Based on discussions at Twitter, my last post may have been too subtle. This graph is less so:
44 thoughts on “Hide The Decline: Trenberth’s Trick”
Comments are closed.
Based on discussions at Twitter, my last post may have been too subtle. This graph is less so:
Comments are closed.
By his logic plateaus are to be expected, ergo the rise in the 80-90 timeframe is a continuation of the 1910 to 1940 rise with a plateau in between. You know, the one that occurred before CO2 had a major effect. When they can definitively explain “why is this rise different from all other rises” I might have a little more respect for their position.
Of course maybe the data is just coming back to the trend that the whole data set shows? (~.6áµ’/Cent)
BarryW–
I think temperatures will resume rising. But Trenberth’s blog post is supposedly a counter argument to people who either think
a) the rate of rise has slowed recently or
b) the rate of rise is not as high as projected by models
And their argument is based on recent trends
But to make his counter argument that we should expect the recently observed 10yr trends as low as they have been (i.e. -0.05C/dec) even if the long term trend is 0.16C/dec sound remotely plausible, he has to ignore recent trends and pretend the recent trends are +0.02/dec. That’s just wrong.
I thought it would be fun to see what + trends existed in the 1800s.
The trend is in C / decade from HADCRUT4
May 1892 to April 1902 was .25C per decade.
http://www.woodfortrees.org/plot/hadcrut4gl/from:1892.5/to:1902.4/plot/hadcrut4gl/from:1892.5/to:1902.4/trend
1893 5 to 1903 4 0.06
1892 5 to 1902 4 0.25
1891 5 to 1901 4 0.23
1890 5 to 1900 4 0.2
1889 5 to 1899 4 0.08
1887 5 to 1897 4 0.02
1882 5 to 1892 4 0.01
1874 5 to 1884 4 0.07
1873 5 to 1883 4 0.16
1872 5 to 1882 4 0.19
1871 5 to 1881 4 0.19
1870 5 to 1880 4 0.21
1869 5 to 1879 4 0.21
1868 5 to 1878 4 0.07
1863 5 to 1873 4 0.12
1862 5 to 1872 4 0.11
1861 5 to 1871 4 0.17
1860 5 to 1870 4 0.19
1859 5 to 1869 4 0.16
1858 5 to 1868 4 0.07
1857 5 to 1867 4 0.14
1856 5 to 1866 4 0.09
Isn’t that pre-CO2?
BarryW
Yeah, this was one of the first things that made me a skeptic. All the rabble about warming at a rate unprecedented in geological history… and its not even unprecedented in this century. That and the consensus gaaabage…
The IPCC and its supporters have a long tradition of cherry-picking not only data but also the literature. But for the economic consequences of their madness their religious crusade (CAGW) would be laughable.
Here are all the 10-year slopes through about May of 2013 for HADCRUT 3 & 4, GISS, UAH, RSS, and NOAA. Five of the six have negative slopes for a number of consecutive months in the last couple of years. (Why is UAH such an outlier though?)
I count about 15 periods since 1840 when we’ve gone through multiple months of negative slopes.
https://dl.dropboxusercontent.com/u/75831381/all%2010-year%20slopes.jpg
Maybe should have mentioned that the slopes are K/century.
Deliberately deceptive is worse than ignorant. This is clearly deliberate and it’s a real travesty.
@lucia
Sure, but what makes the ’70-2000 rise different from ’10-40? There is also what amounts to a 30 yr flat spot between ’40 and ’70. So if these are not associated with CO2 (by their own statements) then we can make the argument that the temps can rise between 0 and ~.14 deg/dec (1910-40) without invoking CO2. The rate for the ’70-00 time frame is ~.16deg/dec so you only have .02 deg/dec that you can argue is in “excess” of natural variation. Now the only real point I’m trying to make is that all that’s going on is cherry picking and hand waving. Just hiding the present decrease is just one of their shenanigans.
Thanks for clearly showing the recent Negative trends.
Suggest showing the year by year graph of the ten year trend for the last century. Somewhere I saw that graph and it very clearly showed the 60 year Pacific Decadal Oscillation discussed by Don Easterbrook, Paul_K and others. In 2000, Easterbrook predicted a change from warming to cooling based on the PDO phase.
The 90 deg 15 year moving trend might show it even clearer.
Hi Lucia, I just left the same comment on the previous thread:
Kevin Trenberth did not present trend lines for those 10-year periods. The horizontal lines are average temperatures for the periods. I confirmed that here:
http://bobtisdale.files.wordpress.com/2013/06/figure-9.png
The graph is from this post:
http://bobtisdale.wordpress.com/2013/06/04/open-letter-to-the-royal-meteorological-society-regarding-dr-trenberths-article-has-global-warming-stalled/
Lucia, you are wrong and unfairly smear Trenberth. Seems like your recreation is missing the Tiljander transform that he has obviously applied. Which obviously excludes the last two trends because they are larger than 0.025
Lucia: PS, Trenberth includes the word “rate” in the first paragraph of his RMS article, but doesn’t use the word “trend” anywhere. There’s no caption for his graph and he never describes the illustration in the text. He leaves it up to the reader to determine that he’s showing the average temperatures for the 10-year periods, not trends.
Bob Tisdale–
That’s even weirder. If that’s all he’s doing, why not just show the average spaced every 10 years. (If he’s going to start in xxx7, show all that. )
Of course that would obviously not engage the point.
Anyway, if that’s what he did, to “rebut” the idea decided to not show that the 10 year moving average has also stalled and that it ends in a decline.
Bob, Lucia,
It sure is weird. Does this mean I can graph the average temperature since 1880 and talk about that plateau?
The best you can say about graphing average temps there is that it doesn’t make any sense when talking about plateaus and trends.
lucia (Comment #116068) says: “That’s even weirder. If that’s all he’s doing, why not just show the average spaced every 10 years. (If he’s going to start in xxx7, show all that. )”
(sarc on) Maybe he’s planning in the future to explain the “big jumps” he mentions are in response to the 1976 Pacific Climate Shift, and the 1986/87/88 and 1997/98 El Ninos. (sarc off)
Well… yes. After all: if you make a graph of data, arbitrarily pick 10 adjacent points, compute the average and draw a straight line showing the average value beginning and ending at the two end points you can show scads of absolutely “flat” lines. But those don’t mean anything.
Let’s take Trenberth’s graph and just go back a little further to let’s say 1940. Oops, now there are some cycles in there.
http://s7.postimg.org/q576usl7f/Global_Temp_NCDC_1940_2013.png
Let’s use some of the natural cycles like the ENSO, the volcanoes and the AMO and see how close we can come to the higher resolution monthly numbers. Looks pretty close.
http://s23.postimg.org/9apy2lj4b/NCDC_Model_1940_2013.png
Now what is the underlying warming trend. 0.06C per decade fairly consistent throughout (a Ln(CO2) trend is a better fit than the linear but …)
http://s21.postimg.org/bzwn4lf9z/NCDC_Warming_1940_2013.png
I made a comment previously at Lucia’s about a similar and related phenomenon to climate models, that being the recent cook survey. My question then was how valid a result can be [97 %] when it matches up exactly with what the authors were looking for in a totally different survey of climate scientists [97%]. * see her response]
Mr Watts has an article on this in a lesser vein re the British Met service predicting the weather wrongly and that on 12 out of 13 occasions predicting that it would be warmer. [none colder?]
All the current climate models predict an increasing temperature with most models having a default 1 or 2 years out of 20 where they will allow a dip to make the model seem more natural.
True models should have a random walk zero input that means the graphs go up or down 50% with a year or 2 in 20 upwards for global warming due to the projected Carbon warming. The fact that no models show a 50% guess for the future of the next 10 years is a beacon of the ineptitude and bias of all global weather models.
It stands out, it cries out that the models are wrong.
No one would offer odds of 20 to 1 that next year will be warmer than last year in any of the next 20 years, we do not know.
But that is the obvious and mistaken conclusion of all the models.
Where is the Honest Climate Model and what should it look like [ Would love JC to make a post on this subject and will cross post at the Blackboard if possible.
Sorry if this is off topic slightly it is about graph validity please delete if you wish
Nice logical analysis. I fear you make the error of assuming Trenberth’s thinking is likewise logically rigorous. It obviously isn’t, since he works from undefined terms reminiscent of the Cheshire Cat. Further proof is provided by his having ‘found’ his hypothesized ‘missing heat’ explaining the pause hiding below 700 meters, without an explanation about how it got there without having been detected above first.
Obviously, in climate science heat does not necessarily travel from hot to cold…
The range of the entire graph is only 0.6ºC. We’re seriously discussing variations in hundredths of a degree? Trenberth too? What are the error margins in the temperature measurements in this series?
Lucia,
Is the first graph actually Trenberth’s or your interpretation of what he was saying? You may want to indicate if the graph is not actually his. (i.s. the first two flate lines – did you add those?)
Sorry, I just looked back and now I see the faint blue lines. The first time I looked at it I did not see them at all, I just saw the last one since the points were red it was easier to see.
I made both graphs. However, the first was an attempt to make the graph Trenberth made– but adding the timing of the volcanic eruptions and the text indicating what Trenberth showed or didn’t show. I may not have done that perfectly– I’m just going by eye.
His graph does have ‘flat’ lines near the 2 points where I show them.
Bob Tisdale:
“Kevin Trenberth did not present trend lines for those 10-year periods. The horizontal lines are average temperatures for the periods.”
Now that seems extremely unlikely, for the simple reason that similar averages for the “rise” periods between the pauses would necessarily also be flat lines. And then what is the point of the whole exercise (except possibly to give a deceptive visual impression)?
Lucia, June 14th, 2013 at 2:14 pm:
“I think temperatures will resume rising….”
Just why do you think that? Atmospheric carbon dioxide today is higher than ever yet there has not been any warming for 15 years. It is obviously incapable of producing that greenhouse warming that is alleged to be the cause of the anthropogenic global warming. That should be sufficient to decide that the hypothesis of global warming by the greenhouse effect of carbon dioxide is invalid. But that is not the only reason for declaring it invalid. There are also other observations and scientific analysis as well that tell us the same thing. When it comes to temperature standstill you probably did not know that in the eighties and nineties there was no warming from 1979 to early 1997, an 18 year stretch. I discovered that by comparing satellite temperature curves with ground-based temperature curves while researching for my book “What Warming?” Satellite data showed a period of ENSO oscillations whose mean temperature stayed constant in the eighties and nineties before 1998. Ground-based data showed a steady warming in this time slot that was called the “late twentieth century warming.” Nobody could find a natural cause for it and that was taken as proof that the warming was anthropogenic. I considered this wrong and said so in the book. But nothing happened when it came out and everybody ignored it. Until last fall that is, when GISTEMP, HadCRUT and NCDC in unison got rid of this fake warming and decided to use the satellite data for the eighties and nineties instead. Nothing was said about it. A nice cross-pond cooperation you might say. Or else somebody got cold feet and decided to cover themselves. Take your pick. I consider this coordinated action as proof that they all knew the warming was fake. But as a result now we have an 18 year uncontested no-warming period beginning in 1979. And we also have the entire twenty-first century as a no-warming period. Between them is a narrow window just wide enough to accommodate the super El Nino of 1998 and its associated step warming. This takes up the entire satellite era and leaves no time for any greenhouse warming at all. It means no greenhouse warming at all for the last 34 years. In view of this fact, what chance is there that any of the warming that preceded the satellite era can be greenhouse warming? Not much, in my opinion. But the scientific argument against greenhouse warming is even stronger. Ferenc Miskolczi studied the absorption of infrared radiation by the atmosphere over time. Using NOAA weather balloon observations that go back to 1948 he discovered that IR absorption had been constant for the last 61 years. At the same time, carbon dioxide in the air went up by 21.6 percent but this had no effect on the absorption of IR by the atmosphere. And no absorption means no greenhouse effect, case closed. This is an empirical observation of nature and it overrides any predictions from theory that do not agree with it. It specifically overrides predictions of climate models that use the greenhouse effect to predict warming. Such predictions have been used to justify “decarbonization” laws and regulations. They have all been passed under false premises and must be nullified.
When Trenbeth says:
“We find exactly the same sort of flat periods in climate model projections, lasting easily up to 15years in length”
Does he specify in which model projections, and/or over which periods these flat periods occur?
I have checked the multi-model means for IPCC AR4 scenarios A1B, S2, and B1 and there are no 10 or 15 year trend of less than 1c/decade after 1975. Scenario B1 does show 10/15 year trends of less than 1c but not until about 2089.
In fact, the only scenario which shows 10/15 year trends of around zero after 2013 is the “commitment” scenario, which of course assumes zero growth in greenhouse gasses.
Of course, he may mean individual models but I am not sure if that would be meaningful if the MMMs don’t show such periods.
Ray–
He means individual model runs. But the paper on which that is based is either (a) a little confused or (b) misleading about what the fact that a few model runs have low trends means in terms of interpreting the range of trends likely to be possible.
Lucia,
Thanks,
If that is the case, it seems to defeat the purpose of using Multi-Model Means in order to obtain the most accurate temperature projections, since doing so removes the 10-15 year flat periods.
Presumably only those models which exhibit 10-15 year flat periods will be used in future.
Of course, this would only work if the flat periods occurred at the same time.
lucia writes “He means individual model runs.”
Smoke and mirrors. That doesn’t excuse the models from the fact they’ve diverged from reality to the point where they’re going outside their uncertainty intervals.
Lucia:
That is so reminisant of a saw tooth mixed to a ramp and some noise that it almost screams at you.
Only if you require the ramp to be a straight line you then disallow any longer natural cycles in the data than 12 months. That seems to be counter intuative.
Ray (#116285) –
I’m sure that Trenberth means individual models. The multi-model mean, by its nature, will attenuate short-term variability because each model will have a different phase.
.
I took a look at 23 CMIP3 model runs from KNMI Explorer. Computed annual average global surface temperature anomalies. Computed the 10-year OLS trends thereof. Flagged all years in which the 10-year trend is less than 0. Excluding historical data (mostly) by starting with the interval 2000-2009, we find 48 occurrences of a negative trend, out of 23*91=2093 model-years, a little over 2%. [The final 10-year trend is 2090-2099]. Eleven of the 23 models had no negative trends.
.
Applying Lucia’s criterion of “flat” as a slope less than 0.025 K/decade, there are 65 occurrences of flat/negative intervals, about 3%. Ten models had no flat/negative trends.
.
So, while Trenberth is correct in saying that 10-year flat trends are seen in model runs, they’re not seen very often.
.
Extending to 12-year intervals, there are 14 intervals (out of 2047, less than 1%) with negative OLS slopes, and 22 with flat/negative trends.
.
As for 15-year trends, there are only 3 negative trends, all in one model, in consecutive years. And only one more with a flat trend.
Harold–
Are you running all the way to 2100? Or ending someplace else?
Lucia (#116301) –
Final trend interval ends with 2099. E.g. for 10-year trends, last trend calculation is 2090-2099, for 12-year, 2088-2099.
Harold–
I”m pretty sure Easterling and Wehner only went through 2000-2030 or something like that. After that, multi-model means for forecasts are non-linear.
HaroldW (Comment #116296)
“As for 15-year trends, there are only 3 negative trends, all in one model, in consecutive years. And only one more with a flat trend.”
Thanks,
Can you identify the model?
Lucia (#116306) –
Granted, forcings are projected to increase, so the intervals later in the 21st century may not be statistically comparable to current ones. Using only 20 10-year trend calculations per model, viz. 2000-2009 to 2019-2028 inclusive, there are 23*20=460 trends computed, of which 12 are negative, 19 are negative or flat.
.
With 12-year trends [2000-2011 through 2019-2030 inclusive], of the 23*20=460 trend calculations, 8 are negative, 9 negative or flat.
.
Edit: Just visited KNMI. Analysis should be redone using “all members” (54 elements), not “all models” (23 elements). Unfortunately, work calls.
Ray (#116307) –
The negative trends are those starting in 2064-2066, for “model 0”. Unfortunately, when downloading “all models” at KNMI, the file doesn’t identify the model by name, only by number.
HaroldW–
I have all the data for A1. I’m just reorganizing a script so that I can discuss a range of start years.
I keep saying I’m going to add data from other models and I likely will finally do so. 🙂
Per note in #116309, I redid calculations using “all members” of CMIP3 20c3m/sresa1b (54 runs) vs. “all models” (which averages some runs hence shows less short-term variability).
.
Using 20 10-year trend calculations per model, viz. 2000-2009 to 2019-2028 inclusive, there are 54*20=1080 trends computed, of which 59 (5.5%) are negative, 78 are negative or flat.
.
With 12-year trends [2000-2011 through 2019-2030 inclusive], of the 54*20=1080 trend calculations, 31 are negative (2.9%), 42 negative or flat.
tty, “that seems extremely unlikely”. Indeed it does, for the reason you say, you could do that over any time interval, and it’s completely meaningless! But it looks like Bob is correct. Kosher Kev is plotting dashed lines that you’d assume are trend lines, but in fact aren’t, they are just flat lines. It’s even worse than we thought.
I have calculated the rolling trends for the individual models used in scenario A1B, (using data from the IPCC website, not KNMI).
Some of the models do show negative trends over 10 years and I thought it would be interesting to see how the trends to the end of 2012 compare with the actual 10 year trend in HadCRUT4.
Most of the models show strongly positive 10 year trends up to 2012, while HadCRUT4 is negative.
I know that Trenberth’s point was that some models show short-term pauses, not that those pauses coincided with actual observations, but I don’t see how we can put any faith in a model which shows pauses but those pauses are in the wrong place.
One model, mpi_echam5, does show a negative trend to 2011 and 2012, which matches HadCRUT4.
When you compare the rolling 10 year trend in HadCRUT4 and mpi_echam5, they do show similar characteristics, although there are obvious differences, particularly in the hindcast period.
One obvious recent discrepancy is that the HadCRUT4 10 year trend to 2001 was +3.9c/decade while mpi_echam5 shows only +0.45c/decade, although mpi_echam5 catches up by 2005, with a 10 year trend of 4.2c/decade.
However, without doing any complicated maths, I would say that model is probably the most accurate, in terms of the current 10 year trend, although I don’t know if that means it can be trusted on longer-term projections. It actually shows an anomaly of 3.44c by 2099, compared to the A1B Multi-Model Mean of 2.80c.
Ray
Because of the type of system, it’s ok that the pauses in the model runs don’t coincide with earth weather.
I’m trying to gin something up to make a new for the blog graph. (The alternatives is too many figures. ) I’m having a little glitch “loopifying” to pull out the data I want. I need to trace the error–b ut the bug is hiding.
But do you have a video?