May 19 2008
7 Year Trends Falsify What?
In what is quite possibly the silliest discussion ever, the Dr. Roger Pielke Jr. has blogged once again about whether the IPCC Temperature Forecasts Have Skill (responses from the usual gang are expected sometime today). He attempts to show that the IPCC temperature predictions have no skill. Forecast verification is important, and there are many ways in which to assess the skill in a forecast. Forecast Verification - Issues, Methods and FAQ, created by the WMO, discusses many of the methods.
Dr. Pielke brings up a good point though - a forecast can be accurate, but have no skill. I’ve discussed this before in relation to the seasonal hurricane forecasts having no skill when compared to climatology. Although, they continue to get press coverage every year, which goes to show that people care more about the making of the prediction than if the prediction means anything. A little over 2 years ago, Dr. James Annan blogged about using skill scores - perhaps suggesting that there is a 2 year periodicity in climate-related Internet memes.
This post keeps with my current style of writing an introduction about one thing and then switching to a completely different topic, which may or may not be actually related to the post title. This post will actually be about cherry-picking, because picking cherries is fun to do.
|
This plot shows the global temperature anomaly as used by Dr. Pielke Jr; the average of the four “main” groups that track global temperatures. He didn’t specify, but I’m guessing he means GISS, Hadley, UAH, and RSS, and not NCDC.
I’ve plotted some cherry-picked trends on the temperature anomalies. The two that seem to be ruffling feathers are the orange and green ones. These are trends from 2001 to present and 2001 through the end of 2007, respectively. The trends during these two time periods is very close to zero. The trends for all time periods is in the legend with units of degrees Celsius per century.
The only trends on the graph represent 2 times when the global temperature has decreased, and once when it has increased extremely rapidly. The two decreasing trends both have volcanic eruptions that have influenced the temperatures, El Chichon (1983) and Pinatubo (1991-1996). The decrease in the cyan trend is likely due to the eruption, but the purple trend may not. The eruption occurred about 2 years before the dramatic decrease in temperature. Since the lifetime for sulfates in the stratosphere from volcanic eruptions is about 2 years, suggesting that the temperature decrease was not due to aerosols.
The point is that 7 year trends don’t tell us anything about anything. In the next 3-7 years there will be a positive ENSO phase which will increase the temperature, much like the current La Nina has decreased the temperature. Unless Dr. Pielke Jr. is going to argue that during that time that the large observed trend falsifies the IPCC prediction by being too large, I don’t understand what his current argument is.
One thing that is not evident from the plot, which is not surprising, is how the temperature in month n is related to the temperature in month n-1. There are 175 instances where the temperature in month n is greater than n-1, 174 times where the nth temperature is less than n-1, and once when they are exactly equal. Yet the temperature trend during this period is obviously positive. For GISS they are 167 (>), 173(<), and 11 (=). This is last paragraph is meaningless, but shows that the month-to-month noise is large compared to the warming signal.
(1 votes, average: 1 out of 5)
Loading ...Related Posts:
27 Responses to “7 Year Trends Falsify What?”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
Atmoz-
Thanks much for the link, but I don’t think your characterization of my post is a fair representation. Here are some clarifications.
I report that the IPCC predictions from 1990, 2001, and 2007 to be skillful. No doubt about it.
The post is part of a series on how to evaluate IPCC predictions (and how not to), not as you assert “to show that the IPCC temperature predictions have no skill.”
I also write, “What does all of this mean for the ability of the IPCC to predict longer-term climate change? Perhaps nothing . . .”
Also, your guess on which 4 obs datasets I used is correct.
Finally, a temp trend higher than the IPCC prediction will always be skillful, as it will always be farther from the naive baseline.
Hope this helps.
Dr. Pielke,
With regard to a temperature trend higher than the IPCC prediction, you’re correct that it will have skill, but it will not be consistent. With higher than IPCC predicted trends, consistency is a more stringent test than skill because the prediction will always have skill.
This entire post can be summed up in this one sentence:
This can be seen in this graph, which shows the temperature trends from the 4 series diverge when calculating the trend over short time spans.
Don’t get me started on “consistency”;-) I think it is a weak test (in many contexts, I’d say “meaningless”) for ensemble forecasts, since the more uncertain the ensemble to greater the likelihood of “consistency.”
If a baseline proves too easy to beat, then I’d argue for a more challenging baseline. In this case perhaps one based on forecasts generated using a very simple climate model, and then establish this as the baseline for evaluating the skill of forecasts from more sophisticated models.
The figure that you point to indicates that trends of any length period tell us something about something, but that uncertainties get larger as time scales get shorter. A minor point, but the figure 3 in my post shows 8+ year trends showing no skill (7 do also).
If the question is “what does eight year skill say about longer term IPCC predictions?” I’d say probably not much.
If the question is “what does eight year skill say about 8 year IPCC predictions?” I say it shows that on such timescales such predictions are pretty poor.
PS. You can call my dad Dr. Pielke, ybut ou can call me just Roger — I do appreciate the gesture. Sure beats some of the other things I’ve been called lately;-)
Atmoz, HADCRUT claims the earth cooled from 1940 to around 1980.
It also warmed from 1910 to 1940.
http://www.cru.uea.ac.uk/cru/climon/data/themi/g17.htm
You prefer to start graphs around 1980, at the end of the cooling and the start of the next warm period.
You cherry pick all the time.
CO2 … supposedly …. kept rising all through the drops and rises.
CO2 doesn’t drive temperature.
[Reply: I typically start graphs at 2 dates: 1979 or 1880. The first corresponds to the beginning of the satellite-derived temperature era. The second is the earliest date when all three of the surface-station based metrics report a temperature. I'd use 1880 all the time except that the only record that has a different trend for the last 25 years is from UAH.
You can say over and over that "CO2 doesn't drive temperature", but this isn't politics. Simply repeating your assertion doesn't make it true.]
couple comments.
1. on model skill I’m betting that Lucia’s Lumpy model has more skill than a modelE. or to hedge my bets more skill per cpu minute.
2. Atmoz. get your butt over to lucias and talk to her and johnV. Its actually a good discussion and your name comes up often.
Steven– I doubt if Lumpy has more skill than Model E. But more skill per cpu minute? Maybe. 0.001/0.00000000..[...]…0001 is a hard number to beat.
I guess we’ll see.
I’m going to be posting more naive projections later this week. One of my visitors has done some econometrics fits. No physics. totally naive. Could be junk– but, in the long run, correct physics should beat naive. That’s the theory anyway.
Oh– Steve. I should have added: It is my goal to manipulate Atmoz into doing the spectral computations to let me create “warmer error bars”. We’ll see if I succeed.
I should have added: It is my goal to manipulate Atmoz into doing the spectral computations to let me create “warmer error bars”.
As much as I hate binary logic, either lucia is a terrible poker player or this is not her goal at all. In the first case, who cares. In the second case, caveat commentor.
Best,
D
Atmoz,
What about 14 year trends (twice the 7 year trends mentioned above)? Is that long enough for falsification? Why I ask? If 2008 anomaly (RSS data) is -0.2 this year, then the trend for the last 14 years (1995-2008) is slightly negative (at -0.014 C per decade). I’m not saying that this will happen, but it shows how cooler global temps in 2008 and 2009 have the capability to expand the cooling trend to before 2001. As shown by the large positive sloped blue line in your above graph, it will take many years of cooling to extend the trend to before 1992 - 1994 (thus, not likely to happen any time soon, or if at all). HOWEVER, there is a real possibility that in 2010 we will be discussing no warming for 15 years (1995-2009). To do so, we would only need +0.1 anomaly for 2008 and 0.0 for 2009. One can only imagine the excuses (er, reasons) that would be flowing from the RC bunch then.
Roger, Lucia, and Steven,
Tell me that I am not the only one who has made this observation (even if not posted). Seems to me that it would important for future discussions on falsification.
Chris,
In 2 years, 14 year trends would not show anything. Please see this image which is linked above in the comments. This was made about 6 months ago. The big dip was caused by the extreme El Nino in 1998. In 2 years, that minimum will occur when calculating 14 year trends. Lucia has a reason for calculating such short trends. Picking years so that the results occur at a minimum is cherry picking at its worst. If you want to talk about 14 year trends, do so now. Or in 2 years, talk about 10 year trends. Or 20 year trends. But don’t change the time frame from year-to-year to get trend results that you like.
Atmoz,
You say you use starting times related to technical developments that allowed what you might call “accurate measurements”
But that doesn’t negate the point that as it happened, by coincidence, both of those starting times began as we were moving out of a colder period that had lasted from roughly 1945 -1975.
So while you use technical arguments it could also be said that to a certain extent this coukd be akin to cherry picking.
[Reply: The link above has the trends calculated from 2 months to about 30 years in increments of 1 month. That's not cherry picking. I could extend that analysis back to 1880, but it would change very little: the asymptote would be around 0.16 instead of 0.2.]
Atmoz,
Two points (please keep separate):
1) If a 15 year negative trend doesn’t disprove the climate models, then what will? 20 years? 30 years?
2) Going in the opposite direction, let’s say that 5, 10, or 15 year trends have no predictability for similar time perioids in the future. If that was the case, if 2008 does indeed have 0.0 anomaly, then what does that say about global warming? If one doesn’t put much faith in past trends (of however duration) predicting the future, then one can only conclude that the climate that ended in 2008 was on average the same for the period between 1979 and 1998. On the surface, the last comment doesn’t jive with global warming.
So, to pin you down, which side of the argument do you agree with, number 1 or 2?
Off-topic, but maybe you can be the first to answer my question: in the last 40 years, for example, have aerosols increased or decreased? Based on a number graphs and presentations that I’ve seen posted on the internet regarding the matter, I’ve concluded that all climate models assume increasing aerosol load (and with no basis in fact). Some of the negative forcings are quite large based on a paper on the GISS website. When I asked the same question on RC, all I got was a link to graph showing climate model results that assumed (predicted?) increasing aerosols over the past 100 years. Funny, but the only real data on the graph went back to the mid-80’s (with slight trend towards less aerosols). When I asked “when did model results substitute for data?” my reply was not posted. When I asked if there was one model simulation that assumed constant or even negative-sloped aerosol loading (i.e., less dimming), I got no reply. Life will not be kind to those guys with each subsequent year of cooling.
Just out of curiosity, Chris, you’ll respond to continued warming by…?
Re the aerosols, have you tried googling? I think maybe half an hour or so would answer your questions. The AR4 would be a great place to start.
BTW, were I an RC co-author I would respond to “when did model results substitute for data?” by ignoring it.
Chris, just to mention also that focusing on calendar year stats can also be a bit misleading. For example, when was the warmest January anomaly? April? January through April? Look it up (or just look closely at Atmoz’ graph and you can probably figure it out). What does that tell you?
Steve,
A few comments to your points above:
1) Regarding monthly vs. yearly data, I once plotted Dec-Feb monthly data (i.e., cold months) and got the same trend as plotting all of the monthly data over 29+ years. I concluded that doing similar exercises for other months, etc. would not bear anything useful.
2) I think your point above is to direct my attention to the outlier year of 1998. Here’s an idea: since 1998 is such a problem for you and Atmoz, let’s delete it from the data table and plot our trends as if 1998 never existed. Interestingly, assuming a 0.0 anomaly for 2008 and -0.1 for 2009, the trend line is flat (0.006 C/dec) between 1995 and 2009 (or 15 years). Atmoz’ cherry-picking argument becomes rather moot, does it not, when the cherry is removed from the dataset? The general question remains the same: at what point can the climate models be falsified?
[Reply: Why would you assume an anomaly of 0 for 2008 and a negative anomaly for 2009? Lucky for you, I've already calculated the trends after removing the 1998 ENSO, so I know you're full of crap.]
3) My response to warming or cooling is the same, i.e., trying to understand the role of CO2, land use changes, the sun, and aerosols in our climate.
4) The role of aerosols is vastly important in understanding the accuracy of the climate models. Large negative forcings are assumed for aerosols. Changes in aerosol loadings (and by type) has a large effect on model results.
5) Despite the point in (4), one would think that there would be ready answers or graphs with respect to aerosols. I have not seen any to date. Of the smattering papers, graphs, etc. that I have seen on the matter, they’re all contradictory.
6) I was told by a poster on RC that the “data” shown in a link in his posted response “proved” me wrong. As mentioned above, it was mostly modeling results.
7) Please see the graph below:
http://en.wikipedia.org/wiki/Image:Climate_Change_Attribution.png
How is it possible for sulfate emissions to be worse in the mid-90s than in the sixties? Since then, there have been two Clean Air Actis (1970 and 1990), the fall of the Soviet Union, and cleaner burning gasoline and/or engines. That was my original question on RC. I thought there would be ready answers. I obviously thought wrong. Thus, my search continues…
In comparing two hypotheses, do keep
http://en.wikipedia.org/wiki/Akaike_information_criterion
in mind: every additional parameter has to explain exponentially more of the variance in order to have the same AIC value.
Once again, both Atmoz and James Annan (and judging by the comments on various blogs, many others too) have misunderstood the purpose of Pielke Jnr’s posts. In case you didn’t notice, he is involved in public policy and science. Therefore, the posts he makes have more in common with what you are likely to encounter in the political arena than the science lab.
It is my belief that Roger is doing you a favour - by using the methods of politics, he is showing you what will happen when your science heads towards politics, and is doing it in advance, allowing you to get “all your little ducks in a row”.
Does he ask silly questions? The wrong questions? Maybe, but only from a climate science perspective, not a political perspective. If you want action to be taken, this is what you will face. If you want action, these are the sorts of questions you better have good answers for. And not just good - readily understandable by someone with minimal training in the sciences.
Is Roger a PITA? Of course - that’s his job. Use this resource or lose it is my advise.
Atmoz,
As noted in my first post, I’m making the anomaly assumptions for the sake of positing a question - i.e., are IPCC models closer to falsification with 14 or 15 yr trends? In other words, when will this debate be over if the cooling trend continues?
Second, I used RSS data for my trendlines. I admit I wasn’t as sophisticated as you in removing the 1998 anomaly via smoothing. I simply removed the data from the RSS dataset. Crude, I know.
Thanks for posting my previous responses.
Chris, my purpose in pointing you to the monthly anomalies was so that you would notice that we had record high temps globally during early 2007. A reasonable person would conclude from this that not much weight should be given to an apparent short-term near-flat trend using annual anomalies.
Re the aerosols, is there some reason you don’t want to look at the AR4? Re the sulfates, your POV seems a bit U.S.-centric. Have you not noticed that we’ve exported our air pollution problem to a great extent?
Up to now, the warmists have had the warmest reception possible by the media. I don’t think they are prepared (or anybody for that matter) for the treatment that they will receive once the media turns on them (assuming future cooling of course). They seem blissfully ignorant of the storm that is heading their way.
—–
Up to now, the Disney cultists have had the warmest reception possible by the media. I don’t think they are prepared for the treatment that they will receive once the media turns on them (assuming future boring movies of course). They seem blissfully ignorant of the storm that is heading their way.
—–
Up to now, the Yankees have had the warmest reception possible by the media. I don’t think they are prepared (or anybody for that matter) for the treatment that they will receive once the media turns on them (assuming future losing seasons of course). They seem blissfully ignorant of the storm that is heading their way.
—–
Up to now, the staring-into-crackberry types have had the warmest reception possible by the media. I don’t think they are prepared (or anybody for that matter) for the treatment that they will receive once the media turns on them (assuming future new shiny baubles coming to market of course). They seem blissfully ignorant of the storm that is heading their way.
—–
IOW: pffffft.
HTH with the quality of the golly-I-must-wish-really-hard argumentation.
Best,
D
Dano,
Did it ever occur to you why Atmoz felt compelled to start this post in the first place? My only point (granted an obvious one) was that there will more of these type of posts if the world is cooler in 2008 and 2009. Atmoz is right. If the warming trend continues in 2009, 7 or 8 years of negative trends does not falsify IPCC projections. However, I get the impression that Atmoz wishes this whole cooling thing goes away just so he can stop refuting this nonsense and get back to things he enjoys best (i.e., posting interesting things about the climate). All I’m saying is that by late 2009, you could be looking at negative trends in the period of 14-15 years (if you don’t believe me, do the numbers yourself with the monthly RSS dataset). Are you prepared for that debate? A few months ago, the debate was if there was a cooling trend at all. Now, its agreed that the cooling trend is 7 years. In 18 months, will it be 15 years? Atmoz may not think that is relevant, but others will. I’m not arguing anything, just looking ahead, that’s all.
Steve,
I’ll curb my comments until I get the chance to see AR4. But what do you want me to do if I find the same unfounded assumptions that I’ve seen so far? I admit my assumption is unfounded, but I’m not justifying a climate model that is dependent on aerosol forcing.
Chris:
Cooling from when?
Today, the rate of warming is less.
It isn’t cooling.
I say again: it isn’t cooling.
The top ten warmest years in the instrumental record have all been since 1995. Unless we wish reeeeeally hard for a re-definition of cooling, then, Chris, you can’t speak intelligently to this issue. Sorry.
I explained this erroneous ‘cooling’ argument elsewhere, here. Note the bolded text in the fifth bullet of that comment:
Why, again, must we look at the temperatures where people live? Because people don’t live in zeppelins, dirigibles, or airships where satellites take measurements. Folks who fail to understand this, IMV, self-marginalize themselves.
Best,
D
Steve,
I suppose you’ve read the post on CA regarding the adjustment to ocean surface temperatures. Everyone over there think the modelers amped up the volume of aerosols to produce the cooling trend in the 1970’s that supposedly is not there anymore. Since the models will have to be tweaked again with less aerosols, that means CO2 sensitivity needs to be dialed down as well (since CO2 and aerosols produce countering effects). So, should I still read AR4 now knowing the aerosol section is bunk?
[Reply: Why don't you read it and realize that not everything is black and white, right or wrong, with us or against us?]
Dano,
To use your test, I plotted the 5-year running mean for the RSS dataset. Assuming the ytd 2008 anomaly holds for the rest of the year, then there has been a cooling trend since 2004. Further, if one assumes a zero anomaly for 2009, that extends the cooling trend back to 2002 (or 8 years). Besides, what is the justification for using a 5-year trend? I live in the present, not 5 years ago (to use your logic). Finally, are not your responses to my simple cooling statistics proving my point about an ensuing meltdown on RC and other sites if 2009 is cooler than 2008? Hey, everybody is happy when you work at a company with a rising stock price. When the stock price starts to fall, watch out below. Bear Stearns just approved a $10 per share sell price to JP Morgan. Only 45% of the employees are expected to keep their job. Plot Bear Stearns stock price for the past year. Again, RC may not know this, but bad things happen when your metrics start going in the wrong direction. If you think I’m full of shit, let’s wait a year and see what happens.
Dano:
>>Today, the rate of warming is less.
Interesting point . .
So, are you now agreeing with me? Do you think the underlying rate of warming has not increased to 2C/century?
/Lurker mode off.
Chris, read AR4. You’ll find the data you want in Chapter 2.
It shows, BTW, the graph you linked to is garbage. The graph shows increasing negative forcing due to Sulfates but AR4 states
“The most recent study (Stern, 2005) suggests a decrease in global anthropogenic emissions from approximately 73 to 54 TgS yr–1 over the period 1980 to 2000,”
Whether people like it or not, emissions are going down and therefore so is the negative forcing.
/Lurker mode on.
I plotted the 5-year running mean for the RSS dataset.
Chris,
people don’t live in dirigibles, zeppelins or airships. They live at the surface. Thank you for analyzing one particular dataset, now what are the trends where people live, grow their food, walk their dogs?
Lucia,
I don’t know what you’re talking about. I don’t read your comments.
Best,
D