Click images for more details



Recent comments
Recent posts
Currently discussing

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > Questions for the UKMO

Hi Richard

“The latest annual update to the decadal forecast is;-“

Many thanks Richard I have already been there and that page does illustrate the issue.

That is a new forecast starting in September 2011 not an update of any previous forecast. What I am trying to get at is how did the previous forecasts perform over their full forecast period? This can only be assessed by considering each annual decadal forecast as a single entity and reporting against it throughout the full length of the forecast period.

The “update” you point to has the previous forecasts, which must have had significant differences, merged into a single ongoing line and as such it cannot be a true representation of the previous individual forecasts. The concern is that this does not allow an overview as to whether subsequent forecasts are showing an improvement in skill.

It must be possible to show the individual annual decadal forecasts updated against actual observations. Could you please point me in the right direction?

Feb 28, 2012 at 5:33 PM | Unregistered CommenterGreen Sand

Hi Green Sand

Actually, I think you've misunderstood (or it's not been explained very clearly on the website!)

In Figure 1 of the decadal forecast page", the right-hand one of the three red "plumes" (the one from 2005-2014) is actually the entire decadal forecast from 2005. So, this does do what you require - it allows comparison of the 2005 forecast (red plume and white curves) over most of its entire period against observations (black curves). Clearly we haven't quite reached the end of the 2005 decadal forecast yet so don't yet have an full assessment of its skill over the entire period.

The red plume represents the uncertainty, such that the observations are expected to lie within the red plume 90% of the time.

You can see that the central estimate was for global mean temperatures to be rising over the last few years, but the lower end of the uncertainty range was for more of a flat-lining. So far, the observations were just about within the lower end of the uncertainty range.

Thanks for asking this excellent question!



Mar 1, 2012 at 9:37 AM | Unregistered CommenterRichard Betts

Hi Richard, many thanks for the reply.

“Actually, I think you've misunderstood (or it's not been explained very clearly on the website!)”

No Richard, with respect I do not think I have misunderstood.

“Clearly we haven't quite reached the end of the 2005 decadal forecast yet so don't yet have an full assessment of its skill over the entire period.”

The page shows and states that Decadal forecasting started in 1985 not 2005. So we have numerous decadal forecasts that we do have a full assessment of the skill over their entire period?

In 1985 we had a certain amount of knowledge and computer power, since then our awareness, knowledge and number crunching power has increased and in the latter case by multiple factors.

Therefore we should (must) know whether this gain in ability has shown through in increased skill? (I use the royal “we” as a concerned member of the human race and not to imply being directly involved in the science)

That cannot be assessed from the chart you reference (ps, why does it now start in 1950?).

It can only be assessed by showing each subsequent Decadal Forecast as single entities over the whole of each forecast decadal span and then assessing if the accuracy of subsequent forecasts are improving or not?

With regard to “the red plume” and the “white line” where does the “2009 decadal prediction of global average temperature and validation of past predictions.” fit in?

It appears the “blue line” from 2009 has now been blended in to the ongoing white line in the chart on the page you reference, though I have to admit that is only by applying a grid and eyeballiung.

We constantly talk about “uncertainties” we are all aware that they exist, but we should by now be quantifying the uncertainties, “we expect to be for 90% of the time” does not do this. Surely we should be targeting and monitoring the forecast means against actual observational data? It is well understood that we are not going to hit the target all the time but we need (must) know if we are getting closer or further away from our targeted mean?



Mar 1, 2012 at 11:27 AM | Unregistered CommenterGreen Sand

Richard, sorry for rambling even ended up a main point.

"All data are rolling annual mean values. The gap between the black and blue curves arises because the last observed value represents the period November 2010 to October 2011 whereas the first forecast period is September 2011 to August 2012."

Why therefore cannot the forecasts be individually updated with HadCRUT3 actual observational data against forecast mean on the same basis? Month in month out?

Mar 1, 2012 at 1:04 PM | Unregistered CommenterGreen Sand

"ended up a main point."

Should be "ended up missing a main point"!

Matron says I must have a nap!

Mar 1, 2012 at 1:07 PM | Unregistered CommenterGreen Sand

Green Sand,

If I may interject a layman's viewpoint of the decadal forecast graph.

The 90% confidence level gives quite a large bandwidth aprox 0.4C generally which is half of the overall rise in temperature over the last 100 years. Commercially, in industry, this would be the catastrophic error margin which would mean if exceeded that you were manufacturing the wrong product.

It is noticeable that the previous two decadal forecasts starting 1985 and 1995 are quite reasonable facsimiles of observed data if the events Mt Pinatubo and 1998 El Nino are respectfully excluded.

With the 2005 decadel forecast, although it has so far resided within the 90% confidence level it is indicative of something major being overlooked as the divergence increases with time.

Unless the MO have managed to model ENSO on a multi year timescale and there is a large EL NINO in the pipeline then this decadel forecast is not going to achieve anything like the previous ones.

My suspicions are that the quietening Sun since the turn of the century is having a greater effect than allowed for. If this is going to have an effect for the next couple of decades then the 2015 forecast will be an interesting release, particularly when AR5 is due and an election year where policies mean points votes.

The thick blue line forecast is basically an EL NINO forecast which I am dubious of even starting in the next twelve months let alone of having any effect, but that is just my two penneth worth I am sure that Richard and colleagues will keep us informed.l

Mar 1, 2012 at 3:06 PM | Unregistered CommenterLord Beaverbrook

Mar 1, 2012 at 3:06 PM | Lord Beaverbrook

My dear Lord you are more than welcome and thanks! Your point re error margins in manufacturing ring very true and I have the scars to prove it.

But they are not yet my concern and neither am I anywhere near any consideration regarding the actual issues of ENSO, Volcanoes, sunspots etc.

At first I feel it is necessary to understand what the forecasts have been and how they have performed.

From the present chart it would appear that previous forecasts that overlapped have been merged to show the single “white line”. PS I doubt that there will have only been 4 – 1985, 1995, 2005 and “the latest”. I would be very surprised if the MO did not carry out an annual update of their decadal forecast, I know I would and Richard did previously point me to “The latest annual update to the decadal forecast is”, though to be fair he could be just using a figure of speech.

By illustration what happened to:-


There is an individual forecast “thick blue line” and the previous forecast “white line, note the overlap another such is shown on the latest chart.

Also there appears to be a variation in the value of the “white line” between 2009 and the present?

I am probably missing something and I am hoping that Richard can help with further clarification.

Mar 1, 2012 at 5:23 PM | Unregistered CommenterGreen Sand

Hi Richard

I am aware that this does not relate directly to the Met Office but it must have a huge bearing on current research.

Does the following sort of report cause the Met Office any concern or is it too way off beam to merit any attention? I refer to this article

I quote from the opening paragraph

GISS and the GHCN are preoccupied with the Arctic, bestowing large “corrections” on the few and far northern stations. The raw data of the Arctic did not tell the story they wanted to hear, and so GISS took it upon themselves to rewrite it.

In general, is this sort of gross adjustment ever warranted and if so, should the fact that it has happened be more transparent instead of being done by a replacement of old data?

Mar 2, 2012 at 3:47 AM | Unregistered Commentermatthu

As this is topical to posts on this thread there is an excellent new post at Climate Etc looking at models.
What can we learn from climate models? Part II

Mar 2, 2012 at 9:25 AM | Unregistered CommenterLord Beaverbrook

Hi Green Sand

The red plumes for 1985-1994 and 1995-2004 are hindcasts - ie: done recently (in 2005) but starting from 1985 (or 1995) initial conditions as if they were forecasts, in order to test the model. As you say, they did OK except for the 1992 eruption of Mt Pinatubo (which would not have been known about in 1985 so was excluded from the hindcast) and the 1998 El Nino. The only actual true forecast in the ref plumes is the 2005 one.

You are right that others have been done starting at other years since 2005 (you found the 2009 one) but by definition they have had even less time to be compared to observations (since they are more recent) so the 2005 forecast remains the best one for testing "out of sample", ie: the one that was actually done as a forecast, not a hindcast, with the longest overlap with observations.



Mar 3, 2012 at 12:07 AM | Registered CommenterRichard Betts

I remember a BBC program many years ago where an MO guy was going into a room where models where being created and saying to the camera something like "of course models only predict what you program them to predict" so we dont really take them too seriously.
Today the MO takes them so seriously that they are willing to try and predict climate as well as weather. I think the following post on TGWPF says it all:

Today In The Annals Of Settled Science
Sunday, 04 March 2012 09:47 Walter Russell Mead, The American Interest
E-mail Print PDF

According to a new study reported in Scientific American and published in Nature Medicine, it turns out that your biology textbook was wrong about the basic facts of the human body:

A study led by Jonathan Tilly of the Massachusetts General Hospital overturns the decades-long idea that women are born with all the eggs they will ever have. It reports that women of reproductive age carry ovarian stem cells, meaning that they can produce new eggs.

So, to recap: after centuries of intensive scientific research we are still learning amazing and surprising things about our bodies. But climate science is settled once and for all and no further questions need to be asked.

Got it?

Body snatchers, doctors and medical scientists have had hundreds of years of available live and dead human beings to study but they still dont know the half of it. However doctors make no predictions about how long it will take them to understand everything, they just get on with finding out more that will help save lives.
The truth about predicting weather and predicting climate is that it is not currently possible because our knowledge is in its infancy. The discussions in this thread are at about the same level as two early adopters discussing whether wood or stone would be best if one was to make a wheel. Then on the back of that discussion to predict that man would invent a nuclear weapon.
Why is it so hard to admit that we dont know enough? Considering the amount of tax payers money being spent on the basis of predictions of climate and weather I believe we deserve better from our public servants.

Mar 4, 2012 at 8:45 PM | Unregistered CommenterDung


From the decadal forecast graph the slope of the 2011 forecast is more vertical for the period 2012-2014 than the 2005 forecast for the same period.

Where could we find what the major difference is for the input variables between the two forecasts?

I did contact the MO enquires as you suggested in regards to the UK temperature accuracy but it seems that this graph is not going to be updated for the general public as temperature is no longer a key performance target, which is unfortunate really.

Mar 5, 2012 at 12:13 PM | Registered CommenterLord Beaverbrook

Hi Richard

"The red plumes for 1985-1994 and 1995-2004 are hindcasts"

Thanks Richard, I missed that, must improve my reading skills! I will need to re-think about “hindcasts”

"You are right that others have been done starting at other years since 2005"

What I actually said was that decadal forecasting started in 1985 so there must be numerous that have run full extent and therefore their subsequent individual performances can be assessed. Assessed very simply forecast against actual, were we better in the 90s than we were in the 80s etc? Are we gaining skill as we gain resource?

"so the 2005 forecast remains the best one for testing "out of sample", ie: the one that was actually done as a forecast, not a hindcast"

But surely they were all originally done as forecasts, how did they do against actual?

But hey ho, as you point out we do have the 2005 decadal forecast up and running, so we will continue to observe. As HadCRUT3 reports as an 1961-1990 anomaly and the decadal forecast is 1971-2000 is it possible to get an official MO “conversion factor”?

The other concern re the 2005 forecast is the imminent arrival of HadCRUT4. I take that 2005 will be reported against HadCRUT3?

Richard I know you are very busy and I am sorry for being a pain, but if you get chance could you get somebody to shed light on the following?

Some days ago you pointed at the following MO page to explain the HadCRUt3 /Arctic issues:-

“Global-average annual temperature forecast”

At the bottom of the page is:-

Figure 3: The difference in coverage of land surface temperature data between 1990-1999 and 2005-2010. Blue squares are common coverage. Orange squares are areas where we had data in the 90s but don't have now and the few pale green areas are those where we have data now, but didn't in the 90s. The largest difference is over Canada.

Why do we no longer have the Canadian land surface temperature data? I am not aware of any of the stations being closed? If they have been could somebody please point me in the right direction?

Many, many thanks for your replies and contribution.


Mar 5, 2012 at 5:02 PM | Unregistered CommenterGreen Sand

Hi Richard

Just wondering whether you missed my post from 5 days ago relating to this report here?

And here is a remarkably similar story from the other side of the world here

I suppose what I am wondering is whether anyone at the Met Office takes these stories seriously?

Do you talk about what is being reported, and if so do you have any concern at all?
Do you wish your colleagues would make the effort to publish a rational response?
Perhaps you ignore the stories and wish they would go away?
Perhaps you don't think these stories warrant investigation at all?
Are these stories not controversial in the slightest?
Or do you think "There but for the grace of God go I ..."?


Mar 7, 2012 at 10:11 PM | Registered Commentermatthu

P.S. And here is yet another report providing clear evidence of more artificial manipulation of the data

Are all of these examples so commonplace that they are now regarded as standard practice in climate science? Perhaps climate scientists look askance at any data that does not show evidence of mainipulation?

Seriously, Richard - what do you and your colleagues think of all these examples? Or do you not discuss this sort of thing openly amongst yourselves?

Are all of these adjustments justifiable and widely publicised so that any interested party can discover why they have been made? Or would you need to submit a FOI request to try to prise the information out?

Mar 7, 2012 at 10:38 PM | Registered Commentermatthu


Just to clarify where I think the misunderstanding lies.
From what I understood Richard has said, the 2005 forecast was the first forecast using the current model. The 1985 and 1995 plumes are hindcasts to validate the model output, but there weren't any previous forecasts prior to 2005 using this model.
The first end of the decadal forecasts will then be in 2014. There have been subsequent forecasts post 2005 but obviously the decadal end of the forecasts are rolling on. The accuracy of this model output against observational data over the full forecast period can only then be assessed post 2014.
That's what I understood from the comments but I'm more than willing to be corrected if I'm wrong.

Mar 8, 2012 at 7:55 AM | Unregistered CommenterLord Beaverbrook

Hi Lord B
I assume you are going back to my very much earlier discussion with Richard and not the points I have been raising in the past few days?

What I would like Richard to address are the high profile instances I have quoted where it is very obvious that temperatures have been adjusted in a very mechanical manner so as to increase apparent warming.

Are climate professionals (including at the Met Office) unanimously and entirely comfortable when adjustments of this nature have been exposed? Are they transparent enough? And if not, where is the voice of science on this issue? All I hear is a deafening silence.

Mar 8, 2012 at 9:07 AM | Registered Commentermatthu

Above, Richard Betts wrote "It is the Arctic that's the issue. The model forecasts covers the Arctic, but the observations do not, so it's not comparing like with like."

Seems to me that the GISS data for arctic warming is badly skewed. At this hyperlink: clicking on the "Make Map" button displays a livid red Arctic. But if the user changes the settings to "Polar" rather than "regular", and "250km" rather than "1200km" a very different image is generated. These utterly different images use the selfsame source data.

The distorted Mercator projection (big fat Greenland) combined with unwarranted extrapolation (it's as if a thermometer in Nice were used to declare UK temperatures) allows GISS to perpetuate the myth that global warming is concentrated in the Arctic and the citizen who judges GW by looking out of the window is being parochial.

Mar 8, 2012 at 10:29 AM | Unregistered CommenterBrent Hargreaves

Lord B
"The first end of the decadal forecasts will then be in 2014. There have been subsequent forecasts post 2005 but obviously the decadal end of the forecasts are rolling on. The accuracy of this model output against observational data over the full forecast period can only then be assessed post 2014.
That's what I understood from the comments but I'm more than willing to be corrected if I'm wrong."

That is now basically my understanding.

My original question to Richard was “Why does the MO not update their Decadal Forecasts?”.

So I am trying to ascertain if the 2005 model forecast is going to be any better than those that went before? I cannot see how hindcasts using the 2005 starting with original conditions can tell me how the original previous forecasts performed. Or am I missing something?

Mar 8, 2012 at 10:45 AM | Registered CommenterGreen Sand

Apologies Matthu, Green Sand, I seem to have got your posts mixed up this morning. Must remember to sequence the coffee first!

Green Sand, I understand what you mean now by comparing the latest model with previous models, but I wouldn't expect that anything will be forthcoming until 2014 when the forecast period ends even though the trend at present doesn't bode well for the model output. Which leads me to log a question from the New Solar Paper thread:


Would you expect then that the extremely complex GCM which not only simulates the Earth's energy balance but also the water cycle and the fluid dynamics of atmospheric and oceanic circulation used by the UKMO to predict weather and climate to be more accurate than a simple climate model?

The former giving an increasingly positive divergence from observational temperatures since the reduced activity of the Sun in the decadal forecasts and the later being used to predict that the reduced activity will not affect Global temperatures to any extent on 100 year timescale against forecasts.

However, if past total solar irradiance variations are larger and climate models substantially underestimate the response to solar variations, then there is a potential for a reduction in solar activity to mitigate a small proportion of the future warming, a scenario we cannot totally rule out.

Might this area be worthy of more attention than something just not to be ruled out?

Mar 8, 2012 at 11:32 AM | Registered CommenterLord Beaverbrook

Lord B

Quantifying solar variations/forcings is one of life's mysteries to me. There appears to be many, many solar “jockeys” each with different mounts and sometimes in what seems to be different races!

Don’t know if you have seen this, apologies if you have, but I would appreciate any comments you may have TIA:-

The magnitude of the solar forcing, which varies about 0.25 W/m2 from solar minimum to solar maximum, is much smaller than the forcing by human-made greenhouse gases. However, the most relevant comparison of the solar forcing is with Earth's energy imbalance, 0.58±0.15 W/m2 (Hansen et al., 2011), because the combined effect of all forcings is less than that of greenhouse gases alone, and much of the greenhouse gas forcing has been "used up" in causing the warming of the past century. It is apparent that the solar forcing is not negligible in comparison with the net climate forcing.

It is under “Solar Cycle” here:-

My initial question amongst many – is the 0.25 W/m2 relevant if “the current prediction for Sunspot Cycle 24 gives a smoothed sunspot number maximum of about 59 in early 2013.”?

Mar 8, 2012 at 2:57 PM | Unregistered CommenterGreen Sand

Green Sand

Yes, I posted that over at Paul Hudsons blog. It seems that warming due to GHG, which is larger than all other forcing, has mostly been used up and the solar forcing, since the turn of the century, has become the dominant.
Or the warming was caused by GHG and the cooling by solar cycles.

Mar 8, 2012 at 7:59 PM | Registered CommenterLord Beaverbrook


Now we have yet more examples of historic trends beig altered.

"Hansen has been deleting cooling trends all over the planet. The image above compares current GISS 1950-1998 trends, with those he published in 1999. He has erased historical cooling trends (blue) in North America, South America, Africa, the Arctic, Antarctica, and elsewhere."

I have seriously lost count now. Are these sorts of adjustment standard practice which no-one ought to be concerned about?

I would like to think that someone at the Met Office does pay some regard to what is going on - perhaps they are even being routinely notified about the necessity for and extent of these adjustments?
The poiint being, of course, if we can't absolutely trust these adjustments, how do we know how rapidly the Arctic has been warming of late?

(Dare one hope we have got to the stage where these adjustments are auditable?)

Mar 8, 2012 at 9:16 PM | Registered Commentermatthu

Hi matthu

I don't think it is "tampering". Analysis methods do get updated. Probably best to read GISS's own information on this, which includes a link to Hansen's paper in Journal of Geophysical Research, and the programs they use in the analysis.



Mar 9, 2012 at 11:50 PM | Registered CommenterRichard Betts


Thank you for the link which has some interesting background material which I will continue to digest.

However, I think the point that is being made is the massive potential for confirmational bias to be a prime motivational force behind many of the adjustments being made. Consider the following two paragraphs from 2001 Hansen et al "A closer look at United States and global surface temperature change"

The U.S. annual (January-December) mean temperature is slightly warmer in 1934 than in 1998 in the GISS analysis (Plate 6). This contrasts with the USHCN data, which has 1998 as the warmest year in the century. In both cases the difference between 1934 and 1998 mean temperatures is a few hundredths of a degree. The main reason that 1998 is relatively cooler in the GISS analysis is its larger adjustment for urban warming. In comparing temperatures of years separated by 60 or 70 years the uncertainties in various adjustments (urban warming, station history adjustments, etc.) lead to an uncertainty of at least 0.1°C. Thus it is not possible to declare a record U.S. temperature with confidence until a result is obtained that exceeds the temperature of 1934 by more than 0.1°C.

Whether 1998 is a record year or not can only have importance in a political context, surely? Yet it appears that tremendous effort might be going into adjusting the data in order to meet a preconceived politcally correct result. Here is another paragraph from the same report:

The U.S. mean temperature in the current GISS analysis is about 0.2°C warmer in the past two decades than it was in the analysis of Hansen et al. [1999] (compare their Plate A2). This is because the warming introduced by inclusion of time-of-observation and station history adjustments is only partially balanced by cooling caused by a stronger urban adjustment.

So one should ask why it was important to introduce time-of-observation adjustments here and not anywhere else in the world? I am not reassured that adjustments are not being made with the prime object of delivering a politically acceptable result.

By 2010 I think I am right in concluding that 1998 had surpassed 1934 but I can't help thinking that if equivalent effort had gone into finding adjustments to raise the 1934 temperature there might have been a different result.

Are you totally unconcerned about this possibility?

Mar 10, 2012 at 7:34 AM | Registered Commentermatthu