HadCrut 3 reported their March anomaly. Here are their current values for the past two months:
2009/02 0.356
2009/03 0.359
You can see how these relate to previous temperatures in the figure below:
I’ve circled the March values in the figure above, you can see March is just a tiny bit warmer than February.
The big question for next month: Will the trend since 2001 for NOAA jump into positive territory? All will all three trends remain negative?
As we are constantly told, these short trends are noisy, so anything could happen. Watching all the snow walloping Utah, I’m betting April comes in cold. But of course, Utah is not the whole world. 🙂
Meanwhile, returning to the discussion of 20 year trends inspired by Deep Climate, recall, that in comments on the previous post, Deep Climate said:
Your second graph (and Chad’s) will look quite different if you use annual data (which makes more sense anyway). It really cuts down on the noise as I’ve already pointed out.
So, I re-computed the showing how 20 year trends vary as a function of end year, this time first computing 12 year month averages based on the monthly data. Here’s the new plot:
Here’s how the plot looked yesterday:
Of course applying 12 month smoothing before computing trends makes a smoother graphs. So would computing 21 year trends. Other than the degree of smoothness, does anyone think these look “quite different”.
To be honest I’m surprised it made a visible difference at all; I’ll have to think about this a bit.
Can I ask whether you have looked at the Easterling and Wehner paper? http://www.cdc.noaa.gov/csi/images/GRL2009_ClimateWarming.pdf
Johanthan,
I have looked at the paper.
For fun, I suggest you read the paper, read the two periods they describe as having “no trend” and then compute the trends using monthly data for that period.
After that, examine when El Chichon erupted and decide what you think of their empirical evidence.
Then, we can discuss the other stuff in that paper. 🙂
Jonathan-I know that RP Jr. has talked about it rather dismissively:
http://sciencepolicy.colorado.edu/prometheus/spinning-probabilities-in-grl-5113
I recall that Lucia commented on it when someone pointed it out earlier, but I don’t know for sure.
On being surprised: Smoothing always smooths. So, I’m not too surprised that using annual average data is smoother.
That said,I may not have done what Deep thinks makes sense. I computed trends suing 12 month smooth data– but reported at ever month; i.e. 240 months’ worth of 12 month smoothed data.
Use 20 months all starting in the same month, the graphs will get noisier again.
As Lucia says, Easterling and Wehner seem to have forgotten about volcanoes. Look at their fig 1 showing temperatures from 1975-2005. Then mark the El Chichon eruption (1982) and Pinatubo (1991) – these correspond well with two major dips in the graph. It is strange that two scientists should make such a simple error (particularly since they do mention volcanic effects). If they had followed Lucia’s blog they would be well aware of this effect.
Lucia, where is the post where you plotted 20th century temps with major volcanic eruptions marked? I think it was on your blog but I can’t find it.
Why don’t you do a new post on this?
PaulM,
If you are interested in just GISS, NASA has a good graph showing temps, volcanoes, and ENSO:
http://data.giss.nasa.gov/gistemp/graphs/Fig.E.lrg.gif
Zeke, your link doesn’t work.
Looks like you gave up on the Caffeine!!
Andrew_FL
go to
http://data.giss.nasa.gov/gistemp/graphs/
scroll to the bottom graph and click on it. It took a couple of tries for it to render in my browser.
kuhnkat-Well now I’ve had a good laugh! They make the volcanoes so tiny they seem cute! Well, I suppose compared to Toba, they are, but why not show GISS’s own AOD record instead?
Lucia,
I’m glad you are finally focusing on long term trends. Really I am. And I think the smoothing is fine (although I usually just use annual points like Pat Michaels did).
But the proper multi-model trend should be about 0.2 deg C/decade for the current period. You’ve got it up around 0.27 or 0.28, presumably because you are using a subset. Chad’s trend (1979 start) is considerably lower than your equivalent, although even his looks a little high to me.
See:
http://deepclimate.org/2009/04/18/20-year-surface-trends-close-to-models/
My graph is at:
http://deepclimate.files.wordpress.com/2009/04/20-year-trend.gif
I was going to wait until I could get my own data set going (I’m new at Climate Explorer party). But now I see your graphs are going far and wide, so I felt I should speak up.
Maybe Chad can supply us both with the right monthly series. Or just use 0.2 – it’s close enough I’m sure. Anyway, it’s the commonly accepted figure for these types of analyses – even Pat Michaels at the Cato Institute used it in attempting to show the so-called “abject failure” of models.
(Actually, the multi-model series probably needs to be smoothed anyway as it has an “artifact” bump up in 2006-8 and a dip in 2009-10 if you look closely at fig. 10.5 in the AR4 WG1).
By that measure, the long term trends seem to have been ahead of the models earlier this decade, but are now a little below. Certainly that’s to be expected with the larger interannual variability of the “real” climate compared to the models.
Deep Climate, you say “even Pat Michaels at the Cato Institute used it in attempting to show the so-called “abject failure†of models.”
Er, I don’t recall him just pulling a figure out his ass but rather using the model data to create confidence intervals. I also, the phrase “abject failure” doesn’t occur a single time:
http://www.cato.org/testimony/ct-pm-20090212.html
Instead he says, more defensibly and modestly, that the models are “in the process of failing”.
Do go misleading people now!
‘scuse the irony, I mean ~don’t~ go misleading people now. Also, what’s the obsession with twenty year trends? The since 1979 trends have way more information, and would, if you use a similar figure, annihilate your claim. How do you “know” the twenty year trends in models average .2 per decade? Why not calculate it first, then pontificate?
Deep, the model trends (and observations) include volcanoes, your 0.2C/decade does not.
The Cato Institute says in their recent ad, presumably written or at least approved by their chief climate “researcher” Pat Michaels:
“The computer models forecasting rapid temperature change abjectly fail to explain recent climate behavior.”
The letter soliciting signatures for the ad was signed by Michaels, so presumably he stands by every word in it. Chip Knappenburger pointed to the Michaels congressional testimony concerning short term trends to justify the statement (see the Realclimate.org thread entitled “With all due respect”). And of course his was one of the 100 signatures.
From 1980-99 to 2011-30, the projected rise for the A1B multi model mean is 0.65 deg C. That works out to 0.21 deg /per decade, but it would be slightly lower in the first part of the period. Hence 0.2 deg per decade seems just about right. At least Michaels and I agree on that.
Deep, when the 20 year trend starts in a period with stratospheric volcanos and ends in a period characterised by a volcanic lull (like the present) the models predict the temperature trend to be higher than 0.2C/decade. The observations just forgot to take their Viagra pill. I think that is what Lucias figure shows.
Deep Climate-now you are playing bait and switch. And what’s with the quotation marks around researcher? You are aware that he has published extensively in the literature, no? And pointing me in the direction of the spin master at “Realclimate”? Surely you can do better than that. And again, you make no reference to Michaels’ statement of “.2 per decade” at all-when did he explicitly state this? But instead of calculating, you go right back to pulling numbers out of your ass and hand waving with them. Come back when you can show your work! One final point-the statement “The computer models forecasting rapid temperature change abjectly fail to explain recent climate behavior.” could be interpreted very loosely-which models? is your .2 figure “rapid”? I think it is safe to say that the really rapid models have not done too well lately. But it isn’t clear in the ad whether they were only referring to that or to other “behavior” of climate-the also reference Douglass et al. (and I don’t feel like getting dragged into that, so :p ).
I didn’t even bother to mark you off for redundantly stating “The letter soliciting signatures for the ad was signed by Michaels” and “And of course his was one of the 100 signatures”. Opps! I guess I did!
DeepClimate–
As I said, I am showing the trends with volcanic eruptions because we are examining a period affected by the eruption of Pinatubo which went off in 1991 and El Chicon. The depression due to the Pinatubo eruption will affect 20 year trends until those ending in 2011.
The trend with volcanos is higher than without, but the 20 year trend ending in March 2009 including runs without volcanos also exceeds 0.2 C/century. You can discover this by downloading all the model runs from The Climate Explorer, as I have.
It’s interesting that you seem to suggest I must use 0.2 C/century. When I was using 0.2 C/century because it’s commonly used, I was lectured by “others” that I should use the downloads from the climate explorer!
DeepClimate:
Wrong. The trends in the multi-model mean for the A1B scenario currently exceed 0.21 deg/decade.
Andrew_FL–
When discussing Deep Climates complaints about Michaels you might want to point out that this has nothing to do with my graphs or arguments. I have no idea why Deep Climate feels the need to jump in here to get in an argument with Pat Michaels. Pat ain’t lurking in my comments!
Lucia, “The trend with volcanos is higher than without”. This had me confused for a minute because volcanoes cause cooling! The higher trend with the volcanoes is because the volcanoes (El Chichon 82 and Pinatubo 91) pull down the left hand end of the graph, have I got this right?
PaulM-
Yes. the 20 year trend with volcanoes has it’s lefthand end pulled down by El Chicon and Pinatubo. This elevate the 20 year trend relative to what one might expect if the eruptions had not occurred. The dips occur 1-2 years after the eruptions themselves, and rises are slower than the dips. Roughly speaking El Chichon’s mostly dies off for trends around 2003-6, Pinatubo will have an effect on computed 20 year trends until a little after 2012-2016. But this is rough.
These affect the 20 year trend for the earth and those in runs including volcanic eruptions. Naturally, they do not affect the multi-model trend for models excluding eruptions.
:blush: Sorry lucia, I would have if I’d thought of it, but since now you have stated explicitly yourself…Yeah, what is the big idea dragging Pat into this, anyway? we are supposed to be discussing lucia’s analysis.
Slight correction to my previous comment: I was confusing the A2 and A1B scenario IPCC projections. The former is 0.66 C (0.21/decade) to 2030 and the latter is 0.69 C (0.22/decade). The current trends would be slightly lower of course. The exact trend probably depends on how it’s baselined and smoothed, but 0.2C/decade is a reasonable benchmark that has been used by every other analyst to my knowledge, including Roger Pielke Jr (who you have quoted from time to time).
I suppose your position is that that’s the wrong benchmark, and we should be using 0.27 or 0.28 C/decade instead. Of course, you’re entitled to your opinion. We’ll have to agree to disagree.
I’m sure Deep would understand what Lucia is saying if the vulcano eruptions and the lull had been switched in time. Right, Deep 😉
Deep Climate-it shouldn’t be a matter of picking and choosing benchmarks which make whoever we want to be right right. This is why I am trying to convince you to actually calculate the 20 year trends in models and then come back. Right now, I get the impression you are speaking Ex Cathedra.
I have used it too. I seem to recall I was criticized for using “about 0.2C/decade” by Gavin, who insists I must use the trends from the underlying baseline.
Obviously, both can be used depending on what one is doing or discussion.
Why do you suppose this?
My position is that it is appropriate to compare like to like. If we are going to examine observed 20 year trend ending in year “x” we should compare them to simulation of 20 year trends ending in year “x”.
I have no idea where you developed the notion that the 20 trends ending this year out of IPCC models would be lower now than 10 years from now. But what you claim is utterly wrong if we look at A1B scenarios.
One of my to-dos is to download the A2 scenarios, which will permit me to check that. I suspect you will turn out to be just as wrong when guessing the trends for A2 scenarios as when you guess those for A1B scenarios.
Have you ever heard the distinction between weather and climate?
It is fruitless to have trends on climate based on data of few years. Don’t you have ANY statistical education?
lucia, regarding Deep Climate’s comment, I would really like to know why he thinks that trends depend on the baseline. This just doesn’t sound right to me.
Micke,
I think the real question is whether you have any statistical education.
I am amazed at the number of people who cannot comprehend that the 30 year rule used by many climate scientists to divide ‘weather’ and ‘climate’ is a number pulled out of a hat and has no physical basis.
I suspect the reason for the number is they needed a number that was short term enough to make scary predictions interesting but long enough to ensure the climate scientists never have to be worry about being proven wrong before retirement.
Mathematically weather/climate are a continuous process that can analyzed at any time scale *provided one includes appropriate confidence intervals*. Generally, the confidence intervals for short periods will be large enough to make it impossible to draw any meaningful conclusions. But that does not mean it is wrong to do the analysis.
Raven, the 30 year convention predates alarm, so I can’t agree with your reasoning. I think that it was just an arbitrary judgment call some scientist made at some point which became sacrosanct-sort of like R A Fisher saying 95% confidence intervals would do for most hypothesis tests-sure, it may indeed be a good cut off, but what makes it the standard? Convention…Which wouldn’t necessary if everyone could just report their dang p values and let the world judge!
Andrew_FL [12905], Raven [12904] and Micke [12901]
Has nothing to do with 30 years. Based on best available evidence the distinction between climate and weather goes like this: when temperatures are up it’s called “climate” and when they’re down it’s called “weather”. When polar ice extent is down it’s called “climate” and when it’s up it’s regional “weather”. Bush fires in Australia are caused by “extreme changes in climate” and absolute snow records throughout North America are due to “changes in weather patterns”.
With a wry 🙂
Yes, any actual weather event is irrelevant storytelling, since it may not be indicative of the Actual Temperature Trend, whose (only) manifestation is a line on a graph. The utter and abject belief in that line is what’s important. What smart person doesn’t get that by now? 😉
Andrew
Joe at Digital Diatribes does a nice job with the charts and illustrates the impacts of differing trend lengths clearly.
http://digitaldiatribes.wordpress.com/2009/04/17/april-2009-update-on-global-temperature-uah/
Andrew– Trends computed based on monthly averages can shift a tiny bit depending on the baseline. This is because there is noise in the baseline used for each month. The effect is so small you would never see it in the graphs.
I don’t think trends computed on annual average values will be affected by the baseline.
tetris-put aside your need need to criticize advocate recipocracy for a moment. Is it not worth both your and your opponents time to try and come up with a more clear definition of what is climate and what is weather?
Andrew_KY-people who make such arguments, I’m sure you understand, don’t understand to important points. They don’t understand that the mean (climate) is ~determined~ by the variability of which it is the average. The second point they don’t understand (perhaps they are to attached to their Playstations to get this?) but weather is what actually matters! Regional scale events and anecdotes actually affect real people-the average of a bunch of anecdotes is only actually experienced relatively rarely.