Commentary from Nature Climate Change, by John C. Fyfe, Nathan P. Gillett, & Francis W. Zwiers
Recent observed global warming is significantly less than that simulated by climate models. This difference might be explained by some combination of errors in external forcing, model response and internal climate variability.
Global mean surface temperature over the past 20 years (1993–2012) rose at a rate of 0.14 ± 0.06 °C per decade (95% confidence interval)1. This rate of warming is significantly slower than that simulated by the climate models participating in Phase 5 of the Coupled Model Intercomparison Project (CMIP5). To illustrate this, we considered trends in global mean surface temperature computed from 117 simulations of the climate by 37 CMIP5
models (see Supplementary Information).
These models generally simulate natural variability — including that associated
with the El Niño–Southern Oscillation and explosive volcanic eruptions — as
well as estimate the combined response of climate to changes in greenhouse gas
concentrations, aerosol abundance (of sulphate, black carbon and organic carbon,
for example), ozone concentrations (tropospheric and stratospheric), land
use (for example, deforestation) and solar variability. By averaging simulated
temperatures only at locations where corresponding observations exist, we find
an average simulated rise in global mean surface temperature of 0.30 ± 0.02 °C
per decade (using 95% confidence intervals on the model average). The
observed rate of warming given above is less than half of this simulated rate, and
only a few simulations provide warming trends within the range of observational
uncertainty (Fig. 1a).
Figure 1 | Trends in global mean surface temperature. a, 1993–2012. b, 1998–2012. Histograms of observed trends (red hatching) are from 100 reconstructions of the HadCRUT4 dataset1. Histograms of model trends (grey bars) are based on 117 simulations of the models, and black curves are smoothed versions of the model trends. The ranges of observed trends reflect observational uncertainty, whereas the ranges of model trends reflect forcing uncertainty, as well as differences in individual model responses to external forcings and uncertainty arising from internal climate variability.
The inconsistency between observed and simulated global warming is even more
striking for temperature trends computed over the past fifteen years (1998–2012).
For this period, the observed trend of 0.05 ± 0.08 °C per decade is more than four
times smaller than the average simulated trend of 0.21 ± 0.03 °C per decade (Fig. 1b).
It is worth noting that the observed trend over this period — not significantly
different from zero — suggests a temporary ‘hiatus’ in global warming. The
divergence between observed and CMIP5-simulated global warming begins in the
early 1990s, as can be seen when comparing observed and simulated running trends
from 1970–2012 (Fig. 2a and 2b for 20-year and 15-year running trends, respectively).
The evidence, therefore, indicates that the current generation of climate models
(when run as a group, with the CMIP5 prescribed forcings) do not reproduce
the observed global warming over the past 20 years, or the slowdown in global
warming over the past fifteen years.
This interpretation is supported by statistical tests of the null hypothesis that the
observed and model mean trends are equal, assuming that either: (1) the models are
exchangeable with each other (that is, the ‘truth plus error’ view); or (2) the models
are exchangeable with each other and with the observations (see Supplementary
Information).
Brief: http://www.pacificclimate.org/sites/default/files/publications/pcic_science_brief_FGZ.pdf
Paper at NCC: http://www.nature.com/nclimate/journal/v3/n9/full/nclimate1972.html?WT.ec_id=NCLIMATE-201309
- Supplementary Information (241 KB) CMIP5 Models
Sorry, I am going to sleep now.
Perhaps WUWT is spending too much time on the the articles written by dr’s
e.g
http://wattsupwiththat.com/2013/09/07/new-paper-says-no-evidence-of-planetary-influence-on-solar-activity/
whilst it should be giving more time and attention to hands-on scientists
I am stunned to find that besides William Arnold, apparently I am currently the only one who found the link between the planets and the warming and cooling periods.
Al, Rich – The ‘measured’ anomalies is actually an attempt at a least-biased trace. I plotted the anomalies for HadCRUT3, HadCRUT4, NOAA and GISS on the same graph and they are very similar. These all use essentially the same raw temperature measurement database. Each group processes the data slightly differently from the others. Each believes their method is most accurate. To avoid bias, each anomaly trajectory is shifted (reference-temperature change only) so its average is the same as the average for HadCRUT4 over the time period for which both are given and then the average from the available values (as-shifted if not HadCRUT4) for each year is calculated. This normalizes the set to a single trajectory which is shown in Figure 1.
The equation works for any one or combination. The coefficients would be very nearly the same but unique for the particular combination. The projection (prediction?) would be for the same combination as used to determine the coefficients. Thus to get a best estimate prediction, determine the coefficients for maximum R2 for the best estimate history.
The first EXCEL file on this was created 25 April, 2013 so I probably used data through February (Hadley is usually over a month late).
Rich – Apparently you still do not see what was done. Application of the energy equation is described more completely in the ‘corroboration’ link in my Sept 6, 1:32 pm post. Although the equation there is a predecessor to the current one, the concept is the same and may help with the equation in the climatechange90 link.
The hockeyschtick link shows a graph that goes back to 1610. I have used the sunspot numbers back to 1610 with the equation (calibrated 1895-2012) and got a very similar graph shape (different proxy factors). You should know that it is a vanishingly small probability that a curve-fit equation to data from 1895-2012 would also fit data back to 1610. But a valid physics based equation should make a fairly good fit.
Dan Pangburn:
In your post at September 7, 2013 at 4:38 pm you say to me
I do. As I have repeatedly attempted to explain to you, it is a curve fit.
Please read my explanations of why that is not informative.
As I said, the fit may be a correct model but there is no reason to think it is correct unless and until it demonstrates forecast skill.
Richard
Henry@Dan
some of the problems you have are intrinsic to the procedure and collection of measurement results, which you must consider
e.g.
I found that SSN from 1895-1927 cannot be relied upon.
Another point is that thermometers were not re-calibrated before 1940
(at least I have not seen calibration certificates from before that time.)
Because of this, I reckon an error of about 0,2 or 0,3 in the anomaly in the historic record is easily possible (which means that your line could go more straight over a longer period)
I could not find any evidence that earth’s temp.s are influenced by more CO2.
you can throw that factor out, as far as I am concerned..
Means temps. are also influenced a lot by earth’s conditions, (volcanic, iron core rotation, etc)
That is why I decided to concentrate on maxima, and I looked only at data from 1974-2012.
In your case, if you looked only at data from 1950-2000, how would the projection be if you go 20 years backwards and 13 years forward?
@richard, dr Brown
clearly, you still seem to think that I was sitting here picking weather stations that in the end would give me a perfect fit. What a silly thing for me to do. How stupid must one be to spend your hobby time fooling yourself that way……
I am saying that you should be able to reproduce my results, if you take another sample of 50 weather stations employing same sampling technique as I did.
Now, what a nice project for a first year statistics class ….
Rich – referring to your explanations in your Sept 6, 5:19 am post.
“This matches the data because the ‘effects’ are tuned to obtain a fit with the anomaly.” Well, close, R2=0.9. But the tuning is only on scale factors. The ‘shape’ of the ‘effects’ is determined from measurements (sunspot number), or the SB function or the saw tooth shape of net ocean oscillations (decided by me because it is unbiased and easy to program), or the logarithmic decline of effect of added increments of CO2.
“Hence, the model demonstrates that those ‘effects’ can be made to match the anomaly, but it does not demonstrate there are not other variables which may be similarly tuned to obtain a match with the anomaly.” I understand. But the ‘other variables’ would have to take away from the other factors already considered at the same fraction that the ‘other variables’ used to keep the same R2. This is how considering CO2 could account for 19.8 % of the 1909-2005 increase without significantly increasing R2.
“The model matches the form of the anomaly.” I wouldn’t call it a model. It is just an equation. And it only matches the form after getting rationally determined coefficients.
“But, importantly, it only explains the opinion of its constructor:” Well, I decided what factors to include in the equation; ocean oscillation, sunspot numbers, SB thermal radiation, CO2 effect, offset.
“it does NOT explain anything about climate behaviour.” It does not address climate at all but it calculates average global temperature with R2 = 0.9.
“Therefore, it does not have a residual of “10%” of climate behaviour which is unexplained.” The number is obviously the difference between 100% and 90%. Seems like I read some place that R2=0.9 means that 90% of the data is explained. What would you call the unexplained 10%? Or would you just not talk about it.
It already has demonstrated forecast skill as described at my Sept 6, 1:32 pm post.
It has also demonstrated back-cast skill as described at my Sept 7, 4:38 pm post.
HenryP:
In your post at September 7, 2013 at 11:00 pm you assert
NO!
We cannot attempt to reproduce your results because you are refusing to say the sampling technique you used.
And, at present, your claims of having selected data which you say is a “random sample” (which is it, selected or random?) induces me to understand that your work is invalidated by your undisclosed sampling technique.
Richard
Dan Pangburn:
Your post at September 7, 2013 at 11:14 pm demonstrates some conceptual problems which explain why you are failing to understand the problem with your model. For example; this
The equation is a model. And if you don’t think it is a model then you don’t think it describes anything.
And you are assuming that the “rationally determined coefficients” are the only applicable “coefficients” but you earlier stated that they are not. You said they are adjusted to provide a good fit. In your post at
http://wattsupwiththat.com/2013/09/05/statistical-proof-of-the-pause-overestimated-global-warming-over-the-past-20-years/#comment-1411064
http://wattsupwiththat.com/2013/09/05/statistical-proof-of-the-pause-overestimated-global-warming-over-the-past-20-years/#comment-1411064
you answer my point concerning the differences between different global temperature data sets.
That answer from you says
YES! You adjust the coefficients to obtain a fit.
As I have repeatedly told you, that is curve fitting.
Your equation has 4 coefficients and an offset which can each be adjusted individually, and you say they ARE each adjusted “for maximum R2 for the best estimate history”.
With that many possible ways to adjust it, the equation could be tuned to agree with almost anything.
And you say of your model
Sorry, but that is NOT true.
Your post at September 6, 2013 at 1:32 pm says
That is NOT a “prediction”. It is a statement that you fitted the curve to the data, but that is not in dispute.
And in that post you also say
That is not a prediction of the future because nobody can accurately predict sunspot number.
So, if I accept your assertions which I quote here, then I am forced to accept that your model has no predictive skill and no use. But, of course, that is true of all curve fitting exercises including yours, and that of Henry P.
Richard
richardscourtney says
that your work is invalidated by your undisclosed sampling technique.
henry says
funny that you keep saying this, as I have most certainly clearly stated this….
but I will copy it and paste it here for you again
1
I took a random sample of weather stations that had daily data
In this respect random means any place on earth, with a weather station with complete or almost complete daily data, subject to the given sampling procedure decided upon and given in 2) below.
2)
I made sure the sample was globally representative (most data sets aren’t!!!) ……
that means
a) balanced by latitude (longitude does not matter, as in the end we are looking at average yearly temps. which includes the effect of seasonal shifts)
b) balanced 70/30 in or at sea/ inland
c) all continents included (unfortunately I could not get reliable daily data going back 38 years from Antarctica,so there always is this question mark about that, knowing that you never can get a “perfect” sample)
d) I made a special provision for months with missing data (not to put in a long term average, as usual in stats but to rather take the average of that particular month’s preceding year and year after)
e) I did not look only at means (average daily temp.) like all the other data sets, but also at maxima and minima… …
3) I determined at all stations the average change in temp. per annum from the average temperature recorded, over the period indicated (least square fits)
4) the end results on the bottom of the first table (on maximum temperatures),
http://blogs.24.com/henryp/2013/02/21/henrys-pool-tables-on-global-warmingcooling/
clearly showed a drop in the speed of warming that started around 38 years ago, and continued to drop every other period I looked//…
5) I did a linear fit, on those 4 results for the drop in the speed of global maximum temps,
ended up with y=0.0018x -0.0314, with r2=0.96
At that stage I was sure to know that I had hooked a fish:
I was at least 95% sure (max) temperatures were falling. I had wanted to take at least 50 samples but decided this would not be necessary which such high correlation.
6) On same maxima data, a polynomial fit, of 2nd order, i.e. parabolic, gave me
y= -0.000049×2 + 0.004267x – 0.056745
r2=0.995
That is very high, showing a natural relationship, like the trajectory of somebody throwing a ball…
7) projection on the above parabolic fit backward, ( 5 years) showed a curve:
happening around 40 years ago. Dr. Brown is right in saying that you have to be careful with forward and backward projection, but you can do this with such high correlation (0.995)
8) ergo: the final curve must be a sine wave fit, with another curve happening, somewhere on the bottom…
http://blogs.24.com/henryp/2012/10/02/best-sine-wave-fit-for-the-drop-in-global-maximum-temperatures/
Now, I simply cannot be clearer about this. The only bias might have been that I selected stations with complete or near complete daily data. But even that in itself would not affect randomness in my understanding of probability theory.
Either way, you could also compare my results (in the means table) with that of Dr. Spencers, or even that reported here, in this post, and you will find same 0.14 /decade since 1990 or 0.13/decade since 1980.
In addition, you can put the speed of temperature change in means and minima in binomials with more than 0.95 correlation. So, I do not have just 4 data for a curve fit, I have 3 data sets with 4 data each.They each confirm that it is cooling. And my final proposed fit for the drop in maximum temps. shows it will not stop cooling until 2039.
In my case I don’t need your’s or anyone’s approval, I merely wanted WUWT to turn (y) our eyes to the planets. Obviously I was not successful.
Henry P:
Approval has no place in science. Falsification does.
Until you state your sample procedure then there is nothing to falsify. Your work is NOT science.
Repeatedly saying you have explained a procedure which you have not explained does nor ‘cut it’.
Nobody can repeat your work because nobody can repeat your sampling procedure BECAUSE YOU HAVE NOT SAID WHAT YOU DID. So your work cannot be evaluated.
By refusing to explain your work YOU are saying that your work is useless and worthless.
I have tried to help you out of this hole of your own making and you have refused to be helped.
Richard
quote from this post
Global mean surface temperature over the past 20 years (1993–2012) rose at a rate of 0.14 ± 0.06 °C per decade (95% confidence interval)1.
quote from henry
you could also compare my results (in the means table) with that of Dr. Spencers’, or even that reported here, in this post, and you will find that I report same 0.14 /decade since 1990 or 0.13/decade since 1980.
(read: I have the results to back me up)
Quote from dr. Brown
What you are doing is also known to be numerically unstable to extrapolation.
(read: I have no results that can confirm this)
Quote from richardscourtney
your work is useless and worthless.
(read: I have no results that can confirm this)
@richardscourtney
So, in your opinion, from these 4 quotes, who is the person that is most likely to be correct about making a prediction on the future of temps. on earth?
go figure
do your maths
it is a simple multiple choice question
the answer is …..
HenryP, I believe what Richardscourtney is trying to say is this – Science (from Latin scientia, meaning “knowledge”[1]) is a systematic enterprise that builds and organizes knowledge in the form of testable explanations and predictions about the universe. Please correct me if I’m wrong but is it testable? Can Richardscourtney do the same experiment given the information you provided? I think he is saying NO he can’t.
Henry you used a psuedo-random sample. Therefore you need to describe it in detail. How many? Per latitude? And how do you know that longitude does not make a difference? How did you quantify that? An easier way would be to just list your stations and let other researchers determine any artifact you may have overlooked.
BBould and Pamela Gray:
Thankyou to each and both of you.
If Henry P could recognise that I am trying to help him then I think he may be more willing to understand the problem.
Richard
BBould says
Can Richardscourtney do the same experiment given the information you provided? I think he is saying NO he can’t.
henry says
I am saying Yes he can and I have given him all the information. And if he (or Dr. Brown) had a whole class of students working on it, he could know for sure in a week.
Perhaps I am expecting that people who react on this post about “statistical proof” know something about stats, i.e. that they studied probability theory, distribution theory, sampling techniques, least square fitting, etc.
It looks to me that there is really no such qualified person here.
HenryP: No, what you are expecting is for people to take your word at face value rather than have it tested. That is what I see as do apparently many others.
richardscourtney:
At September 8, 2013 at 6:47 am I wrote to you saying
At September 8, 2013 at 7:28 am you have written
NO! That is NOT a “Quote from richardscourtney”.
It is a misrepresentation of my words and a nasty response to my attempt to assist you.
Richard
Ouch! Obviously, my last post was intended to be addressed to Henry P and not myself.
Sorry.
Richard
Richarscourtney and Dr. Brown: There is an extremely interesting discussion over at Dr. Spencers Blog, in the comments section of “Revisiting Wood’s 1909 Greenhouse Box Experiment, Part II: First Results” By a poster named Konrad. He puts forth that GHG actually tends to cool the planet and he is doing a good job with his argument something I’m sure you both would be interested in. I don’t have any conclusions on this I’m merely an uneducated observer.
One of Konrad’s quotes: “I can show through empirical experiment that without radiative gases our atmosphere would super heat and most of it would boil off into space. You should acknowledge that the AGW hypothesis depended on the misapplication of SB equations to a planet with liquid oceans covering most of the surface and a gaseous atmosphere in a gravity field.”
And he does have the experiments to back this claim up which makes it even more interesting to me anyway.
Pamela Gray says
And how do you know that longitude does not make a difference? How did you quantify that?
Henry says
the argument here is that earth turns every 24 hours and we are looking at average yearly temperatures, or rather the change in annual temperatures, so longitude does not matter. This is not a matter of stats but of physics, i.e. understanding that the amount of exposure time from the sun is constant so we measure only the difference in exposure actually coming from the sun.
BBould:
re your post at September 8, 2013 at 8:07 am
Thankyou for that. Can you give a link, please.
I would like to read it but will probably not get involved. Roy is good so never has need of my help.
Richard
Yes he is but he isn’t taken part in it. The comments are toward the end and start With Konrad.
http://www.drroyspencer.com/2013/08/revisiting-woods-1909-greenhouse-box-experiment-part-ii-first-results/#comments
Bbould:
Thankyou for that. I need to give it more thought before concluding anything, but it is interesting.
Richard
8) ergo: the final curve must be a sine wave fit, with another curve happening, somewhere on the bottom…
Just because your curve looks likes the flight of a ball, does not mean the thing you’re modelling is a ball in flight – it may be a butterfly.
@BBould re sunshine heating water. (Note : I’m no scientist.)
I’d point out that hairdryers don’t produce much (visible) light, only IR.
You can see the bottom of your swimming pool because it reflects visible light. If you layed roofing slates on the bottom, it’d reflect less and warm more.
In lakes, you often can’t see the bottom because the water is muddy, or full of life, and whatever it is that blocks the transmission of light will have some opportunity for converting it to heat.
I remain impressedby how little heat is transferred to the water from the hairdryer.
I have had a similar experience when trying to heat water using a heat gun. Almost no transfer of heat into the water and the gun operates at 450degsC. If you persevere for a while a small amount of heat will transfer but I think that the fan forcing simulates weight and fools the surface tension into allowing in a very little heat. I’m convinced that surface tension blocks heat.
Sleepalot:
I your post at September 8, 2013 at 8:27 am you provide a clear and succinct explanation of the reason why curve fitting is often misleading when you write
I write to ask if I may use your sentence in future, please?
Richard