Most of you will not be surprised to see that the first figure I wanted to examine was this comparison of temperatures to the AOGCM model spread. This seems to be the cmp5 model/observation comparison for individual runs.
Quick notes:
- Comparison uses 1961-1990 baseline. The AR4 projections were all expressed relative to 1980-1999 mean. This is not an important change but it’s different. That means I’ll need to show both at year end.
- Observations are in the low end of models projections.
- They are using color to show that spread is partly due to models. (Earlier graphs shows everything in black and grey. I suggested this on twitter or at a blog. I’ve been using color this a long time. I’m going to claim credit even if I don’t deserve it.)
Text connected with this figure seems to include
There is likewise very high confidence that models simulate realistically the global-scale surface temperature increase
over the historical period, especially the last fifty years (Section 9.4.1.3.1, Figure 9.8).
and
9.4.1.3.1 Global surface temperature and humidity
56 Figure 9.8 compares the observational record of 20th century changes in global surface temperature to that
57 simulated by each CMIP5 model. The frequency and magnitude of the interannual variability in most of
58 these simulations is generally similar to that of the observations although there are several exceptions. The
Second Order Draft Chapter 9 IPCC WGI Fifth Assessment ReportDo Not Cite, Quote or Distribute 9-26 Total pages: 218
gradual warming evident in the observational record, particularly in the more recent decades, 1 is also evident
2 in the simulations, although again there are some important differences among models. The interannual
3 variations in the observations are noticeably larger than the multi-model ensemble because the averaging of
4 individual model results acts to filter much of the variability simulated by the models. On the other hand, the
5 episodic volcanic forcing that is applied to many of the models is evident in the multi-model agreement with
6 the observed cooling particularly after the 1991 Pinatubo eruption. Because the interpretation of differences
7 in model behaviour can be confounded by internal variability and forcing, some studies have attempted to
8 identify and remove dominant factors such as ENSO and the impacts of volcanic eruptions (Fyfe et al.,
9 2010). Efforts such as these can reduce trend uncertainties and thereby improve our ability to evaluate
10 simulated changes with observations. In summary, models broadly capture the observed historical changes in
11 global surface temperature, and in particular the warming of recent decades. Both model formulation and the
12 applied external forcings (see Chapter 10) influence this level of agreement.
13
14 [INSERT FIGURE 9.8 HERE]
15 Figure 9.8: Observed and simulated annual mean global average anomaly time series of surface air temperature. All
16 anomalies are differences from the 1961–1990 time-mean of each individual time series. Top: single simulations
17 currently available for CMIP5 (thin lines); multi-model mean (thick red line); different observations (thick black lines).
18 Vertical dotted brown lines represent times of major volcanic eruptions. Observational data are HadCRUT4 (Morice et
19 al., 2012), GISTEMP (Hansen et al., 2010), and NCDC (Smith et al., 2008b) and are merged surface temperature (2 m
20 height over land and surface temperature over the ocean). Top, inset: the absolute global mean surface temperature for
21 the reference period 1961–1990, for each individual model (colours) and the observations (black, (Jones et al., 1999).
22 Bottom: single simulations from a variety of EMICs (thin lines). Vertical dotted brown bars represent times of major
23 volcanic eruptions. Observational data are the same as for the top panel.
In all this verbiage mentioning that “models broadly capture the observed historical change”, I see no mention that the observations are in the lower range of AOGCM predictions. I’ll have to look further. Has anyone found any such mention?
Of course we don’t know what the projections will ultimately be. But I thought I’d show this particular graph as we are bound to see similar things again. I also read to look into the EMICs (“Earth System Models of Intermediate Complexity (EMICs)” )models which appear to be in better agreement with data? (Or not. But cursory inspection at least shows the observations more or less in the center of the models in the beginning of the 21st century.)
More later. 🙂
Now call me Dr. Suspicious but I think the mark of success of a model is when it has the same values as the real data. If my model runs are all above or all below the actual line, then its a bit of a fail.
Perhaps someone could take these people to a shooting range and give them rifles with scopes.
Doc Martyn–
It will be interesting to see where 2012 appears when they add it to that graph. I’m glad they added the colors so idiots would stop “patiently explaining” that the spread in the model runs from different models is what we expect from “weather”. (The notion that it’s “weather” had been promulgaged at RC and in peer reviewed papers. It’s obvious nonsense– but using “grey” for the spread permits people who want to convince those with too little time to download the graphs that that spread might be “weather”.
Lucia,
“very high confidence that models simulate realistically the global-scale surface temperature increase over the historical period, especially the last fifty years”
What are these folks looking at, cause it’s not the graph. Are they delusional? Since changes are stated relative to 1961-1990, I don’t see how that qualifies as simulating “the last 50 years”. Maybe the last 21 years… no wait, they could tune the models to match up to 2001, and after that, by the purest of coincidences, the models look like crap.
.
I have not seen this chapter. Do the models continue to use, shall we say, ’boutique’ net forcing to generate those graphs? I rather suspect they do… kludges are bad habits to break.
SteveF–
I don’t know what they did in the early period. Owing to the fact that ZBblock had just updated, AND someone in china wanted to get access and I had made a “rule” due to weird things Google Chrome does (it sometimes spoofs referrers by default. Really!) I ended up dealing with issues. But… I would not be surprised if they do what they always did: Let the modeling groups use custom forcings in the 20th century and then follow SRES for projections. I’ll check after I get back from visiting my mother-in-law.)
Even in the “reference period” 1961-1990, it looks to me like the models are (by and large) lower than observations at the start of the period, and above observations at the end. And of course in the 21st century, models have a much higher trend than observations.
While one can say that “models broadly capture the observed historical changeâ€, it doesn’t suggest that one should expect accuracy in forecasting, when even hindcasting is challenging.
“There is likewise very high confidence that models simulate realistically the global-scale surface temperature increase
over the historical period, especially the last fifty years (Section 9.4.1.3.1, Figure 9.8).”
Wouldn’t it be more correct to say that “the multi-model mean simulates realistically the global-scale surface temperature increase”, since the individual models don’t seem to.
Does the second graph indicate that in AR5 the models may have been re-aligned to match observations?
In Ch9 on p26 they say that most climate models show more troposphere warming than observations, then discuss controversy in the literature, and cite McKitrick McIntyre & Herman 2010, Santer et al 2012, Po-Chedley & Fu, all of which say the models have a warm bias (yes, even Santer et al have now admitted this, though comically they don’t cite MMH, or even their own 2008 paper where they said the opposite!)
On p 27 they say there is high confidence that most models overestimate warming in tropical troposphere, and the cause of this bias remains elusive.
Ray
Do you mean figure 9.8(b)? I think those are “EMICS”. That’s a different sort of thing.
It’s refreshing to read this admission on 9-7:
It would be even more refreshing they just admitted that that modelling centers generally do not describe how they are tuned at all!
I trained all my models to match the 1961-1990 lineeshape. Actually the models are quite crap at this and so my average runs 0.45 degrees too cold. The moment I allow my models to run free, they shoot up and run too hot.
Training cold and future too hot. What about hindcasting? In the period i pretent I have not trained my model to (but really have) all my models are hot, damned hot.
So past; hot, training period, cold, future; hot.
What does this say? All your models are crap, really really crap. The level of crapness is even better that carp when you look at the individual runs, the coldest hindcast is the hottest future.
If someone attempted to model any biochemical or chemical process and got this; then presented it to me I would throw the drive at them If I found that the fit was more than 4 variables I would tell them to get employment in a field better sited to their skills; a sewage farm worker or zoo keeper.
Well… except to the extent that the model fails to match those features it is tuned to. Then we can know it’s likely not very good.
For steveF
page 9019
Suggests the modeling groups still used what they wanted, but the spread might be tighter.
Generally, this document is better at admitting many obvious things that was the AR4. We see admissions that hindcasting skill is not a stringent test in several places. For example here:
More for SteveF”
So the 20th century emissions and etc. changes more constrained that previous. (Obviously, we’ll want to read further details.)
Yes, the models are crap. But that said….
The notion of plotting all these model runs on one graph along with the model average is bizzare. What on earth does the modeling average represent?
The spread does not represent variance in the ‘data”. Models are not “data” and model results have no variance. Each model represents a different version of reality. It captures reality, or not. The only reason to plot more than one model on a graph with data is to figure out which model is right and which is wrong Presumably this contract would lead to an even better model.
Lookit: two wrogns don’t make a right and neither does the average of two wrongs and this observations scales.
Besides. They are, each and every one, curve fits.
lucia:
Really? That’s crazy. I’ve never liked Chrome so I don’t know too much about it, but… why would it do that?
To me, the most remarkable factor in the CMP5 vs data first graph is how rapidly the CM5 model mean is diverging hotter than the observed temperature after 2000 – especially when compared to the relatively close match prior to that. I look forward to Lucia’s quantitative evaluations of of such difference.
This major divergence strongly suggests to me that the CMP5 models are influenced by the “argument from ignorance” – i.e., that they are missing major physics such as natural cloud declines while consequently strongly overestimating climate sensitivity to CO2. e.g., see:
“Ryan Eastman, Stephen G. Warren, A 39-Year Survey of Cloud Changes from Land Stations Worldwide 1971-2009: Long-Term Trends, Relation to Aerosols, and Expansion of the Tropical Belt Journal of Climate 2012 ; e-View doi: http://dx.doi.org/10.1175/JCLI-D-12-00280.1â€
Eastman and Warren show about 1.56% decline over the 39 years from 1979 to 2009 in global average cloud cover (~0.4%/decade), primarily in middle latitudes at middle and high levels.
Declining clouds if from a natural cause, may have caused most of the observed global warming. That raises the major challenge of distinguishing natural vs anthropogenic causation.
Interesting to look at ~1883 (Krakatoa?). The models seem to show much more cooling than the temperature record. I wonder if that is because they all over estimate aerosol cooling.
Brandon: Read this:
http://googlewebmastercentral.blogspot.com/2012/03/upcoming-changes-in-googles-http.html
Here
*”simplifying” means “instead of sending you the correct referrer, we will pretend the user came from the top of the domain. So, on my end, I know they came from google but from a page on which there is no link to my site. That is: it looks fake because– the “referrer” does not point to the page that actually referred them.
* “modern browser such as Chrome.” means “with Chrome”. (As far as I can tell no other browser buys into this “simplified referrer notion”.)
Note also:
If ever search engine were to buy into this, and all browsers decided to be “advanced” oin this way, it seems that to learn which search terms brought traffic to your site from google, you would have to use google web master tools. You couldn’t get it form the referrer. Imagine if bing, and other search groups do it to. We then have to get our google referrers from google, bing referrers from msn and so on. Meanwhile, in the name of “privacy” they still knwo where you came from and if someone like me wants to learn what search term you used to come to my blog I must let them watch you. I let them do this by adding a google tracker to my blog. (Mind you– I do have a google tracker on my blog. But this new “method” would sort of mean that sites who really want to know the search terms that bring people to their site have to let google peak at all their users visits!!)
I would think that privacy advocates would be horrified by this decision by google– since google is one of the main groups whose ever present monitoring bothers them!!
Anyway… with respect to me… I did see some weird behavior from things that were leaving “impossible” google referrers. I banned that. Humans– those using Chrome — contacted me. I tweaked my rule. But I was very puzzled… Very. So I had some conversations to figure out “what the heck”.
As far as I can tell: Chrome is automatically spoofing referrers by default under some circumstance. One visitor said he hit my “rule” even when he cam from climate audit or bishop hill– but possibly he is mistaken and the “ban” was just something I had not yet cleared up. Whether goggle decision is “good” or “bad”, it is “weird/non-standard”. I didn’t anticipate it and I needed to know it to create speedy not intensive “rules” that catch bots without banning real surfers.
lucia, thanks, that’s interesting. The idea Google is expressing makes sense. They want to avoid having to redirect users to a page to strip out potentially private information. It just seems like their implementation is weird.
Question (I’m not a maths person).
Is there a (technical) reason why the model spread appears to narrow for the reference period in the first figure?
HR–
Yes. It’s due to the fact that every single realization is set to have an average =0 during that period. Outside the period, they can spread apart more. You can create an even tighter more dramatic narrowing by choosing a short time period for the baseline.
lucia (Comment #107442)
“Do you mean figure 9.8(b)? I think those are “EMICSâ€. That’s a different sort of thing.”
Thanks,
I haven’t encountered “EMICS” before.
How do they relate to the other IPCC climate model scenarios?
Does “intermediate complexity” mean they are less complex than the models used for the normal scenario projections?
kdk33 (Comment #107449)
“Lookit: two wrogns don’t make a right and neither does the average of two wrongs and this observations scales.”
I think in accounting that would be known as “compensating errors”.
I have always thought it dubious that they can claim that the multi-model mean is accurate, when individual models are mostly wrong.
Lucia said
“…. the EMICs …. models which appear to be in better agreement with data”
Scientists tend to be precise or precisely imprecise.
For AOGCMs the legend says “single simulations currently available” which says to me they used all the available data.
For EMICs the legend says “from a variety of EMICs” which says to me something else!
Hmmm, is anyone else bothered by how similar the past reconstructions are in overall shape, despite being rather far apart in temperature?
Why is “the models resemble each other” a good thing, exactly?
I’ve been running a home experiment the last couple of days and I’ve been changing everything that I can think of which might show my hypothesis to be wrong, the data spread is scattered as a result, but the only similarities are due to the particular constraints on the overall experiment… not because I’ve been trying to get each test run to resemble every other run… that’s not science, is it?
Hi Lucia,
It is good to see a clear acknowledgment that hindcasting is a “necessary but not sufficient condition” to establish confidence in the reliability of forecasts. However, it is already very well known that the test on average surface temperature turns out to be an extremely weak measure of a model’s predictive reliability. Any individual model can achieve a reasonable match of surface temperature by being “out-of-range” on other key inputs or key outputs
…not to mention having a vertical temperature profile which is at odds with observations.
Additionally, the recent paper by Armour 2012 http://web.mit.edu/karmour/www/Armour_JCLIMrevised_2col.pdf finally offers a coherent explanation for why the future aggregate global radiative response in the GCMs is curved rather than linear in the temperature domain. It is due in the models to the large latitudinal variation of effective feedbacks AND the very different response times in the different latitude bands. On present evidence, this phenomenon accounts for around 50% of the ECS values calculated in the models relative to a quasi-linear projection of radiative response to temperature from the historical period.
This strongly suggests to me that it is more important to compare surface temperature observations against models by latitude bands rather than as a simple average.
I would dearly love to see a grown-up conversation about model reliability in the AR5, but I suspect it is still too early to expect the appropriate level of scientific humility. Still, I wait to be pleasantly surprised.
It would be even more refreshing they just admitted that that modelling centers generally do not describe how they are tuned at all!
It would be additionally more refreshing if they admitted that the tuning is accomplished through tweaks to some of the multitude of the, generally ad hoc, algebraic parameterizations that carry whatever limited connection with physical reality the models weakly attain. And that several of these tweaks are associated with ad hoc parameterization modeling of phenomena and processes that are among the fuzziest ( 2 difficult to perceive clearly or understand and explain precisely; indistinct or vague ) of all those in the models. And that such tuning / tweaking is prone to lead to ‘right answers for the wrong reasons’.
Not only do the EMIC seem to match the obs. much better they seem to match each other much better as well. There’s a lot less ‘noise’ in the runs. A feature of the working of the models or data selection or something else? Who Knows?
HR
The MIC’s in EMIC is “models of intermediate complexity”. I need to check further, but I suspect they are of the “N-box” type. It wouldn’t be surprising if they had no “weather/internal variability” at all.
Paul_K
I’d say it’s much better than the AR4. Moreover, the figures are organized to let people see warts even if the text doesn’t come out and admit them. For example: notice the “top” panel of figure 9.8(a) I show as my first figure in this post. That looks to be an explosion that shows the range of absolute temperatures for the models. Clearly, they all differ from each other and by quite a bit– and anyone with 2 working brain cells knows they can’t all be “right”.
To find equivalent information in the AR4 you had to dig up the supplement, find their graphics of the entire earth, and mentally integrate the ‘color’ — mentally thinking of the little color legend provided– and then based on your mental idea decide what the spread in model mean absolute temperatures of the earth might be. Meanwhile, with respect to describing these differences the language was suitably opaque and was worded to “spin” the notion that what we were seeing was somehow “good” agreement.
Of course, they are still saying things like “broad agreement” between observations and models. But what the heck is their definition of “broad agreement”? As far as I can tell it means little more than:
1) Both the models and observations say it’s cold at the poles and warm at the equator.
2) Both the models and observations say it’s warmed over the last 100 years or so. The difference in the rate of warming is less than 1 order of magnitude. (It’s a bit less than that even– but some calculations would need to be required. )
3) Both models and observations agree that the absolute temperature is such that during the reference period, ice formed at the poles, and vast expanses of liquid water exist on the earth.
As far as I can tell that’s the definition of “broad agreement”. In contrast, under this definition, “disagreement” would occur if the models said the earth climate resembled mars or venus. But it’s not at all clear how far off they might need to be before anyone would admit “disagreement”.
On the other hand– there is a lot of discussion about how one might quantify disagreement.
Lucia,
Thanks for the excerpts. So if the “second phase” covers 2006 to 2100, then does that mean modelers could tweak the models for a better fit 2001 to 2005? Sure sounds like they could. But the fact that they must zero over the same period makes tweaking more complicated; any tweaking has not reduced the post 2001 divergence from the historical record.
The question in my mind is how long it will be before the modelers become too embarrassed by the obvious and growing divergence from the measured temperature, and start changing cloud parameters to reduce feedbacks and get closer to realistic behavior. That clearly has not happened so far… looks like it will be ‘kicking and screaming all the way’. Unfortunate.
Paul_K,
I don’t think a grown-up conversation is possible yet; too much political and personal baggage for that. But since it is pretty clear that there will be continued (and growing) CO2 emissions for at least two decades, with ever higher forcing, I am cautiously optimistic that conversation will happen, and probably sooner rather than later, unless there is a sudden increase in the rate of warming, which seems very unlikely. One can look at the advancing ages of influential activist-scientists (who will strongly resist reductions in estimated climate sensitivity) to get an idea of when the grown up conversation will happen; my guess is within a decade, and maybe less.
Die Klimazwiebel has a couple of posts linked up to a Spiegel article covering the outcome of Doha.
http://klimazwiebel.blogspot.com/
The article closes with this:
http://www.spiegel.de/international/world/top-researchers-call-for-an-end-to-united-nations-climate-summits-a-872992.html
(Krauss is one of the regular contributors to Klimazwiebel.)
SteveF,
I agree with you. The problem is that a huge amount of political and economic damage can be done in 10 years. It always takes far longer to build than to dismantle.
Paul
Lucia,
I think some have suggested that the threshold of true disagreement is when all runs of all the models are above the measured temperature. So by that logic, if any model is plausibly correct, based on at least a single run consistent with the measured temperature, then we are required to accept that they are all plausibly correct, and to agree that the model mean is the best estimate of the future. That is the bizarre world of climate modeling…. the modelers seem quite beyond embarrassment.
Paul_K,
” The problem is that a huge amount of political and economic damage can be done in 10 years. It always takes far longer to build than to dismantle.”
Agreed, and so reasonable people should do their best to impede the processes which lead to that damage… both political and ‘scientific’.
John_M
Also, oddly enough, discussing local efforts to prepare for the local effects might have the effect of demonstrating how much preparation would cost and what is required. If preparations were very costly — resulting in larger taxes– and the requirements were unpalatable (like ugly things to control storm surges), this might convince some people to lower CO2.
Of course, the danger is that people might also see they can deal with the worst aspects locally in the same way they deal with many other issues
But as it stands…. well… we just have these large cumbersome meetings that get lots of press coverages and results in ‘kick the can’ agreements that are nothing more than to meet again in two years and come up with something we can agree on at that future meeting.
Hi Lucia,
re #107473
I think your slightly sardonic tone is appropriate. It is like putting gloss paint straight over the top of cracking wood. I liked this one:
Perhaps. Perhaps.
Moreover, the figures are organized to let people see warts even if the text doesn’t come out and admit them.
We should always celebrate any movement towards scientific integrity. My pessimistic side however says that if the IPCC really wanted to offer a fair summary, they would bring up-front in the summary for policy-makers some clear conclusions about the evident GCM problems and the residual uncertainties, rather than offer a sop to us nerds (who are occasionally willing to delve into the small print or try to squeeze sense from a picture with inadequate supporting information) in the form of an under-explained picture. In your specific example, can you estimate the impact that “running cold” has on the latent heat calculations in any given model? Is it a small effect? Is it OK to ignore it? What level of uncertainty does it introduce? Should we be worried about the integrity of forecasts based on these large apparent differences in absolute temperatures?
Rhetorical questions, obviously. I don’t have a clue, and I suspect that the AR5 authors for the most part don’t have a clue.
But, as you say, we should celebrate the fact that at least they have included the figure rather than suppress it, well so far at least.
Paul
Lucia,
Yes, it will be interesting to see if this attitude takes hold and what the ultimate impact will be.
The view in the article is reminiscent of Pielke Jr.’s “no regrets” policy…i.e, it doesn’t matter if Sandy was x.yz% stronger because of climate change, protecting the NY/NJ coast and NY Harbor against storm surges is the right thing to do.
Perhaps this is not surprising, since the article reflects the thinking of some of the Klimazwiebel guys, who are like Pielke, meaning while they are not skeptics, they aren’t violently opposed to skeptical viewpoints (climate blue dogs?).
Of course, like Pielke, their moderate position can cause the likes of Gavin Schmidt to have a hissy fit.
http://klimazwiebel.blogspot.com/2012/12/i-have-new-paper-out-co-authored-with.html?showComment=1355524251653#c4195303094709502094
Lucia,
“If preparations were very costly — resulting in larger taxes– and the requirements were unpalatable (like ugly things to control storm surges), this might convince some people to lower CO2.”
Yes, it might, and that is a perfectly normal political process. The two flies in the cooky dough are that 1) the costs are local and highly variable (Chicago probably benefits from warmer temperatures, Venice will suffer from rising sea levels), and 2) the advocacy of scientists-cum-green/leftist politicians has lead to projections which are easy to discredit… as your regular comparison of models with reality shows.
.
There are real problems which need to be addressed (sea level increase being the most obvious). But outrageous projections like 1-2 meters sea level rise in 88 years and 3-4C warming in 88 years, which appear designed (a la Stephen Schneider) only to force specific, extreme policies on the global public (Distribute global wealth evenly, and stop using fossil fuels now, regardless of cost!) are so obviously and demonstrably wrong that reasonable steps toward reducing future warming, and adapting to the consequences of future warming, are never even discussed. The wages of the political corruption of science are heavy indeed.
Paul_K–
Remember that there is “The IPCC”, but there are also “Authors” and contributors. Plus each group consists of individual people.
Reading the AR5, my sense is there are at least some authors who really truly want to play it straight. I know that previous comparison graphics used “grey” to show the spread. The current one uses the colors which I think I first introduced at this blog. I know I suggested the colors to Ed Hawkins and stated why it’s useful. And I know he thought it was a good idea.
Now… suppose an author who is interested in not hiding stuff comes across the idea– and then — that same author hears of or himself comes up with the nifty idea of elaborate by adding the graphic showing the spread in temperatures above. etc. That shows stuff. That author can almost certainly introduce the graphic and defend it. Because it’s hard for someone who wants to obscure to come up with any excuse for why the colors should be tweaked to make the graph more confusing etc. Or to explain why the information showing how the models disagree on temperature spreads should be removed.
So… the figure gets in. Because it is just as compact as any more vague figure, communicates more, is accurate and so on.
But now… for the words. Can that same author who went to the trouble to get this good graph get text in that says the difference between in the mean temperature of the earth in the hottest and coldest model is 2 or 3C”? Or to put that 2-3C next to the size of the projected increase over the century? I’m sure that all sorts of people will highlight this… and say that sentence takes up too much room… and it’s unnecessary… or confusing… .(In some sense, it sort of can lead to confusion. Because no matter what the temperature of the earth is, a very rapid rise — or drop– would be difficult for people to adapt to. But presumably, that is a point that can be made actively instead of by “hiding” information about the models.)
Anyway, the result could be precisely what we are seeing: Much better graphics. Long sections discussing how one would test models. But tap dancing away from anything that comes out and admits that models are hardly precision tools and our ability to project is not all that good. You’ll end up with a sentence that is just gush– like “they all broadly agree” and then ‘so we can have confidence that models are useful” and such like. (Note: what they might be useful for is omitted. Obviously, they are useful for something. Mallets and hatchets are both useful. But it is generally not helpful to write prose in ways that suggests that a mallet might be an appropriate tool if our objective is to chop wood. Yet that’s the sort of prose we get out of these committee written reports (and it’s the sort of gush we get out of people like Gavin Schmidt.)
Well.. and obviously, the sea rising feet isn’t a local problem for the Chicago area. We might need to worry about Lake Michigan levels varying– which they have in the past. They were once low. Buildings got built. Nice plush apartments with underground parking garages. The lake level then rose and some parking garages were flooded. (I think this happened at Lake Point Tower.) Now the lake level is low again.
Both changes can result in costly problems even in a city show building plan called for large strips of parks along the lake. ( Quote from Wikipedia ““The Lakefront by right belongs to the people,” wrote Burnham. “Not a foot of its shores should be appropriated to the exclusion of the people.†The plan recommended expanding the parks along the Lake Michigan shoreline with landfill, which was done in the early 20th century. Of the city’s 29 miles (47 km) of lakefront, all but four miles (six kilometers) are today public parkland. The plan also provided for extensive lakefront harbor facilities, which became unnecessary with the city’s development of facilities in the Lake Calumet region.”)
Having much of the lakefront park was a wise decision– and for reasons that go beyond those advocated by Burnham way back when!
John M (Comment #107482),
Thanks for the link. It is a very interesting exchange. Gavin gets his panties up in a wad because he recognizes (quite correctly) that scientists who publicly advocate policy lose scientific credibility because they are judged by reasonable people to be subject to bias. What he does not seem willing to recognize is that scientists (like most everyone!) who hold strong views on specific politics and policies, even if they do not discuss them publicly, may be just as subject to bias.
.
I particularly liked the references to RealClimate’s harsh moderation policies and how it is a climate advocacy site. Gavin wisely didn’t take that bait. But those comments must have fried Gavin’s as… er, upset him… since they are so obviously true. Lemme-see, Gavin spends loads of time working on a climate advocacy blog, but is in no way an advocate. For sure. 😉
re: “Because it’s hard for someone who wants to obscure to come up with any excuse for why the colors should be tweaked to make the graph more confusing etc. Or to explain why the information showing how the models disagree on temperature spreads should be removed.”
We’ll see how it shakes out in the end. Editors and authors of prior assessment reports seemed to have no problem filtering out meaningful data that was contrary to the religious doctrine message, despite objections. I wouldn’t get your hopes up.
Lucia,
FYI.
The technical summary document uses the all-one-color graphic (all yellow lines) instead of the colored lines like above. Kind of weird though… the technical summary is labeled “first draft” in some places but appears to imply “second draft” in others. Not sure what to make of that. But if the technical summary stays in that form, then all that new transparency about differences between models disappears for the non-technical types who will not ever get into the nitty-gritty of the individual chapters.
DocMartin
Re: “So past; hot, training period, cold, future; hot. #107445
What does this say? All your models are crap, really really crap. The level of crapness is even better that carp when you look at the individual runs, the coldest hindcast is the hottest future. . . .If I found that the fit was more than 4 variables . . .â€
Thanks for that reality check and reminder that:
Attributed to von Neumann by Enrico Fermi, as quoted by Freeman Dyson in “A meeting with Enrico Fermi” in Nature 427 (22 January 2004) p. 297
Now just how many variables are “tunedâ€, “adjustedâ€, “fit†or otherwise manipulated in these global climate models? Especially for the clouds? (seeing that clouds contribute ~97% of the uncertainty?)
Why is it then that phenomenological models with fewer parameters and better forecasting/hindcasting and predictive skill are so disparaged?
Re:David L. Hagen (Comment #107495)
December 15th, 2012 at 9:14 pm
Hi David,
I suspect that we may define “phenomenological model” a little differently; it is not of great importance, but I would use the term “empirical model” to describe Scafetta’s model.
The main value of Scafetta’s paper is in demonstrating that the GCM’s do not model the observed natural frequencies. In this, it does a good job – along with several other good papers on the subject.
However, the use of Scafetta’s “fit” to the data as a method of predicting future behaviour is fraught with problems, at least if it is extended beyond half the wavelength of the longest certain cycle modeled; in this instance this means that if the assumptions are valid, one might get a moderately good projection for 30 years or so. (Scafetta himself recognises this.)
Beyond this, the long-term trajectory is seen in the temperature response only at a frequencies lower or much lower than this longest certain cycle. Scafetta doesn’t know what the long-term trajectory is. His model uses a quadratic fit to this trajectory in the temperature domain, which is equivalent to assuming a long-term linear fit to incremental flux forcing (this integrates into a quadratic form for cumulative flux forcing). I would argue from the data that this is not the best fit to the observed data, but that is not my key point here. The key point is that the longer-term projection – in Scafetta’s model or my own – is based entirely on this low frequency behaviour, and we don’t have sufficient information to know what it should look like with or without anthropogenic influences. Hence, we can’t use this type of empirical model for projections beyond about 30 years. It has no coherent basis until it is underpinned by a solid assumption about the shape of the low frequency data.
The competing assumption adopted by the GCM fraternity is that there are no long-term cycles in the data, and hence the low frequency trajectory is controlled entirely by known or unknown exogenous forcings. This does at least form a coherent assumptive foundation for predicting temperature in 100 years and beyond. Of course, if the assumption is incorrect, then so are their predictions, all else being equal.
David Hagen,
“I would argue from the data that this is not the best fit to the observed data, but that is not my key point here.”
I should have said that I would argue from the data and the physics that Scafetta’s low frequency model is not the best fit to the data. His quadratic fit in the temperature domain implies a straight line fit of incremetal flux forcing in time. This does not sit well with the notion that it is part of a continuous oscillatory cycle, where we would expect to see some curvature. Alternatively, if the counterargument is that the straight-line is part of a sawtooth oscillation, then it is unpredictable anyway.
You do not need Scafetta to tell you that, they will tell you that themselves. The natural cycles are not ‘frequencies’ anyway.
I remember noticing this years ago, http://www.ipcc.ch/publications_and_data/ar4/wg1/en/spmsspm-projections-of.html in the AR4 WG1 summary, the projections of precipitation map has stippling, but is low quality.
I thought perhaps it was just unavoidable because the larger version it was resized from couldn’t be saved to that size without taking up extra space.
Original 226 kb version: http://www.ipcc.ch/publications_and_data/ar4/wg1/en/fig/figure-spm-7-l.png
Resized 103 kb version in the SPM: http://www.ipcc.ch/publications_and_data/ar4/wg1/en/fig/figure-spm-7.jpeg
My 162 kb version the same size, resized in GIMP without reducing the quality: http://i341.photobucket.com/albums/o396/maxarutaru/figure-spm-7-l.png
I have a sub 90 kb .jpg version somewhere which is just as clear as the .png, but I could never figure out why they resized it with such low quality.
That’s rather similar to choosing different colors than ones which would make a chart more legible, isn’t it?
Lucia, did you find a statement as to the magnitude of natural variability in your readings? It was stated in Ch 10 of AR4. If you look at your first graph, the models are running hot at about 2.8C/century from year 2000. If they did not run the HadCrut4 to 2012 as it appears, that makes it about 3C/century hot from year 2000. They are showing a spread that already falsifies the magnitude for expected natural variance by 2030 as indicated in AR4. Was this commented on anybody?
Evidence of the effects of human influence on the climate system has continued to accumulate and
4 strengthen since the AR4. The consistency of observed and modeled changes across the climate system,
5 including regional temperatures, the water cycle, global energy budget, cryosphere and oceans, points to a
6 large-scale warming resulting primarily from anthropogenic increases in greenhouse gas concentrations.
effects of internal variability become more significant in masking or enhancing externally
45 forced changes. Observational uncertainties for climate variables, uncertainties in forcings such as aerosols,
46 and limits in process understanding continue to hamper attribution of changes in many aspects of the climate
47 system, making it more difficult to discriminate between natural internal variability and externally forced
48 changes.
This is from the attribution summary. I have to ask what consistency of observed and modeled changes are they talking about. Well only 128 more pages to find out.
The only reason the models “match” the overall pattern of temperature is that there is a big jump down of about 0.4deg in about 1962 (volcano?) by all models, which they then take forever to recover from. Willis discussed somewhere how the models are way too sensitive to volcanos.
Craig Loehle (Comment #107505)
“Willis discussed somewhere how the models are way too sensitive to volcanos.”
I wouldn’t limit it to volcanos. 😉
I thought that RC smuggling in “weather” in this context was in response to the fact that “natural variability” or “uncertainty” are now seen as the rhetorical property of the heretics.
Tune oh so bravely
Natural variations
Become mere weather
Craig Loehle (Comment #107505)
That was the source of my comment way above (107455)
The models appear to require a large volcanic adjustment to Pinotuba in order to match the temperature records. But then you look back to circa 1893, the time of Krakatoa, and the models show a distinct cooling which is missing from the temperature records. To me that seems to imply that the models may be significantly over estimating the volcanic cooling.
Re: Craig Loehle (Dec 16 08:02),
Mount Agung erupted in 1963, but it wasn’t noted as having particularly high sulfate aerosol emission (see this graph ). Certainly not in the class with Pinatubo (1991), El Chichon (1982), Krakatoa (1883) or Tambora (1815). 1965-66 was an El Nino event, but not a particularly strong one. OTOH, the GISS stratospheric aerosol forcings show a large peak that can’t be due to just Agung as it starts before the eruption. Tuning anyone?
Re: John Vetterling (Dec 16 10:25),
When you go back that far, the instrumental record has a lot of uncertainty.
SteveF, ““Willis discussed somewhere how the models are way too sensitive to volcanos.â€
I wouldn’t limit it to volcanos.
Some of the modelers appear to be pretty sensitive also.
Re: DeWitt Payne (Dec 16 11:56),
Got that El Nino/La Nina thing backwards. There was a La Nina in 1963-64 but there were El Nino events just before and just after.
Dallas: “Some of the modelers appear to be pretty sensitive also.”
is this like how people and their pets start to look alike?
very funny!
Craig,
I think the modelers are indeed pretty sensitive, because they know how the sausage is made. Lucia noted above that the climate model inter-comparison studies use the same assumed forcing for all models, at least post 2001. The key question is if that assumed forcing is consistent with the current aerosol offsets and net GHG forcing that the second order draft itself describes (with much lower aerosol effects compared to AR4) or some much lower current net forcing level similar to AR4; I rather strongly suspect the later, and we are seeing something akin to apples versus oranges.
SteveF, ” I rather strongly suspect the later, and we are seeing something akin to apples versus oranges.” We at least have to up to pineapples by now. Annan and Hargreaves had an interesting comment in their discussion paper on Climate of the Past, something like their models fit the supposedly unreliable proxy data too well and the “reliable data” not so well.
I am still waiting for the BEST Global land and oceans absolute temperature. My best guess is 17.5C if they use the marine air instead of “SST” which not a sea “surface” temperature.
Paul_K
Thanks for your comments and observations.
Re: “The competing assumption adopted by the GCM fraternity is that there are no long-term cycles in the data, and hence the low frequency trajectory is controlled entirely by known or unknown exogenous forcings. This does at least form a coherent assumptive foundation for predicting temperature in 100 years and beyond. Of course, if the assumption is incorrect, then so are their predictions, all else being equal.â€
On the fitting only being accurate to ½ the longest frequency, wouldn’t that longest frequency correspond to a period with twice the longest length of data (from Nyquist)?
E.g. Scafetta identifies “9.1, 10–10.5, 20–21, 60–62 year periods†using direct temperature data back 160 years to 1850. He is able to hindcast/forecast using half that period.
Furthermore, he finds natural causes for those cycles, so they are not just “fitted†data.
There is growing evidence from numerous authors on climatic fluctuations, providing greater cautions over these GCM assumptions. E.g.,
Alexander et al find significant impact of the 21 year Hale solar cycle.
WJR Alexander et al., Linkages between solar activity, climate predictability and water resource development, Journal South African Institution of Civil Engineering Vol 49 No 2, June 2007, Pages 32–44, Paper 659″ http://nzclimatescience.net/images/PDFs/alexander2707.pdf
D’Aleo and Easterbrook (2011) address the ~ 60 year oscillations.
Aleo, J. and Easterbrook, D.J., 2011, Relationship of multidecadal global temperatures to multidecadal oceanic oscillations: in Easterbrook, D.J., ed., Evidence-Based Climate Science, Elsevier Inc., p. 161-184;
From available empirical evidence, the GCM’s assumption of “no long-term cycles†is wrong beyond ~ 100 years because of the observed Little Ice Age, Medieval Warm Period, Roman Warm Period, Minoan Warm Period, cooling from the Holocene Climatic Optimum, the rapid warming from the last glaciation, and the geological records from the previous interglacial periods. E.g.,
Akasofu (2010) approximates the fluctuations causing the Little Ice Age as a linear rise in temperature from the LIA.
Syun-Ichi Akasofu (2010), On the recovery from the Little Ice Age, Natural Science, Vol. 2, No. 11, 1211-1224, doi:10.4236/ns.2010.211149;
(Girma approximates that underlying rise from the LIA with a parabolic curve.)
Humlum, O., Solheim, J.-K. and Stordahl, K. 2011. Identifying natural contributions to late Holocene climate change. Global and Planetary Change 79: 145-156.
The fit 4000 years of data to find 3 periods of 2804-year, 1186-year and 556-year.
On longer periods, Loehle & Singer find ~ a 1500 year cycle in the data.
Craig Loehle and S. Fred Singer, Holocene temperature records show millennial-scale periodicity. Canadian Journal Earth Science Vol. 47 pp 1327-1336 (2010).
Similarly, see:
Late Holocene ~1500 yr climatic periodicities and their implications Ian D. Campbell, Celina Campbell, Michael J. Apps, Nathaniel W. Rutter and Andrew B. G. Bush, Geology v. 26 no. 5 p. 471-473 doi: 10.1130/0091-7613(1998)0262.3.CO;2
J.R. Petit et al. (1999) discuss the ~ 100k year cycle of glaciation. Etc.
J.R. Petit et al. (1999) Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica, Nature, Vol 399, 3 June 1999, 429
http://www.daycreek.com/dc/images/1999.pdf
So how should GCM’s be constrained to incorporate evidence of these natural oscillations?
I find the potential to descend into another glaciation in ~ 1500 years to be a far greater risk than a few degrees of warming – if the GCM’s were accurate. Can we generate enough global warming to prevent that next glaciation?
Paul_K
Correction: “On the fitting only being accurate to ½ the longest frequency, wouldn’t that longest frequency correspond to a period EQUAL TO the longest length of data (from Nyquist)?”
New projections? Very slight mean cooling until 2027. Then moderate warming for 30 years (not quite as much as 1970-1998) then, soon after that maximum expect long term cooling for 500 years.
Until you understand what happens on Venus you will never understand what happens on Earth.
The small amount of solar energy absorbed by the Venus surface would very easily exit the surface the next night by conduction (diffusion) and radiation. Then, when this small amount of energy is back in the atmosphere there is plenty of carbon dioxide to radiate it away. There is absolutely no possible way by which such energy would be trapped in the surface and somehow add hundreds of degrees. The problem is, if you follow the “First School of Thought” (the IPCC bluff) then you are at a complete loss to explain Venus temperatures, because, if you think like the IPCC it is because you have been subjected to Ignorant Promulgation of Chaotic Consensus.” You need a paradigm shift to the “Second School of Thinking” in my paper.
Please respond to this comment on another thread..
Something that would improve AR5 hugely would be some sort of graphical colour-coding of the text to reflect the reviewer’s agreement – basically if each reviewer went through the sections that they were involved in producing and gave a 1-10 agreement mark (1 disagree 10 agree) to each statement then we would have a much truer view of the ‘averaged’ consensus.
It would be particularly valuable if applied to the summary.
“E.g. Scafetta identifies “9.1, 10–10.5, 20–21, 60–62 year periods†using direct temperature data back 160 years to 1850. He is able to hindcast/forecast using half that period.”
he is just doing a fourier anaylysis without even realising it. It means nothing in physical terms.
DeWitt Payne (Comment #107509),
There’s more detail on the stratospheric aerosol input here: http://data.giss.nasa.gov/modelforce/strataer/
You can see that Agung’s diagnosed effect on AOD was heavily-weighted towards the Southern Hemisphere, so it wouldn’t be as apparent at the NH Mauna Loa site.
bugs
Re: “he is just doing a fourier anaylysis without even realising it. It means nothing in physical terms.”
On what basis are you distinguishing between “physical” causes and non-physical causes and with what evidence?
Have you looked at the physical causes Scafetta discusses relating to these frequencies?
What reasons do you have for implying that they are not “physical”?
bugs.
Scaffetta refuses to share his data or his code. Even his co -author craig Loehele has condemned this behavior. But you won’t find hagen or others raising a stink about his mann inspired behavior.
Don’t expect skeptics to demand that scaffetta do real science.
They are happy with cyclomania.
Lucia
Any comment on the graph found in AR5 on page 1-39?
Steven Mosher
tu quoque
I’m still waiting for your references to your “better one” than Scafetta’s.
Similarly, there have been many requests for your code supporting your On Volcanoes and their Climate Response.
Or are the above not “scientific”?
What data/method details do you need to reproduce his results beyond what Scafetta provides in his paper and supplemental file?
http://people.duke.edu/~ns2002/pdf/ATP3533.pdf
http://people.duke.edu/~ns2002/pdf/ATP3533_supplement-file.pdf
DeWitt Payne:
I’m not sure this was tuning. My understanding is GISS uses the data described in this paper to calculate its volcanic forcings (at least, for data prior to 1980). That paper describes how a single series was created piecemeal from many different data sets. Piecing together disparate data sets can introduce all sorts of peculiarities, and I’d wager that’s what you’re seeing.
For example, according to that paper there’s a significant change in data set at 1960. If you look at the results from the paper, you’ll see almost immediately after that change the data’s changes. The volcanic forcing was practically non-existent for three years prior to 1960. A few months into 1960, it shoots up by ~1000%. By 1962, it has increased from 0.0002 to 0.013. For the rest of the record, it never drops below 0.0016.
I don’t have the underlying data used in the Sato paper I linked to above, and I’ve never delved too much into this record, but it has obvious problems. Even without reading the Sato paper it is easy to see differences in the series which arise from how the various data sets were combined. I’d assume problems are with how the data series was created before assuming they’re from tuning.
David.
The better one is in the Results paper. Or you could look at Dr. Pratts newest one. Its easy to do better.
The code for Robert Rhode’s work? Check the comments, I posted the monte carlo people asked for. The rest of the code was posted ages ago. Its on the berkely site many people use it. Not too bad for unpublished work.
Scaffettas work? Even Steve McIntrye couldnt make sense of it and I’ve been unable to reconcile his supplements with the final paper. His results are not reproduceable. period.
@Lucia
I haven’t posted here for a while, so maybe my question @ 107537 got lost in the dust.
To repeat my question, one which for conversational purposes probably extends to others commenting on this thread: what do you make of the graph on page 1-39 of the AR5 draft?
Right or wrong, it’s now out there now and can’t be called back -other than on some very heavily corroborated data to the contrary. E.g.: does it fit with the HadCru4 data as per the October Met Office HadCru4 based “bulletin” [since they refuse to call it a report] ?
tetris–
YOu mean in chapter 1? With this discussion below:
“Figure 1.4: [PLACEHOLDER FOR FINAL DRAFT: Observational datasets will be updated as soon as they become 4 available] Estimated changes in the observed globally and annually averaged surface temperature (in ï‚°C) since 1990 5 compared with the range of projections from the previous IPCC assessments. Values are aligned to match the average 6 observed value at 1990. Observed global annual temperature change, relative to 1961–1990, is shown as black squares 7 (NASA (updated from Hansen et al., 2010; data available at http://data.giss.nasa.gov/gistemp/); NOAA (updated from 8 Smith et al., 2008; data available at http://www.ncdc.noaa.gov/cmb-faq/anomalies.html#grid); and the UK Hadley 9 Centre (Morice et al., 2012; data available at http://www.metoffice.gov.uk/hadobs/hadcrut4/) reanalyses). Whiskers 10 indicate the 90% uncertainty range of the Morice et al. (2012) dataset from measurement and sampling, bias and 11 coverage (see Appendix for methods). The coloured shading shows the projected range of global annual mean near 12 surface temperature change from 1990 to 2015 for models used in FAR (Scenario D and business-as-usual), SAR 13 (IS92c/1.5 and IS92e/4.5), TAR (full range of TAR Figure 9.13(b) based on the GFDL_R15_a and DOE PCM 14 parameter settings), and AR4 (A1B and A1T). The 90% uncertainty estimate due to observational uncertainty and 15 internal variability based on the HadCRUT4 temperature data for 1951-1980 is depicted by the grey shading. Moreover, 16 the publication years of the assessment reports and the scenario design are shown.”
1) I’m mystified by their baseline claim. They say both “Values are aligned to match the average 6 observed value at 1990.” and “Observed global annual temperature change, relative to 1961–1990,” So are the projections relative to the average projected in 1990 and the temperature relative to the average from 1961-90? That (a) would be a weird choice and (b) doesn’t match the graphic.
2) I have no idea what the grey shading is supposed to mean. Is that spread the full variance including the part from the trend? And the volcanoes? Or what? (I know it reads “The 90% uncertainty estimate due to observational uncertainty and internal variability based on the HadCRUT4 temperature data for 1951-1980 is depicted by the grey shading”.) Basically: Uncertainty is supposed to be uncertainty in something. What is this uncertainty in?
3) Irrespective of (2) — the whiskers seem to be ” the 90% uncertainty range of the Morice et al. (2012) dataset from measurement and sampling, bias and 11 coverage (see Appendix for methods).” So, I’m seeing observational uncertainty double booked? Why would you double book and show the measurement uncertainty in both whishers and then a grey region. If that’s what they did it is both (a)wrong and (b) silly.
As for this specific:
“does it fit with the HadCru4 data as per the October Met Office HadCru4 based “bulletin†[since they refuse to call it a report] ?”.
I don’t know. I guess I’d have to rebaseline to exactly 1 year also. In anycase, I don’t have the FAR, SAR and TAR ginned up. So… dunno.
I guess I can read the appendix. But right now, that seems to be one of the worst least informative graphics ever created!
I also haven’t figured out what the gray region is trying to represent. But beyond that, for the previous ARs’ projections — why include all the scenarios? [E.g. FAR BaU and D] The assumptions of Scenario D did not occur, so the conditionals associated with it are irrelevant. Including that in the colored region gives a vague impression that the FAR BaU projection was not unreasonably high compared to observations.
lucia, haroldw, tetris
If have not really looked at this closely but do recall that the first posting below did say ignore the gray. That said, FYI there is some commenting/discussion at WUWT:
buthttp://wattsupwiththat.com/2012/12/14/the-real-ipcc-ar5-draft-bombshell-plus-a-poll/
http://wattsupwiththat.com/2012/12/18/dr-david-whitehouse-on-the-ar5-figure-1-4/
mwgrant–
WUWT said ignore the grey. But the IPCC doesn’t and tetris asked me my opinion of the graph. It looks like a dumb misleading graphs. Number of choices seem to be an attempt to make projections look less overly-warm than they are. But I can’t be entirely certain because some things are too vague.
Maybe I need to read the appendix. Or something. But it should be possible to get a general understanding of the message in a graph in chapter 1 without hunting down an appendix to discover what it’s supposed to even mean!
HaroldW (Comment #107564),
Actually, the radiative forcing of FAR Scenario D is much closer to widely accepted GHG forcing estimates since 1990 than the BaU scenario. Mainly this was due to:
1) BaU methane increasing more than observed. Partly this was probably due to BaU assuming no transition from coal to natural gas for power production.
2) BaU assuming only limited curbing of Montreal Protocol gases – CFC concentrations continued to increase significantly from 1990 in this scenario.
3) Better understanding of CO2 RF since the FAR reduced the accepted 2xCO2 value by ~15% from 4.4 to 3.7W/m^2.
lucia
Regrettably graphics sell. We all know that. There is a less than subtle difference between making one’s point easy to grasp and and making it appealing. Just as we are quickly favorably biased in some manner to ‘attractive’ individuals, we are biased toward ‘attractive’ graphics. Take it in and right or wrong and move on. (‘Attractive’ is not a good word here–‘high-techy’-ness is part of the picture– but it is the best I could come up with for now.)
One of the downsides to desktop technology is has been the explosive growth in the use of bad graphics. I think it was initially the ease of making any graphics [EXCEL. ugh!!!], but now we can go further and use techy bells and whistles with ease. Looks become an obsession [objective] at the expense of thought and analysis [objectivity]. Still the flaw is in the hands and mind of the user and now the superb tools. Oh well.
Regarding interpretations of the IPCC graph, the choice of 1990 as an absolute start point rather than using a longer average of the period will tend to make the projections look slightly worse than they were given that 1990 was a local maximum and likely above trend. Tricky one though, given the volcanic activity around that time.
lucia
and last paragraph…
Absolutely no argument there. And indeed it is backwards when one has to read ‘volumes’ to for even that initial understanding of the graph(s).
Paul_S–
Whether that choice makes them look worse or better depends on consistency. Are the models relative to model 1990? With earth relative to what?
Some– especially- earliers models don’t have volcanoes. So, the effect might be to pin “the earth” to a high value (1990) and the models to a low value. But did they do this? I don’t think so because neither models nor data are T=0 at 1990. I have no idea what the text under the graph is trying to explain.
Paul_S (#107567)
1) methane — the overestimation is also present in Hansen’s 1988 prediction of future RF. I have some sympathy for this; methane concentration was rising at that time, and I think it was logical to extrapolate from the current trends. Not sure about the explanation that the lowered trend is due to reduced coal usage for power generation; that seems implausible but I don’t have figures with which to argue.
2) CFCs — again, the overestimation is also present in Hansen 1988. The Montreal Protocol had been signed and I think there was no reasonable expectation that CFCs would continue to grow.
3) “Better understanding of CO2 RF since the FAR reduced the accepted 2xCO2 value by ~15% from 4.4 to 3.7W/m^2.” You’re conflating two things here. If you’re saying that FAR overestimated temperature rise in part due to an excessive value of sensitivity, I’d agree. But that doesn’t justify a reduced RF. What it seems to come down to is something like, “if we reduce the predicted RF & reduce the sensitivity, then the predictions can be made consistent with observations.” Well, yes — if you adjust the numerator & adjust the denominator, you can get the quotient you want.
Bottom line, what the FAR predicted is what it predicted. We did not encounter the strong decarbonization hypothecated in Scenario D. Any policy-maker would have used the BaU range to justify mitigation costs.
It is indeed valuable to go back and see *why* FAR (& SAR, TAR, AR4) over-estimated; that’s how we learn. But if the purpose of the chart is to show *what was predicted*, then it needs to show *what was predicted*, and that does not include hypotheticals which did not occur. It may not look as pretty, but pretty should not be the goal here.
Lucia,
The description clearly states that all projections are aligned to be equal at 1990, using the average of the observational data as reference – i.e. they average the HadCRUT4, GISTEMP and NOAA 1990 values relative to 1961-1990 and make that average the starting point for all model runs. The y-axis values aren’t important, I assume they’re just picked for numerical consistency with other temperature anomaly graphs in the report.
Of course, it’s possible 1990 was a local maximum due to deterministic factors – solar maximum, recovery from El Chichon eruption – rather than internal variability so it may not be such a bad choice.
PaulS–
I don’t know that 1990 is good or bad. It happens to not be the reference for the AR4 projections which makes it “wrong”.
I would think consistency with what was actually projected would be the more rational choice.
That’s a little difficult to do because in the early reports they don’t entirely define “relative to” when making projections. But in the AR4, it’s relative to the mean from 1980-1999. To the extent that the earlier reports are less specific, it might have been better to stick with showing all projections “pinned” at the mean temperature for that time period at the center of that time period (which is near 1990. )
With respect to the FAR, SAR, TAR: If– for some mysterious reason- on feels compelled to rebaseline to a specific year 1990 is likely a better choice than the trough due to Pinatubo because the authors do at least in someplaces say relative to that year. I don’t think they were projecting relative to the temperature any other single year. So, I don’t think that choice makes the models look “worse”– since it’s close to what those making projections probably meant to communicate as their projections. (If you have more words from those documents, we can see better.)
Certainly, if the authors chose to rebaseline to the dip after Pinatubo, they would be criticized. But I don’t think you can argue that pinning the pre-AR4 to 1990 makes the models look “worse”– at least not if you mean “worse than what the authors of the article told people to expect”.
With respect to 1990: It appears the choice of 1990 likely doesn’t make the models look better or worse than the true projections.
I need to make a better graph. But here is an early one.
The graph above is lagging averages (and not all the models.). So, the 1990 temperature is shifted to 1991. Rebaselining to “match” 1990, rather than the mean from 1980-1999 means sliding the red graph up or down to match the blue in the appropriate year. Almost no sliding would be required if one picks 1990 as the single year.
I’ll have to gin up a better graphic to double check. But at least– if that graph and it’s description is not to far off, choosing 1990 for a “single year” baseline for the AR4 doesn’t make the models look ‘worse’ if by “worse” we mean “worse than what we would conclude if we tested the projects as described when they were made”. It also doesn’t make them look better.
But…as I said: I need to update with all the models– and make sure that graph really was lagging and so forth.
HaroldW (Comment #107573),
Methane:
Not sure about the explanation that the lowered trend is due to reduced coal usage for power generation; that seems implausible but I don’t have figures with which to argue.
I’m not saying that’s the the reason for the lower trend in reality, but that is one of the differences between BaU and Scenario B, along with improved efficiencies, carbon monoxide controls, deforestation controls, most of which actually occurred to some extent.
CFCs:
The Montreal Protocol had been signed and I think there was no reasonable expectation that CFCs would continue to grow.
That’s hindsight talking. At the time there was much uncertainty, wailing and gnashing about the prospect of adherance to these controls. In any case, continued (albeit slower) growth of CFCs was an assumption of the BaU scenario. Under your terms this assumption did not occur, hence it should not have been included in the graph.
CO2 RF:
You’re conflating two things here. If you’re saying that FAR overestimated temperature rise in part due to an excessive value of sensitivity, I’d agree. But that doesn’t justify a reduced RF.
Not sure what you’re trying to say here. The 2xCO2 RF estimate reduced due to better understanding and application of the radiative physics involved. This has nothing to do with sensitivity, other than interpretation of the stated sensitivity ranges of the upper, lower and best estimates (given that ECS is defined as the response to the RF of 2xCO2).
As I remember, the 15% drop in RF was because the radiative transfer parameterizations in many of the models produced forcings that were too high. That was found in one of the CMIP’s when some forcings were calculated line-by-line and compared to the modeled forcings and has since been corrected. The number of CPU cycles for even a low resolution band model for RT is still too high for use in a GCM.
Paul_S (#107578)
Re: CO2 RF – Yup, my error, apologies. FAR evaluated CO2 forcing as 6.3 W/m2 per factor of e, or 4.4 W/m2/doubling. As you say, that’s been reduced by ~15%. And that is a factor in overestimation of temperatures.
Re: CFC’s
We agree that overestimation of CFCs was a factor in FAR over-estimation of temps.
Going back to the AR5 figure — what is it trying to say? All ARs have considered a variety of scenarios, from BaU to strong mitigation. One can certainly compare observed temperatures to projections based on mitigation scenarios — but what does that mean? The purpose of the graphs, and comparing the scenarios, was to advise policy-makers: if you do nothing, we think X will result; if you take measures to produce Y% reduced emissions, we think Z will result. There are many reasons why X didn’t happen, and understanding them is important to improve future predictions. And one can retrospectively compute, given our current knowledge (e.g. CO2 RF) and measurements since (e.g. CFCs, methane) that the FAR methodology would result in lower projections, possibly even consistent with observations — I haven’t done the exercise. But this is “Texas sharp-shooting.” The key question is “how reliable are current predictions”, not how accurate one can make them in retrospect.
HaroldW,
Making predictions is hard, especially about the future. Explaining why your predictions were wrong is easier. 😉
.
I just wish that instead of arm waves and explanations for past predictive failures, people who believe the IPCC a useful organization would recognize that history has consistently shown past IPCC predictions are not accurate… and the most rational expectation is that the next round of predictions will turn out to be a whole lot less than accurate. Yes, predicting the future is hard.
.
But a fair question is: are the sign and magnitude of errors in past predictions unbiased? These ought to average out to near zero. It is here that the IPCC seems clearly biased. They almost always (not always, but nearly so) predict warming and consequences of warming to be worse than what actually happens. This reduces their credibility… a lot. IMO, just as it should.
@Lucia,
Thx for your answer. I thought it sloppy as well, but even with that the divergence between the data and the models stands out.
I have been out of it for quite a while, but will be returning now – it looks like a good time to do so.
Quick eyeball of this graphic attempting to follow the traces: it appears to me that the only lines ending up “below” the observed temperature are due to models that are also far above the average in the early period; in other words, they have an overall very small trend over the full period. The opposite seems to be true of those models which end up far above observe: they have an overall larger trend. I scanned the comments above and have not seen this considered. Perhaps I am all wet.
It is very hard to follow these traces.
Is there an Excel data sheet (or other data file) of this data so I/we can break this down more clearly?
Regarding projections and the validity or otherwise of “sensitivity” If anyone wishes to ask questions about my paper, or if you believe you have an alternative explanation for the Venus surface temperature, please post your question or response below this post as I wish to keep all discussion on the one thread.
Doug–
I endorse your desire to keep all comments on your paper on one thread. I think it should not be this one. To enforce your preference I decree all discussion of your paper off topic to this thread. (If Roy decides your thread jack of his post violates his policy, I advise you start your own blog and host discussion of your paper there.)
This should come as no surprise to anyone who has compared IPCC model projections with observed temperatures:
http://www.agu.org/pubs/crossref/2012/2012GL053650.shtml
Lucia,
“I’m glad they added the colors so idiots would stop “patiently explaining†that the spread in the model runs from different models is what we expect from “weatherâ€. (The notion that it’s “weather†had been promulgaged at RC and in peer reviewed papers. It’s obvious nonsense– but using “grey†for the spread permits people who want to convince those with too little time to download the graphs that that spread might be “weatherâ€.”
Those “idiots” were mostly correct… internal variability dominates systematic differences between models on short timescales up to at least two decades in projections… a few years ago there was a neat online tool published in bams for visualizing this which you can probably dig up… since you don’t reference any graph it’s hard to tell what you’re referring to, but IPCC AR4 fig. 9.5a shows the output of a total of 58 runs from 14 models, multiple runs per model mean further variance is accounted for by internal variability “weather”… you might want to cut the “idiots” crap. It’s unconvincing when you passively let your commentators claim the AR5 models “look like crap” without any accounting for internal variability which will *correctly* create differences between model runs and observations… let alone those which claim models and observations should have “the same values”…
And the dog-whistling insinuations in #107484 are really disgusting…
“Reading the AR5, my sense is there are at least some authors who really truly want to play it straight.”
“Now… suppose an author who is interested in not hiding stuff comes across the idea”
“it’s hard for someone who wants to obscure to come up with any excuse for why the colors should be tweaked to make the graph more confusing”
“I’m sure that all sorts of people will highlight this… and say that sentence takes up too much room… and it’s unnecessary… or confusing”
“that is a point that can be made actively instead of by “hiding†information about the models.)”
“tap dancing away from anything that comes out and admits that models are hardly precision tools”
“The current one uses the colors which I think I first introduced at this blog. I know I suggested the colors to Ed Hawkins and stated why it’s useful. And I know he thought it was a good idea.”
Using colors for differentiating models can be found in almost every graph in AR4 chapter 8 titled “Climate Models and their Evaluation”.
“Can that same author who went to the trouble to get this good graph get text in that says the difference between in the mean temperature of the earth in the hottest and coldest model is 2 or 3Câ€? Or to put that 2-3C next to the size of the projected increase over the century?”
You mean like they did in AR4 chapter 10, figure 10.5?
Lazar
I don’t know what lucia is talking about. IPCC temperature projections have always been exceptionally accurate. Always.
Brandon–
So… for the AR4 they adjust the observations for volcanic eruptions but don’t adjust the projections– which include volcanic eruption? Strange choice. And… uhmm.. wouldn’t it be nice if they showed the projections using the baseline actually chosen by those making the projections?
Sorry… but the AR4 didn’t make projections pinned to 2000. They make projections relative to the mean from 1980-1990. I don’t know if the baseline change Tamino and Stephan used to make the comparion makes a difference. But when you are comparing absolute values, the changing the baseline changes the level of agreement. It very odd to make a baseline change when trying to decide if the projections were accurate! (Ok… well… actually, I know if their baseline change makes a difference. We all know that subtracting out the error makes the projections look not so bad.)
Lucia,
“If you examine the graph you can see that not withstanding the fact taht internal variability is large, the observations are skirting outside the range consistent with models. […] What you can see is that the observations are inconsistent with the warmer running models and the model mean”
And how do you “see” all that?
“Dog whistling? Insinuations? I think my statements are said in frequencies audible to human ears. Moreover, the fact that you object to them doesn’t make them “disgustingâ€. They are what I think– and I’m saying it rather directly.”
Evidence-less insinuations against the integrity of others are viewed as disgusting according to universally shared ethics across most of the friggin’ globe. And when those insinuations are encapsulated in statements like “now… suppose an author who is interested in not hiding stuff”, that’s called dog-whistling… not honest straight-talking like you pretend it is.
“If you think what I am saying is incorrect feel free to say why you think I’m mistaken.”
Factually? Please ask someone else to search for your pony / prove your negative.
“using your eyeballs, it’s pretty easy to see that the realization is skirting the lower edge of the range, the spread is not dominated by “internal variability—
And how exactly do you “see” what proportion of the variance is internal variability to determine that the spread is “not dominated”?
“at least some of the models”
How many is “some”? How do you and SteveF generalize from “some” to “the models look like crap”, “the models don’t look so hot”?
“Everyone here knows that the individual realization from the earth isn’t going to match the model mean– and also knows that the individual realization isn’t going to match a model”
Then you might advise your first commentator that his standard for “success” is unreasonable… “the mark of success of a model is when it has the same values as the real data”… which ordinarily would be the mark that the results have been forged by an unusually dense undergrad, for practically any model across any field of study. At least then your calling others idiots on this issue would look a tad more authoritative and less like the usual self-serving PR.
And I recall it took a lot of headbanging with you and others over at CA for this standard knowledge to sink in ref. Douglass Christie Pearson Singer, what TCO called a ‘Box-Hunter-Hunter fixable f*up of an error’.
“Uhmm.. no. The information I am discussing is the absolute temperatures which is not in AR4 Figure 10.5.”
I get it, the rascally scientists despite showing the metric which has the most relevance regarding uncertainty in projections for a given emissions scenario, the systematic differences in model sensitivities, they were clearly rascally dishonest scientists for not including the absolute temperatures of simulated climates, the relevance of which for projections and justification for which you have not provided… and may be approximately nil and nil judging from AR5 Fig. 9.43 “Equilibrium climate sensitivity (ECS) against the global mean surface air temperature of CMIP5”. Are there any other arbitrary unjustified metrics you can think of to similarly imply scientists are dishonest? The following are already in AR4 Ch. 8…
Outgoing SW radiation
Outgoing LW radiation
Zonal oceanic heat transport
Zonal oceanic wind stress
Sea surface temperatures
Climatological patterns of precipitation
“” mean sea level pressure
“” surface air temperature
ENSO power spectra
Climate sensitivity
Transient climate response
Feedbacks
Cryosphere seasonal cycles
Is the process by which you measure dishonesty similarly scientific as to how you eyeball variance components?
So when compared against polynomial extrapolations made by people who aren’t climate modelers, the IPCC projections work “exceptionally well.”
Got it.
Lazar
By keeping my eyes open as I focus them on the graph.
First: There is nothing “insinuating”. An insinuating means “an indirect or covert suggestion or hint”. There is nothing indirect or covert about what I said. I stated it flat out.
Second: “Reading the AR5, my sense is there are at least some authors who really truly want to play it straight.”
I am relating what “my sense” is. That is my sense. The “evidence” that is my sense is that I (that would be me) perceive that I sense it. No other evidence is required to support my claim that I sense this. (You are being pretty absurd here. Really.)
Your question about “how [we] generalize & etc” suggests you are utterly confused about thought and communication in general. SteveF is expressing his diagnosis of how the models look based on examining the graph. I am expressing mine. My impression is not a conclusion based on his opinion. We don’t have to explain how “some” relates to anything.
If you have a different opinion of the models and you may think that graphs shows the models are splendiferous. And if one uses the “method of the eyeball” on the appearance of that graph, the models look pretty bad. That’s my impression.
Well… you might want to quote the full comment
Had he not followed the bit you quote with discussion how how the observation compares to all runs I might say his standard is unrealistic. But given what he actually wrote, it’s clear that he understands that the question is whether or not the observations falls in the spread. I don’t think he is that off the mark.
Perhaps you should try to avoid quote-mining to find sentences which, taken out of context, might suggest someone doesn’t understand the correct standard for testing models.
Possibly if you avoid this sort of quote mining, you might look a bit more authoritative rather than appearing to be a shill who is spooning out PR.
I have no idea what you are talking about. Possibly you can supplement your possibly faulty recollection with some quotes. That would help us understand what the heck you are talking about and what misconception you think I (or others) had.
FWIW: I don’t think I am responsible for whatever faulty statements anyone other than I might have posted at Climate Audit!
Uhh… First: I don’t know why you think that showing what you consider to be the metric with the “most relevance” means that anyone is allowed bury well known information about other relevant metrics. Even if you don’t think failing to show the inability of models to agree on the earth’s absolute temperature is important– I do.
Second: I didn’t say they were “rascally” for failing to include absolute temperatures. You quoted my paragraph discussing the figure in the draft AR5 that shows the absolute temperatures. Then you suggested that the absolute temperature appeared in Figure 10.4 in the AR4. I pointed out that you were wrong on this point.
My pointing out that your claim that absolute temperatures appear in figure 10.5 the AR4 is totally wrong is not any sort of insinuation that the scientists who created that graph are “rascally”. It is simply observing that your claim that information appears in that graph is wrong.
As for the decision to not bury that data in the appendices in the AR4:
I realize it may not bother you that the information indicating that the models estimate of absolute temperatures differ dramatically was buried in the AR4. But I don’t see why I’m not allowed to observe this fact and note that I am glad it is not buried in the AR5. My prefering this revealed front and center is not an accusation of “rascalliness”. I simply think this is important information revealing uncertainty in the models. I applaud whoever thought of a simple graphical way to include it in the AR5– and I’m glad it’s there. Unless some decision is made to re-bury it, the information will not be presented to decision makers in a way that makes this uncertainty obvious.
I have no idea why you are asking about the list of other information. That other more complex information (e.g. power spectra) that is difficult for lay people to interpret is included in more prominent parts of the AR4 hardly means that I can’t criticize their choice to hide obvious, easy-for-lay-people to understand shortcomings of models in places like appendices!
Lazar just wondering a) what your native language is because it obviously isn’t English, b) why you insist on using overly affected terms like “dog whistling”, “search for your pony” and “rascally” in a technical discussion (Lucia’s tongue-in-cheek comment flew over your head on this matter I see) and c) what you actually meant by this: “Is the process by which you measure dishonesty similarly scientific as to how you eyeball variance components?” Is that English?
I’m approaching “pretty certain” I can safely ignore anything you have to say and in the process I won’t have wasted any time and I won’t have lost an opportunity to learn anything new.
Cheers.
Lazar, I am unbathed in the intellectual brilliance that you are obviously subjected to and so could you help me on a minor point.
This ‘weather’ thing is a little strange to me. As I understand it the global reconstructed temperature is calculated by taking the daily individual maximum and minimum temperatures and dividing by two. Then these are averaged the world over, making allowances for changes in altitude.
Now according to the modellers moving lard chunks of atmospheric energy around, will not change the overall temperature of the whole system. Hence the modellers love of ‘annual’ and ‘global’.
This weather variability you now regurgitate like a dog after eating a skunk has not be defined. So, Lazar, could you give the me and the rest of the class a explicit written and mathematical description of weather variability?
Carrick–
I’ve never even heard the idiom ““search for your pony†and he used “dog whistling” incorrectly. (That said– it appears the problem is that s/he may not know the definition of the word “insinuation”.)
As for your question about what Lazar means by this:
I also have no idea what he means to ask. But I suspect this was intended as a rhetorical question and that he thinks posing it communicates something. What he intends to communicate I don’t know. It is my experience that people who try to make ‘points’ using rhetorical questions are often just trying to avoid stating their point directly because if the stated their claim or point directly, everyone would immediately see the point was either a) wrong or b) stupid — and often it is both.
If Lazar wants to know how I look at variances he can go back to previous posts. I don’t think he wishes to do that. He seems to prefer by arguing that he “recalls” that I (and others) might have said something he thinks was wrong somewhere else. What I might have said, where I said it, what it meant and even if it was wrong is left entirely up to the reader to imagine. I might suggest he is hoping we would figure it out by reading his mind– but honestly, I think that would be impossible because even he doesn’t know precisely what he thinks I (or others) said, where we said it, how we said it. He merely has a vague recollection that it might have been wrong.
(The fact that here in comments Lazar gives us proof that he can’t even interpret the information content in Doc’s comment in the first comment right here on this thread means that most readers are going to suspect his recollection that I — or others– said something wrong at Climate Audit might not be entirely accurate. He may be misinterpreting what was said and so his recollection might be wrong.
Mind you– I’m sure I have said things that are wrong form time to time. I may even had done so at climate audit. Moreover, I’m sure “others” have said things “wrong”. But Lazar sure is a lazy thinker, arguer etc. to use as “evidence” of his position an claim that somewhere I said something wrong at Climate Audit!)
Lucia (#107850)
To quibble slightly, I don’t think that the inter-model variation in absolute temperature is going to be presented to policy-makers. While it is shown in AR5 WG1 SOD Figure 9.8(inset), it doesn’t make the brief list of GCM limitations in the chapter’s Executive Summary, nor does it appear in the SPM in any form.
.
It would be interesting to see the temperature equivalent of this panel of [AR5 WG1 SOD] Figure 9.4: “Bottom left: Mean absolute model error with respect to observations”. Is anyone aware of something like this?
.
[And I just noticed that the caption for the bottom right panel of Figure 9.4 is “Bottom right: multi-model mean error relative to the multi model mean precipitation itself.” Why on earth would one figure the error relative to the *model* values, rather than the actual (observed) ones?]
Harold– It might not be in the summary for policy makers. Nevertheless, the information is in that figure. It isn’t mentioned in the discussion of the figure. Still… I think it’s an advance.
Do they just mean the variance about the multi-model mean? Maybe “error” is the wrong word. There are difficulties with language choice since differences between a model mean and the observations isn’t necessarily error. The model-mean is closer to the “expected value” of the observations. The observation is an individual realization. By nature of the system, we don’t anticipate these need be identical and a non-zero difference doesn’t correspond to “error” in the strictest sense. (It would be error if someone actually predicted the mean would happen. But that’s not what anyone predicts.)
Lucia:
Neither term is included in my Handbook of Mathematics and Physics, hence it is possible that he used “dog whistling” correctly based on how he learnt the idiomatic phrase.
I suppose this illustrates the dangers in using idiomatic expressions in general, when technical terms exist.
(I’m sure Lazer is trying to reduce the number of provable errors he/she makes by obfuscating his/her arguments with idiomatic expressions. Nonetheless it’s clear that numerous substantive errors still exist in his/her posts.)
Carrick–
Here’s what wikipedia has under “dog whistle politics”
http://en.wikipedia.org/wiki/Dog-whistle_politics
But Lazar seems to be complaining about statements I made that would mean the exact same thing to all groups. There is no coded secret meaning that would have a different resonance to any targetted subgroup. This is actually proven by the fact that he can pull out quotes whose literal meaning is precisely what he and everyone else would take them to mean. He may not like what I said, but there is no “dog whistle” there.
Well… yes. His/her attempt to avoid admitting that he was simply wrong when he suggested that the AR4 figure includes absolute temperatures by a lame attempt to change the conversation? And accusing me of making accusations I didn’t make? And then– after accusing me of things I did not do and subsequently challenging me to defend myself for doing those things (which I did not do)? Just lame. Totally lame.
Lucia (#107856): “Do they just mean the variance about the multi-model mean? Maybe “error†is the wrong word.”
While you are correct that the observation represents one realization of a stochastic process, the comparison is being made over a 25-year interval (1980-2005). So one would like to see the models come fairly close to the observed average precipitation over that interval, much as one would like to see each model correctly replicate the average surface temperature.
.
My interpretation of the words for Fig. 9.4 bottom right panel, by the way, is that they are computing (MM-O)/MM as a percentage, where MM is the model-mean annual average precipitation (at a given lat/lon) and O is the observed annual average precipitation (also a function of lat/long). [Subscripts omitted for clarity.] Top right panel is clearly (MM-O). It just seemed much more natural to me to compute (MM-O)/O if one wants to see a relative mis-estimation. I also think that the chart would be improved by a symmetric color-code — e.g. white representing -10%..+10%, &c.
Somehow I get the impression that the pony searched for is the one buried in the pile of horse-sh*t. On the off chance that any of you is unfamiliar with the joke, I can try to reconstruct it.
Turns out to have been one of Ronald Reagan’s favorites.
English may be Lazar’s native language. I had no problem understanding his sentence. It is a complex snarky rhetorical question to illuminate the supposed lack of ethical/scientific behavior. The problem is that he left out commas. The thought appears to be:
Is the process, by which you measure dishonesty, similarly scientific, as to how you eyeball variance components?
with “as to how” meaning “the same way.”
John F. Pittman–
The rhetorical question was also loaded because it pre-supposes I was accusing people of dishonesty. It’s true that I think earlier graphics were designed to reveal what whoever made them wished to reveal and obscure what they did not wish to highlight. Similarly, in the AR4, those graphic that permit someone to perceive the disagreement in absolute temperatures were buried in the little read appendix — and shown in an obscure way that did not permit readers to quantify the discrepancies. (WEll… unless they waited 3 years for the data, downloaded it and processed it themselves.)
But I didn’t suggest this was dishonest. Communicators always emphasize what they think is “important” and don’t communicate that which they think is either (a) unimportant or (b) confuses people and might prevent them from understanding that which is “important”.
I think those who wish to obscure information that clarifies just how uncertainty model are poor judges of what is important. But I didn’t accuse them of dishonesty.
I also think that whoever came up with the better graphics that communicate the deficiencies in models in a compact visual did well. And I think with this new graphic it will be difficult for those who think letting people easily detect the huge spread in results from different models so they can see the differences are not ‘weather’ but rather disagreements between models is a good thing. So, the new graph will prevent people from becomeing “confused” and believing the AOGCMs have high levels of accuracy or robustness.
I did not have problems with your statements either.
I remember when you posted about the absolute values and weather including the bizzarro world analogy. I agreed with you then and now, the actual temperatures should have been displayed prominently. I also think the lack of explanation in AR4 wrt TOA for the different temperatures unjustifiable, IMO. I always chuckle when I see AR4 model anomalies graphs and think of your statement of whichever planet some of them are, it is not planet Earth.
Of course, the other statement that brings me laughter is when someone is explaining how the models (plural) are physical, and I start thinking of bizzarro world, and some of the others.
lucia:
There are so many questionable (or worse) things about the link I posted and the F&R paper it links to.
lucia:
Speaking of odd, I really ought to look through the code linked to in this post.
I decided to do a quick run of that code before heading out for the night. The first thing I looked at was the results of Tamino’s lag calculation. I figure if he’s updating his data by only a year or so, we should expect the lags to be fairly consistent. In other words, they should be close to the earlier (MEI,Volcanic,Solar):
They turned out to be:
I have no idea why CRU would suddenly have a lag for volcanic forcing half that of the others so I’ll have to try to look into that tomorrow. However, the main thing I find interesting is every lag for solar forcing has increased. Tamino switched from TSI to sunspot counts for this version of his code, and it would seem that makes a significant difference.
Brandon–
I only scanned the paper. But didn’t it claim they used the exact same fit they used before? Maybe they mean the same method– but they get to change the parameters each time? Or I misremember?
Certainly changing from TSI to sunspots or vice-versa would not even constitute the same method though.
lucia:
The paper says:
Since the method they used calculated the lags and forcings, their 2012 paper won’t use the same values.
The post I linked to was written after the paper we’re talking about. I assume the paper used TSI and the changes in the code in that blog post haven’t been implemented (in a paper) yet. That’s completely consistent with what Tamino has said, and it makes the most sense.