tag:blogger.com,1999:blog-9959776.post1661757906732709134..comments2024-02-15T04:42:41.606+00:00Comments on James' Empty Blog: Assessing the consistency between short-term global temperature trends in observations and climate model projectionsJames Annanhttp://www.blogger.com/profile/04318741813895533700noreply@blogger.comBlogger99125tag:blogger.com,1999:blog-9959776.post-2741177124535114612010-12-21T06:31:48.606+00:002010-12-21T06:31:48.606+00:00I think the Heartland presentation was basically t...I think the Heartland presentation was basically the same material as the GRL manuscript, though I am commenting now from memory. I don't think Chip's assessment in that linked comment is unreasonable, though it's just one possible perspective.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-19080837666032324672010-12-20T11:56:34.172+00:002010-12-20T11:56:34.172+00:00Can you compare your description above to the desc...Can you compare your description above to the descriptions of the material recently discussed here<br />http://klimazwiebel.blogspot.com/2010/12/tony-gilland-time-to-move-on-from-ipcc.html?showComment=1292515463147#c3286621<br />(what was submitted to GRL? and what was presented at Heartland?)Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-9670267182673781982010-06-28T04:41:02.240+01:002010-06-28T04:41:02.240+01:00cce,
I agree that in principle accounting for all...cce,<br /><br />I agree that in principle accounting for all the various factors should give a more "pure" comparison of the underlying models and climate system. Another possible tweak would be to back out the ENSO effect from both models and data, so as to get closer to the forced response, and we will probably do at least some of these things.<br /><br />OTOH these models (including the representation of forcings) were presented as state of the art and the temperature time series are also routinely referred to as representing global mean temperature, so I don't think it is invalid to perform the comparison on that basis. Solar should certainly be mentioned as a contributor to the discrepancy, however.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-54857686749508799372010-06-27T06:06:02.304+01:002010-06-27T06:06:02.304+01:00There was some mention of the solar cycle in this ...There was some mention of the solar cycle in this thread, but I think it was shortchanged. The last solar maximum occurred somewhere around 2001 and the solar minimum was last year (I think). Most estimates put the solar cycle worth about 0.1 degrees from peak to trough. Correct me if I'm wrong, but if model runs of the "future" leave out the solar cycle, and the "future" began in 2000, then we would expect observations to be about 0.1 degrees cooler now than they would be if the solar cycle didn't exist, correct?<br /><br />Another point. Not that it would be easy, but it would be better to apply masks to the model output and calculate those anomalies based on the same area covered by the various temperature analyses. That would be a more apples to apples comparison.ccehttp://laymans-guide.comnoreply@blogger.comtag:blogger.com,1999:blog-9959776.post-59977731342937490782010-06-09T03:03:15.392+01:002010-06-09T03:03:15.392+01:00Well the slope over say 50 years is a function of ...Well the slope over say 50 years is a function of CO2 concentration increase, sensitivity and ocean heat uptake. The final equilibrium (assuming concentration stabilises) will be proportional to the sensitivity, but obviously depends on the stabilisation level. The simple models can simulate this sort of behaviour pretty well IMO. <br /><br />When you look at the 10y time scale, there is lots of natural variability to consider, superimposed on this forced response.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-83198627308599905612010-06-08T22:19:24.979+01:002010-06-08T22:19:24.979+01:00say 'Charney' sensitivity in 100 years? a...say 'Charney' sensitivity in 100 years? and that's what the simple presentation is showing, the outcome of the more complicated and expensive models, but not the variability during short time spans?Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-68511577926354911502010-06-08T22:02:09.839+01:002010-06-08T22:02:09.839+01:00I think not exactly, Hank. Remember that what the...I think not exactly, Hank. Remember that what the models are trying to project is the likely response to the parts of the system known to change quickly. As the longer-term feedbacks kick in (generally assumed to require >100 years, although that assumption looks increasingly shaky), there will be a substantial divergence, and as I keep mentioning studies of the Miocene and especially Pliocene warm periods give us a pretty good idea of where we'll end up (the equilibrium conditions, anyway, bearing in mind that the really unpleasant effects would be associated with the transient) if we stay on the present path.Steve Bloomhttps://www.blogger.com/profile/12943109973917998380noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-56262199848303402352010-06-08T05:14:18.492+01:002010-06-08T05:14:18.492+01:00I think part of the answer is that over a sufficie...I think part of the answer is that over a sufficiently long time span, the trend ends up at the climate sensitivity number, right?<br /><br />And in between, there's noise and maybe flipflops as the system sorts itself out?Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-16184829492959971042010-06-08T04:59:53.846+01:002010-06-08T04:59:53.846+01:00Sorry, I got called away on an urgent holiday. Tha...Sorry, I got called away on an urgent <a href="http://www.flickr.com/photos/56924251@N00/4679703547/" rel="nofollow">holiday</a>. Thank you for all carrying on in my absence. Was there anything in particular more I needed to say?<br /><br />Lucia, with reference to those descriptions of the projections, there is a good reason why they are all stated on a 100y time scale - notably, that over such a long time we can basically ignore internal variability, but over 20 years the IPCC only said the trend was likely to be about 0.2C/decade (and even then they did not precisely quantify what that meant). And that statement says very little about the <10y trend.<br /><br />Hank, I think you have found your own answers :-)James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-65816932620233882222010-06-07T16:08:15.232+01:002010-06-07T16:08:15.232+01:00Magnus points out a quote-mine smorgasboard over a...Magnus points out a quote-mine smorgasboard over at DC's place.<br /> <br />> warming is almost exclusively <br />> confined to the dry, cold, <br />> anticyclones of Siberia and <br />> northwestern North America....<br />> Warming of this air mass type<br />> may, in fact, be benign or even<br />> beneficial ....<br /><br />See? It's how you spin it!Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-52701553248322849202010-06-07T12:59:23.366+01:002010-06-07T12:59:23.366+01:00Dude you're late, everyone else went home. I&#...Dude you're late, everyone else went home. I'm finishing off the booze.achttps://www.blogger.com/profile/08817718132877704613noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-75165568762559742072010-06-07T11:40:52.805+01:002010-06-07T11:40:52.805+01:00If some one missed it, some interesting remarks.
...If some one missed it, some interesting remarks.<br /><br />http://deepclimate.org/2010/06/06/michaels-and-knappenbergers-world-climate-report-no-warming-whatsoever-over-the-past-decade/Magnushttps://www.blogger.com/profile/01617272924116099306noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-43651432199158689402010-06-05T16:08:42.376+01:002010-06-05T16:08:42.376+01:00> a figure (which evidently you
> consider ...> a figure (which evidently you <br />> consider to be a "cartoon").<br /><br />"Cartoon" was the picture from the first IPCC report that Wegman reverse-engineered, deriving an imaginary data file that he then analyzed. <br /><br />If I understand what James said about the SPM figure, you should have gotten the data file instead.<br /><br />Arguing over words doesn't help.<br /><br />Bye.Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-18921475738630231552010-06-05T10:12:48.555+01:002010-06-05T10:12:48.555+01:00Lucia said "Because testing whether the MMM i...Lucia said "Because testing whether the MMM is consistent with an observed trend does not involve assuming that the MMM will accurately predict any individual trends. I don't know why you think it does."<br /><br />Because if A is an estimator of B, then there is no reason to expect A to be consistent with C = f(B,D).<br /><br />I don't know how to put it any more clearly than that, if the MMM is not intended to be an estimator of the observed trend (just the forced component) there is no reason to expect it to be consistent with the observations. However, I am willing to be corrected on this point if you can give a good reason why it should be consistent with the observations (other than in the sense that the observations lie within the spread of the ensemble).<br /><br />And yes, I do understand the point of the Santer test; the question is why is it (eqn 12) is superior to seeing if the observations lie within the spread of the ensemble, which AFAICS is the obvious test for consistency of the ensemble.Dikran Marsupialnoreply@blogger.comtag:blogger.com,1999:blog-9959776.post-84602881930586356192010-06-05T03:09:57.853+01:002010-06-05T03:09:57.853+01:00Dikrum
>>If we both agree that the MMM is no...Dikrum<br /><i>>>If we both agree that the MMM is not intended to be an accurate prediction of the observed trend, why are you expressing interest in statistical tests that determine if the MMM is consistent with the observed trend?</i><br />Because testing whether the MMM is consistent with an observed trend does not involve assuming that the MMM will accurately predict any individual trends. I don't know why you think it does. <br /><br />Once again: Do you understand what Santer did in the paper you asked James about? Do you understand what the {s(bo)} terms in his various equations are supposed to describe? (In words, not numbers?)luciahttps://www.blogger.com/profile/12342621789338198739noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-72419173463548098562010-06-05T02:33:41.232+01:002010-06-05T02:33:41.232+01:00>>But this has nothing to do with the paper ...>>But this has nothing to do with the paper being discussed, does it?<br />No. But since you are are trying to make points about things that have nothing to do with the paper being discussed, you should anticipate others might respond to the things you say. <br /><br />Than you for telling me you think the AR4 was clear. But what is your actual answer to the question I asked? Is it that you think a section discussing uncertainty in estimating <i>sensitivity</i> describes <i>projections</i>? <br /><br />Like you, I also think the AR4 is quite clear. Unlike you, who quote text <br />*does not not describe projections (it discusses sensitivity instead), <br />*does not contain the word "projections", and <br />* comes from a section that does not contain the word "projections" in the heading,<br /><br />I will quote from a text that <br />* uses the word "projections", <br />* specifically discusses "projections" of temperature (not sensitivity which is a property, not a projections) <br />* comes from a section that discusses "projections" and finally<br />* is in the single most visible portion of the AR4: that is the supplement for policy makers. <br /> <br /><br />With no further ado, this is a quote from the section entitled "<b>Projections</b> of Future Changes in Climate" in the SPM of the AR4: <br /><br /><i>"Best estimates and likely ranges for global averagesurface air warming for six SRES emissions marker scenarios are given in this assessment and are shown in Table SPM.3. For example, the best estimate for the low scenario (B1) is 1.8°C (likely range is 1.1°C to 2.9°C), and the best estimate for the high scenario(A1FI) is 4.0°C (likely range is 2.4°C to 6.4°C). Although <b>these projections</b> are broadly consistent with the span quoted in the TAR (1.4°C to 5.8°C), they are not directly comparable (see <b>Figure SPM.5</b>). The Fourth Assessment Report is more advanced as it provides best estimates and an assessed likelihood range for each of the marker scenarios. The new assessment of the likely ranges now relies on a larger number of climate models of increasing complexity and realism, as well as new information regarding the nature of feedbacks from the carbon cycle and constraints on climate response from observations. {10.5}</i><br /><br />Note that SPM.5, illustrates "the projections" in the AR4. One could compare these to "the projections" from the TAR-- also provided in the form of a figure (which evidently you consider to be a "cartoon"). Comparison of the two "cartoons" would permit those who read both documents to see the form of the projections is not directly comparable--(as stated in the supporting text). That is to say: examiniation of the figure showing <i>the projections</i> permits you to diagnose a feature of <i>the projections</i>. <br /><br />So, it seems very clear to me that figure and table that the authors of the AR4 tell us describe their projections actually describe "the projections". <br /><br />In contrast, a section you quote which discusses the uncertainties in estimating the magnitude of a property of a quantity called "sensitivity" doesn't come close to telling anyone what the "projections" of anything might be. The sensitivity could be any number whatsoever, without knowledge of all sorts of other things like. at a minimum, heat uptake by the oceans, current forcings, and the entire time evolution of forcings one could not even use knowledge of the sensitivity to create projections.<br /><br />I'm pretty sure you understand the difference between a estimate of the magnitude of sensitivity and a projection. So, I am amazed that you would quote such a thing when trying to communicate what you think the projections are!luciahttps://www.blogger.com/profile/12342621789338198739noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-2424261253597648172010-06-04T23:32:09.472+01:002010-06-04T23:32:09.472+01:00to answer Lucia's question, the AR4 explanatio...to answer Lucia's question, the AR4 explanation seems clear to me. A brief excerpt follows.<br /><br />But this has nothing to do with the paper being discussed, does it?<br /><br />"http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch10s10-5.html<br /><br />10.5.1 Sources of Uncertainty and Hierarchy of Models <br /> <br />Uncertainty in predictions of anthropogenic climate change arises at all stages of the modelling process described in Section 10.1. .... Probabilistic estimates of climate sensitivity and TCR from SCMs and EMICs are assessed in Section 9.6 and compared with estimates from AOGCMs in Box 10.2."Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-29213905407093848742010-06-04T22:27:36.196+01:002010-06-04T22:27:36.196+01:00> Hank--Are you suggesting ....
No. That was a...> Hank--Are you suggesting ....<br /><br />No. That was a question for your coauthor about what you understood.Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-4945542805227375142010-06-04T21:55:29.445+01:002010-06-04T21:55:29.445+01:00"(I would accept my calc is a bit of a hand-w...<i>"(I would accept my calc is a bit of a hand-wave, but I don't think it is really our job here to try to do a detailed uncertainty analysis on the obs to challenge the figures that the originators of these analyses have actually published, such as Brohan et al for HadCRUT.)"</i><br /><br />It seems that the job you have set yourself is to reconcile (or not) the short term variability in the observations with short term variability in the models. The observational error is a component of that variability and if you don't account for it properly, you run the risk of getting the wrong answer.<br /><br />The models typically show their greatest rate of warming at high latitudes, particularly in the Arctic. HadCRUT3 systematically under-represents this area so is very likely to have a trend that is too low. The same is true, to a lesser extent, for the NCDC analysis.<br /><br />If accounted for properly by subsetting the models to where there are observations in HadCRUT3, you won't get a slight broadening of your distribution as you claim, you will probably narrow the range (there's a lot of variability at high latitude) and almost certainly shift the whole thing down (the Arctic warms faster than the rest of the globe).<br /><br />It would be interesting to see the radiosonde estimates of the lower tropospheric temperatures as well to give a more comprehensive estimate of the uncertainty on the upper air trends (a la Santer et al. and numerous similar). The trends in these analyses probably lie closer to the middle of your distribution and if you exclude them it might look like cherry picking.<br /><br />NebuchadnezzarAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-9959776.post-43813756470461055232010-06-04T20:57:35.408+01:002010-06-04T20:57:35.408+01:00I wrote "As I have pointed out the MMM is not...I wrote "As I have pointed out the MMM is not intended as an accurate prediction of the observed trend, just of the forced component of that trend, so there is no reason to expect them to coincide."<br /><br />Lucia said "Yes, you pointed this out and I agreed with you that the MMM is not intended to be an accurate prediction of the trend. I have never thougt it was intended to be an accurate prediction of any individual realization of a trend."<br /><br />If we both agree that the MMM is not intended to be an accurate prediction of the observed trend, why are you expressing interest in statistical tests that determine if the MMM is consistent with the observed trend? As James said "... rejecting the "nill hypothesis" that the MMM coincides with the truth would be a pretty worthless exercise."<br /><br />It seems at least one of us is not explaining themselves clearly.Dikran Marsupialnoreply@blogger.comtag:blogger.com,1999:blog-9959776.post-39209072971129413492010-06-04T20:17:41.690+01:002010-06-04T20:17:41.690+01:00Dikram
>>As I have pointed out the MMM is no...Dikram<br /><i>>>As I have pointed out the MMM is not intended as an accurate prediction of the observed trend, just of the forced component of that trend, so there is no reason to expect them to coincide.</i><br />Yes, you pointed this out and I agreed with you that the MMM is not intended to be an accurate prediction of the trend. I have never thougt it was intended to be an accurate prediction of any individual realization of a trend.<br /><br />It is worth nothing that the test in Santer are why the test in Santer is designed to test whether the MMM is consistent with the <i>forced</i> component associated with the observed trend. <br /><br />You do understand this, right? You do understand why Santer's method is better than Douglas's method, right? <br /><br />>>It seems to me a mistake to concentrate on any point estimate,<br />Of course. I don't know anyone who is currently concentrating on a point estimate.<br /><br />Hank--<br />Are you suggesting that use of SCMs as one of the processing steps to creating projections based on an ensemble of GCMS make the resulting projections presented to the public and policy makers in the "projections" sections of the IPCC report not projections? <br /><br />BTW: I am familiar with SCMs. What I want to know is whether you think the projections actually presented to the public and policy makers in the form of figures, text and tables the sections of the AR4 with headings like "projections" are not "the IPCC projections". If the material presented to the public in these sections are "not the projections", what is?<br /><br />And if the IPCC does not include its projections in the report itself, do you think they should include their actual honest to goodness projections in future reports so that the public can learn what the projections actually are?luciahttps://www.blogger.com/profile/12342621789338198739noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-89253622294184946022010-06-04T19:24:13.234+01:002010-06-04T19:24:13.234+01:00Lucia,
"But it seems to me that James claime...Lucia,<br /><br />"But it seems to me that James claimed that the models actually failing a properly constituted test would be unimportant and I was respoding to that claim."<br /><br />My reading of James' comment was that the question answered by the "properly constituted test" was uninteresting, presumably because basic reasoning is enough to tell you the answer to the question without the need for performing a test. As I have pointed out the MMM is not intended as an accurate prediction of the observed trend, just of the forced component of that trend, so there is no reason to expect them to coincide.<br /><br />This is one of the difficulties of frequentist statistics, an unbiased estimate of the forced component of the trend cannot be expected to be an accurate estimate of the observed trend on this particular Earth, just the best estimate on average over a large sample of alternate Earths with the same forcing, but different realisations of the internal variability. Sadly we can't observe these alternative realities to find out if the MMM actually is unbiased, which rather limits the conclusions we can draw from the one Earth we can actually observe.<br /><br />I didn't say that "no one cares about detecting whether the multi-model mean is high or low". It is obviously high at the current time (you don't need a statistical test to see that). However that may just mean that the MMM is a good estimate of the forced component of the trend (which is what it is supposed to do), but the unforced component is sufficiently dominant to push the observed trend well into the lower tail of the ensemble. Or it could be that the unforced component is small meaning the MMM is biased, however, AFICS there isn't a way of telling which explanation is more plausible.<br /><br />It seems to me a mistake to concentrate on any point estimate, rather than considering the distribution of plausible outcomes, especially if you are going to ignore the stated uncertainty of that point prediction and substitute a different one.Dikran Marsupialnoreply@blogger.comtag:blogger.com,1999:blog-9959776.post-51752753283627993272010-06-04T19:03:45.485+01:002010-06-04T19:03:45.485+01:00James, finishing the digression in hope it suggest...James, finishing the digression in hope it suggests a way to clarify what's being done -- from poking a bit, I would encourage you or some other climate scientist to write an explanation of what's being done in papers using SCMs and why it's done (time, expense do matter!).<br /><br />Here as an example (perhaps relevant to the RC thread "On Attribution" about using local fingerprints):<br /><br />http://adsabs.harvard.edu/abs/2009AGUFMGC44A..06H<br /> <br />Probabilistic Predictions of Regional Climate Change<br />American Geophysical Union, Fall Meeting 2009, abstract #GC44A-06<br /><br />"We present a methodology for quantifying the leading sources of uncertainty in climate change projections that allows more robust prediction of probability distribution functions (PDFs) for transient regional climate change than is possible, for example, with the multimodel ensemble in the the CMIP3 archive used for the IPCC Fourth Assessment. ....<br />... The scaling uses a simple climate model (SCM), with global climate feedbacks and local response sampled from the equilibrium response, and other SCM parameters tuned to the response of other AOGCM ensembles. Use of the SCM allows efficient sampling of uncertainties not fully sampled by expensive GCM simulation, including uncertainty in aerosol radiative forcing, the rate of ocean heat uptake, and the strength of carbon-cycle feedbacks. Uncertainties arising from statistical components of the method, such as emulation or scaling, are quantified by validation with GCM ensemble output, and included as additional variance in our projections...."<br /><br />Gotta go, summer field work calls.Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-79558520005403480362010-06-04T18:35:11.423+01:002010-06-04T18:35:11.423+01:00James, where you say:
"... the IPCC figure y...James, where you say:<br /><br />"... the IPCC figure you refer to never even claimed to present the GCM projections, rather it presents the output from a simple climate model (SCM) which was tuned to emulate the GCMs. This precludes the possibility of that graph representing short-term natural variability in any way whatsoever."<br /><br />You're saying that the simple climate model is used to produce a picture -- an illustration, is that correct?<br /><br /><br />I gather Lucia can't use that picture to do short term analysis because it doesn't contain the information she's looking for -- is that right? <br /><br />It's easy to find papers mentioning using SCMs in this way for various purposes -- perhaps a topic explaining it would be generally helpful.<br /><br />Recall the problem creating "the digitized version of the 1990 IPCC curve" --mistakenly assuming it was more than a cartoon, deriving imaginary data from it, then testing that data. (Wegman, Barton hearing)Hank Robertshttps://www.blogger.com/profile/07521410755553979665noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-85056284346258131142010-06-04T17:22:58.299+01:002010-06-04T17:22:58.299+01:00Dikran Marsupial,
>>Douglass et al. made a ...Dikran Marsupial, <br /><i>>>Douglass et al. made a claim that is not supported by the test they actually performed, and as I pointed out, that is what caused the controversy.,</i><br /><br />I agree it did not. The reason is that the models did <i>not</i> pass a test that was constituted to detect a discrepancy between the multi-model mean and the observations., <br /><br />But it seems to me that James claimed that the models actually failing a properly constituted test would be unimportant and I was respoding to that claim. <br /><br /><br />I don't think <i>no one</i> cares about detecting whether the multi-model mean is high or low. There are many arguments we can have about whether or not bias has been demonstrated. Some people may think other questions are more interesting. <br /><br />But even if the notion that the multi-model mean is not perfect is widely accepted among specialists, at any given time, policy makers and the public certainly care whether or not the big fat central line on graphs indicating what to <i>they</i> are to expect is biased high or low. If data are indicating either situation, this is something that interests the public even if scientists think discussions of this are snooze inducing and should be greeted with yawns. <br /><br />By the way, on another note: Singer sought me out to discuss his response to the Santer paper. <br /><br />The conversation involved us mostly disagreeing on technical points, him thanking me for discussing his response and offering to thank me formally by adding a note to his paper when published. I asked him not to add any such formal thanks to the paper. (I'm not sure there is anything approaching a manuscript yet. If there is, I didn't see it.) Then, we went to lunch.luciahttps://www.blogger.com/profile/12342621789338198739noreply@blogger.com