Bishop Hill The GISS graph mystery
Jun 26, 2015 Climate: Models There are lots of people getting excited by a new animation put out by Bloomberg, which seeks to persuade people that only carbon dioxide can explain the temperature history of the last century or more. It's nothing new - just a prettier version of arguments that have been put forward in the past. I have to say I am greatly amused by the fact that the models stop in 2005. I wonder why that could be?
The simulation was put together by Gavin Schmidt and Kate Marvell of GISS, using GISS Model E2, a climate simulator with a relatively low TCR of 1.5 but a rather strong aerosol forcing of -1.65 Wm-2. However, the IPCC's best estimate of aerosol forcing is only -0.9 Wm-2 and the recent Bjorn Stevens paper put the figure at just -0.5 Wm-2. What this means is that had the GISS model had an aerosol forcing in line with recent best estimates, it would have warmed much too quickly. The resulting embarrassment would have been greater still had the model data not ended ten years ago. I really would like to know why this is.
Still, it's a pretty graph.
Bishop Hill
Barry Woods points out a tweet by the Met Office's Peter Thorne, who says that the GISS Model E2 simulations stop in 2005 and were not extended by most groups.
I wonder why.
Bishop Hill
Nic Lewis emails to say that the GISS webpage is wrong. The aerosol forcing behind the Bloomberg page was in line with the AR5 best estimate, although still much stronger than the Bjorn Stevens estimate. However, the TCR was also even lower, at only 1.4, similar to the Lewis estimate (1.35) based on the AR5 aerosol figure.
So we have a model that is not very sensitive to carbon dioxide, and which appears to reproduce past temperature history. But if Stevens' aerosol estimate is correct, then it's still too sensitive.
Reader Comments (66)
Correct url:
http://www.metoffice.gov.uk/hadobs/hadcrut4/data/versions/previous_versions.html
Just how much further from the actual real world measurements must these models/adjustments get before the elastic band snaps and the Emperors Clothes are revealed as balls and buttocks.
A better question at the end might - So what?
Nothing there proves that the influence is neutral in cost if not benign.
With the Progressives collapse of Electricity Generation, how long will it be before the UK suffers its first Ice Cream free summer?
There will be riots on the beaches from Brighton to Blackpool, when people realise that flying to guaranteed sunshine is better value, as Greendy leaders have demonstrated for years.
Ending that Bloomberg graph 10 years ago in 2005 is inexcusable.
Amancalledchuda, could you give a link to the actual plots and data? I find it is quite easy to make errors when plotting the data at woodfortrees, especially when editing and re-plotting data on multiple graphs.
Having said that, many people are misled by two things:
1) OLS (ordinary least squares fit) linear plots are relatively sensitive to changes at either end of the data.
2) The natural variability of the temperature as measured is large relative to the trends under discussion.
So, to illustrate, plotting normalised HADCRUT4 temperature data, with linear fits, from 2001 to 2014.2 and from 2001 to 2014.3 can apparently produce a huge change from negative to positive trends, larger than in your question. Shown here.
How can this be? How can one or two months produce such a catastrophic change in long term trends? Can it be true? Are we going to die?
Yes it can be true. And I certainly expect to die. But not from graphs like that. Or from global warming, which is patently not even unusual, never mind catastrophic.
The graph with identical raw data can be plotted like this. And probably should be. Did you see my trick in the first graph? The difference is normalising the data. This has the advantage of getting everything, linear trends included, significantly visible on the same scale. But at the expense of distorting your perception of the original data, especially if you chose to not even plot the original data on the same graph. William M. Briggs, statistician to the stars, has exhorted people to "just look at the data". This is something certainly not done much in the MSM, and, sadly, not often enough by some self-proclaimed science-communicators. Even worse, scientists and scientists in-training make this mistake too often.
When we just look at the data, before someone has made their choice of how to influence the way we see it (or at least earlier in the process) we only see this: In the period chosen, the temperature went up over the last two months or so, but certainly nothing to write home about compared to what has been seen before in recent history. As any sports fan knows, it is easy to dig into even simple data like that and find "a record" if there is a desire to find one.
And that last paragraph sums up global warming in a nutshell.
Amancalledchuda -
I've done the same calculation (trend since Jan 2001) in Excel, and it agrees with Woodfortrees. The trend, expressed as you do in K/century, has increased by .03 to .04 each month for about a year, with only the cooler Nov 2014 as an exception. As of March 2014, the value was ever so slightly negative.
Last value in the HADCRUT4 file is for April 2015. The OLS trend line for May 2015 is 0.52; the reported temperature anomaly will certainly be higher (Jan-Apr 2015 are 0.655 to 0.69). Graph. So the trend will take another step upwards.
As you and michael hart say, the numbers remain small.
Correlation is not... ah, you know.
http://www.tylervigen.com/spurious-correlations
The Bloomberg graphs are comprehensively trashed over at WUWT
Schrodingers Cat, climate advice is based on market forces. Bloomberg like things going up.
Climate advisors always assume things go up
People who bet money on the climate going up, need investors to improve their sustainability.
Somehow, it all went terribly wrong. Hockey stick models aren't always what they are cracked up to be.
4F, Faked, Fiddled, Fudged and Forged.
A quick post-script to my response above to Amancalledchuda -
With a little algebra, one can work out the change in slope ("trend") of an OLS linear fit, due to extending a uniformly-sampled data set by one additional point. The new slope is equal to the old slope plus an increment proportional to the difference between the new point and the extrapolation of the old OLS fit at the new sample point. The constant of proportionality is 6/((N+1)(N+2)), where N is the number of points in the original set.
For the case of HadCrut4 from Jan 2001 through Apr 2015, N=172. The extrapolation of the current OLS curve is about 0.52 K, while the May 2015 report will be in the neighborhood of 0.67 K. Applying the above formula, the change to OLS slope will be about (0.67 - 0.52 K) * 6 / (173*174) = 2.99*10^-5 K/month. Converting to K/century, expect a change in your metric of about 0.036 K/century next month.
The give away is the 1880-1910 baseline used in this animation. You have to be extremely cautious with anomalies beacuse they 'measure' just the differential of temperature. You can't distinguish changes due to CO2, land change, urbanisation, or natural variation.
The GISS people are being dishonest if they claim that "only carbon dioxide can explain the temperature history of the last century or more".
My latest paper, published in October last year, showed that the ENSO and occasional equatorial volcanic eruption could account for HadCRUT4 variations from 1950 to 1986 and after that there's a reduction in total cloud cover then a decrease in low level cloud but increase in mid and upper level cloud.
I concluded that there was very little, if any warming, left to explain after taking all these into account.
So much for the GISS model (a.k.a computer games).
By a strange coincidence, the 2009 Climategate emails exposed the REALITY world and religious leaders had hidden from themselves and from the public since 1543:
http://junkscience.com/2015/06/28/you-are-known-by-your-allies-big-popey/comment-page-1/
In hindsight, “Coincidence seems to be God’s way of remaining anonymous!”
Gavin's tweets seem to be a bit prickly lately. I think you may have touched a nerve, Bish!