Recent Comments

Prev  829  830  831  832  833  834  835  836  837  838  839  840  841  842  843  844  Next

Comments 41801 to 41850:

41801. Steve Metzler at 09:41 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       @35 Kevin C:

       Very well stated. Why can't the contrarians understand this, no matter how many times it's patiently explained to them? A hiatus in surface temperatures just means that the excess energy the Earth is accumulating has been displaced somewhere else. The trend over the past 30 years or more is still relentlessly upward, and the next big el Nino event is going to make 1998 look like a walk in the park.

       My fav explanation of the TOA radiative physics is here, in a guest post by Spencer Weart at RC:

       A Saturated Gassy Argument

41802. Gary_g at 08:41 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       Math is not my favourite past time but I thought Engineer's example is what is misleading him. It has two problems. It is an example which uses integers to give an integer result, and it includes no iteration (the result of one cycle is fed into the next). As soon as you use floating point values and iterations any equation will go out of alignment within about 3-5 cycles. The opening chapter of Chaos and Fractals: New Frontiers of Science (2004) by Peitgen, Jürgens & Saupe does a nice job of showing this. It uses runs through iterative equations using diferent calculators that handle floating point rounding differently and within 3-5 iterations they rapidly get out of alignment. Just when you think you might be able to retain the illusion that one calculator might be correct they use the example of two mathematically equivalent equations with one calculator, and the same thing happens, within 3-5 iterations they are hopelessly out of alignment. The odd thing is that whilst a simulation is in a sense quite accurate it's not temporally precise.

       That's the problem when you try to impose 19th century mechanistic thinking onto statistical mechanics. Once you admit statistics into reason you are saying that when you repeat an experiment you don't get the same result (if you did no need for statistics) but you might get an intelligible pattern. The use of the notion of falsifiabilty by Engineer is also misguided. It simply shows that academic logicians have not caught up with the math.

41803. John Hartz at 07:53 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       Here's the lead paragraph of a very informative article posted on the website of the Lawrence Livermore National Laboratory.

       LIVERMORE, Calif. -- By comparing simulations from 20 different computer models to satellite observations, Lawrence Livermore climate scientists and colleagues from 16 other organizations have found that tropospheric and stratospheric temperature changes are clearly related to human activities. 

       The article includes a sophisticated animated graphic. I highly recommend that everyone particpating in this comment thread check it out,

       A human-caused climate change signal emerges from the noise by Anne M Stark, Lawrence Livermore National Laboratory, Dec 11, 2005

        

41804. Rob Nicholls at 07:10 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       "This is not science: it is mumbo-jumbo" seems to sum up perfectly Nigel Lawson's writings on the subject on climate change. Lord Lawson has surpassed even the wonders of his hilariously named book "An appeal to reason" with this article. It's almost a full house of climate change denial untruths. Calling AR4 "grotesquely flawed" is just beyond parody.

       I expect the torygraph to print this kind of nonsense; it saddens me more that Lawson fairly regularly appears on BBC TV as some kind of authority on climate change, and always gets away with the same tired old rubbish - "no warming since year x"

       [make x more recent as soon as sample size is enough for trend to reach statistical significance]". I've never seen him challenged on this on TV.

       Unfortunately the BBC are doing a great job of making sure the IPCC's urgent message is lost by putting people like Lawson on almost every time global warming is discussed.

41805. julesdingle at 06:58 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       Scientists, governments and thinking people accept the IPCC report but Lawson is appealing to the popular gallery, preaching to a right wing choir. The hope is that a popular uprising of voters will defy the science, unfortunately there is an element of capitalism who wish to keep business as usual whatever the cost.

       A political statement perhaps [in defiance of comments policy but I trust this is within topic to stay]. Lawson's free market politics appears to behind much of the attempted discrediting of AGW/IPCC, now I don't know if Lawson or the right in general selfishly sell the dream of wealth for all to retain it for the 1% or believe that capitalism is the only way to rid the world of poverty. What is clear is the predominately right in the US, Europe, UK lean towards CC denial because dealing with it is political. Capitalism [and the 1%] has done rather well out of fossil fuels [as well as state sponsored scientific research- think iPhone and lots of other cool stuff].

       Decarbonising need not be a socialist or back to nature and sandals green dream but whilst capitalism seems stuck with business as usual and refuses to think alternatives the only strategy is to get the voters to support the old system. The worst option is those of acceptance of CC using the fear for our children's future against the deniers weapon of fear of present prosperity.

41806. Jose_X at 06:17 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       OK, KR, I'll try again.
       
       Engineer, the long comment above (@59) may make a few things more clear if this next shorter version fails:
       
       Let's perform a card draw experiment one time. We get a 2. You seem to be saying that the observed result of 2 is way below the average of 5.5. What others here say is that the 2 is within the range of the model's 95% window (if just barely). The model predicts an integer anywhere from 1 to 10, specifically, with at least 95% of the time coming in the range 2-9.
       
       Would you say that the model for the card drawing is wrong because the single drawing of 2 is at the edge of the 95% window? So why would you say that a climate prediction is wrong if the measured earth temp lies at the edge of the 95% window?
       
       Note, a "prediction" can be derived from the projection by taking the projection and replacing the parameters that were not known back when the projection was made.
       
       Note, a prediction necessarily has an error range. An experiment on a simple system can yield a tight error range prediction (ie, a small "prediction interval"). Those models claim high accuracy. The climate model does not claim high accuracy. However, to say the climate is wrong you must show that the data does not easily fit within the wide error range used by the models. You seem to think that the climate models have to have a narrow range or they are wrong. Not so. While a wide range might mean the model is useless (eg, my useless model predicts the global temp this year will be between 0 C and 100 C), that in itself doesn't mean the imprecise model is wrong. In a sense, a model can be "imprecise" yet "accurate".
       
       Note, the climate models are not that imprecise since they predict, contrary to what contrarians predict themselves, that it's very likely (over 95% confidence) that the temp in 2100 will be higher than where we are now. A useless model would peg that probability at 50%. And most contrarians peg it much less than 50%, likely making them less than useless (ie, wrong).

41807. funglestrumpet at 05:56 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       Perhaps Lawson's position would be more easily understood if he were to disclose where the funding originates for his Global Warming Policy Foundation. His errors that are highlighted in this article could be down to ignorance, I suppose, but considering their consistent thrust, it is difficult not to see them as deliberate in nature. With that in mind, I sincerely hope that the aforementioned funding for his G.W.P.F. does not have the fossil fuel industry as its source. For were that the case, it would be difficult to imagine that his behavior would not be brought to the attention of The House of Lords Commissioner for Standards.

       The House of Lords code requires that “Members of the House shall base their actions on consideration of the public interest.” Trying to ensure that climate change continues unabated, which his statement:

       “So what we should do about it [climate change] – if indeed, there is anything at all we need to do – is to adapt to any changes that may, in the far future, occur. That means using all the technological resources open to mankind – which will ineluctably be far greater by the end of this century than those we possess today – to reduce any harms that might arise from warming, while taking advantage of all the great benefits that warming will bring.”

       seems intended to achieve is hardly in the public interest.

       Even if there is no fossil fuel element influencing Lawson’s motives, he has to realize that his peerage is intended to enable him to protect the U.K., its Head of State and its people, and he is rewarded accordingly. Seeing as that reward comes from the public purse, if he is to continue in the role his peerage designates, the least he can do is take the trouble to get his facts straight.

       If he is no longer up to doing that, even on issues as potentially important as climate change, perhaps he would be better advised to resign his peerage and give younger blood the chance, especially seeing as they are likely to suffer more from the effects of climate change than someone of his advanced years will. Accordingly they will be far more motivated to do a proper job of protecting their country (and their family) than Lawson can claim to have if his comments on the I.P.C.C. are any guide.

41808. Composer99 at 05:23 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       Lawson, like so many denialists before him (and surely after), is trying to sell us gold in Busang. Unfortunately, as with the real deal, too many are willing to try and get a piece of the action.

41809. KR at 05:00 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       Jose_X - No offense, but... Brevity is the soul of wit. 

41810. Jose_X at 04:47 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       engineer:
       
       You said (@44) your problem was in part in not agreeing on the value being placed here to differentiate between projections vs predictions ("but why is this distinction between prediction and projection even necessary? the terms are interchangable") yet your main dispute is that the models fail to live up to reality. I think you understand projections but are forgetting what a prediction actually means, and this is why you think the models are being falsified by the data.
       
       For reference, I'll quote 2 other commenters.
       
       Leto addressed your statement ("being on the high side 114 out of 117 times is indicative of a bias") by saying,
       
       > You can observe that process 117 times, or 117000 times, and it does not mean that the model has been falsified unless you choose time points that are truly independent.
       
       Mammal_E had earlier said,
       
       > Let's say I have a model that simulates the outcome of process of drawing cards from a shuffled deck.  I run the model once, and it generates a 3 of diamonds.  I have an actual shuffled deck, and draw a 10 of clubs.  The model and reality disagree.
       
       First, let me adjust the example above from Mammal_E. The model is instead run 117 times (not once) and the average is (assuming we ignore the face cards) 5.5. Now let's perform the draw experiment one time (to match the "single" time that the Earth performs its experiment). We get a 2.
       
       You seem to be saying that the observed result of 2 is way below the average of 5.5. [You stress this point by using the extreme example of getting a certain card 20 times in a row to start the experiment.]
       
       What others here say is that the 2 is within the range of the model's 95% window (if just barely). The model predicts an integer anywhere from 1 to 10, specifically, with at least 95% of the time coming in the range 2-9.
       
       Would you say that the model for the card drawing is wrong because the single drawing of 2 is at the edge of the 95% window? Even if we had drawn a 1, we'd still follow the model.
       
       Maybe you understand this example above and it improves your understanding of the climate modeling, but let's add more details.
       
       Now, can we run the experiment on the earth more than once? Well, I think this is part of what might be confusing you. To address this, we can map the card analogy in at least two ways to the earth system.
       
       In way one, we treat a single year as a sample pick. Note that since the average temp doesn't change that fast year to year, this example is not a good model for independence (like the card pick is), but it does get at the basic point of whether the climate model is correct or not.
       
       So let's go back to mid 1970s (since that appears to be around when the temps started a notable upward trend from which they have not recovered). The question would be, has the actual earth temp for each year since 1975 been within the 95% model range close to 95% of the time? I won't answer that, but you can consider that question for yourself first before marrying the analogy of picking a card 20 times in a row. While the modern models didn't exist back then, we can take a modern model and hindcast. I think the models are calibrated that way and in fact the answer would be close to "yes". I don't know the specifics, but unless you do, I don't see how you would argue that the model (or a particular set of projections) is off by a lot or even by a little. Does the data fit the models' range most of the time? What data not matching the 95% boundary nearly 95% of the time are you using to claim that the applicable predictions are "falsified"?
       
       A second way of looking at this is to look at linear trends across time periods of a certain size. This is similar. Here we might look at all the earth data for 15 year subintervals since 1975 and see how those trends match the projection trends. That would provide fewer data points, but its the same general idea in the limiting case (of us being able to have a very long superinterval of data points).
       
       [Note that Leto and others referred to the earth providing a single draw, but that is if we zoom in on a single subinterval. And this single draw was constrasted to the numerous model simulation runs that are used to calculate the expected range for that single subinterval.]
       
       Do you agree with the above? Did it help?
       
       If you are still wondering about predictions vs projections:
       
       First, don't assume that a prediction gives an exact value in physics. While in a textbook we might calculate the final position of a particle as "x=27.98". In dealing with observations, such a prediction would come with an error interval.. always! Mammal_E called this by its technical name, a "prediction interval," in comment 28.
       
       Because the earth system is complex, the prediction interval for average global temp for the planet has a somewhat wide range for any given year as based on the model projections. This contrasts to very narrow widths for some physics predictions that rely on simple well understood systems. Regardless, for the model to be a good one, we'd want near 95% of the "independent" observations to lie in the particular model's own 95% prediction interval. You have not shown that the earth climate has been off the 95% yearly model range in significantly more than 95% of the years.
       
       As for projections, that is like a wide set of predictions (parameterized set of implied predictions). This parameterization ("vagueness") is necessary because we don't know many of the x (sub i) variables in the future, so we provide distinct prediction graphs for several potential x values as a way to convey a general feel for what is expected to happen.
       
       I do think you understand projections. You appear to agree that we want to look precisely only at the actual "prediction" for the actual x values that are today known but weren't back when the projections were made. You agree, I think, that to judge an earlier projection/model, we want to first pin down the observed variables and then treat the resulting statistics of the numerous model runs as the relevant prediction we are judging.
       
       Anyway, I think you forgot that all predictions come with error ranges. To show the climate models are wrong, you can't judge them by the narrow ranges used in simple Newtonian mechanics examples but must judge by the wide boundary claimed by the models for predicting the complex earth system.
       
       Now, you might think that using a wide error range means the model is whimpy. Yes, if we had an error range of +/- 100 C, then that model is useless as any temp we'd observe would almost surely fit in there. If you want to make that claim of whimpiness, do so, but that is a different claim than to say that the models are wrong.
       
       And as for being whimpy, the current models predict a 95% range for 2100 that lies entirely above our current temps. In contrast, most contrarians would have a range (if they believed in using error ranges to more properly quantify their guesses) that would have a lower end way below our current temps. Also the mean of the models lies several degrees above our temp today while most contrarians would have a mean below the current temp. ["most contrarians" is a vague notion, true.]

41811. John Russell at 04:31 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       So Lawson rubbishes Dr. Rajendra Pachauri by calling him“a railway engineer and economist by training, not a scientist”?  Consequently I guess this makes it legitimate to point out the fact that Nigel Lawson is "not even an engineer, just an economist, not a scientist"; which I guess by Lawson's account means his own opinions are worth even less.

       This leads nicely into my other point: the Lawson family seem partial to coming out with 'deniatribes'. Witness Nigel's son, Dominic Lawson, in last weekend's Sunday Times (behind a paywall but also re-printed in The Australian). In this article, entitled "A warm consensus, but the planet is not following suit", he makes an incredibly stupid comparison between the consensus of the IPCC's climate scientists and the economists who, almost to a man, failed to foresee the economic crash of 2007. How he can arrive at the view there is any equivalence between climate research and the opinions of economists beggars belief. 

        A warm consensus, but the planet is not following suit - See more at: http://www.theaustralian.com.au/news/world/a-warm-consensus-but-the-planet-is-not-following-suit/story-fnb64oi6-1226729516410#sthash.2t38mD92.dpuf

        

       A warm consensus, but the planet is not following suit - See more at: http://www.theaustralian.com.au/news/world/a-warm-consensus-but-the-planet-is-not-following-suit/story-fnb64oi6-1226729516410#sthash.2t38mD92.dpuf

          

41812. jja at 04:01 AM on 3 October 2013
       Lawson, Climate Change and the Power of Wishful Thinking
       

       Perhaps the IPCC would have more success forwarding their report if they included a breakout discussion, as a subset of the executive summary, that showed how the uncertainties contained within the carbon cycle climate feedbacks (fig. 6.27) were not included when projecting the fossil fuel emissions required to yeild the RCP 8.5 scenario (fig. 6.25).

        

       from the discussion for figure 6.25, page 6-55

       climate impact on carbon uptake by both land and oceans will reduce the compatible fossil fuel CO2 emissions for that scenario by between 6% and 29% between 2006 and 2100 respectively (Figure 6.27) equating to an average of 157 ± 76 PgC

        

       and

       Compatible emissions would be reduced by a greater degree under higher CO2 scenarios which exhibit a greater degree of climate change (Jones et al., 2006).

        

       Figure 6.25 only shows 1 standard deviation.  The uncertainty in cumulative land-based carbon cycle uptake in figure 6.24 is +/- 250 PgC by 2100 (+/- 918 gigatonnes of CO2).
       
       f this uncertainty was adequately addressed in a discussion that included an ECS of 4C for 2XCO2, then it would be clearly shown that we are on track for locking in 4C of warming before 2050 on our current emission trajectory.

41813. KR at 03:24 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       MarkR - Regarding the McIntyre link and the draft Fig. 1.4 discussed, that figure showed the range of projected model trends +/- observed HadCRUT temperature 2σvariability, not +/- the model variability. 

       Models are currently running high - and seven years ago, as Tamino points out, they were running low. However, the periods for which they have been high or low with regards to observations are too short for statistical significance. 

41814. MarkR at 02:03 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       This 'real world realisation' which affects baselining is also important for short term trends.

       Firstly, you can use Kevin C's trend tool to get a 95% confidence interval on the trend from 1997, and find it's between -0.07 and +0.19 C/decade.

       But that's a purely frequentist estimate of the probability based on assuming we know nothing about the noise and that it's completely random. But we do know something about the noise: we know solar activity is much lower than expected and that there has recently been a trend towards more La Ninas if you choose to start in 1998.

       Foster & Rahmstorf (2011) tried to address this, but I understand that there are continuing difficulties with that sort of assessment. Kosaka & Xie's new paper is interesting: they found that according to the model, the real world realisation of El Nino and La Nina means that the model does a good job of reproducing the recent changes.

       The fact that most of the models are running high does not show that most of them are wrong yet. Based on the fact we've seen a negative trend in the ENSO index, we know that if the models are right then most of them would overestimate the warming trend since 1998 because the model average should be close to no trend in ENSO activity.

       If we had enough models and could only select the runs that featured a similar trend in ENSO, then we'd have a better idea of whether the models were actually overestimating warming or not. Kosaka & Xie's paper suggests that they might not be.

41815. Philippe Chantreau at 01:52 AM on 3 October 2013
       Dueling Scientists in The Oregonian, Settled by Nuccitelli et al. (2012)
       

       For starters, I believe that the Trenberth quote is inaccurate and I would ask for the original source. As I recall, the "travesty" applied to missing energy in the overall budget, which is an area of expertise of Trenberth. I'm sure that Trenberth elaborated on that and that there is context.

       If you look at the ARGO website, they state very clearly that the period of observation for ARGO data is still too short to calculate a trend. "The data is dominated by interannual variability" per ARGO website. There is no way to calculate an OHC trend except by using data before the deployment of ARGO, so your interlocutor is disingenuous.

       I am also pretty sure that claiming that Levitus used "a model" is a wild misrepresentation. Levitus, Antonov and their collaborators have been studying this for years and I doubt that anyone knows the observational data better than them. Perhaps your interlocutor is of the opinion that correcting for errors as Levitus and Antonov did, notably by using Wijffels et al, 2008, is "using a model."

       The truth is that Levitus is the most knowledgeable in the matter and his papers have hundreds of cites, some over a thousand cites. I don't have the time to dig deeper but I believe that, if you do the digging, you can refute each and every one of your interlocutor's claim. The most obvious is that Argo does not show a cooling trend because the time series are too short to show any trend.

       As for NOAA, their site is not available at the moment due to the government shutdown, so digging through their references is not possible.

       D&K has been looked at here and elsewhere and their wild claims of "step changes" are a little too much like magical thinking.

       To make a long story short, yes your interlocutor is misrepresenting the science but placing a big burden on you to show that he is. Anyone who is not scientifically litterate following the discussion will get the impression that some science says one thing, some say different and they'll go where their emotions/ideological preferences take them anyway. Typical modus operandum of the obfuscators these days.

41816. MarkR at 01:51 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       Bob @3

       I checked through the McIntyre link but can't find the full description of the simulation setup. It appears to be a single run from 1900 onwards, with known forcing data until 2000 and then RCP forcings after. 

       Generally, GCMs generate their own natural variability, like El Ninos. An individual model might be in La Nina or El Nino or neutral in any one year, but the ensemble average tends to be equivalent to a 'neutral' year. Similarly for other sources of natural variability.

       If you match up a single year against the ensemble averge, then you can effectively shift the temperatures by any amount you want, just by artificially selecting your start year. 

       In the worst cases of the 97/98 El Nino you could shift your temperatures up or down by 0.4 C. That's why baselining is typically done over a longer period over which the natural 'noise' averages closer to zero.

       Alternatives would be to initialise the model with the 'real' climate state at the start point, or to only select those models which match the most important 'real' states, but these are time consuming and/or cause you to lose data.

41817. KR at 01:48 AM on 3 October 2013
       Why is the IPCC AR5 so much more confident in human-caused global warming?
       

       empirical_bayes - I'm a bit puzzled by your post. If the probability distribution function for data or models is taken from previous information, there are still (although unlikely) high-sigma deviations possible in projections, temperatures outside observed ranges due to observed variances. Standard deviations don't have hard limits. 

       I'll note that your colored ball/urn example requires that you assign (not estimate, not without additional assumptions) probabilities for the frequency of unobserved categories - and category presence not the same thing as looking at the variance of a single variable. 

       That said, the Bayesian realm of statistics is not a field I am expert in, and I may well be incorrect...

        

41818. Jubble at 01:20 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       For what its worth, I have shared this article with the UK Daily Mail.

41819. Dikran Marsupial at 00:55 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       Tom Curtis - A five year baseline period?  That is quite, err... unusually short!

       To be fair, there are those that like to use an even shorter baseline ;o)

       

       Same idea though (click on the graph for a debunking).

41820. Tom Curtis at 00:12 AM on 3 October 2013
       IPCC model global warming projections have done much better than you think
       

       John Oh @51, looking at Christy's graph, as reproduced by Spencer shows that the observed record lies below the multi-model mean not just in the recent, so-called "hiatus" years, but over the whole record.  The apparent disagreement, therefore, consists entirely in Christy using a low baseline to creat a visual appearance of disagreement, where little disagreement actually exists.

       Having said that, it is interesting to see how he accomplishes this legerdemaine.  The graph indicates that it indicates "departure from the 1979-83 average".  That means that for both observational series, and for the multi-model mean, the average over the period 1979-83 equals zero.  Despite that, there is already a marked discrepancy between observations and multi-model mean in that period.  Specifically, for most years the obeservations are below the multi-model mean, but in 1983 they are well above it.  Indeed, because the average over that interval is set to zero, because the observations are well above the multi-model mean in 1983, the other years need to be below the mean to achieve the same average over that period.

       1983, of course, was an unusual year.  Specifically, it was unusual because of the significant volcanic eruption the year before (El Chichon), which as a forcing shows up in the multi-model mean as a dip in temperatures.  It was also unusual for possibly the strongest El Nino on record, with an SOI reading of -33.3 in Feb, 1983 (compared to the -28.5 in March 1998 for the more famous 97/98 El Nino).  Unlike volcanic forcings, however, El Nino warmings do not show up in the models - or at least, they do not show up in the models on the same year for all models, with the result that in the multimodel mean they are cancelled out. 

       So, Christy has forced a low baseline for the observational records by including in the baseline period a known, very large warming perturbation which he knows to be reflected in the observations, but which cannot be reflected in the models.  To ensure that this lowers the baseline sufficiently, he then makes the baseline as short as possible to ensure the effects of the 1983 El Nino in distorting the graph are not diluted (as they would have been had he used a more appropriate 1970-2000 baseline).

       And to top it all of, knowing the so-called "hiatus" is predominantly a consequence of recent ENSO variations, he has chosen a data set which shows a heightened ENSO effect relative to the surface temperature record.  I assume he has done this because an honest comparison would be too damaging to his case.

41821. empirical_bayes at 00:02 AM on 3 October 2013
       Why is the IPCC AR5 so much more confident in human-caused global warming?
       

       I also think some assessments, while they may be well-meaning, fall into a trap of statistical methodology.  In particular, surface temperature profiles from, say, HadCRUT4, express a single realization of surface temperature development on Earth.  It's highly probable that even if the entire system were magically reinitialized to the state it was in 1980, with exactly the same GHG forcings and solar radiation vs time, it would track a different path of temperature.  What we observe is but one realization.  

       Now, you can try to assess the so-called "internal variability" by looking at ensembles of temperature subsets, appropriately adjusted for serial dependency, but classical ("frequentist") techniques will never produce temperatures outside of the observed range.  This is also true of climate models, e.g., 37 models from CMIP5. 

       What's needed is a Bayesian extension of both data sets, perhaps using predictive posteriors and weak priors, or priors initialized with paleoclimate results.  (I'm thinking of Geisser and Eddy, 1979, "A predictive approach to model selection", http://dx.doi.org/10.1080/01621459.1979.10481632, or the paper on the same subject by Gelfand in the compendium, Gilks, et al, Markov Chain Monte Carlo in Practice.)

       Facts are, and to use a very rough analogy, if you try to estimate the proportions of differing colored balls in an urn by sampling, knowing the possible colors but not knowing how many balls there are in the urn, a frequentist assignment will give you zero as the proportion for any colors you have not observed.  A Bayesian assessment assigns some non-zero probability to colors which have not been observed, so after the data, they have some proportion, possibly small.  Temperatures which have not been observed are not impossible and, to the degree to which models try to minimize error against all possible futures, they go for that, not a specific path.

       There are also problems with trying to estimate magnitude of "internal variability" without taking such an approach, but the analysis is more involved there, and includes serious questions of identifiability without making unrealistic assumptions.  More on that some other time. 

       I have not done the calculation so cannot assert, but I have done toy problems that are pretty analogous, and what happens is that there is a much greater overlap between the range of possible temperatures HadCRUT4 and others suggest and CMIP5 and others, and, so, assertions of incompatibility stand on less evidence than other kinds of analyses indicate.

41822. KR at 23:51 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       John Oh - What "exaggeration" are you referring to? Are you making a claim that IPCC data is in error, and if so on what grounds?

       As I noted on that Spencer thread, satellite temperatures have their own issues, many of which are not acknowledged by the collectors of that data. From “Temperature Trends in the Lower Atmosphere – Steps for Understanding and Reconciling Differences” 2006, authored in part by the very John Christy who supplied the data in your linked blog post:

       “On decadal and longer time scales, however, while almost all model simulations show greater warming aloft (reflecting the same physical processes that operate on the monthly and annual time scales), most observations show greater warming at the surface.

       These results could arise either because “real world” amplification effects on short and long time scales are controlled by different physical mechanisms, and models fail to capture such behavior; or because non-climatic influences remaining in some or all of the observed tropospheric data sets lead to biased long-term trends; or a combination of these factors. The new evidence in this Report favors the second explanation.”

       [Emphasis added]

       Radiosonde data (intended for short-term weather analysis, not climate studies) has consistency/calibration issues, and the quite complex satellite data analysis has been repeated updated due to various errors. 

41823. John Oh at 23:06 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       http://www.drroyspencer.com/2013/04/global-warming-slowdown-the-view-from-space/ < see link.

       The satelite figures differ with IPCC numbers, and show steady, historic increases in temperature. The exagerated numbers again come from the IPCC.  So can this be explained or is this also IPCC figures that are being misrepresented?

41824. Dikran Marsupial at 19:27 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       engineer wrote "I think I'm being misunderstood.".  The reason that I suggested that you dial back the tone of the discussion is that quite often it is difficult for individuals to know if they are being misunderstood or whether they themselves misunderstand something.  This is a classic example as in my post I explained the difference between a projection and a prediction, but in your reply you ignored this point and return to "but if a model disagrees with nature then there is something wrong with the model. And that applies to any model.".  The distinction between a prediction and a projection is very important in this situation because the reason for the disagreement may be because the scenario (the X) is not a sufficiently accurate representation of reality, and that needs to be taken into account.

       If you think that someone is using a subtle distinction between words to evade a point, then the onus is on you to do your best to understand the distinction when it is explained to you, rather than just ignore it, which leads the discussion to becoming ill tempered.

       I am a big fan of Hanlon's razor ("never attribute to malice that which can be adequately explained by thoughtlessness/stupidity"), which I generalise to "always try to view the intentions of others in the best light that is consistent with the observations".  So rather than assume that someone is being evasive, assume there is some subtle point that you don't understand and help them to explain it to you.

41825. Drewd006 at 18:54 PM on 2 October 2013
       Dueling Scientists in The Oregonian, Settled by Nuccitelli et al. (2012)
       

       I have come across this blogger who is claiming: "The oceans are cooling just like the air is, as proven by the measurements of the 3,000 Argo buoys; the oceans are cooling at all measured levels, and have been since the buoys were launched"
       
       I cited: Levitus et al. 2012, Lyman et al. 2010, Von Schuckmann et al. 2009, Trenberth 2010, Purkey & Johnson 2010, and Trenberth & Fasullo 2010.
       
       And he response saying:

       "NOAA have just used Levitus's paper, we can forget them as they simply estimated the OHC using a model; there were no measurements (only ARGO after 2003).

       Lyman et al's paper has been debunked by R. S. Knox and D. H. Douglass and by NODC OHC data.

       Trenberth 2010; HAHA! This is the guy who said; "“The fact is that we can’t account for the lack of warming at the moment and it is a travesty that we can’t.”

       HE is debunked by FOUR other papers; Willis, and Loehle, and Pielke, and Von Schuckmann." 
       
       Are there any validity to this mans claims, if he is misrepresenting the science I would love to know.
       
       Thank You

41826. rkrolph at 16:05 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       On the chart above titled "Global Average Surface Temperature Change" there isn't a clear distinction between RCP8.5 and RCP2.6 until around 2060, which means we could still be arguing for about 45 more years whether the climate is changing per a low emissions or a high emissions scenario, based on model projections.  So it seems to me that arguing about exactly how accurate the models are shouldn't be the point to get caught up on.   

41827. DSL at 14:02 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       I stand corrected.  Thanks Mammal_E, and I agree that a post would be a nice thing, though the comment is surely at least halfway there.

41828. scaddenp at 13:28 PM on 2 October 2013
       Models are unreliable
       

       You would also need to separate out the other forcings at play if trying to establish k by simple fits. I would recommend Chpter 10 of the newly out AR5 WG1 which has a section on estimation of TCS and ECS, along with references to the papers which attempt this from various observational sets. The section of estimation from the instrumental record looks like it would interest you most.

41829. WheelsOC at 13:00 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       In fact, mousing over the underlined word "projection" anywhere on this page brings up a cogent explanation in the upper right of the screen.

41830. mikeh1 at 12:42 PM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       There is a discussion of statistical language at the Australian Bureau of Statistics web site which compares the terms projection and forecast. This may help Engineer get over his apparent belief that the distinction was invented by climate modellers.

       "A projection is not making a prediction or forecast about what is going to happen, it is indicating what would happen if the assumptions which underpin the projection actually occur."

41831. Bob Loblaw at 10:37 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       engineer: "the terms are interchangable"

       No, they are not. If you have a model that states a + b = 12, a projection is saying "if a = 4, then my model says b = 8". An observation that b=14 is not a falsification of the model, if a<>4. A prediction would be to say that a will = 4, and b will = 8.

       A series of projections might cover a range of values of a, stating what the model says b will be for each value of a. In modeling terms, this is call a sensitivity analysis - seeing how a change in an input parameter affects an output value, or how sensitive b is to changes in a. If you want to talk about the probability distribution function of output values (e.g., there is a 95% chance of an outcome in range X), then you also need to look at the probability distribution of the input parameters. You can't just take N runs of unknown input distribution and assume that it fits your expectations. At least, not if you are doing good science.

41832. engineer8516 at 09:37 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       @Rob, ok, I think understand what you're saying, but why is this distinction between prediction and projection even necessary? the terms are interchangable. Adding a distinction seems like adding confusion for no reason.. "This is why climate modelers don't make predictions; they make projections, which say in scenario 'x', the climate will change in 'y' fashion." when I read it the first time it seemed like a dodge to the standard of falsifiability. Apparently, I'm the only one that thought that.

41833. scaddenp at 09:37 AM on 2 October 2013
       Models are unreliable
       

       But you would only expect temperature rise to be linear if transient climate sensitivity was same as equilibrium climate sensitivity. Emperical determinations of ECS suggest it is reasonably robust over quite a temperature range. However this is an observation, not an assumption.

41834. scaddenp at 09:34 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       I think it would also be helpful to say climate models say IF you get this forcing, THEN you get this result. They do not however predict what the forcings will be.

       Climate models also make no pretense at having any skill at decadal-level surface temperature trends, only on climate trends where most certainly the models can be tested.

41835. Rob Honeycutt at 09:18 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Engineer @41...  But this is exactly the same thing as what Dana is stating, just using different words.

       "Prediction" would be saying stating what the temps will do, which seems to be the view most "skeptics" seem to take toward models.  Modelers don't do that.  They, 1) make a variety of "projections" based on different emissions scenarios, and 2) the projections say that surface temps will likely run in the bounds of the model runs.

       None of that suggests anything about the falsifiability of the projections.

41836. engineer8516 at 09:03 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       @Rob,  "What is expected is that surface temps will continue in the general direction of the mean but stay within the bounds of the model runs." I completely agree.

       My point of contention was this statement: "This is why climate modelers don't make predictions; they make projections, which say in scenario 'x', the climate will change in 'y' fashion.", which in my opinion (-sloganeering snipped-).

       Moderator Response:

       [DB] Your opinion has been voiced before, and addressed by multiple parties.  To then repeat said opinion without the added value of new information based in the science is pointless sloganeering.  Such sloganeering is a waste of everyone's time here, which is why the practice is listed as banned by this site's Comments Policy.

       [JH] Plus, it's a violation of the prohibition against excessive repetition.

41837. Leto at 08:55 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       "Being on the high side 114 out of 117 times is indicative of a bias."

       "if a model disagrees with nature then there is something wrong with the model."

       "You're dealing with probability which is dependent on the sample size and a sample size of 1 is insufficient. However, if your math dictates that you should get a 10 of clubs 5% of the time but you do the experiment and draw 10 of clubs 20 times in a row off the bat, a probability that amounts to 0.05^20 = 9 * 10^-25 %...then it's probably time to revisit your model. But the key here is that the model is still ultimately judged by nature, which is how it's supposed to be."

        

       engineer,

       Some contrarians have implied that a single result at odds with a model falsifies the model, which is obviously wrong - like blaming a card-strategy program for not picking a single deal correctly. Your statements are more sophisticated and imply that the situation is instead one of repeated bias, predicting the 10 of clubs 5% of the time but instead getting it 20 times in a row. This card analogy breaks down, though, because there have not been 20 independent runs of the real-world climate. There have not been 117 runs either.

       If the climate models made predictions about 117 independent runs of the real-world climate, and temperature estimates were high 114 times, then yes, the models would be wrong. That is not what has happened, though. We have had a single real-world "deal", which has produced surface temperatures within the predicted range. You can observe that process 117 times, or 117000 times, and it does not mean that the model has been falsified unless you choose time points that are truly independent. It is like watching a single bridge hand being played out, noting the progress of the hand 117 times, then saying the card-strategy being played was wrong because it had been tested 117 times and found wanting.

       Not only that, as soon as we start to consider the factors that account for the variability (volcanoes, el ninos, etc), we see that the models are basically accurate, and the variables that were part of the unpredictable "shuffle-and-deal" are affecting global surface temperature in agreement with the models.

       The ensemble mean does not represent any single model run very well - a fact well known to the modelers, but easily misunderstood by those predisposed to misunderstand. Personally, I think it would be useful if the models gave predicted intervals for "el nino years", "la nina years", and so on. Then we would see that, for the actual "cards dealt", the models have performed well.

41838. Albatross at 08:45 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Hi Mel @36,

       We at SkS are aware of the paper by Fyfe et al..  If you look at their figures and TS.3 Fig. 1 from AR5 that posted above, you will see that the AR5 figure was probably generated by one of the authors of Fyfe et al.. 

       
       These short-term descrepancies are of little or no consequence for policy makers-- they are concerned with the long-term projections (which have thus far been very accurate). Importantly, Fyfe et al. also acknowledge that in the long-term the models have been almost spot on in predicting the amount of warming (see their Fig. 2).

        

       Studies such as the one by Fyfe et al. are of academic interest and for improving our understanding about the climate system's (and models') intricacies.   Fyfe et al. looked at the 1993 through 2012 interval when the models were running cooler than the observations. Here is a science brief on Fyfe et al. that is easier to digest.

        

       Ask yourself this Mel, why are the contrarians making all the fuss about the 1993-2012 window, but are completely silent about the 1984-1998 window when the models were running too cool?

       
       I suspect that you know the answer, it has to do with confirmation bias and fake skepticism.

41839. Albatross at 08:33 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Everyone,

       Please, comments such as

       "Let me ask you this, if a particle is discovered that can travel faster than the speed of light, is Relativity Theory wrong?"

       are really not helpful.  Debating "what ifs" does not address the subject at hand.  If one wishes to play that game, then one should also entertain the distinct possibility that temperatures could rise a lot more than theory suggests.

       The contrarians here have had the fact that the tempeature observations lie within the uncertainty bounds of the simulations explained to them several times now.  Several figures showing the agreement between the models given the bounds of uncertainty have been offered. It has been explained to that no model is perfect-- yet that they insist on using trying to using this to make a straw man argument. Funny too how for contrarians the uncertainty is always biased to the low side.

       This sort of behaviour is all too typical of contrarians.  No amount of data, facts or explaining will change their "opinion", their "feeling" or their "belief". 

       While we fiddle, the energy imbalance arising because higher CO2 levels (from human activities) is adding the energy equivalent of four Hiroshima A-bombs a second to the planet's climate system.  

41840. Rob Honeycutt at 08:23 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       engineer @34...  If the surface temps fall off the model mean, that does not suggest the models have a problem.  That's what everyone here is trying to convey.  What is expected is that surface temps will continue in the general direction of the mean but stay within the bounds of the model runs.

       If surface temps started to track outside of the range of the model runs, then that would be an indication something was not being accounted for.  But the whole point of the article here is, this is not what's happening.  Surface temps are well within the bounds of where the models project they will be.

41841. Mel at 07:58 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Nature Climate Change recently published a paper entitled ‘Overestimated global warming over the past 20 years’ by Fyfe, Gillett and Zwiers. The paper compares CMIP5 results with observations and the general conclusion is pretty much summed up in the title. Could SkS please comment on these findings in relation to the discussion here.

       I suspect this paper will be used to challenge the utility of climate models in general.

41842. Kevin C at 07:54 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Anyone played contract bridge?

       So as declarer, I look at my hand and dummy and see that whether we make the contract or not depends on a finess, which depends on the card split in the opponent's hands. And on the basis of the statistics of the split I calculate that I've got an 80% chance of making the contract.

       We play it out, and go down on a bad split. Was my calculation wrong? No, it was exactly right. I just got the 20% split. If we were to play the hand 100 times with all the known cards in the same places and the rest distributed randomly, we'd make the contract around 80 times.

       Someone else will have to reframe that for poker or blackjack players.

       So, climate.

       Thought experiment: Imagine a system with a linear component which is varying slowly, and a large chaotic signal overlayed on the top. We can model the physics of the system. But when we run the real system the result we get will contain the linear component, plus a random realisation of the chaos. If we run the real system many times, we can average out the chaotic component and just see the linear part. But if we can only run the system once, we're stuck with that one run.

       Suppose we have a perfect model. We run it once, and it also produces the linear response with the chaotic signal on top. We can run it lots of times and average, and we'll see just the linear signal. But that doesn't look like the real run, because the real run has the chaotic part too.

       But neither do any of the individual runs of the perfect model look like the real run, because the chaotic component is different.

       So even with a perfect model, we can't reproduce the real run either with an individual model run or with an ensemble of model runs. And yet that's with a perfect model, a model which is right.

       That's exactly the problem we're facing here. Because weather (or more generally the internal variability) is chaotic.

41843. engineer8516 at 07:48 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       @ Mammal_E

       You're dealing with probability which is dependent on the sample size and a sample size of 1 is insufficient. However, if your math dictates that you should get a 10 of clubs 5% of the time but you do the experiment and draw 10 of clubs 20 times in a row off the bat, a probability that amounts to 0.05^20 = 9 * 10^-25 %...then it's probably time to revisit your model. But the key here is that the model is still ultimately judged by nature, which is how it's supposed to be.

       It seems you disagree with my statement "if a model disagrees with nature then there is something wrong with the model." Let me ask you this, if a particle is discovered that can travel faster than the speed of light, is Relativity Theory wrong?

41844. Mammal_E at 07:11 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       @Rob Honeycutt --  you interpret my comments correctly.  I'll see if I can gin up something that is brief and clear.  I see this issue a lot with people misinterpreting population simulations, e.g., projecting the fate of endangered species.

41845. Rob Honeycutt at 07:03 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Also, if I'm interpreting your comments correctly, you're saying what I was saying (albeit more precisely than me) about climate modeling being a boundary conditions problem.  I'm equating boundary conditions to your statements about prediction intervals.

41846. Rob Honeycutt at 07:00 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       I agree with Tom.  Mammal_E, you're providing great insight into what are pretty common misinterpretations of what climate modeling is about.

41847. Tom Dayton at 06:42 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       Mammal_E, I think it would be great for you to write a post expanding your explanation of prediction interval versus confidence interval, if it includes some graphs.

41848. Mammal_E at 06:28 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       engineer --

       "if a model disagrees with nature then there is something wrong with the model."

       Let's say I have a model that simulates the outcome of process of drawing cards from a shuffled deck.  I run the model once, and it generates a 3 of diamonds.  I have an actual shuffled deck, and draw a 10 of clubs.  The model and reality disagree.  Is that sufficient information to conclude that something is wrong with the model?

41849. Mammal_E at 06:24 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       franklefkin --

       @19: Who specified that it had to be 90%?

       @26: I think there may some confusion in terminology here.  The range containing 95% of model realizations (from the 2.5th percentile to the 97.5th percentile, not -- as you might be thinking -- from the 0th to 95th percentile) corresponds to what is properly called a 95% prediction interval, not a confidence interval.  Confidence intervals reflect uncertainty in estimated parameters (e.g., population mean), whereas prediction intervals denote uncertainty in individual observations (e.g., a single sample from the population).  Since the behavior of the Earth system only happens once in reality, it should be compared with a model's (or group of related models) prediction interval. 

       
       The big difference is that a confidence interval always shrinks with more higher sample sizes (e.g., the estimate of the mean becomes less uncertain), but the prediction interval does not (although the interval boundaries do become more precise with higher sample sizes).  Running more model runs will tighten up the confidence interval around the ensemble mean, but will not generally tighten up the prediction interval. 

       If intervals are based on decent sample sizes, a very large fraction of observations will fall outside the confidence interval (for the mean) but within the prediction interval (for observations).

41850. engineer8516 at 06:13 AM on 2 October 2013
       IPCC model global warming projections have done much better than you think
       

       "Please dial the tone back a bit and try to see the value in the contents of the posts to which you are replying, rather than merely trying to refute them."

       I think I'm being misunderstood. My comment was just directed at the statement, "This is why climate modelers don't make predictions; they make projections, which say in scenario 'x', the climate will change in 'y' fashion." I was commenting on a statement that looked like it was arguing differences in the definition of words as a way to gloss over inaccuracies in a model's predictions. I don't want to get into a philosophical discussion, but if a model disagrees with nature then there is something wrong with the model. And that applies to any model.

Prev  829  830  831  832  833  834  835  836  837  838  839  840  841  842  843  844  Next