermometer in the water in the bucket. This meant that the water had a bit of time to cool before the temperature was taken. In contrast, the modern method is to measure the temperature directly at an intake, so it is the actual temperature of the sea. As you can imagine, if the cooling of water with the "bucket" method were not taken into account, the early 20th Century sea surface temperatures would appear to be cooler than they really were, so the warming since then would appear to be greater than it really was.

Dec 9, 2011 at 12:26 AM | Unregistered CommenterRichard Betts


Have to agree with RB: L&S does not support the 'internal variability/random walk' brigade (see my earlier comment at Dec 7, 2011 at 2:29 PM). Eg Koutsoyiannis, Douglass etc. But then, nothing does. Unphysical arguments should not be applied to physical processes. Like the climatically significant effects of CO2 forcing on a centennial scale.

Dec 9, 2011 at 11:41 AM | Unregistered CommenterBBD

"the water had a bit of time to cool before the temperature was taken"

Really? How long are we talking about? Water holds heat rather well and I shouldn't have thought that a gallon or so would cool significantly in the time taken to bring it up on deck and read a thermometer, unless there was a very large difference between the water and air temperatures. Might it not also be warmer on deck?

I'm not convinced by this, especially as water intakes are often further below the surface than the water obtained by dunking a bucket. There may be more significant differences in the temperature layers near the surface, surely?

Dec 9, 2011 at 4:48 PM | Unregistered CommenterJames P

Thanks for your reply, Richard. Again, I see what you mean, but also point out that the paper leaves it as an open question whether the climate ingredient missing from the models is low frequency internal nonlinear interactions, or low frequency climate forcings of some kind. If low frequency climate forcings, then the nature of those forcings is also unknown.

I also notice that you didn't comment on my points regarding (A) The effect of changing parameterizations on model results, (B) The circularity in 20th C attribution arguments claimed by Curry and (C) Implications of the climate regime scaling for the boundary-value vs initial-value question.

Dec 9, 2011 at 5:05 PM | Unregistered CommenterPhilip


Thank you for your reply.

I'll try to do better. Kennedy 2011 uses assumptions for which not much evidence is put forward.

McIntyre stated it this way:

" The new HadSST3 dataset still contains some seemingly arbitrary assumptions. They assert that 30% of the ships shown in existing metadata as measuring SST by buckets actually used engine inlet and proceed to reallocate the measurements on this assumption:

It is likely that many ships that are listed as using buckets actually used the ERI method (see end Section 3.2). To correct the uncertainty arising from this, 30+-10% of bucket observations were reassigned as ERI observations. For example a grid box with 100% bucket observations was reassigned to have, say, 70% bucket and 30% ERI.

The supposedly supporting argument at the end of Section 3.2 is as follows:

It is probable that some observations recorded as being from buckets were made by the ERI method. The Norwegian contribution to WMO Tech note 2 (Amot [1954]) states that the ERI method was preferred owing to the dangers involved in deploying a bucket. This is consistent with the rst issue of WMO Pub 47 (1955), in which 80% of Norwegian ships were using ERI measurements. US Weather Bureau instructions (Bureau [1938]) state that the \condenserintake method is the simpler and shorter means of obtaining the water temperature” and that some observers took ERI measurements \if the severity of the weather [was] such as to exclude the possibility of making a bucket observation”. The only quantitative reference to the practice is in the 1956 UK Handbook of Meteorological Instruments HMSO [1956] which states that ships that travel faster than 15 knots should use the ERI method in preference to the bucket method for safety reasons. Approximately 30% of ships travelled at this speed between 1940 and 1970.

The safety reasons arise, according to Kennedy, from the severity of the weather. It is not just the speed of the ship but the weather conditions that are likely to affect what happens in practice. As Kennedy reports, around 1955, the bucket contribution drops to 40%, presumably for safety reasons relating to WW II. After that bucket use rises to a peak around 1965 after which there is a general decline in its use. In seems questionable that there was a 30 plus or minus 10 % drop in the use of bucket measurements, the evidence for which is a HMSO recommendation and the record of ships that could travel in excess of 15 knots.

Kennedy 2011 says that the start of the general transition from bucket to use of inlet engine temperature measurement was between 1954 and 1957. Paltridge, not cited in Kennedy, says the change was not until after 1960.

If one sits in a dinghy and dangles a thermometer in the sea, one will end up with a reasonable measure of temperature. If temperature is measured by scooping up some water in a bucket before measuring the temperature on deck the measurement will have more error than that done when sitting in the dinghy. There are more things that can change. Intuitively, I would say that measuring temperature by bucket is likely to have less error than measuring by inlet engine use.

I know the reasons put forward for making the adjustments. But they may add more error than the plus or minus error one gets by using a bucket.

Dec 9, 2011 at 5:18 PM | Unregistered Commentersam


You are already aware of my confusion over sec 3.1 of Kennedy.

Thank you for your kindness, patience and unfailing civility.

I came across this post at wuwt: "Buckets, inlets. SST's (sic) and all that.."
Among the comments is this:

"Jon Jewett says:
May 31, 2008 at 10:30 am

I am a retired merchant marine engineer. I made my living on merchant ships from 1966 to 1999, when I retired as Chief Engineer of a 32,000 SHP steam turbine powered container ship.

A ship moves through the water at between 12 knots and 22 knots. (Twelve knots for an old Liberty Ship from WWII, 22 knots fairly representative of new ships.) At even 12 knots, the speed is fast enough to cause turbulent flow of the sea water and thereby prevent a “warm” layer forming along the ship. In addition, very little of the ship is a significant heat source; only the area in way of the machinery spaces. Most of the ship is thermally inert cargo space that may be warmer or colder than the sea temperature. The main cooling water pump (Main Circulator) takes suction through the hull at a depth of 15 feet to 60 feet (depending on the ship) below the surface of the water. In addition, a Main Circulator will pump up to several thousand gallons per minute.

In short, the concept that the water would be heated by a measurable amount is far fetched at best. The statement indicates that assumptions were made about the measurement process and there was no meaningful investigation before these opinions were voiced.

Actually, there is cause for concern in using these numbers. Typically, either alcohol type bulb thermometers or bimetallic type thermometers were used. They were calibrated at the factory before they were installed, but never after that as long as the readings were “close enough”. It was not until the late ‘60s that thermo-electronic (e.g. RTD and Thermister) methods started to enter the fleet. It is to be expected that, particularly with the older readings, each one may be on the order of up to 10 degrees off, with a random distribution.


Steamboat Jack"

I began to wonder how much weight, if any, could be placed on such anecdotal evidence. It should not be too difficult pethaps to obtain more of it.

There is this post at Judy Curry's site, Climate Etc: " Unknown and uncertain sea surface temperatures" which I think you would find interesting.

This subject will surely come up again once BEST releases its work on SSTs. McIntyre will not have forgotten the questions he raised.

May I suggest that you spend more time at climateaudit than you have been doing? There is a series of posts on BEST land surface temperatures which I am sure would be of interest if you have not already seen it. Here are the titles: BEST Data "Quality"; Help Robert Rohde locate Argentina; BEST Menne slices and Closing Thoughts on BEST.

There is a post on wuwt by Willis Essenbach which explores the possibility of the thumb on the scales of a part of the land surface temperature record: "The Smoking Gun at Darwin Zero." Dec 8 2009.

McIntyre has analysed all of Mann's work. Mann's 1999 study which you first mentioned is the subject of a good post on this site by Andrew: " A good trick to hide a decline". It uses Mcintyre's analysis to tell a lot about the way Mann approaches his work.

You mentioned Folland. He was aware of the original Briffa graph proposed for TAR and discussed the problem this might cause in diluting the message. He knew that data had been deleted from Briffa's work to "hide the decline".

There is an excellent post by McIntyre called: "Hughes and the Ababneh Thesis." It seems to me to throw light on some of the matters of my initial questions. It shows again why paleo studies are problematic and it helps resurrect ( if that was needed) the MWP.

I looked for this on William Briggs's (statistician to the stars ) site "The CRU "climategate" proxy code.."

"Sources of reconstruction uncertainty

Here is a list of all the sources of error, variability, and uncertainty and whether those sources—as far as I can see: which means I might be wrong, but willing to be corrected—are properly accounted for by the CRU crew, and its likely effects on the certainty we have in proxy reconstructions:

Source: The proxy relationship with temperature is assumed constant through time. Accounted: No. Effects: entirely unknown, but should boost uncertainty.
Source: The proxy relationship with temperature is assumed constant through space. Accounted: No. Effects: A tree ring from California might not have the same temperature relationship as one from Greece. Boosts uncertainty.
Source: The proxies are measured with error (the “on average” correlation mentioned above). Accounted: No. Effects: certainly boosts uncertainty.
Source: Groups of proxies are sometimes smoothed before input to models. Accounted: No. Effect: a potentially huge source of error; smoothing always increases “signal”, even when those signals aren’t truly there. Boost uncertainty by a lot.
Source: The choice of the model m(). Accounted: No. Effect: results are always stated the model is true; potentially huge source of error. Boost uncertainty by a lot.
Source: The choice of the model m() error term. Accounted: Yes. Effect: the one area where we can be confident of the statistics.
Source: The results are stated as estimates of β Accounted: No. Effects: most classical (frequentist and Bayesian) procedures state uncertainty results about parameters not about actual, physical observables. Boost uncertainty by anywhere from two to ten times.
Source: The computer code is complex. multi-part, and multi-authored. Accounted: No. Effects: many areas for error to creep in; code is unaudited. Obviously boost uncertainty.
Source: Humans with a point of view release results. Accounted: No. Effects: judging by the tone of the CRU emails, and what is as stake, certainly boost uncertainty.

There you have it: all the potential sources of uncertainty (I’ve no doubt forgotten something), only one of which is accounted for in interpreting results. Like I’ve been saying all along: too many people are too certain of too many things."

I have little to say on the uncertainty of models. That they are unable to model convection is well known. They may never be able to model convection. One might never be able to say anything much about their reliability.

I think it is you who should have been referring literature to me. Forgive me, please, for my presumption. I know that you visit climateaudit less frequently than you visit here. I know that you have said you are not that familiar with paleo stuff. Make what you will of the posts I mention. There is no need to make any reply. When I posed my questions I thought I might not contribute much if anything to what followed. It was a strange thread for me. Not a lot on topic and disfigured by frequent outbreaks of incendiarism round which one had to step. Having squabbled at the back the bad boys went outside for a real fight.

Best of luck Richard. Embrace McIntyre. He will talk to you if you wish, I am sure.

Dec 11, 2011 at 10:50 AM | Unregistered Commentersam


I have no doubt you mean well - and we all create the odd rambling, lengthy post - but it would be good to:

1. Keep the posts shorter and focused, and
2. Refer to WUWT, CA, Judy's blog. Sks, RC etc with a synopsis and a link

Many of us read the blogs you mention and, while it is great to call attention to relevant content, yours could be a bit tidier.

I'm no paragon myself. And I wish I could work out the bloody HTML block quote thing.

Dec 11, 2011 at 11:01 AM | Unregistered CommenterGixxerboy


Thanks for the constructive criticism, all accurate.

I am not a frequent poster anywhere, nor do I intend to be. You may have guessed the first part of that already.

When I raised my questions I did not think it would necessarily mean my involvement in a thread and perhaps I should not have involved myself. I did and found myself reading more than I have done. My participation in this blog and others is not that deep. In posting I put my time constraints ahead of clarity. Rude to the readers perhaps (apology, Richard) but rudeness may be a characteristic of this thread.

This is my final post I think. I heaved in the kitchen sink. I am sorry about the clumsiness. That said, I have learned a bit more about the state of play of things even if I have not always conveyed it to readers adequately. SSTs seem to rest on very shaky assumptions. Engine inlet measurements may have a cooling rather than warming bias if ship engineers are generally in agreement with Steamboat Jack.


Dec 11, 2011 at 1:35 PM | Unregistered Commentersam


I'm no paragon myself. And I wish I could work out the bloody HTML block quote thing.

It's a doddle. In this example, the chevrons ('horizontal Vs') are replaced by square brackets. To get it to work, just use chevrons instead of square brackets. Try it out me on the other discussion thread later - I won't care if it goes a bit wrong.

[blockquote]This text will appear as an indented, coloured quote.[/blockquote]

Dec 11, 2011 at 7:08 PM | Unregistered CommenterBBD


Do please stick around. I know it's tough finding the time but you seem to have an interest. It's a shame you've encountered rudeness. BH certainly wasn't like that when I first was drawn here, in fact it was the civility and standard of debate that attracted me.

As you might have seen, some of us (including the Bish himself) would like to improve things. It's not easy - we all have to do something positive. Perhaps, if you stay involved, you can help :-)

Dec 12, 2011 at 5:43 AM | Unregistered CommenterGixxerboy

Hi sam

Sorry for the silence again - I've been travelling.

Thanks for your further comments. While I think that including a correction for the known bias of using buckets, even if it is an estimate and hence uncertain, is better than ignoring it altogether. However I agree that there are probably further improvements that could still be made, and I doubt if Kennedy at al will consider their latest paper to be the last word on the matter (even though it is a major improvement over previous work).

The same thing is probably also true of the palaeo reconstructions - the authors have done the best jobs they can so far, but there is still room for further improvement especially in quantifying and narrowing the uncertainty.

The key thing will be how this is all dealt with in IPCC AR5. I'm currently at the IPCC Working Group 2 lead authors' meeting and there has been much discussion of the clear presentation of uncertainties, especially with reference to the IPCC guidance note on treatment of uncertainties. There is a lot of thought and effort going into this now, as in AR4 there were some inconsistencies in how it was done. However, things are already improving: careful reading of the IPCC Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation ("SREX") shows that uncertainties are indeed already being treated more consistently than in AR4. Hence I am optimistic about AR5 being clear about the estimated probability of different changes, and about which conclusions have high confidence attached to them and which have lower confidence.

Dec 13, 2011 at 5:19 AM | Unregistered CommenterRichard Betts

Dec 9, 2011 at 5:05 PM | Philip

Hi Philip

Sorry I didn't get to your other points yet.

On the circular argument, I do agree that there is a risk of circularity if the only justification for the usefulness of the models is the comparison against past change in a retrospective manner, but fortunately this is not the only source of (cautious) confidence in the models. Comparison with current climate (irrespective of any actual change and its causes) is important, and so is the fact that earlier versions of the models made projections that were shown to be fairly reasonable.

On changing parameterisations, yes if we tune certain parameters within realistic ranges then this does affect the climate sensitivity quite substantially. Although results of some attempts at this have ended up with ranges of between about 1 and 10 C for climate sensitivity, although the higher extremes are regarded as very unlikely. Climate sensitivity can of course also be estimated from observational data, and the resulting probability distribution can help constrain the models.

On the boundary value vs. initial value problem. I think Roger Pielke Snr has been shown to have a point when he said future climate projections were not just a boundary problem, as the initialisation is important for forecasting (or attempting to forecast!) internal variability for decadal forecasts. However, for the longer term projections the exact initial values still seem to become less relevant because chaos means that the predictability of the internal variability drops off a after a few years, so the approach is to use a large ensemble of projections and initialise the models with a range of different initial values so that different trajectories of variability are followed and the range of possibilities at any one time can be explored. However in the longer-term, the external forcing such as increasing GHG concentrations (i.e.: the boundary conditions) become relatively more important if the forcing continues to increase.

Dec 13, 2011 at 5:43 AM | Unregistered CommenterRichard Betts

Thanks for your answers, Richard!

I've not been able to access the body of the 70s paper you mentioned that contained the correct forecast. Are you able to make a copy available to me?

On the boundary value vs. initial value problem. ... If the climate regime didn't exist, then I agree that it would be possible to flatten out longer-term fluctuations by averaging. However, from the L&S paper it looks as though a climate regime does exist, and hence for times greater than 10 or so years, the fluctuations increase with scale, and so cannot be averaged away. Whatever the source of these fluctuations, their existence means that on the climate relevant scales of 100s or 1000s of years, the temperature will nearly always be trending either upwards or downwards, and hence an averaging process is not robust. Since AR4 models are calibrated to approximately reproduce 20th C temperatures, and since they do not reproduce the longer-term fluctuations, they must therefore almost certainly either over-estimate or under-estimate the impact of CO2 over this period.

Dec 14, 2011 at 10:18 AM | Unregistered CommenterPhilip


You are over-focussed on the models. Don't forget paleoclimate. There's plenty of corroboration for the emergent value of climate sensitivity derived from modelled studies. See figure 3a for example, in Knutti & Hegerl (2008). Or the commendably clear approach taken by Hansen & Sato (2011).

Or re-visit Annan & Hargreaves (2006) - discussion on JA's blog here.

Dec 14, 2011 at 11:22 AM | Unregistered CommenterBBD

BBD, I'm not forgetting paleo-climate, it's simply that at the moment we are discussing models. Again, let's wait and see what Richard has to say.

Dec 14, 2011 at 11:31 AM | Unregistered CommenterPhilip


I finally found a copy of Sawyer's 1972 paper. I agree with you that it's a very good paper, but disagree that it provides a convincing argument in support of the "very likely" assessment of the statement "[m]ost of the observed increase in global average temperatures since the mid-20th century is very likely due to the observed increase in anthropogenic greenhouse gas concentrations."

Reading the paper, I see that Sawyer's estimate is based on the 1967 result of Manabe and Wetherald, who take into account the warming effect of CO2 and the associated water vapour feedback to obtain a 2.4 C estimate for the 2xCO2 rise; Sawyer obtains his 0.6 C value by assuming a 25% rise in CO2 levels by 2000. However, he is also very careful to point out that the Manabe and Wetherald calculation (and hence his estimate) does not include the effect of other important processes:-

1/ The effect of increased water vapour on cloud cover, which constitutes "a negative feedback to counterbalance the positive feedback arising from water vapour". He points out that, a "change of only 1% in the average cloudiness would produce a change in temperature of almost 1 C."
2/ The thermal inertia of the oceans, saying that "the oceans impose a substantial lag on the response of world temperature to such changes as I have discussed here."
3/ The warming or cooling effect of changes in the Earth's permanent cover of ice and snow.
4/ The effect of changes to circulation systems resulting from a change in temperatures.

As far as I know, neither #3 or #4 require Sawyer to modify his estimate for changes by year 2000. However, this is not the case for #1 and #2. If they are both accounted for, then Sawyer's 0.6 C estimate must be reduced by some unknown but potentially significant amount. Since the actual observed rise from the early 1970s has also been ~0.6 C, then I think a more reasonable conclusion from Sawyer's paper is that natural causes also made a significant contribution to the warming, in contradiction to the IPCC's statement.

Dec 17, 2011 at 11:02 AM | Unregistered CommenterPhilip

Hi Philip

Really sorry to take so long to notice this excellent post. Thanks for reading the paper.

I think #3 would require the projections to be altered and would be a positive feedback through decreased surface albedo (northern hemisphere snow cover and Arctic ice extent have both generally declined in the long-term average, exposing the darker underlying land / ocean surface). So a positive feedback opposing the negative ones you mention.

Jan 11, 2012 at 11:25 PM | Unregistered CommenterRichard Betts

Hi Richard,

No problem, thanks for your reply, I appreciate it very much. At the risk (certainty!) of being a pain, can I also point you to my comment above @ Dec 14, 2011 at 10:18 AM?

Regarding Sawyer, unfortunately he doesn't quantify any of the additional factors over the period in question, so I can only guess whether he had thought that ice/snow-feedback related albedo changes would have had a significant impact. My point is that Sawyer's 0.6 K looks like being fortunate rather than anything more significant, and this makes his paper far less convincing to me as an argument against a natural causation of late 20th century warming.

Jan 12, 2012 at 7:00 AM | Unregistered CommenterPhilip


this makes his paper far less convincing to me as an argument against a natural causation of late 20th century warming.

Why does the RF from anthropogenic CO2 appear to have no effect on climate in your view of physical reality? I mean, does it just disappear, or what?

Jan 12, 2012 at 3:44 PM | Unregistered CommenterBBD


In my view, the RF from anthropogenic CO2 does have an effect on climate, absolutely no doubt about it.

Jan 12, 2012 at 6:04 PM | Unregistered CommenterPhilip

In all this time, I can't actually recall asking you what estimate of equilibrium climate sensitivity to a doubling of CO2 you favour. Would you mind indicating a value?

Jan 12, 2012 at 8:45 PM | Unregistered CommenterBBD

I'm very happy to tell you what I currently think, thank you for asking - although to be honest I'm still busy trying to form my own detailed picture about this issue. One thing I am confident about, is that sensitivity is associated with a linearization of the equations of motion, and that therefore is only meaningful whilst that approximation remains valid. Therefore, I'm doubtful about the meaningfulness of the idea of a CO2 doubling sensitivity, but happy with the kind of transient sensitivity treated by Spencer and Lindzen. With this caveat in place, then inside the linear regime, I'd expect sensitivity to be low as per those authors; outside, I'd expect system variables including temperature to change quickly under forcing before settling down again into a new equilibrium. I'm well aware that a low sensitivity within the linear regime may not help very much if CO2 emissions end up pushing the system away into an unfortunate place ... which is just another way of explaining why I would still feel happiest to see a decarbonization of the energy supply!

Jan 12, 2012 at 10:21 PM | Unregistered CommenterPhilip

The rest of us are happy enough with a transient sensitivity that leads to an equilibrium sensitivity. It is the latter value that will help most here - if you would be kind enough to provide your preferred estimate. I don't want to put numbers into your mouth.

Jan 13, 2012 at 8:34 AM | Unregistered CommenterBBD

I'm not sure I understand your point this time, although I appreciate that you don't want to put numbers into my mouth. If you reread my previous response, you'll see that I've already given you my answer.

Jan 13, 2012 at 1:37 PM | Unregistered CommenterPhilip