All quiet on the climate front

The personal loans suppose to be important for people, which want to organize their own organization. As a fact, this is not hard to get a term loan.

According to my exploration, thousands of persons all over the world get the loan from different creditors. Thus, there's great possibilities to find a financial loan in any country.

That is good that we are able to take the loans and it opens up new chances.

William,
My calculations assume the IPCC projections don't include natural variability. Natural variabiilty would widen the bars compared to what the IPCC showed. The weather data do include natural variability, but the the central tendecy for the trend, and it's uncertainty don't.

So, I am comparing like to like.

[The central trend includes natural variability, of course. How could it not? There is less of it, if you use 10 years, but its still there -W]

William:
Could you elaborate on the precise mistake you believe I made?

I computed the best fit central tendency for the underlying climate trend in temperature over time. I computed uncertainty intervals for that central tendency in the climate trend.

I used a standard technique for differentiating between the uncertainty in individual data points (aka weather) and the uncertainty in the trend (aka climate).

As far as I can see, we both agree the IPPC graphic does not show variability due to "weather noise".

Though you suggest I incorrectly assumed the uncertainty intervals in IPCC graphic include the weather noise, that is simply untrue. I assumed the uncertainty intervals in the IPPC graphic do not include weather noise.

So, my analysis is based on the assumption you say is correct.

I can perfectly easily discuss the issue of weather noise vs. climate noise sometime next week. But if you will clarify what error you believe you detect, that would help me address the specific issue that currently concerns you.

[Yes. I think your mistake is in comparing the trend since 2001, which inevitably includes weather noise, with the IPCC projections, which don't. We both agree that IPCC doesn't include the noise (that is, the IPCC projection ranges contain no allowance for weather noise). The disagreement therefore lies in you thinking that a trend since 2001 contains no weather noise? I am having a slightly hard time believing that you believe that! You've fitted a line through some data points. That reduces the noise, but (of course) doesn't remove it. How could it? -W]

Sorry, I forgot to answer your question about what happens if I use 10 years. I prefer to limit myself to one choice. However, it turns out David Stockwell happened to do the analysis for 10 years.

Using IPCC terminology, he finds, using a somewhat different technique, that, depending on the data set, the IPCC projection of 2C/century is rejected with a certainty ranging form "likely" to "very likely".

So, this result is not heavily influenced by the choice of year. Nichec Modeling.

I agree with you that, until such time as the temperature trend turns upward, people will point out that the IPCC projections predicted more warming than actually occurred. If the temperatures remain flat, or turn down, you will hear people point out the IPCC projections are much warmer than what happened.

So, this result is not heavily influenced by the choice of year.

Yes it is. If you choose 9 years, the trend is ~2C/century. If you choose 10 years, the trend is ~0C/century. Just because you cherry picked two time intervals when the trends agree doesn't mean they agree for all time intervals in between.

Atmoz:
I didn't mean to suggest the magnitude of the trend doesn't jump around. Obvioulsy it does.

However, if we ask:

"Are the empirical data consistent with the IPCC projections?"
That answer is not particulary sensitive to the number of years selected.

The two values for best fit trend you posted were 0C/century and -2C/century. Both of these fall inside 95% uncertainty intervals range I posted are -3.2C/century to 1.1C /century. It is notable that neither your two values, nor my uncertainty bands includes the IPCC projection of 2C/century.

"the IPCC projection of 2C/century"

It's probably worth emphasizing that a 2C century-end projection doesn't imply that each individual decade is expected to average 0.2C...

Those are tildes in my post, not minus signs. The current 9-year temperature trend [GISS, RSS, and UAH] is around +2C/century. This is well within the uncertainty bands of the IPCC projection of +2C/century.

Re recent results, I thought these should have gotten a lot more attention. Not that it's all a done deal by any means, but it does appear to involve an unusually large number of pieces of the puzzle fitting together. The implications of enhanced ocean circulation as warming proceeds seem rather large and ripe for discussion by someone who knows what they're talking about (hint).

Another recent big deal is the work (by lots of people) heading toward much higher SLR projection, but I suppose we can't call this surprising.

Arctic sea ice news: NSIDC and NASA have scheduled an internet press conference to discuss current ice conditions and implications for the coming season. It sounds as if anyone can listen to the audio webcast starting at noon EDT this Tuesday, plus there's a link for background information that will be made available that morning.

As far as I can tell conditions remain very mushy. Jeff Masters noted on his blog a few days ago: "For the first time on record, the edge of thin first-year ice has pushed beyond the North Pole. IF we get another relatively warm and sunny summer in the Arctic in 2008, we will likely see Arctic sea ice loss surpassing last year's astounding collapse."

NASA has just come out with the past winter numbers:
http://www.sciencedaily.com/releases/2008/03/080314175834.htm
Notables: coolest world average since 2001.
February was much warmer than January.
The record J-08 snow cover, is now over, snow cover is supposedly now slightly below normal.
I thought it notable, that although 07-08 seemed cold to us, it was rated as the 16th warmest!

I was previously making a prediction that the high NH snow cover would give us a negative SW forcing this year. Of course spring snowcover is much more important than winter, as insolation is much greater later in the season. Of course an analysis of SW snow forcing would require that we don't just look at the extent of the cover, where it is, how clear the weather, and what sort of terrain
have a major influence. Clearly snow covering desert or step will have a much greater effect, than snow in a heavily forested area. In any case, it looks like any anomalous cooling SW forcing is likely over.

William,

"I think your mistake is in comparing the trend since 2001, which inevitably includes weather noise, with the IPCC projections, which don't."

Some obvious questions are:

1) Why don't they?
2) What on earth do they include if not weather noise?
3) What would they look like if they did include weather noise?

Re (1) I think people could complain that it makes the IPCC projections look more certain than they are, at least in the short term. And they may have a point.

[Probably best to start with 2. Actually it would be best to start by reading the linked bit of IPCC, but I imagine thats too much trouble for anyone. What they are trying to include is uncertainty from the different GCMs. As to 1, because it varies according to the timescale you're looking on. 3, it would depend on how you drew it. If you started from 2001, then the point for 2002 would have to have a large error bar, 2002 rather less, and so on -W]

Ian says that 'It's probably worth emphasizing that a 2C century-end projection doesn't imply that each individual decade is expected to average 0.2C...' But the IPCC's projection was NOT for 2C/century, it was for an underlying trend of 0.2C/decade. This is from the SPM of AR4:

'For the next two decades, a warming of about 0.2C per decade is projected for a range of SRES emission scenarios. Even if the concentrations of all greenhouse and aerosols had been kept at year 2000 levels, a further warming of about 0.1C per decade would be expected' (p. 12).

[Again, in the absence of weather noise -W]

Thanks William. I had downloaded the IPCC link and read that page but to be honest it is very obscure to anyone not intimately acquainted with the models.

Am I right then in thinking that the uncertainty range in the graphs refers to uncertainties in the models themselves (outputs?). Is this the spread of results from different runs of the model? In other words, is this uncertainty about what the models actually predict, rather than how close they get to the obs?

[Yes, this is uncertainty in the climate sensitivity of the models. To make it worse, I think that graph (from memory) isn't the actual models themselves but a simpler model fitted to have the same sensitivity as the various different models -W]

To use an analogy, is this anywhere close: Suppose we had ten dice and 9 of them were fair and the 10th not. Suppose for the sake of argument that the bias in the 10th die also increases over time. The 9 dice represent weather noise. The 10th represents the forcings. Let's say we have a model of the bias in the 10th die but there is uncertainty in that. The graph is is then a prediction of the overall trend in all of the dice, but the only uncertainty that is shown relates to the prediction regarding the 10th one. To get the real uncertainty (if you want to compare to obs) you have to include the uncertainty from the other dice also.

[I wouldn't put it like that. We have a family of curves going into the future, of the nature of dT=CO2*f*t, for various values of f, where t is time. IPCC is graphing that. Missing from that is the weather noise, which corresponds to the throw of the dice for each year -W]

We all know you cannot separate weather noise from a genuinely new trend so we are in guesswork land. A flattening trend is plausible bearing in mind the extent of the uncertainties. But I have to wonder about just how long this supposed long-term trend is. According to some scientists (eg Trenberth on a nature blogpost) it is now generally agreed that prior to 1980 or so, natural variability dominated. Is that now the general consensus view? If it is then the long term trend really isn't that long and the last 10 years of it represents a not-insignificant third of the total. For balance I hope you are just as cutting with those scientists who like to insist that "the cooling is masking the warming" so it's worse than we think.

[I'm surprised by what you say Trenberth says. Could you quote exactly? -W]

I wonder if the source of this Trenberth attribution is from Karl and Trenberth's (2003: p. 1720) "Modern global climate change" (Science, 2003, 302: 1719-1723):

Already it is estimated that the Earth's climate has exceeded the bounds of natural variability (2), and this has been the case since about 1980.

[Ah, maybe. Thats a bit different -W]

Yes.

Ian Castles said "But the IPCC's projection was NOT for 2C/century, it was for an underlying trend of 0.2C/decade."

I was quoting "2C/century" from Lucia's comment above, not from the IPCC.

But out of curiosity following your SPM quote, I looked at the IPCC reports. The part you quoted seems to be condensed from Chapter 10, section 10.7.1: "Climate Change Commitment to Year 2300 Based on AOGCMs" (begins on p. 822). While talking about committed warming and SRESs, it eventually gets to:

"The committed warming trend values show a rate of warming averaged over the first two decades of the 21st century of about 0.1°C per decade, due mainly to the slow response of the oceans. About twice as much warming (0.2°C per decade) would be expected if emissions are within the range of the SRES scenarios."

This 2 decade bit shows up in the middle of talking about long-term trends, and to me it sounds more careful ("trend ... averaged over" etc.) than the SPM quote. Given the special 20-yr GCM runs for the report, I'd be inclined toward a more charitable interpretation of the SPM quote; that is, its language can't be simultaneously brief and heavily qualified. Unfortunate wording in the SPM, perhaps. But also not the single, overarching projection of the IPCC.

In any case, this is a side point. As William points out, the IPCC comparison analysis looks only at recent temp measurements, while turning a blind eye to weather noise comparisons, and to the physics & chem of GW. In fact, we know enough about the physics to predict warming even if we didn't have any temp measurements of the recent past.

Atmoz-- Yes. You're right about the trend at 9 years, I don't hunt around for maxes or mins. In fact, I advocate against this. I also do this on excel and to dealing with autocorrelation is a bit clumsy. Also, I strongly prefer to test predictions excluding data that were available before the predictions were made. And including 1999 data violates that rule.

Anyway, William had asked specifically about 10 years. When I posted, I'd run the year with the criterion I picked, and then looked at David's numbers. David had happened to pick 10 years. The 10 year makes IPCC look not good. The 11 year trend-- not so good.

But yes, the 9 year trend, associated with selecting the relative minimum temperature after whatever the heck happened in 1998 is positive.

Taking monthly data I've been using, I pulled out data to get the exactly 9 years. Correcting for serial autocorrelation, I get 95% confidence interval for the underlying trend of

-0.2C/century < m < 3.9 C/century, with a central tendency of 2.1 C/century.

So yes, right now, if we eyeball the graph and find the strongest possible positive trend, -- which includes data that were "in the pocket" before the IPCC made its projection, the IPCC projections aren't falsified. (Frankly, I never expected them to falsify if we tested their predictions against data available before they made their predictions! )

However, notice that even permitting the IPCC to include 'past" weather when testing their predictions, the uncertainty intervals include a zero trend.

Also, the 95% confidence intervals for the 9 year data set, overlap uncertainty intervals if we start in 2001. (Basically, the lower half of the 9 year trend overlaps the upper half of the 'post 2001 band. )

In the case of the 9 year trend, and the "since Jan 2001" trend, the enormous span of the uncertainty bands is due to weather noise.

Weather noise (or any noise) is the main function of noise in a statistical test it to make it take more data to falsify or truly validate a hypothesis. (Truly validating requires failing to falsify with a test with sufficient power to be meaningful. So, you want the "beta" error to be pretty much as low as you insited for "alpha" error to falsify.)

So, when the power of a test is weak, failing to falsify means almost nothing. In contrast, it is quite notable that, with such a short amount of data, the recent data collected since the time the IPCC made its projections, don't contain the IPCC projections.

This ought to be a very rare thing.

Since it appears lots of people want to see what happens to the "falsification" or "validation" as a function of number of years, tomorrow I'll write an EXCEL macro so I can make some graphs so people can see precisely where the top and bottom of the uncertainty intervals land if we calculate back from "now".

But, William, your weather/ climate idea is incorrect.

By this method, the "trend" is supposed to detect the underlying climate trends that are consistent with a particular realization of weather. The best fit trend is the most likely climate trend based on the weather that actually occurred. The confidence intervals span the possible climate trends that are consistent with the weather that actually occurred.

[As far as I can tell, you still haven't understood. The IPCC trends don't include weather noise. Its not at all surprising tha there would be periods when a short-term trend is negative. Its somewhat surprising that a 10-year trend would be negative.

Your "best fit trend" is... a best fit trend to the last 10 years. Thats all it is. Its only the most likely climate trend if you throw away all other pieces of information, which would be unwise. You appear to be attempting to do this as a pure stats exercise with no physical understanding, which is not a good idea -W]

So, both weather and climate are involved. And everything you state here about what the IPCC means by their chart is precisely what I thought they meant: it doesn't include weather.

FWIW, this method is exactly the one used to decree that 0C/century is not consistent with previous realizations of weather!

This is a very standard way to test the IPCC projections: The question is: will the trend I found, applying the criteria for selecting start dates I selected, stick? If the IPCC are correct in their projection, the trend won't stick. This event, which should be very rare will reverse when the weather changes.

But otherwise, until, and unless the flat-ish spell breaks, the IPCC predictions are looking suspiciously like they overpredict.

Of course you will hear people say this because that's the way science works. After someone makes a prediction, one tests to see if it matches the data.

If the predictions start to look like they are failing, you are going to hear about it.

Lucia,
To my mind, your analysis demonstrates that 'until and unless the flat-ish trend breaks, the IPCC predictions are looking suspiciously like they overpredict.' This is an important conclusion.

However, Roger Jones, an IPCC Coordinating Lead Author from Australia's CSIRO, argues that the IPCC predictions under-predict. In a comment incorporated in a compilation which is to be discussed under Agenda Item 4 at the Panel's Plenary Session at Budapest on 9-10 April, Roger argues that 'the AR4 is not very well suited to decision-making' because misleading projections were made 'for the early 21st century especially (e.g., 0.2C per decade for the next few decades - we are already above this rate and accelerating)' ( 'Future of the IPCC', p. 106; the same comment from Roger Jones is also reproduced on pps. 92-93).

If there's flat in the 15 years trend, then that would be something indeed.
http://scienceblogs.com/stoat/upload/2007/05/5-year-trends.png
(The bottom one)

[I'd forgotten about that. See-also http://scienceblogs.com/stoat/2007/05/the_significance_of_5_year_tre.php -W]

So what was the weasel wording in Trenberth's words that I missed? It seems straightforward to me. It's amazing that you can never tie anyone down in climate science. People keep shifting their story:

Loop While (1):
A: Most solar physicists say the sun's activity can account for most of 20th century warming.
B: Ah yes but not those last 30 years.
A: 70 years out of 100 is still better correlation than for CO2.
B: Not if you include those aerosols.
A: But the aerosol parameters were obviously picked to match the climate record.
B: Um...
A: Can we compromise on mainly natural variability up to 1980 and thereafter mainly AGW.
B: Sounds fine to me.
A: Do you think temperatures are plateauing - it's been 10 years now.
B: It's just noise in a 100 year AGW trend but give it a few years and it'll rocket upwards.
End Loop:

Some manage to end the loop by the following "pulled it out of my arse" gambits:
HadleymanB: It natural variation but give it a few years and it'll rocket upwards.
DimmingmanB: Aerosol cooling masks the warming but give it a few years and it'll rocket upwards.

[I don't think your first (A) is true -W]

Well, given that reply, one wonders why bother to answer. However ...

My take (FWIW) is that if this supposed (Karl and) Trenberth claim on a nature blog is a repetition or some bastardisation of the line from the Science paper noted above (which you still have not confirmed whether 'tis or 'tisn't), then what I think they were saying is that climate varies naturally over time scales from decades to centuries, and within this natural variation there are extremes. Directly attributing things like temperature changes or increased storminess to human activities is complicated. However, those natural variations have been assessed and observations have shown that climate change and its spatial pattern extended beyond the bounds of this natural climate variability in the 1980s. It doesn't mean that anthropogenic climate change didn't exist before the 1980s (nor that it could not necessarily be attributed), it's just that it was more difficult to discern with confidence because it was within the bounds of natural climate variability.

If Karl and Trenberth meant anything different to this assessment, then W. will no doubt be along shortly to enlighten us both.

[You've actually sourced & quoted exactly, so we can analyse yours. The idea that T change rose out of the level of variability at about 1980 is a common one, and I think that is what Trenberth is saying. So that quote says nothing at all about the auses of pre-1980 change, for which you need a more subtle analysis, which has subsequently been done -W]

Tamino is a statistician specializing in time series analysis. I know enough Fischer-Pearson style statistics to vouch for the quality of his work. Here are links to four of his threads discussing temperature trends:

http://tamino.wordpress.com/2007/12/16/wiggles/

http://tamino.wordpress.com/2007/08/31/garbage-is-forever/

http://tamino.wordpress.com/2008/01/09/dead-heat/

http://tamino.wordpress.com/2008/01/24/giss-ncdc-hadcru/

P. Lewis
Yes he uses that phrase in more than one place but I couldn't remember the exact wording. Anyway it says to me that you shouldn't include pre-1980 values in any purely AGW warming trend calculations. The Science paper references the IPCC report "Climate change: The scientific basis" from 2001 where Karl was a lead author so that makes it semi-official. I cannot however find that phrase in the report. In any event the plateau period started in 2001 (or even warmish 1998 if you average it with coolish 1999/2000) so maybe they regret writing that now.

It's frustrating when scientists choose to interpret the same data in different ways to argue whatever they like. For example:
Michael Mann et al. "On Past Temperatures and Anomalous late-20th Century Warmth"
"As it is only the past few decades during which Northern Hemisphere temperatures have exceeded the bounds of natural variability, any analysis (SB03) that considers simply '20th century' mean conditions, or interprets past temperatures using the evidence from proxy indicators not capable of resolving decadal-timescale trends, can provide only very limited insight into whether or not recent warming is anomalous in a long-term and large-scale context."

So the same phraseology was used to debunk the Soon and Baliunas (SB03) solar theories. As I indicated with my gentle satire above, it's common to use the "from 1980" argument to dismiss the solar connection but then use the "entire 20th century" argument to dismiss the recent downturn as "noise". Tricky yes, ethical no!

James G.: "Most solar physicists say the sun's activity can account for most of 20th century warming." Wrong.

William--
It would be nice if your graphs related to a period similar to the one we just have. Then we could count and compare statistics.

Can you find a 30 year trend with no major volcano eruptions during the period, the most recent one more than 5 years prior to start year, upward trend of 2 C/century. Then show that 30 year period has as many flat spots?

As we all know, the variance, and serial correlation of residual errors of the 5 year trends will be much larger if a period with regular volcano eruptions is compared to one with no major volcano eruptions. The tendency is to see large 'downtrend' relative to the mean after the volcanic eruption and large 'uptrend' afterwards. And of course, the autocorrelation in the serial errors results in even more dramatic excursions that one might expect based only on the elevated variance.

The result, your 30 year period would be expected to have much, much more frequent flat spots than a period similar to the one we just had.

We know this based on both:
a)known properties of random distributions is that
b) the effect of volcanic eruptions postulated by climate modelers.

[I'm afraid I don't have convenient access to the appropriate tools any more. But we're straying from the original point -W]

Analogy: roll a fair dice a large number of times. Sometimes you'll get the sequence 6-5-4-3-2-1. Does that mean the next number is zero? Well, of course not.

Bad analogy. Better: hear for years from the doomsayers that global warming is out of control and wicked, wicked humanity is the cause. Then observe nearly a decade of little or no temperature increase. Does that mean the doomsayers are full of sh*t? Well, of course.

[Sigh. You can take any 10-year time series and it will never tell you anything definitive about the underlying trends (well, alright, thats not quite true: it can put some bounds on the underlying trends. But they are unlikely to be narrow enough). People knew this in the 1970's... -W]

Re the cherrypicking discussed above, what Lucia has done is define it in a certain way and justified her analysis as not cherrypicking because it doesn't fit that precise definition. This is only an attempt to trick the unwary, who having only ever seen a Rainier might be convinced that a Bing is actually a plum. Waiting for natural variability to produce an end-point suitable to support a short-term "cooling" analysis isn't too supportable to begin with, but doing so while failing to note the lack of such analyses at times when opposite conclusions would be drawn from the short-term trend brings me face to face with the fact that I only know the names of two varieties of cherry. Beyond that, failing to examine other similar short-term flattish spots in the record of the last thirty years or so makes me realize that there's something going on here that mere reference to stone fruit cannot encompass.

I think we've got a burgeoning Climate Audit on our hands.

I had managed to forget about this similar and rather more authoritative analysis until John Cross blogged about it just now. Adding 2007 wouldn't change much.

William-

Lucia is saying that recent trends are inconsistent with the IPCC prediction. You are saying that they are in fact consistent. She has showed her work. You have showed a little blog attitude and nothing more.

If you want to teach us something, then please just show the quantitative basis behind your claims that recent trends are consistent with IPCC predictions, rather than just asserting that claim without evidence. I'd guess you won't, but if you do we'll highlight your work on our site.

[I've corrected one misstatement of mine. I didn't mean to assert that recent trends *are* consistent: only that Lucias analysis, which is missing a key variable, doesn't show inconsistency. I fully agree that I haven't done the opposite analysis. I'm not even sure how one would.

The spread on the IPCC graph is based on climate sensitivity uncertainty generated from a range of GCMs. The spread in Lucias is based on weather noise from one realisation of climate. Do you understand that these are different, and don't directly compare? -W]

See my previous, Roger. That should make you happy.

William. I have developed hundreds of models, including models for climate (not earths though).

When you have an almost perfect validation of a model, done with independent data, I suspiciously trust it.

When you have not validated your model, you certainly do not defend it when contradicted by real world data. You do contemplate what might be wrong.

It is much to difficult to make models for you to have such confidence in them. Mathematics is all fun, but the link between your set of partial differential equations to reality needs to be established. The IPCC have not done it nor anyone else, time is due.

[I think you're missing the point. You do realise that the IPCC figure being used isn't directly comparable to real-world data, don't you? -W]

Where has this "more subtle analysis" been done and why doesn't Trenberth know about it? Shall I assume it isn't settled science yet and if not then the 1980 number, being "common" is still the best start point?

If however you are talking about the famous Hadley modeling exercise which tried to show that natural variance alone couldn't account for 20th century temperatures then that was prior not subsequent and they recently admitted that they had not properly included for natural variance. They didn't admit that they faked the aerosol component too though that has since been concluded in several peer-review papers. It remains a great example of mere speculation being presented as clear evidence.

By the way, you can calculate a nice rising trend from the sunspots from 1900. It has a little deviating "noise" at the end but then it's the century long trend that's important isn't it?

William, I was surprised to see your assertion at Prometheus that 'the point that ... Lucia seem[s] to have missed is that the projections in the IPCC graph don't include uncertainty due to weather noise.' Far from missing this point, Lucia states explicitly that 'The IPCC climate projections don't include weather noise' and that 'That's what makes the appropriate basis for testing!'. If you really believe that her analysis is 'missing a key variable', you need to do better than this.

[If Lucia knows that the IPCC graph is missing weather noise, why is she comparing it against weather noise? It makes no sense -W]

She's not comparing the IPCC climate predictions with the weather noise, William. She's comparing them with the CLIMATE trends that are estimated to be consistent with the observed weather.

[She is comparing something with no weather noise (IPCC) to something with weather noise (her trends). Or do you think some magic has removed that noise? -W]

William-

Given a time series of global temperature of length N, please explain exactly how you would calculate a global temperature trend for this time series.

[I'd use least squares. You're missing the point, again -W]

Calculating the trend in global temperature isn't the issue - the trend calculations can be done in a variety of ways for a variety of time periods. The sticky point is how to compare the calculated trend to the IPCC projections.

[Yes -W]

William-

OK, once you have your trend calculated with least squares, please tell how then you would then compare that with the trend presented by the IPCC. Please be specific.

[Ah, finally the point: its not at all clear how you do compare this to the IPCC trend. Thats what I've been saying all along. As a rough guess, you need to add appropriate weather-noise-trend-uncertainty to the IPCC line. That would be about the same as Lucias trend range. If the two ranges then overlap, they are consistent (at that level of stat sig) -W]

William-

This is a lot of effort for you to show that your complaint is already covered in comments Lucia (and I) have already provided. Here is what I actually said in my post:

"Those wanting to quibble with her analysis would no doubt observe that the uncertainty around IPCC predictions for the short term is undoubtedly larger that then IPCC itself presented. Lucia in fact suggests this in her analysis, making one wonder if uncertainties are indeed larger than presented, why didn't the IPCC say so?"

Lucia kindly put up another post explaining in detail the limitations of what she has done and why the critiques, yours in particular, are simply wrong.

http://rankexploits.com/musings/2008/inadequate-reasons-to-for-suggesti…

Your complaint is in fact with the IPCC. There is certainly some calculable probability of the consistency among the uncertainties in IPCC modeled and subsequent observed trends.

Consider that if we simply assume that uncertainty in the IPCC modeled trend is the same size as the uncertainty bounds calculated by Lucia for the observed trend (as you have suggested), then given that the modeled trend does not fall within the 95% CI of the observations then the overlap cannot be more than about 25% (see the figure in the link above to understand why this must be so). So even under your own assumptions, it seem clear that the probabilities fall in favor of observed and modeled trends being inconsistent.

Is this for sure? No.
Does it falsify global warming? No.
It is worth asking? Yes.

When you turned in your climate scientist credentials, I sure hope you didn't turn in your willingness to ask and answer interesting questions, in favor of tossing snarks.

[Hi Roger. Please save the gratuitious impoliteness for your own blog. Lucia makes any number of mistakes. Her first 2 points are strawmen - I haven't put my 5-y trends forward as relevant to this issue. They were done ages ago for another reason.

As to weather noise, Lucia says "This trend is a straight line and does not include "weather noise"". Lord knows what she means by that. The value of the trend is of course affected by weather noise.

Lucias estimate of the trend is m, +/- some error which contains weather noise. The IPCC's estimate is missing this component, so you need to add it to the IPCC uncertainty, as I said last message. You seem to be agreeing with this. So our rough-and-ready test for inconsistency is that her range, and the IPCC range plus her estimate of weather noise, overlap. They do indeed overlap, according to your guess above. So, we don't have inconsistency demonstrated.

This is what I've been saying all along.

My complaint isn't with the IPCC, because their figure isn't intended to be used for this -W]

Roger,

"Consider that if we simply assume that uncertainty in the IPCC modeled trend is the same size as the uncertainty bounds calculated by Lucia for the observed trend (as you have suggested), then given that the modeled trend does not fall within the 95% CI of the observations then the overlap cannot be more than about 25% (see the figure in the link above to understand why this must be so). So even under your own assumptions, it seem clear that the probabilities fall in favor of observed and modeled trends being inconsistent."

The figure in the link above is a hypothetical, so it in no way shows that the overlap cannot be more than 25%. The degree of overlap matters and has implications for saying that the IPCC "prediction" has been falsified with 95% confidence (which is what Lucia does say).

Frank-

Yes, it is a hypothetical. See if you can draw two distributions, in which the central point of one is outside the 95% bound of the other, and show an overlap of >25%.

I have not read Lucia saying what you have attributed to her. She did say the following:

"The central tendency of the IPCC projection, m=2.0C/century, falls outside the 95% uncertainty intervals for trends estimed based on data collected since 2001."

William-

You complaint about impoliteness is quite funny given how you started off this conversation with the snarky "pointless statistics without physical understanding."

Your suggestion that two probability distributions need to have no overlap in order to be inconsistent is just not how statistics works. Distributions of forecasts and observations always overlap (with rare exceptions).

If your complaint is that Lucia has not shown this overlap to be significant at the 5% level given uncertainty in the IPCC prediction, then you should just say so. Of course she can't do this because the IPCC doesn't provide the relevant data. What she has done is provide a compelling analysis indicating that observations are inconsistent with IPCC predictions at a relatively high probability (we can quibble what it is, but it seems that we all agree that it is >75%).

So much for "pointless statistics without physical understanding."

William-

You complaint about impoliteness is quite funny given how you started off this conversation with the snarky "pointless statistics without physical understanding."

Your suggestion that two probability distributions need to have no overlap in order to be inconsistent is just not how statistics works. Distributions of forecasts and observations always overlap (with rare exceptions).

If your complaint is that Lucia has not shown this overlap to be significant at the 5% level given uncertainty in the IPCC prediction, then you should just say so. Of course she can't do this because the IPCC doesn't provide the relevant data. What she has done is provide a compelling analysis indicating that observations are inconsistent with IPCC predictions at a relatively high probability (we can quibble what it is, but it seems that we all agree that it is >75%).

So much for "pointless statistics without physical understanding."

William-

You complaint about impoliteness is quite funny given how you started off this conversation with the snarky "pointless statistics without physical understanding."

Your suggestion that two probability distributions need to have no overlap in order to be inconsistent with one another is just not how inferential statistics works. Distributions of forecasts and observations always overlap (with rare exceptions) -- the issue is always the size of the overlap.

If your complaint is that Lucia has not shown this overlap to be significant at the 5% level given uncertainty in the IPCC prediction, then you should just say so. Of course she can't do this because the IPCC doesn't provide the relevant data. What she has done is provide a compelling analysis indicating that observations are inconsistent with IPCC predictions at a relatively high probability (we can quibble what it is, but it seems that we all agree that it is >75%).

So much for "pointless statistics without physical understanding."

William: [She is comparing something with no weather noise (IPCC) to something with weather noise (her trends). Or do you think some magic has removed that noise? -W]

Comparing noisy data to predictions of underlying trends stripped of noise is done all the time. It is hardly as if weather is the only system that includes trends an noise. So do manufacturing processes, turbulent mixing problems, econometric problems, and even traffic modeling.

So techniques to do this are standard. The technique is not to graph a graph of the projections and splatter dots on the graph and eyeball the graph.

Rather, techniques suited to testing simpler problems are taught in undergraduate statistics. These include t-tests, and linear regression.

If it weren't for the relatively long integral time scale of weather compared to what we are trying to test, this test would have been amenable to the standard OLS treatment used in undergraduate lab course everywhere! To deal with the autocorrelation, I had to look up the modification.

But the fact is comparing noisy data to noise-free predictions is exactly what the regression is used for: We compare a realization with noise to a prediction with no noise!

I don't see why this concept of testing a prediction for a central tendency (with error bars) using noisy data, is bothering you.

When this is done, the IPCC graphic fails.

As I see it: The problem with the IPCC's graphic is that it either:

a) underestimates the influence of natural weather variability such as ENSO or PDO on climate trends or
b) over-estimates the rate of warming.

Roger Pielke Sr. suggested (a) some time ago, and you criticized him. The remedy for (a) would have been easy to implement. All the IPCC would have had to do was incorporate the uncertainty due to ENSO and PDO or anything else into their estimate of the uncertainty in the climate trend.

The more appropriate graphic, reflecting the would be qualitatively similar to the graphics provided. There would be no "wigglyness" or "spaghetti lines" to show weather.

The appropriate graphic would still come to a point at time=0, but the angle near time=0 would be less acute. Heck, for all I know it would have included 0C/century near zero.

But, as it stands, because they appear not to have included the uncertainty due to known phenomena like ENSO and PDO, so the projections falsify to a 95% confidence intervals using standard statistical techniques designed to compare noisy data to noise-free trend predictions.

This falsification is not due to my mis-interpretation of their graph. It is due to their failure to publish uncertainty intervals that reflected the true uncertainty in climate trends.

[Hi Lucia. I thought we were talking to Pielke Jr, not Sr. I think its probably Jr, as Sr is unfailingly polite.

You say one poss is that IPCC "underestimates the influence of natural weather variability such as ENSO or PDO on climate trends". Have you still not understood that the IPCC graph you're using contains no allowance for this at all? It doesn't underestimate it - it simply ignores it entirely -W]

William- Sorry for the duplicates. TypeKey gave me an error, we're familiar with that on our site. Please delete the duplicates, thanks.

[As far as I can tell, I can't. They are property of your blog -W]

Roger,

Point taken, I thought you were referring to the degree of overlap of the error bars.

Re Lucia's statement, it appears in the comments of the post you linked:

This falsification says that given the length of the trend and the weather data that actually materialized after the projection was made, the IPCC projections are falsified to a 95% confidence level. That means: Given the weather we have had since the IPCC made the projections, there is a 5% chance their projections for the climate trends are correct.

William: [She is comparing something with no weather noise (IPCC) to something with weather noise (her trends). Or do you think some magic has removed that noise? -W]

Comparing noisy data to predictions of underlying trends stripped of noise is done all the time. It is hardly as if weather is the only system that includes trends an noise. So do manufacturing processes, turbulent mixing problems, econometric problems, and even traffic modeling.

So techniques to do this are standard. The technique is not to graph a graph of the projections and splatter dots on the graph and eyeball the graph.

Rather, techniques suited to testing simpler problems are taught in undergraduate statistics. These include t-tests, and linear regression.

If it weren't for the relatively long integral time scale of weather compared to what we are trying to test, this test would have been amenable to the standard OLS treatment used in undergraduate lab course everywhere! To deal with the autocorrelation, I had to look up the modification.

But the fact is comparing noisy data to noise-free predictions is exactly what the regression is used for: We compare a realization with noise to a prediction with no noise!

I don't see why this concept of testing a prediction for a central tendency (with error bars) using noisy data, is bothering you.

When this is done, the IPCC graphic fails.

As I see it: The problem with the IPCC's graphic is that it either:

a) underestimates the influence of natural weather variability such as ENSO or PDO on climate trends or
b) over-estimates the rate of warming.

Roger Pielke Sr. suggested (a) some time ago, and you criticized him. The remedy for (a) would have been easy to implement. All the IPCC would have had to do was incorporate the uncertainty due to ENSO and PDO or anything else into their estimate of the uncertainty in the climate trend.

The more appropriate graphic, reflecting the would be qualitatively similar to the graphics provided. There would be no "wigglyness" or "spaghetti lines" to show weather.

The appropriate graphic would still come to a point at time=0, but the angle near time=0 would be less acute. Heck, for all I know it would have included 0C/century near zero.

But, as it stands, because they appear not to have included the uncertainty due to known phenomena like ENSO and PDO, so the projections falsify to a 95% confidence intervals using standard statistical techniques designed to compare noisy data to noise-free trend predictions.

This falsification is not due to my mis-interpretation of their graph. It is due to their failure to publish uncertainty intervals that reflected the true uncertainty in climate trends.

William says that the IPCC graph that Lucia used, which is from section 10.5 of the AR4, contains no allowance at all for internal climate variability: 'It doesn't underestimate it - it simply ignores it entirely.'

The graphic certainly doesn't IGNORE internal climate variability. The title of the section in which it appears is 'Quantifying the range of climate change projections'. The authors state at the outset that 'The effects of internal variability can be quantified by running models many times from different initial conditions, provided that simulated variability is consistent with observations' (p. 797). They explain that 'The use of ensembles of AOGCMs developed at different modelling centres has become established in climate prediction/projection on both seasonal-to-interannual and centennial times series' (p. 805).

So, as Lucia says, the graphic either UNDERSTATES natural climate variability or overestimates global warming.

[I'm not at all convinced that the graph contains allowance for int var. Reading theheading of the section isn't enough. You'd need to read up on the graph itself. My recollection is that it doesn't, but I'm willing to be corrected. If no-one else will read it carefully then I guess I'll hav to myself. Where is JA when you need him? -W]

This is the text from WGI that describes lucia's graphic:

The dark shaded areas in the bottom temperature panel represent the mean ±1 standard deviation for the 19 model tunings. The lighter shaded areas depict the change in this uncertainty range, if carbon cycle feedbacks are assumed to be lower or higher than in the medium setting.

[Yep, that seems right. So it doesn't underestimate the weather noise; it ignores it entirely, except insofar as its present in the model fitting, which is not a lot -W]

The problem with the IPCC's graphic is that it either:

a) underestimates the influence of natural weather variability such as ENSO or PDO on climate trends

The answer is obviously a.

Roger,

"If your complaint is that Lucia has not shown this overlap to be significant at the 5% level given uncertainty in the IPCC prediction, then you should just say so. Of course she can't do this because the IPCC doesn't provide the relevant data"

The fact she doesn't have it doesn't mean it isn't important. Besides, Lucia's claim is that she doesn't need it so she seems to think the degree overlap is not even relevant.

"What she has done is provide a compelling analysis indicating that observations are inconsistent with IPCC predictions at a relatively high probability (we can quibble what it is, but it seems that we all agree that it is >75%)."

It is only >75% if we must go on these stats alone (which is still not 95%). In other words her estimate is only the most likely trend if the data on which it is based is all we know. But it isn't. In fact we have observations prior to 2001, so not only do we have several other models which are based on physics that tell a different story, we have other statistics models based on more data that tell a different story too. So there is reason to consider that her best fit of the slope is not the most likely trend in the real world.

Including this information would move the likely trend up, or put it another way would widen the error bars on her estimate. Quantifying that is another matter of course.

I said: "It is much to difficult to make models for you to have such confidence in them. Mathematics is all fun, but the link between your set of partial differential equations to reality needs to be established. The IPCC have not done it nor anyone else, time is due."

and you replyed: [I think you're missing the point. You do realise that the IPCC figure being used isn't directly comparable to real-world data, don't you? -W]

First of all that appear to be wrong, the trend IS extracted from the weather noice and is compared to the forecast.

Second, my point was that you are not humble enough with regards to the models in question. Since they are not validated at all they would not be allowed to be used for any regulated engineering practise for example.

Your words "pointless statistics without physical understanding" shows that you do not at all have any experience with establishing the physical connection to a model (or possibly that you are careless with your words). The IPCC has not done enough of the "pointless statistics" to even begin to validate their models.

What regards the "physical understanding" anyone skilled in the art of modelling knows that theory has its boundary. What for example, would be your equation for describing plancton reactions to decreased low level cloudiness? And how do you know if it matters in your model or not?

Your faith in the models are not warrented by their qualities.

[You are reading things into my post that aren't there. I haven't expressed any faith in the models in this post. All I've done is point out that Lucias cnclusions aren't justified by her analysis. As judged by her most recent comments, she still hasn't realised that the IPCC figure doesn't include any weather noise.

As for the IPCC not having done the stats... I think that shows you haven't read the report. Do. There is lots of stats in there -W]

I should be more clear. The graph "understates" internal variability because it does not consider internal variability at all.

Why do you folks think this uncertainty should be included? Especially in a graphic with the primary purpose of showing the difference in carbon cycle scenarios 30 years out and beyond. The graph's subject is radiative forcing. But would be very confusing and awkward to including internal variation on a graph that is trying to show differences in radiative forcing decades form now.

[I agree. The graph in the IPCC report is there for one reason. Lucia is trying to use it for another. You can't do it in the way she does -W]

There are ways of doing this if you have some idea of what the variability aka noise is using a Monte Carlo simulation. See, for example, W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C: The Art of Scientific Computing, 2nd Ed. (Cambridge University Press, Cambridge, 1992) pp. 656-706.

This will produce a best estimate for the model as well as an estimate of the variance.

William, I am talking about validation of the models. I have read the IPCC reports and they are all completely missing model validation. Lucia have done one small but necessary part of a model validation and the models failed it. You claim this doesn't matter because you think wrongly that Lucia is comparing apples and oranges. But even if I would accept your logic you still would be wrong to say that it doesn't matter. It does matter as a failure in any one aspect of a model leads you to reject the model, even if it would be different trendlines within natural variability. The IPCC models have already failed many tests and here is another one where they also failed. The models should have been rejected years ago. This of course does not refute AGW as such, but modelers need to get back to work.

[Its hard to take you seriously when you say that "I have read the IPCC reports and they are all completely missing model validation" -W]

AR4 WG1 Chapter 8 would seem a good place to start reading.

Of course you could always show two flat lines with a step change in the middle.

It's hard to take any computer modeler seriously when they don't understand that hind-casting is not model validation. Comparison of predictions with reality is validation.

"You do realise that the IPCC figure being used isn't directly comparable to real-world data, don't you?"

Imagine for a moment that the IPCC "projections" had actually matched reality and it was a skeptic who wrote the above comment! Somehow tells me he would be scoffed at.

If you think that AR4 WG1 Chapter 8 contains any validation attempts, please point them out. I could not find any.

Just to give a hint of the effort I would make to try to kill a model before investing anything in the outcome of it:

I would like to see quantitative analysis of residuals, Chi's for example.

I would like to see a test of sufficiency of the functional part in the form of scatter plots of residuals versus predictors on all important variables and their temporal and spatial distributions.

I would like to see examined the non-constant variation across the data, i.e. the scatter plots mentioned above and residuals against time

I would like to have examined the input data to have it scrutinized for correlation to data it is supposed not to be correlated to.

I would like to see tests for the independence and normality of errors

Unfurtunately cross-validation is not possible at this time as there is no way to know if a certain data set has affected the development of a certain model; however data published after the predictions may serve this purpose.

I also think that if a model fails one variable, it's prediction capabilities of all coupled variables should also be viewed as null.

But, one can also assess a model merely by looking at postdictions vs measured data. Looking at figure 1.1 in chapter 8 as above, it is appearant that models fail the warming from 1910 to 1944 and the subsequent cooling, until Agung, indicating that models do not incorporate true natural variability. Also, models give a much to high responses to volcanoes, indicating to high sensitivity.

The failings to validate is bad in itself. Attempts to downplay others attempts at validation is purely political. In this light, the IPCCs and its followers extremely high confidence in model output is incomprehensible.

AVK, you have not been looking. Google is your friend, and as lucia has shown, there are lots of naive and incorrect ways of approaching the issue

It seems people are ignoring a lot of issues like serial correlation. Weather "noise" is not strictly random; there are substantial decadal-scale variations due for example to ocean circulations. A period of 10 years is too short to say anything at all about the mean state of climate -- much less climate change! Witness the fact that WMO climate normal periods are 30 years and not 10 years.

Yes, Eli Rabett, there are indeed many naive and incorrect ways of approaching this issue. Take the IPCC, for example, which says in its latest Assessment Report (2007) that 'Six additional years of observations since the TAR ... show that temperatures are continuing to warm near the surface of the planet' (p. 683).

Eli Rabett,

so tell me: would you put your childrens health at risk because an unvalidated model tells you that is safer than not risking them? Because that is what the "unpolitical" IPCC tells us to do: put poor people in danger because of a potential risk. I would not, because I've seen thousands of models fail, even after I convinced myself that it was a good model.

Raymond Arrit, 10 years is a short period, but that is the only data we have to do cross-validation against. The fact that the IPCC models fail this test along with scores of others tells us the models are poor. (Unfurtunately it does not tell us wether there is a true risk for catastrophic climate change or not, but that is another issue.)

AvK, one has to weigh risks. There is no risk free choice, and I don't notice you beggering yourself for the poor, so put away the Croc tears.

As to how many years of data we have to check against the predictions of three dimensional models try 20. Moreover trying to pretend that the models only provide a single metric is dangerously naive. And yes Ian, the world continues to warm.

Eli, the six (now seven) years of observations since the TAR was published do NOT show that mean global surface temperatures were continuing to warm in that period. The truth of this statement does not depend on what happens in the rest of the 21st century, nor on the accuracy of the observations made so far.

Eli Rabett,

so at least you admit that your proposed actions hurt the poor. But you do lack humility to such an extent as that you think that you can weigh risks based on falsified models and the assumption that your proposed policy produces the intended consequences.

The fact that the models provide more than one metric is a fact that makes it easier to falsify. This has been done and as they stand they have been falsified, in spite of the IPCCs deliberated attempts at avoiding validation.

As for your snips, you don't have a clue as to the passion I feel for helping the poor, nor for creating a sustainable society. The fact that you are ready to make such assumptions provides an answer for why you are prepared to believe those utterly poor models: to you, being righteous is more important than doing right.

so at least you admit that your proposed actions hurt the poor.

The inability to understand simple written communication is, apparently, a defining characteristic of skeptics nowadays. Lucia has trouble with it and now AVK does too.

These folks are hell bent on falsification and want nothing to do with actually learning anything or contributing to the science. They sure can squawk, however. Gotta give 'em that.

"AvK, one has to weigh risks. There is no risk free choice, and I don't notice you beggering yourself for the poor, so put away the Croc tears."

It sounds pretty clear to me. Rabett weighs the risk between hurting the poor (through the economy) and hurting ourselves in the future if Hansen is right.

But perhaps I should have written "...MAY hurt the poor." That way, I would not have allowed Boris to focus the discussion on something else than validation.

By the way, Lucia is not a sceptic, according to her own words. She's just honest, a rare gift nowadays it seems.

[Lucia talks like a sceptic. Oddly enough, you seem to be judging honesty by whehter people agree with your position or not. That says more about you than them -W]

Reasonable people should not fight over whether the past seven years' temperature data present weather noise or a falsification of the IPCC's climate models. Instead, everybody should set some falsification criteria for their own positions for the future.

For example: I agree that, if global warming returns to the 2C per century trend by 20XX, the reason is obviously the buildup of anthropogenic greenhouse gases. And the other side: I agree that, if global temperatures remain flat or the trend becomes negative by 20XX, the reason is obviously natural and the IPCC's AGW-based scenarios have been wrong.

Could we agree on the year when this might be settled?

2100 seems to be a magical number.

Ian Castles --- You should read this:

http://www.skepticalscience.com/Comparing-IPCC-projections-to-observati…

based on a one page paper in last May's issue of "Science".

David Benson, Thanks for your suggestion that I read the recent 'Skeptical Science' posting on 'Comparing IPCC Scenarios to Observations', which draws on Rahmstoorf [sic] et al (2007) to argue, inter alia, that the IPCC's 2001 projections 'underestimated temperature rise with observations warmer than all projections.'

Unfortunately, the authors of Rahmstorf et al misunderstood the basis of the 'projections' for 1990-2000 in the TAR. As explained in the Special Report on Emissions Scenarios (2000), 'the 1990 and 2000 emissions scenarios were standardized in all the SRES scenarios, with emissions diverging after the year 2000' (Box 5.1, p. 243).

The SRES went on to explain that 'The standardized scenarios share the same values for emissions in both 1990 and 2000', and that 'The 1990 and 2000 emissions estimates for all gases, except SO2, were set to be equal to the initial values in the unadjusted four marker scenarios.' So, for example, the projections for emissions AND FOR TEMPERATURE for the A1FI scenario, which was not a marker, were set BY ASSUMPTION as equal to those of the average of the four markers.

As an indication of the possible scale of the effect of this procedure, the A1FI scenario assumed that the global energy supply from coal would increase by 30% between 1990 and 2000, compared with an increase of about 5% for the A1B and B1 scenarios.
.
So when Rahmstorf et al conclude that ALL OF the IPCC projections underestimated the temperature rise, they are not in fact evaluating the performance of models against observations - at least as far as the 1990-2000 period is concerned (the greater part of the comparison). In fact, the comparison is invalid in the context in which it was used in the article published in 'Science' on 4 May 2007.

The poor are us Av, so stop trying it on. You know, as in no man is an island, no man stands alone. Of course, you and your ilk like to think you can survive behind your moats.

The poor are gonna get hammered if even the optimistic parts of the IPCC WGII and III working groups are the path the planet follows. Moreover, we could help the poorer people on this planet and deal with man made climate change.

You remind me of Roger Bate. He sold the tobacco companies on a campaign to attack the World Health Organization over DDT to stop them from mounting an anti-smoking effort. Stop trying to offer your lying choices.

seems like the bottom line is if it cools its weather if it warms its part of a long term trend.

how long does the cooling has to go before we can call it a trend? 20yrs? 30yrs?

seems like the warming from 70's to 2000 is unquestionably a long term warming trend right? so 30 yrs is the magic number.

[People are trying to understand what is going on. The observations are one component of that. Theory and models is another. To understand, they have to be consistent. Theory and models say warming, with some possible weather noise. 20 years of cooling would be very hard to reconcile with that theory. But a theory predicting cooling isn't easy either -W]

William, we know the inputs have such large uncertainties that they can easily show cooling in a sensitivity analysis. These cooling results are only rejected on the basis that warming is expected; IOW circular reasoning. A theory predicting cooling is in fact quite easy, as follows;
a) Cooling aerosols are now counteracting the CO2 warming.
b) Indirect solar effects might be greater than we thought.

Scenario a) has, as you know, already been postulated by Rasool, Schneider and others and was rejected when the earth started warming up. A similar argument was even used by Hansen during the coolish year 2000.
Scenario b) has been rejected purely on the basis of the warming of the last 30 years of global temp data. However there are a few regional temperature plots which back-up a solar link (Arctic, US48, Ulster observatory) and Svensmark even finds two confirmatory global plots. Since the IPCC admits to poorly understanding the role of the sun, and multiple books and papers have been written about the sun-climate connection, it's not exactly a big leap in faith.

Reconciling Eli and Avk, of course there are good intentions behind the rush to judgment over CO2 but there are also many good reasons to think that CO2 cuts could do far more harm than good so it's important to be correct. I wish everyone would consider the recent "warming soot" idea instead: the science seems to support it and it's a lot cheaper to fix.

[For (a), the numbers don't work. R+S was wrong when published, see footnote 3, I think it was. Something weird would have to be going on for aerosols to be cooling more than CO2 is warming - not only don't the numbers work in theory, they wouldn't fit the T history etiher. People have looked at solar a lot, and it doesn't fit. If you make the indirect amplification too high, then apart from anything else you ought to see the 11y cycle much more clearly. At the moment its near invisible -W]

JamesG stated "... CO2 cuts could do far more harm than good ..." Not for the oceans. Acidification is bad for many marine organisms.

[Lucia talks like a sceptic. Oddly enough, you seem to be judging honesty by whehter people agree with your position or not. That says more about you than them -W]

-Lucia says she believes in AGW. But reported upon the results she got none the less. I did not include you in the bunch of dishonest people in climate science if you got that impression.

Eli Rabett, are you born with a stone in your chest?
"The poor are us Av, so stop trying it on. You know, as in no man is an island, no man stands alone. Of course, you and your ilk like to think you can survive behind your moats."

If you think that your tenure in a nice american university gives you any perspective on how poor people suffer today, here is my advice: buy a plane ticket, go to Ethiopia, watch the poor state of people. Speak to them and notice how they all tell you that they are much better off today than 15 or 20 years ago, thanks to ECONOMIC DEVELOPMENT and GLOBALISATION. Then tell them that you want to take from them their littles goodies like malaria pills, mosquito nets and water treatment because you think that flying flowers from the New Flower to Europe contributes to CO2 emissions. Watch how they are really impressed by your concern for the well-being of their offspring, and how greatful they will be that the few survivors without the mosquito nets and water treatment will really be so well off thanks to you.

"The poor are gonna get hammered if even the optimistic parts of the IPCC WGII and III working groups are the path the planet follows. Moreover, we could help the poorer people on this planet and deal with man made climate change."

-Yeah, like in the shape of government aid I presume? You know, the kind of aid that on average hurt the poor? That creates corruption both home and abroad? You know what? When I get my boots dirty traveling to Mota on a bumpy bus with the smelly but nice locals, I HAVE YET TO MEET A DARN LEFTIST US ACADEMIC ON HIGH HORSES. Because they are like you: prefer to talk instead of doing something.

"You remind me of Roger Bate. He sold the tobacco companies on a campaign to attack the World Health Organization over DDT to stop them from mounting an anti-smoking effort. Stop trying to offer your lying choices."

-You are possibly the worst person I have ever debated with. You're evil.

William,

What seems obvious is that co2 is not the dominate temp driver. If it were we would see a clear monotonic temp rise since CO2 has been rising consistently. Thus if the models don't track with the last ten years then they clearly don't contain forcings stronger than the co2 forcing.

Thus to rely on current models to forcast the effect of co2 on the temp for the next 50-100 yrs is just silly (as any high school science student can see).

[If you want to forecast the climate-weather for the next few years, then Smith et al is available, as Jmaes has pointd out. If you're interested in the next 50 years you need a different analysis, errm, the one that IPCC has provided. You cannot use the present trends to deduce that CO2 isn't the dominant driver on long timescales -W]

No Av, Eli simply doesn't take your cheap brand of nonsense. Evidently you never had your particular smokescreen blown back in your face.

Human caused climate change will hurt the poorest people worst. The world is rich enough to help the poorest right now and protect them and us from a coming disaster.

Occasional flotsam peeks out from behind the moat and tries to convince all that they should not be inconvenienced because horrors it might hurt the poor.

Mostly, those who try it on have a history of screwing the poor in every way possible, usually on the grounds it is a good thing for the poor to learn how to be screwed so they can stand on their own two legs. And then, of course, horrors, governments might do something.

Having grown up in a United States where people were suffering from all sorts of deficiency diseases, malnutrition and basic bad health do tell me young Av about those government programs that hurt the poor like the aid offered by food stamps, like the education programs that brought the WWII generation out of the Depression and more. You and your ilk were just standing there doing nothing when this was happening.

Eli didn't make it up about Roger Bate, Roger came up with that little goody all by himself while working for one of those oh so Annie Randy like places. Simply following in S. Fred Singer's paw prints blowing smoke in everyone's face.

BTW, if Eli is the worst person you ever exchange pleasantries with, you have lead a very sheltered life. We suggest you come out from beneath the moat.

"You do realise that the IPCC figure being used isn't directly comparable to real-world data, don't you?"

Makes one wonder, what was the point of publishing projections which couldn't be validated? The public is looking to the IPCC to statistically quantify the range of conditions that could happen outside their window, not inside their computer models. If the subsequent real world data can't prove the IPCC's projection right or wrong on even a decadal time scale then it was a pretty useless prediction.

As another poster alluded to, it's ridiculous that we find ourselves in this situation of arguing if it's even possible to test the projections. All IPCC forecasts/projections/predictions should be simultaneously accompanied by a detailed description of the data and tests that can be used in the future to validate them. If this means that embarrasingly large confidence intervals are needed,then so be it.

[The answer is obvious: not all the IPCC figures are about model validation. On just a few occaisions they do talk about other things. They are not responsible for the misuse of their figures by other people -W]

Eli. Your forth person ramblings are becoming less coherent by the day. Perhaps it is because of the world warming.
Are you seriously comparing the poor in America 60 years ago to the poor in Africa now?

Makes one wonder, what was the point of publishing projections which couldn't be validated?

Who says it can't be validated? The point is to predict climate changes, not weather patterns. Who knew that people would be so obtuse as to not be able to grasp this point?

Are you seriously comparing the poor in America 60 years ago to the poor in Africa now?

Reading comprehension again.

The libertarian teeny government club have never helped the poor before and they won't help the poor now. But they will use the fear of the poor being hurt to stall changes they don't like.

Boris is literate.

If you don't believe in AGW, then you can live happily ever after in the now. In the now, many Africans are extremely poor. It's great gWb finally came out of his coma and noticed. Lomborg can get "passionate" conservatives to spend billions on them just to delay doing anything at all about CO2 mitigation. I get it. And when it stops serving a political purpose, making it seem like you give a rat's butt about the poor, it will all stop.

William, If I am to understand your defense of the IPCC "projection" it is this. Lucia cannot compare her estimate of the observed trend, with the Ipcc GCM trend estimate because the former contains weather noise and the latter doesnt.

Surely, then this will be the case 24 years from now. that is,
in 24 years the ipcc chart from 2001 will still contain no weather noise, and if lucia computes a 30 year trend from observed data it will still have weather noise. If in 30 years
lucias trend is lets say zero, your objection, namely that the two trends cant be compared, would still hold. Does that seem odd to you? In short, it seems like making a projection that is never subject to falsification.

In anycase I suggested that lucia might put the error bars from the observational data onto the ipccc trend and see what results, but that's a sketchy method at best

[In very general terms, weather noise declines with sqrt(n_years), which is why looking at longer trends makes sense. So for 30 years there is hope.

I've already suggested that a rough and ready test is tob assume that Lucias error bars are appropriate for the IPCC too, so the test is, roughly, do her errors bars overlap with the same imposed on IPCC. If she has done this, I don't see it -W]

Take the IPCC global mean temperature figure reproduced in Lucia's original posting, and pick a year, say, 2007. Was the IPPC saying that the global mean temperature in 2007 would be expected to be around 0.3 C higher than the average for 1980-2000? If so, the actual observation did not bear this out and Lucia makes a very convincing case that this difference between observation and "prediction" or "projection" is significant and did not happen just "by chance" because of the variability in the trend of the actual observation attributable to "weather". Are the above statements wrong? If so, could someone explain why, preferably without making assumptions about my motives or intelligence?

[The IPCC fig used by Lucia is saying, effectively, that if you took a large number of earths then on average T is expected to increase from ~1990 to ~2010 by ~0.3 oC. But we only have one earth, and I've been trying to explain with very limited success, the IPCC figure doesn't tell you very much about how you might expect the T to change on that one earth. Or rather, it tells you that an increase is expected (and has indeed occurred: by my rough reading its gone up about 0.25 over that period, but I don't have exact numbers to hand) but doesn't tell you how unlikely an increase by less-than-expected is -W]

It's very simple. Over a number of years (experience says ~20-30) climate variability averages out. You can compare long term averages to the prediction. What you can;t do is compare a single year, or even two or three to a projection which does not have such variation in most cases.

You could try an approach such as using a Monte Carlo simulation. See, for example, W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C: The Art of Scientific Computing, 2nd Ed. (Cambridge University Press, Cambridge, 1992) pp. 656-706.

This will produce a best estimate for the model as well as an estimate of the variance, but even then you can't look at a single year.

Eli,
Thanks. So the IPCC Figure says that the mean temperature averaged, say, over 20 years with 2007 as the mid point, is expected to be around 0.3 C higher than the average for 1980-2000. In other words since the IPCC figure was produced in 2001, we won't be able to say if the extrapolated bit of the curve is accurate or realistic till at least 2021. Of course there are many people, while not questioning the physics linking CO2 to AGW, are not sure about the accuracy (skill) with which the GCMs forecast future global temperatures. I can see that if the trends discussed by Lucia continue even for a few years, it would be impossible to persuade such people or governments to take corrective actions which require really drastic changes in current lifestyles.

steven mosher --- Tamino just put up a proper comparison of the actual temperature data on his 'Open Mind' blog. Makess it clear that the error bars for Lucia's data are so wide as to agree with the 34 year trend.

Gautam --- Hadley Centre issued a prediction (last fall?) that 2008 CE would be about like 2007 CE and to expect warming in 2009 CE. I can only surmise that this prediction is based, in part, on the current strong La Nina.

Sorry David, When Tamino publishes his real name
and posts his analsysis files for download, then I'll
look at it. Nothing against him, I just prefer people who
share the open philosophy.

Also, I have an outstanding question that William has avoided. Namely, what does the IPCC "forecast" "predict" or "project" ? My understanding of their writings, and my understanding of his writings is that .2C warming for the next two decades is the Average of model outputs. and that this average contains no weather noise. So, one would like to know what is the RANGE ( max to Min) of all the model outputs from say 2001 to 2008. And then since this doesnt include weather noise as William argues, then one would like to see weather noise imposed on this. What you will find
is not a climate consensus, but rather a huge error band.
Simply, a NAIVE forecast ( straight line from the past 30 years) is likely to outperform the GCM consensus from 2001 to 2011 and probably out to 2021.

You would expect the added physical understanding of the models to allow them to outperform mere statistics. Hmm

Now to Eli's point "experience shows that climate variability evens out over 20 to 30 years?"

I refer you to Robock. I refer you to 1/f noise. I refer you to hurst like phenomena. But yes, "evens out" was the precise mathematic term we were searching for. Like oblate spheroid.

Having said all this, I'm convinced that the warming trend WILL recover. As Lucia says events that happen at 5% rates do happen at 5% rates. How to put this another way. AGW is true, but the IPCC wasnt exactly precise in what it meant by its projections. It needs to be.

steven mosher --- Up to you, of course. However, the data is readily available and Tamino carefully explains just what he did, including pointing out why enough data is a requirement for trend analysis. Easy for you to set up the calculations yourself.

How come people who readily blames other for not listening to the scientific consensus on climate change does not even want to understand the true scientific consensus on how our economies work?

This is it: markets work so as to create efficient economies; economic development and globalisation have lifted billions out of poverty. Technical development driven by the markets have taken technology to a point where we are sustaining more than 6 billion people on earth with less environmental impact than we were a mere billion. This whole process is now at risk because Eli Rabett and his like lacks the willingness to take in elementary knowledge of how economical dynamics work.

Now Eli Rabett tells us that halting this progress is no risk because government education programs and food coupons "worked" in the US depression?

Eli and his like, nor the government he adores, have created development. Development takes place anywhere where the government won't stop it. The process that led to billions fewer poor people were created by the ones Eli describes as the big evil, the people staffing big and small businesses over the world. Engineers, economists, salesmen, managers...

I could quote the rabbit: "You and your ilk were just standing there doing nothing when this was happening." but that would be too kind. Because Eli is just not standing by when other people solves the worlds big problem: he wants to interfere with them because he think that he knows better. But he don't.

sorry David, but lucia and tamino are not looking at the same thing. Lucia averaged the 4 temperature series. She then looks at the trend since 2001 and compares this to the trend projected by the ipcc in 2001, which is .2C decade.

Yes the error bars on a trend line are wide when the time peroid is short. but they are not wide enough to capture the .2C trend at 95% confidence.

Finally it has nothing to do with AGW being true. it's merely testing the projected trend from the ipcc against the observed data since the projection was made. by the end of the year I suppose the warming will return and the ipcc projection will be in the ballpark. until then its not

william, you dont need 30 years. suppose ipcc project a 10C warming per decade. one would hope that after 6 years such a wild projection would be rejectable.

steven mosher --- Thank you for the explanation.

By the way, Hadley Centre predicted (last fall, I think) that 2008 CE will be about like 2007 CE, but that 2009 CE will be noticably warmer.

David,

2008 will not be like 2007. Not even close. 2008 will be the coldest year since 1997. A naive empirical forecast would suggest that it will be as cold as 1994.

Hopefully, the warming in the "pipeline" will compensate for this.

I have asked (both here, at Open Mind and RealClimate) for people to specify some future values for global temperature development that would constitute a falsification of their own pet theory or prediction of climate change. I don't seem to be getting anywhere with this, although I expected the concept of falsification to be appealing to scientific minds.

Now that I think of it, I remember Gavin Schmidt at RC stating explicitly that, the global temperature record of 1998 will be broken in five years, that is by 2012. Maybe, if the prediction fails, he will join the denialists.

[I don't have a pet theory, it seems that only the skeptics do. Perhaps you're asking the wrong question? -W]

Dodo --- Science proceeds by confirming (or disconfirming) hypotheses compared to other hypotheses. See

http://en.wikipedia.org/wiki/Bayes_factor

for a good method to do so.

It was predictable that AvK would invoke the Free Market Fairy. As Joe Gregorio pointed out:

How to fix global warming? Well, first deny it exists for seven years, then when the weight of scientific evidence is completely overwhelming claim it can be fixed, by the free market, even though the free market hadn't solved it over the previous seven years.

What works are regulated markets Av. Government regulated markets. Mom Rabett remembers what happened before we put the brakes on your wonderful "majic market" It was called the Great Depression and before that other depressions and crashes came more often than the El Nino. Today, well, as Gregorio put it

the Free Market Fairy was the justification for taking a chainsaw to lending regulations in 2003 which precipitated the housing bubble, subsequent collapse, resultant credit crunch, and impending recession.

Same thing happened with the gogo savings and loans in the 1990s. I will mess with Texas if I want, I bought the damn thing in the S&L bailout. You could, if you wanted go read Paul Krugman in the NYTimes today

America came out of the Great Depression with a pretty effective financial safety net, based on a fundamental quid pro quo: the government stood ready to rescue banks if they got in trouble, but only on the condition that those banks accept regulation of the risks they were allowed to take.

Over time, however, many of the roles traditionally filled by regulated banks were taken over by unregulated institutions -- the "shadow banking system," which relied on complex financial arrangements to bypass those safety regulations.

Now, the shadow banking system is facing the 21st-century equivalent of the wave of bank runs that swept America in the early 1930s. And the government is rushing in to help, with hundreds of billions from the Federal Reserve, and hundreds of billions more from government-sponsored institutions like Fannie Mae, Freddie Mac and the Federal Home Loan Banks.

Given the risks to the economy if the financial system melts down, this rescue mission is justified. But you don't have to be an economic radical, or even a vocal reformer like Representative Barney Frank, the chairman of the House Financial Services Committee, to see that what's happening now is the quid without the quo.

Last week Robert Rubin, the former Treasury secretary, declared that Mr. Frank is right about the need for expanded regulation. Mr. Rubin put it clearly: If Wall Street companies can count on being rescued like banks, then they need to be regulated like banks.

And, the evil government that creates nothing, oh yes, things like social insurance, introduced in the 19th century by Bismark, guaranteed retirement income which has lifted the elderly out of poverty, and in civilized countries universal health insurance which has helped both rich and poor alike reach a ripe old age. Terrible, just terrible.

Still a lot better than listening to the free market fairy fans whine.

Tamino's done a post on this over at Open Mind now:

http://tinyurl.com/386239

William: Why does the solar connection always have to pass tests that the CO2 connection is exempt from? If the last 10 years are weather noise for CO2 then the warming from '85 to '98 could have been weather noise for the solar connection. That 13 year period is, arguably, the only non-fit for the solar theory. PDO shift anyone? Anyway, what does this 10 years of noise consist of? Could it be that we don't really know what drives climate yet. Clearly there is something out there more important than mere CO2, despite what the IPCC or Hansen try to tell us.

[I think the answer is that you seem to think that climatology is done by curve-fitting; it isn't. Theory, models and data all have to fit together. For solar, it clearly doesn't, for reasons that have been repeatedly explained. Sometimes people simplify this to just talking about the obs, but thats not it. The solar theory doesn't fit anywhere -W]

Assuming 1980 again was the start of the clear CO2 warming signal (apparently IPCC endorsed according to Trenberth, Mann etc), then we only have 27 years of data to judge with. So until now every scary scenario has been based on less than the optimal 30 years - starting with Hansen in 1988 (8 years of warming). So the upshot of teh 30 year requirement is that all previous predictions were invalid anyway. At this point in the loop, William would go back to the supposed 100 year trend and I'd point out the 100 year solar trend.

[Re 1980 - I thought we'd done that a few comments back. No, 1980 *isn't* the start of the CO2 warming trend -W]

James G asks about tests. Well, the first test is that your ideas about what will happen fit in with your knowledge of basic science. Greenhouse gases affect the radiative balance of the atmosphere, so if you think there is no effect you would have to find a counterbalancing physics. Good luck.

Eli I didn't say there was no CO2 effect but the extent of the effect is still uncertain. It is indeed quite plausible that the effect is very small. Neither is it basic physics as you suggest: We have here 3rd hand radiation with multiple interactions and fudge factors. But I don't need to prove anything because I can simply ask if the observations verify the prevailing theory. Obs are the ultimate arbiter, having ruined more than one great theory, and it seems they tell us that the forcings and feedbacks are not at all as was thought: Water vapour and clouds giving less and soot giving more. No doubt the models will be changed at some point to reflect the obs - as they always are - after a lot of foot-dragging and bruised egos.

JamesG --- from "Global Circulation Model Development"

http://books.google.com/books?id=vnYeHl6AvgkC&pg=PA128&lpg=PA128&dq=Jas…

one reads that testing the GISS model against data from LGM gives the same climate sensitivity of 3 +- 1 K that is obtained from current data. There is also lots of radiative physics studies. So there is little chance that the CO2 effect is wrongly understood.

Eli Rabett

"even though the free market hadn't solved it over the previous seven years"

Are you seriously suggesting that if the markets didn't "solve" the "climate crisis" in seven years, that is proof that regulation is better than market driven development? If we turn that on its head, does such a short period prove that the Kyoto protocol and the accompanied regulation is not working?

Secondly, you are suggesting that it was a lack of regulation that put banks in the mortage mess, but it was in fact the government who leaned onto banks to achieve laxer lending standards.

Qouting Stan Liebowitz:
"From the current hand-wringing, you'd think that the banks came up with the idea of looser underwriting standards on their own, with regulators just asleep on the job. In fact, it was the regulators who relaxed these standards - at the behest of community groups and "progressive" political forces."

So, in 1995, US authorities told banks that they had to start giving mortgage to poor areas to avoid accusations of discrimination. The only way for the banks to profit from this risky debt was to bundle it with good debt and pass it on.

Unregulated businesses as hedge fonds for example, have created much less turbulence even though they take on higher risks.

In short your position seem to be:

unvalidated economic models have created emissions scenarios, which when fed to unvalidated climate models yield very high temperatures in the future, which gives us cause to destroy the only force so far that has both reduced environmental harm and lifted people out of poverty, because listening to scientific consensus in economics are wrong, even though we should listen to the majority of climate scientists in spite of them not presenting as a strong consensus as the economists in support of economic freedom.

I do find your ramblings a bit incoherent. And I still think that you represent an evil ideology. Lots of people will suffer because of you and your ilk.

Oh but Av, the free market fairy knows everything. Av. questions whether the emission scenarios or observations, the answer is simple, and as Thomas Knutson put it to Pat Michaels:

Should we trust models or observations?" In reply we note that if we had observations of the future, we obviously would trust them more than models, but unfortunately observations of the future are not available at this time.

Av does not appear to realize that the special emission scenarios were created to cover the possible (not even likely) range of possible futures not to predict the future.

And, oh yes, Av dear, lots of people ARE suffering because of you and your ilky free market fairy (sorta like we do know what is happening now). Regulated markets can work. Unregulated markets are a race to the cliff.

Eli, I find your evidence as persuasive as the IPCC case for a CO2 lifetime of 200 years, a climate sensitivity of 3°C and future emissions of more CO2 than could plausibly be brought to market...

[I'm stepping in to terminate this discussion, which has veered too far into economics for me (errm, and you're on poor ground ranting about CO2 lifetime and clim sens, not to mention future emissions). Eli has a blog; you can discuss it there if you like -W]

Just for the heck of it, Real Climate has a good summary of how long increases in CO2 atmospheric concentration last.

The short of it is, that a slug thrown into the atmosphere rapidly dilutes into the ocean and soils, decreasing the increase by a bit more than 1/2. That is why the observed increase is less than would be predicted by just calculating the emissions. There is then a relatively rapid exchange ofthe CO2 between the oceans and the atmosphere, so that any particular CO2 molecule in the atmosphere would only stay there ~ 5 years, but the increase in concentration lasts a very long time. The limits on that are set by mixing of the increase into the deep ocean (hundreds of years) and rocks (thousands of years). 200 years lowballs how long the last two processes take