Friday
Jan242014
by 
Bishop Hill
Bishop Hill HadCRUT 2013
Jan 24, 2014 Climate: Surface The HadCRUT global temperature anomaly for 2013 is 0.486. If so it should be outside the 5-95% bands on Ed Hawkins' famous graph.
That will be more standstill then.
Reader Comments (93)
It looks like the 8th warmest year since HADCRUT began and the 8th coldest year in this millenium. Catastrophe looms.
I am still amazed by the precision of the HadCRUT, better than 1/1000th of a degree Celsius, what a wonder.
http://physics.nist.gov/Pubs/SP811/sec07.html#7.9
Despite the 'homogenisation' of the temperature data by CRU the climate is just not doing wot its spose to.
Perhaps HAD needs to find another crew?
The UKMO forecast for 2013 was:
based on a 2012 figure of 0.45C. The actual of 0.49C only just above their lower tolerance indicative of the fact that the model is running too warm but the WMO figure, which is an average across three data sets will produce a higher actual figure for 2013 than HadCrut.
So are the models running too warm or the data too cool?
And yet again a vague MO "forecast" is wide enough for what passes for an "actual" to squeak in at the bottom of the range.
Bing! model validated!
LB
I am surprised at you! The models are right they are just waiting for the world to catch up:-
Global average temperature forecast for 2014
Deja vue? Sure looks like they expect a continuation of the "standstill"?
The forecast of 0.71C for 2013 was stunningly unrealistic (the warmest year so far being 2010 with 0.547C and a total amount of only two months over 0.7C so far - February 1998 and January 2007). As this absurdity was given for the upper end of the forecast, I wonder whether they considered their 0.43C for the lower end as unrealistic as the upper end? And the result is almost there at the lower end (what people with enough common sense could have predicted). These guys are not for real (scientivists?). What do they smoke?
I predict that 2014 anomaly will be between 0.4C and 1,5C.
"I am still amazed by the precision of the HadCRUT, better than 1/1000th of a degree Celsius, what a wonder."
To be fair HadCRUT reports that taking into account all uncertainties, the anomaly is within approx. +/- 0.1C of 0.486C (i.e.somewhere between 0.385 and 0.591). That's a lot more informative than some time series I could mention.
Greensand
I suspect as you do that the ENSO neutral territory has persuaded them that El Nino is not knocking at the door, which is why they have the same figures for 2014.
I had a bet on with Richard Betts, a pint, that the actual would be below the forecast for 2013, it will be interesting to see how that pans out.
Does anyone keep tabs on the HadCRUT dataset to check whether the raw data is massaged, tenderised, liquidised and generally cooked up like some of the others?
And it's also interesting to see that on decadal scale all three of the first years of the decade have been colder than the first three years of the previous decade.
2001 2002 2003
.437 .492 .503
2011 2012 2013
.406 .448 .486
The averages 2001-2003 0.477C and 2011-2013 0,446C.
It's only the first 3 years out of 10 of course, but we'll see how the drumbeat (and Gavin's absolute certainty) about "the next decade will be warmer" will sound in 7 year's time...
The 13 real-time forecasts made by UKMO for 2000-2012 are discussed by Folland et al., 2013, in GRL. This is a different issue than the climate projections vs. observations (e.g. Ed Hawkins' graph), but since some comments refer to the UKMO forecasts it is worth taking a look at this.
They note the the apparent warm bias in the forecasts (though discuss also the possibility of a cold bias in the observations). The correlation between the actual forecasts (i.e. not hindcasts made after-the-fact, but the forecasts issued in advance) and HadCRUT is 0.74 over 2000-2010 and 0.75 over 2000-2011. RMSE of 0.07 degC and a warm bias of 0.06 degC.
I love it when each year the MSM readies the headline: "last year was the nth warmest year since records began" (where n is supposed to be terribly frightening). I like to use the following scenario to show that this scare tactic says very little:
Suppose for a period of 100 years each year is 0.01C warmer than the last. Every single year for those 100 years would notch up a "warmest year ever" headline. But by the end of a hundred years the average global temp would be 1C higher than at the start. Who'd have serious cause to worry?
@ Jan 24, 2014 at 12:27 PM | HG54
If the anomaly is within approx. +/- 0.1C, then you should never give more than one significant digit after the decimal point. It is, and always have been, bad practice to do otherwise. It gives a false impression of accuracy, which we know is not there. In the link I provided before the SI rules are very clear about it.
And that is a very arguable precision, a more realistic one would probably be ±1°C
9 warmest years are all basically the same - inside a difference of 0,064C (2010 being the warmest with 0,547C and 2007 being 9th warmest with 0,483C). So (especially with an error margin of +/- 0,1C) it's absurd to rank them.
Has anyone noticed that the Met Office have stopped updating their graphs of HadCrut4 and HadCET? See http://www.metoffice.gov.uk/hadobs/hadcrut4/diagnostics.html, frozen end 2012 and
http://www.metoffice.gov.uk/hadobs/hadcet/index.html, frozen end 2013.
I suspect they are trying to “hide the decline”, i.e. to suppress the fact that natural climate variability appears to have initiated a cooling temperature trend, globally and nationally, around 2005, as this runs counter to the IPCC’s predictions of global warming which they so publically endorse.
I’ve written to the Met Office asking for an explanation.
This "correlation" is largely a function of the vagueness of the forecasts and therefore says nothing useful.
Doug Brodie -
I don't understand your complaint. As far as I can see, both of those charts show the 2013 data point, which of course is quite recent.
HaroldW:
If you look at the title of the HadCrut4 graph, it says 1850-2012. I think they stopped updating it at the end of 2012. Similarly, the HadCET graph says 1772-2013. I remember looking at just last December and it always gave the current date. I’ve got a saved copy with the title 1772-25th Sep 2013.
@NW "This "correlation" is largely a function of the vagueness of the forecasts and therefore says nothing useful."
As I understand it, the correlation given is between each year's single "central estimate" forecast and the subsequently "observed" HadCRUT temperature anomaly. So, the width of the forecast uncertainty range (i.e. the vagueness) doesn't affect the correlation.
@DougBrodie: the title of the HadCRUT4 graph says 1850-2013 when I look at it. Have you tried refreshing the page?
Sven, not only the nine warmest, but the last seventeen years cannot be stated to be warmer than each other if the accuracy is ± 0.1°C.
Tim Osborn,
You're not the CRU chappy, are you?
When you say "a warm bias of 0.06 degC" is that over the course of a decade?
@ patagon
the precision to 3 decimal places stems from
global temperature change and its uncertainties since 1861
ck folland pd jones and chums
http://www.math.ualberta.ca/~shen/Sam_Papers_pdf/foll_grlet_2001.pdf
In this paper it is claimed (incorrectly) that as measure error on thermoter is 0.2c (yeah right)
and there are 60 measurements a month to get daily tmax tmin then the error on estimating montly mean
is 0.2/sqrt(60)
try looking at min max daily values for a real weather station and the error in estimating montly mean
is more like +/- 0.5C
so all the "warming" is within error
the reason why the folland paper talks about +/- 2 stddev is
+/- 1 stddev has a 68% chance of containing real value
+/- 2stddev has about 95% chance
so 0.8C warming with +/- 1C is conclusive prove that we dont need to do anything!!
which means if the whole warming claim is based on 0.8C
Jan 24, 2014 at 12:51 PM | Unregistered CommenterSchrodinger's Cat
They won't release the raw data. I think it might be on cards like the NOAA ones but yes, they adjust as they all do. It's almost certainly why we are seeing a stall and not a decline.
Jan 24, 2014 at 4:34 PM | Unregistered CommenterJeremy Shiers
I had this argument with RC a few years ago . All I got back was the usual foul mouth reply + it's an anomaly and therefore can be accurate to 1/1000°C. Fing idiot.
Wow! I knew climate change could make things hotter and colder, I just never expected it to happen simultaneously.
@ Stephen Richards
Shows the value of peer review if no one picks up what is taught in first year undergrad science courses
Also shows Phil and chums have no shame in displaying gross errors in basic statisitics
I seem to remember Doug Keenan challenged Phil to sit a stats exam
and Doug would pay him £500/minute for sitting exam
Phil declined - busy schedule no doubt
regarding RC my recommend is not to argue with stupid people
they'll just try and drag you down to their level
@ Tim Osborn, Jan 24, 1:00 PM
The link to Folland et al. is much appreciated.
@ HG54, Jan 24, 12:27 PM
I agree, the uncertainties given for HadCRUT are helpful and important.
@ Jeremy Shiers, Jan 24, 5:09 PM
You remember rightly—strictly, the exam was to be in time series.
Its another game with numbers pay rises running at 2.5 percent CPI inflation rate 2.4 percent.
Thanks David!
Remind me he who flips coins is called a........?
Tim Osborn,
OK, the 'Folland et al' link is now working for me. (I couldn't connect to the server until now.) The abstract, which is the only bit that a pleb like me can read, says:
"We discuss 13 real-time forecasts of global annual-mean surface temperature issued by the United Kingdom Met Office for 1 year ahead for 2000–2012."
And:
"This observational cold bias was mainly responsible for a statistically significant warm bias in the 2000–2010 forecasts of 0.06°C."
0.06 degC in a year? That would extrapolate to an error of 6 degrees per century... which is an order of magnitude greater than the totality of "Global Warming". Or have I completely misunderstood?
@ Jeremy Shiers
Many thanks for the link, I'll have a look at it.
If the error is 0.2/sqrt(60) = 0.0258, then they should not go beyond the second significant digit, i.e. 0.48 ± 0.03
That is the thermometer error that you mention. I suspect that the error introduced by choosing a lapse rate to interpolate from station altitude to grid altitude is far larger, specially to a huge 5° x 5° grid.
Tim Osborn,
I feel perhaps that ridicule is coming my way.
"This observational cold bias was mainly responsible for a statistically significant warm bias in the 2000–2010 forecasts of 0.06°C."
Perhaps that means that the compound error over the decade is 0.06°C. Meaning that the models are only wrong by an extrapolated 0.6°C per century, which is all of the warming we've experienced, give or take.
Is that it?
It's hard to tell when the scientific method hides behind a paywall. Is this the way that the discovery of truth is really supposed to work? I think it's pretty shameful.
http://xkcd.com/1321/
The last time I looked at the smoothed data from the met office the figures were
2007 0.476 now 0.476 diff nil
2008 0.472 now 0.473 diff +0.001
2009 0.468 now 0.470 diff +0.002
2010 0.464 now 0.469 diff +0.005
2011 0.460 now 0.470 diff +0.010
2012 0.456 now 0.471 diff +0.015
Can anyone explain why the figures for '08 to 12 are increased?
James
The annual forecasts are produced taking into account the current years weather. The central forecast figure for the following year is then measured against the actual temperature. You would expect some years the forecast will be higher than actual and some years lower than actual. The UKMO forecasts are consistently higher and over the period measured give a RMS error of +0.07C. So whatever the forecast that UKMO make, 0.57C for 2014, if you deduct 0.07 you would be closer to actual every time.
The GRL paper is suggesting that the reason since 2004 for the error is that the actual temp is 0.06C lower than it should be due to missing data in the Arctic and so the data needs adjusting up rather than the model adjusting down.
The annual error between forecast and actual is not fixed at 0.07C, it's basically the average over the period and so cannot be extrapolated into decadal or century figures.
The last time I looked at the smoothed data from the met office the figures were
2007 0.476 now 0.476 diff nil
2008 0.472 now 0.473 diff +0.001
2009 0.468 now 0.470 diff +0.002
2010 0.464 now 0.469 diff +0.005
2011 0.460 now 0.470 diff +0.010
2012 0.456 now 0.471 diff +0.015
Can anyone explain why the figures for '08 to 12 are increased?
Just thinking...
Global temperature varies monthly , with a peak during the boreal summer due to increased solar radiation over the northern hemisphere, which has a larger ratio of land to sea than the southern hemisphere. The amplitude of this oscillation is about 5 times larger than the current yearly temperature anomaly.
Now, if global temperature has increased due to atmospheric LW radiation, caused by increased CO2 and H2O, this is a different mechanism and should show a different pattern. The amplitude of the oscillation should decrease, with a higher minimum value.
Has this been observed?
My Lord,
"The annual error between forecast and actual is not fixed at 0.07C, it's basically the average over the period and so cannot be extrapolated into decadal or century figures."
Why can the annual average not be extrapolated? If there is an average error of 0.07C per year, then the average error can be precisely expressed as being 7 degC per century. That's just a restatement of the numbers.
My Lord
P.S. I do have a degree in mathematics, so I'm hoping for incredible gymnastics in your response.
James
I don't have a degree in mathematics but if the same bias was apparent for a hundred years or a thousand years it would still be an error of 0.07C as the forecast is for one year only and the bias not compounded but 'reset' each time as the forecast is based on actual temperatures not previously forecast temperatures.
@Patagon Jan 24, 2014 at 5:50 PM
The common error of divide by sqrt(N). "They" prate about the "Law"(?) of Large Numbers. Applies only if the observations are i.i.d.
Independent? Have "they" never heard of auto-correlation? (pace the thought that it's [3,1,0] and not [1,0,0]) And I also thought that "tele-connections" were a big part of Climate "Science"™. "They" use it to justify the tying together of Darwin and a pub in the middle of the Nullarbor. I think that kills the first i.
Identically distributed? Worldwide? The equator (in the middle of a rain-forest at ~0 ft. elevation) versus the highest point in Antarctica? Puh-leze. Tell me "they" are joking.
Common errors in Climate "Science"™. But then basic statistics was always beyond them!
Focusing on the global average temperature is so yesterday. Why worry about a hiatus in average temperature when there are other more obvious concerns.
At the World Economic Forum in Davos a leading climate change expert Al Gore noted that:
"these extreme weather events which are now a hundred times more common than 30 years ago are really waking people's awareness all over the world"
http://www.theguardian.com/environment/2014/jan/24/climate-change-al-gore-davos-haiyan-sandy
Lord Beaverbrook,
OK. Let's just agree that we live on different planets.
Lord Beaverbrook
Just in case you haven't seen it take a look at the "Supporting Information" especially:-
Small but significant warm bias in issued global temperature forecasts
It appears the authors are convinced they can improve forecasts by tuning the observations? Reckon even I could do that:-)
I found the following interesting,
I wonder when? The time served cynic in me can't help feeling that the change will only take place when all concerned are sure it will provide the required answer. Time will tell
Tim Osborn
Many thanks for joining in and providing the link to Folland et al
accuracy to three decimal place for something with some many unknown factors and so many problems !
Its a wonder these people do not do the lottery every week given their 'abilities' how can the fail to win .
Basic rule , you can throw as much maths as it as you like , but your chances of being able to give a valid figure which is more accurate then the means of measuring what the figure represents it is the square root of FA.
Greensand
I had the impression that all would be amalgamated into a single WMO data set which the UKMO have increasingly quoted but that doesn't include satellite data as yet I don't think.
Obviously the observed data is not as it should be, the models say so and they are never wrong.
This warm bias we have been discussing with Richard Betts for many years but I can't recall whether it was prior to 2004, can you?
Lord Beaverbrook,
LB,
It turns out that I'm not OK with us living on different planets. I want to understand:
"I don't have a degree in mathematics but if the same bias was apparent for a hundred years or a thousand years it would still be an error of 0.07C as the forecast is for one year only and the bias not compounded but 'reset' each time as the forecast is based on actual temperatures not previously forecast temperatures."
1) Surely 0.07C/year = 7C/century. That's just numbers.
2) You are saying that if a model overestimates temperatures every year (on average) by 0.07C, that equates to the model overestimating by 0.07C every century, and also equates to the model overestimating by 0.07C every millennium. Because the error is 'reset' each year. Either I am totally misunderstanding you, or you are talking the most ridiculous nonsense.
I don't rule out the "me misunderstanding" explanation. God knows I have my moments. But wouldn't that imply that if I was overpaid each year by 7p, after a thousand years I would have gained an extra 7p.
it's good to know that the HadCRUT team can still manage to chisel out a tenure with their meagre talents.
Lord Beaverbrook
Don't think I can help LB. We have retrospectively discussed the periods prior to 2004, but I did not get involved until just prior to the damned 2008 Act. Carrying out due diligence on potential investment opportunities that came recommended as "sure fire" because of the forthcoming Act. Found a crock, such a crock that I am still here! Shows you what I know, a lot of money has been made on the back of the Act. However I am still more than happy with my original decision.