The Blackboard

[...] blogger, Lucia plotted long term GISS trends and got some interesting [...]

Isn’t this just an artifact of fitting a linear trend to non-linear data?

1998 was a step change. In 2001 there were not enough points after 1998 to affect the trend. By 2008 the 1998 step change was completely factored into the trend but we are now seeing the effect of the last 8 years with no trend.

If I am correct I would expect to see the trend to continue to decline for a few years even if the temps starting rising again.

Raven–
Yep. Part of the shape is just an artifact. For example: Why is there a dip on the 20 year trends around 1998-2000? Well, some variations are just wiggles. But we know that Pinatubo erupted in 1991, depressing temperatures with the center of the dip about 2 years later (1993).

How does this affect a 20 year trend? Well, it causes a dip in the 20 year trend. But, oddly, the deepest part of the “dip” occurs considerably after the temperature dip. This is because the trend continues to drop as temperatures rise coming out of the dip until the point where the end temperatures rise above the trend line.

So, to some extent, the “dip” in 1999 is pinatubo. Similarly, the recent peak is the 1998 El Nino.

Slightly OT Lucia, but have you done your “How many months does this month’s number change in the past 1200 months” sum yet?

I always look forward to that with GISS.

KevinB,
I suspect you are asking me something I thought about doing before Dad got sick and I went to Florida. Now I don’t even remember what it is. Could you refresh my memory?

Lucia-can you make a similar graphic for NOAA and HadCrut’s GMST data sets? I know that you are doing GISS because it just came out, but I have a hypothesis about what they will show.

Lucia,

Given the astounding degree of data corruption Anthony Watts demonstrated during his surface stations audit across the US, compounded by GISS’ known propensity for “adjusting” their data on an ongoing basis, arguing with Deep Climate over the AGW/ACC meaning of differences of 0.014C or 0.018C in a tells me a] we’re getting caught up in his message, and b] we’re drifting off into the realm of “how many fairies we can get onto the head of a pin”?

Andrew_FL-
I’ll wait until their data sets update. That should happen within a few days.

tetris– The fact is, the 20 year trends reached a peak and are down. A sizeable fraction of the behavior DeepClimate describes can be attributed to Pinatubo. If there is no warming an eruptions would which would cause 20 year trends to dip when they did, then rise, then fall and flatten.

To some extent, those trend we see are what we would anticipate if the “volcano swings” are superimposed on modest warming. That said, smoothed graphs like that difficult to interpret.

Deep Climate wants to pick a few points off the graph, and say the trends speak for themselves. They don’t– knowing whether or not volcanoes erupted, suddenly makes you realize that if those graphs speak, their language passed through the tower of Babel.

Lucia said:
“The 20 year trend ending with the month of Dec 2000 is higher than the 20 year trend computed now.”

I think you meant the other way around, no?

“So, to some extent, the “dip” in 1999 is pinatubo. Similarly, the recent peak is the 1998 El Nino.”

No. The 1998 peak is El Nino, 1999 dip is La Nina.By the way, that’s why I usually use annual temps for these kinds of analysis – cuts down on the noise. The GISS long-term trend curves computed annually show much less local variation.

Anyway, It’s interesting that you consider a slightly rising long-term trend an “artifact”, while a short-term point-to-point trend is somehow more meaningful. We’ll have to agree to disagree on that one.

I’ve actually gotten some good ideas for a post of my own out of all of this.I have a lot more to say (I’m sure you’re surprised at that!), but I’ll leave it for now until I have time for a fully considered post.

Deep,

The 1998 El Nino shows up as a step change in the record. i.e. between 1997 to 2002 the baseline jumped by about 0.2 degC. This shows up as an increasing trend as more of the post 1998 years are factored into the fit.

Since then we have seen temperatures flat line and that is now showing up as a decreasing trend.

However, the more important observation is that despite the increasing trend the long term trends are still significantly lower than the mean model predictions. This suggests that the climate models do not accurately represent the earth’s climate.

Obviously, the question is why and there are many possible reasons. Unfortunately, the climate science community seems be wedded to this idea that they ‘know’ CO2 sensitivity is ~3 degC per doubling which means they will look under every rock for any explanation other than acknowledging that CO2 sensitivity could be lower than they thought.

IMO, this unwillingness to revisit the sensitivity question as a ‘possible’ explanation really undermines the credibility of the climate modellers.

Lucia
How does 1972-1983 compare with 1997-2008?
http://home.casema.nl/errenwij.....recast.gif

Lucia
“If, like Deep Climate, I thought the bulleted points spoke volumes by themselves, I would have start scratching my head and wonder what the dip from 1999 to 2001 meant. And what does the fact that both trends have been declining since 2004 mean?
I don’t know the answer to the first. But it seems to me the answer to the second is, “Deep Climate’s talking point is pretty weak.””

It seems to me that you have no interest in finding out.
Are you actually going to try and find out? Do you have a plan? Or do you like just leaving it hanging… A mystery for everyone to ponder…

What does it mean that the trends have been declining since 2004 actually mean? What conclusions do you draw? Are you able to actually investigate something?

Find out what Nathan? Why should anyone try to gain understanding of what’s going on using smoothed graphs (i.e. 20 year averages) when it’s well understood that smoothing reduces the information in the original noisy graphs?

Other than that: The goal of this blog post is not to explain those particular graphs. It is merely to provide more complete descriptions of the trends than those provided by Deep Climate and to illustrate that his explanation of the sub-set of data he picked appears in adequate when we look at all the data.

Sometimes, knowing that an explanation someone (in this case Deep) claims is “obvious” actually seems to be wrong is useful. This is true even if the correct explanation is continues to elude us.

we can see all 20 year trends ending after 2005 fall below that 2C/century

appears in adequate [sic] when we look at all the data

It seems you have little to go on but a few cherry-picked years (first comment above), which backfired in your face (second comment above). The “ending after 2005″ is certainly not much of a dataset or conclusion, and your advocacy is apparent for all. Why do you do feel the need to promulgate such pre-disposed ‘analyses’?

After reading some of your recent comments in prior blog postings, I’ll soften my last statement (missed the 10 minute window), and monitor you a little more closely.

However, I am don’t believe the explanation for the temperature increase during the 20th century is rebound from the glacial period or even the little ice age. I admit I can’t prove that theory wrong, but it strikes me as less plausible than the theory that it was mostly caused by greenhouse gases.
…
FWIW: I think warming will resume.

Will–
The projections/predictions were published in 2007 and the underlying model runs are based on SRES officially published in 2001. Given these publications dates, the fact that the 20 year average falls noticeably below the projections/predictions after 2005 is, in my opinion, rather notable. Those making projections didn’t even select the method of weighting model or model runs until sometime around 2005 or 2006. So, it’s hardly surprising that the “projections” weren’t too far off up to 2005!

I don’t know why you would think observing the 20 year trend falls below the projected trend after 2005 is backfiring in my face.

I still stand by what I said which is: “… these sorts of eyeball methods aren’t very useful for discriminating whether the IPCC projection of “about 2C/century” from the AR4 is on track or not,”

So, the fact that the 20 year trend is declining may not tell use much– but it is running below the IPCC projection. This is worth noting when we read claims that warming is even faster than projected (which we sometimes read.)

When the HadCrut and NOAA data come in, I’ll repeat with those data because another visitor requisted that. I’ll add the 20 year trends based on the models since that might be interesting. (I haven’t looked at that–but it would be interesting to compare.

FWIW: I don’t feel any “need” to promulgate anything. DeepClimate jumped in here to explain that these particular 20 year metrics have some sort of deep meaning. I don’t think they do– but if they did, we would need to consider everything they tell us. Right now, the 20 year temperature trend is dropping and falls below the most recent IPCC “predictions/projections”.

If DeepClimate is going to materialize on blogs and explain that a continued rise in the 20 year trend “proves” global warming has not stopped, he should rethink that strategy. It’s true there is little evidence that warming has “stopped” in the sense that it won’t resume soon. But DeepClimate’s “proof” is dangerous. For one thing, it falls apart because the 20 year trend are not rising (which tells us very little.)

Desperation to shore up an ailing hypothesis will always lead one to overstep the bounds of rationality. Go, Deep, Go. Nathan’s here to help with his incredible analytical mind.

Peole here are talking about temperature trends as if these AGW. Temeprature had trends before significant human carbon and will ahve trends after these are abated. The link from observed trands to AGW is being done by the computer models. If teh predictions of tehse models can be shown to be questionable then this also raise questions about the attributions taken from these models in respct to the human cause of any temerpautre trend.

So teh issue is not the size of an trend but the capability of teh current models in matching an observed trend.

I have a feeling that I am stating the obvious here but eh disucssion seeems to be all about trends and not about model accuracy.

Tag–
I generally focus on model accuracy. I just posted graphs that include the 20 year trends as projected by models. See:
http://rankexploits.com/musing.....an-models/

For a larger context here is a graph of the 10,30, and 50 yr trends for Hadcrut. Notice only the 10 yr trends exceed the 2 deg/Cent value. The trends have been increasing, however, while overlayed by a sixty year cycle.

link

Straight line trends are interesting to be sure.

But, global temperature seems be very non-linear on nearly every time scale.

Applying standard statistical tests (of, for example, confidence intervals) makes implicit assumptions regarding the temperature data. The assumptions may be true or false.

Thus, the use of straight line trends to reach defensible conclusions in the case of AGW need more than “statistical” justification — they need “scientific” justification — i.e. their applicability should be proven empirically.

Then, there is the issue of published temperature data quality. In the current context, historical thermometer data is suspect. And, satellite data is calibrated to thermometer data to an extent.

None of this is news to most people here. These thoughts run through my mind when people debate AGW on the basis of straight line trends.

What I like about this site though, is that Lucia is mathematically objective and many posters attempt objective agreements and rebuttals. Consequently, in my mind, this site adds value to the discussion of AGW. And, I visit often.

I don’t understand the specific meaning of “trend”. The meanings which I try to apply don’t match the discussion.

Scooter–
I fit straight lines using least squares.
Here are definitions of “trend” from the web:

# tendency: a general direction in which something tends to move; “the shoreward tendency of the current”; “the trend of the stock market”
# course: general line of orientation; “the river takes a southern course”; “the northeastern trend of the coast”
# drift: a general tendency to change (as of opinion); “not openly liberal but that is the trend of the book”; “a broad movement of the electorate to the right”
# swerve: turn sharply; change direction abruptly; “The car cut to the left at the intersection”; “The motorbike veered to the right”
# vogue: the popular taste at a given time; “leather is the latest vogue”; “he followed current trends”; “the 1920s had a style of their own”

The common usage of the word “trend” permits us to refer to these lines as “trends” for those specific periods.

Yeah, I tried the meaning of “fit a straight line through the data”: http://en.wikipedia.org/wiki/T.....st-squares

But I’d expect that to produce a single line through the data. The graph does not show a single line, which contradicts that expectation. Also the phrasing suggests one trend per year, and I’m not seeing the many expected lines (1988-2008, 1987-2007, 1986-2006). My hypothetical meanings are not matching the results.

Scooter:
I fit straight lines through data from year A to year b. The slope of that line is the trend, and indicated at year ‘b’ (the end year). I repeat for a variety of years.

So, this is a plot of a large collections of trends fit to a graph that is not shown here.

Calculated linear-least-squares trends act like band-pass filters with a phase delay when there are oscillatory components in the time series. The shorter the chosen trend-length, the lesser the delay, but also the greater the trend variability from year to year. With very short lengths, we get something approaching a smoothed time-derivative. In the face of known climatic cycles of centennial and longer quasi-periods, 20-year trends are simply much too short to tell us anything reliable about long-term expectations. At best, they register the intermediate swings, rather than persistent climatic trends.

AGW’ers have jumped on the upswing that global temperatures experienced from the mid-70’s to 1998, without ever noting that the preceding downswing was almost as severe. With all the arbitrary “adjustments” (“bi-linear” homogeneity, bucket, etc. ) they never truly adjust for UHI. This produces an UHI and adjustment-corrupted surface record that conceals the preceding downswing. That’s why satellite measurements are so important! Alas, they’re still too short to discern true climate trends.

Lucia
What you are doing just doesn’t mean anything.
“They just don’t appear consistent with warming at a rate of “about 2C/century” projected by the IPCC in the AR4. ”

To assume that the IPCC trend of 2C/Century means that the decadal trend will always be 0.2C/decade is… dumb.

“Why should anyone try to gain understanding of what’s going on using smoothed graphs (i.e. 20 year averages) when it’s well understood that smoothing reduces the information in the original noisy graphs?”
What? So if you want to find a long term signal, you use the shortest term data you can find?

o assume that the IPCC trend of 2C/Century means that the decadal trend will always be 0.2C/decade is… dumb

Who here assumed this? Not me.

As for long and short trends: You are unhappy if I note the smoothed, long term trends fall below the IPCC “about 2C/century”. You are unhappy if I show the IPCC “about 2C/century” is not consistent with the short term trend. If you are blond, can I ask, “Goldilocks, what length trend is just right? “

Lucia

Well it looked like you were implying it as you are saying if the trend is below 0.2C/decade (on your graph) then it is below the 2C/ Century.

I am not “unhappy” about it. I think it’s kind of comic actually.
So what if the trend to 2008 doesn’t fit the 2C/Century trend? The Century has barely started.
Why is it important that the trend to 2008 doesn’t fit?

lucia (Comment#12834)

More to the point: “you can’t get there from here”. Only taking trends 20 yrs or shorter do you see trends above 2C/century, even using all of the Hadley data (ref Comment#12790).

BarryW
Did the IPCC say that the trend for last Century was 2C/Century?

Nathan–

Did the IPCC say that the trend for last Century was 2C/Century?

I have no idea what substantive point you are trying to make or what specific statement you are trying to rebut.

But to answer your question– which I suspect you intended as rhetorical: Read the the next blog post. It shows the 20 year trends based on IPCC projected during the period of we are discussing which includes portions of the 2C/century.

As I said in this post:

In closing, do note: While these sorts of eyeball methods aren’t very useful for discriminating whether the IPCC projection of “about 2C/century” from the AR4 is on track or not, we can see all 20 year trends ending after 2005 fall below that 2C/century.

I could have added “and fall below the 20year trends based on projections based on models including volcanic eruptions since at least 1979.”

Feel free to say “So what?” In this post, I’m just decribing the data.

Nathan (Comment#12841)

There is a 60 yr cycle in the trends of the historical data and we are at the peak of a cycle. The trends are going to go down for the next 30 yrs before starting back up if historical observations hold true. If we are to get to 2C/Century the average rate has to get to that value. So if you believe that the rate is going to go well above .2C/decade for the last four decades great, but at 2060 the trends should be starting down again.

I’m not arguing that the temperatures will continue going down or the rise stop, just that the projection of 2C, or even close to it, doesn’t seem to be reasonable. The average rate, on the other hand, may reach that but not anytime soon.

Hadcrut is finally out:
2009/03 0.359

Lucia,
Barry said “you can’t get there from here” – which I assumed meant you can’t get to a 2C trend/Century from where we are (and he seems to have confirmed that) – For some reason Barry thinks it’s going to get colder.

Nathan (Comment#12914)

I think the question is not if it’s going to get colder (cooler?) but how much. There’s an underlying upward trend but there is also this cycle (which I’m assuming is related to the PDO). How big an effect that has I don’t know, but I think it’s a real problem for the modelers if they haven’t taken that into account. If the models don’t show that cycle in the output then they’re going to be off when the PDO goes negative (which is seems to be doing). The PDO is also going to bias any long term trend if you’re only catching part of a cycle.