A free-market energy blog
Random header image... Refresh for more!

A Cherry-Picker's Guide to Temperature Trends (down, flat–even up)

Accusations of cherry-picking—that is, carefully choosing data to support a particular point—are constantly being hurled around by all sides of the climate change debate. Most recently, accusations of cherry-picking have been levied at analyses describing the recent behavior of global average temperature. Primarily, because claims about what the temperature record says run the gamut from accelerating warming to rapid cooling and everything in between—depending on who you ask and what point they are trying to make.

I am often asked as to what is the “right” answer is. What I can say for certain, is that the recent behavior of global temperatures demonstrates that global warming is occurring at a much slower rate than that projected by the ensemble of climate models, and that global warming is most definitely not accelerating.

Choice of Cherries

But as to questions concerning just how far beneath climate model predictions the rate of warming is, or for just how long the average temperature of the world has not warmed at all, the answers depend on several things, among them the dataset you want to use and the time period over which you examine—i.e., which cherries you wish to pick.

Figure 1 illustrates the various cherry varieties that you have to choose from. It shows the global temperature history during the past 20 years as compiled in five different datasets (three representing surface temperatures, and two representing the temperatures in the lower atmosphere as measured by satellites—the latter being relatively immune form the data handling issues which plague the surface records).

cherry-pick_fig1_small
Figure 1. Global temperature anomalies from September 1989 through August 2009 as contained in five different data compilations. The GISS (Goddard Institute for Space Studies), NCDC (National Climate Data Center), and CRU (Climate Research Unit) data are all compiled from surface records, while the RSS (Remote Sensing Systems) and UAH (University of Alabama-Huntsville) data are compiled from satellite observations of the lower atmosphere.

To give you some guidance as to which cherries to use to make which ever point you want, I have constructed a Cherry-Pickers Guide to Global Temperature Trends (Figure 2).

 

cherry-pick_fig2_small

Figure 2. Cherry-Pickers Guide to Global Temperature Trends. Each point on the chart represents the trend beginning in September of the year indicated along the x-axis and ending in August 2009. The trends which are statistically significant (p<0.05) are indicated by filled circles. The zero line (no trend) is indicated by the thin black horizontal line, and the climate model average projected trend is indicated by the thick red horizontal line.

It shows the current value (though August 2009) of trends of various lengths from all of the five commonly used global temperature compilations. I compute the trends as simple linear least-squares fits through the monthly global average temperature anomalies for each dataset (from Figure 1). Each point in Figure 2 (for each dataset) represents the trend value for a different length period, beginning in September in the year indicated along the horizontal axis and ending in August 2009.

Starting in September of particular year and ending in August of this year produces a trend with a length expressed in units of whole years. For example, a trend starting in September 1999 and ending in August 2009 include 120 months, or 10 complete years. The values for the 10-yr trend for each dataset are plotted on the chart directly above the value on the horizontal axis labeled 1999. If the trend value is statistically significant at the 1 in 20 level (p<0.05), I indicate that by filling in the appropriate marker on the chart.

I also include several other items of potential interest to the cherry harvesters; first is the dotted horizontal line representing a trend of zero—i.e., no change in global temperature, and second, the thick red horizontal lines which generally indicates the average trend projected to be occurring by the ensemble of climate models. Bear in mind that red line only represents the average model expectation, not the range of model variability. So it shouldn’t be used to rule out whether or not a particular observed value is consistent with model expectations, but does give you some guidance as to just how far from the average model expectation the current trend lies (a cherry picker is not usually worried about the finer details of the former, but, instead, the coarser picture presented by the latter).

General Conclusions

Here are a few general statements that can be supported with using my Cherry-Pickers Guide:

• For the past 8 years (96 months), no global warming is indicated by any of the five datasets.

• For the past 5 years (60 months), there is a statistically significant global cooling in all datasets.

• For the past 15 years, global warming has been occurring at a rate that is below the average climate model expected warming

And here are a few more specific examples that the seasoned cherry-picker could tease out:

• There has been no (statistically significant) warming for the past 13 years. [Using the satellite records of the lower atmosphere].

• The globe has been cooling rapidly for the past 8 years. [Using the CRU and satellite records]

Or on the other side of the coin:

• Global warming did not ‘stop’ 10 years ago, in fact, it was pretty close to model projections. [Using the GISS and NCDC records beginning in 1998 and 1999]

• Global warming is proceeding faster than expected. [Using the GISS record staring in 1991 or 1992—the cool years just after the volcanic eruption of Mt. Pinatubo]

I am sure the more creative of you can probably think of many others.

Judging the Cherry Pickers

Another use of my Cherry-Pickers Guide besides choosing your own analysis, is to check and see what level of cherry-picking was required to support some statement of the behavior of global temperatures that you saw somewhere.

For instance, in a recent post over at RealClimate.org, Stefan Rahmstorf used about 10-yr to 11-yr trend in the GISS dataset to support the idea that global warming was proceeding pretty much according to plan, concluding “the observed warming over the last decade is 100% consistent with the expected anthropogenic warming trend of 0.2 ºC per decade, superimposed with short-term natural variability.”

A quick check of my Guide would show how carefully Rahmsdorf’s selection was made. Trends a few years longer or a few years shorter that the period selected by Rahmstorf would not have borne out his conclusion with as much conviction.

Another example of careful data selection can be found in recent claims made by Richard Lindzen who is fond of stating that “there has been no statistically significant net global warming for the last fourteen years.” A quick check of my Cherry-Pickers Guide shows Lindzen to be particularly crafty because there is no support for such a statement in any of the five datasets. So how did he arrive at that conclusion? By using annual data values instead of monthly data. Using fewer data points (14 annual values instead of 168 monthly ones) doesn’t affect the actual trend value so much, but it does affect the statistical significance of the trend. The fewer data points you use, the less significant the trend is. So by using annual data (from the CRU or satellite datasets), Lindzen is able to cite a 14-yr temperature trend that is not statistically significant.

The statements by Rahmstorf and Lindzen are not wrong, per se, but neither are they particularly robust.

So next time you encounter some claims about what recent temperatures tell us about global warming, or want to make one yourself, check my Cherry-Pickers Guide to get a full appreciation for the degree of grounding that such statements enjoy. And for those folks who want to push the envelope a bit, you’ve got to hope that your audience doesn’t have access to my Guide—otherwise, someone, somewhere, is sure to call you on it!

45 comments

1 Andrew { 10.12.09 at 3:15 am }

I get the weird feeling I prompted this somehow. Perhaps it’s because I observed quite some time ago that, as you show above, the trends in both the satellite and CRU data sets for the past twelve years, or 144 months, are negative, although only ever so slightly and thus not different from zero. I think that’s important because it means we are a mere three years away from the point at which just about everyone will have to admit that something is amiss-the 2008 STOCR threw that figure (15 years) out there anyway.

Will the records make it all the way to fifteen without the trends picking up? I think, personally, that it is hard to say. The reason I think that it could is that the recent El Nino that was supposed to kick warming back into high gear is floundering, and we are barely struggling to get out of the minima that won’t die.

On the other hand, you would think that such a string is unlikely. You’d be falling for the gamblers’s fallacy, though.

2 Steve C. { 10.12.09 at 9:53 am }

I’m not a scientist or a statistician, but I’ve been analyzing numbers for 30 years. Any time I see someone say that actual trends pretty much match what they projected, I’m skeptical. Working with real world data on real world problems just isn’t that simple. And here I’m talking about problems where the variables in a work environment are fairly limited. Add in the complexity of trying to project temperature, in a world where we don’t have any more than a rough guess as to the interaction of many factors, not to mention how those influences might change over time; well I just don’t see how anyone could claim anything more than “we predicted a general uptrend and that’s what our reading of the data indicates”.

Any manager who went to the capital committee with that sort of justification for investing millions of dollars would be ignored.

3 PaulD { 10.12.09 at 10:12 am }

Your graph is remarkably helpful. I am sure you are not looking for new projects, but is there any chance you could extend it back even further?

4 How to deal with a global warming heretic « Take Highway 12 { 10.12.09 at 10:42 am }

[...] in the article, I found “Master Resource: A free-market energy blog.” Which led me to this fascinating post, showing how easily climate data can be cherry-picked to support either side of the discussion. [...]

5 anthill { 10.12.09 at 6:15 pm }

Gavin debunks this piece at Real Climate

http://www.realclimate.org/index.php/archives/2009/10/a-warming-pause/comment-page-7/#comment-138126

When will you be issuing a correction to those you have mislead?

6 Bob Murphy { 10.12.09 at 8:55 pm }

Chip, now I have to demote your Waxman-Markey analysis down to the *second* best MasterResource post ever. This is just great.

I second PaulD’s request, though: Any chance you can go back farther? I realize the satellite data might drop out. But it would be helpful to get a bigger picture.

7 rbradley { 10.13.09 at 8:08 am }

Chip is off to the hills and will return later this week–I’m sure he will address some of the above questions then.

8 goodspkr { 10.13.09 at 11:06 am }

anthill { 10.12.09 at 6:15 pm } Gavin debunks this piece at Real Climate

Ant, how come when skeptics point at a single month of data we are told it’s just weather, but when Gavin does it, it’s debunking.

I read this and it seems to be exactly what Chip is talking about. He is cherry picking his data to come to the conclusion he wants you to have.

First using GISS data which somehow over the past ten years moved from .15 degree difference with UAH data to .4 degree difference (I’ll give you one guess which way).

And why does GISS and NOAA keep changing the rules of how they measure things? No urban heat effect. No satellite data used by NOAA.

Gavin is a bad example of a scientific blog. I saw his defense of the Hockey Stick with some of the most ridiculous other hockey sticks (he used CO2 as one). He tends to go off in his straw man arguments often misquoting people.

So I would read it again and then come back and read this so you can see he’s cherry picking.

9 Snorbert Zangox { 10.13.09 at 11:28 am }

I suppose that one could quibble with the details and minutiae, as Gavin did, but the fact remains that the atmosphere has not warmed appreciably in the past decade or so. In addition, the Argos diving buoys tell us that the oceans have not warmed appreciably during the same period. The folks over at RealClimate.org tell us that we need to be patient; the warming will return – with a vengeance – soon.

All of this implies that some believe that the climate system is accumulating heat somewhere. My question is where is the heat? It is not in the atmosphere, either in the form of warmer temperatures or in the form of higher absolute humidity, nor is it in the ocean.

The carbon dioxide concentration has continued to accumulate at about the same rate that it has for the past century. The climate models clearly have no mechanism for describing the accumulation or loss of the heat that should have entered the system because of the increasing carbon dioxide concentration.

If my facts are correct, the models are woefully inadequate.

Can anyone offer an explanation?

10 Cherry Ping-Pong « Cruel Mistress { 10.13.09 at 11:49 am }

[...] a lot of tit-for-tat going on in the blogosphere over the alleged cherry-picking of data (also here and original, criticized post here).  I’ll remain agnostic on the empirical question, as [...]

11 The Blackboard » Adding Apples and Oranges to Cherry Picking. { 10.13.09 at 11:57 am }

[...] October, 2009 (10:57) | Climate models Written by: lucia Recently, Chip Knappenberger commented on some cherry picking by both Richard Lindzen and the fellows over at RC. Roger [...]

12 Nick { 10.13.09 at 12:17 pm }

One problem, its not time that is claimed as the cause of warming.

The x-axis needs to be log (co2 concentration). We have a pretty good model from Hawaii so long as you use annual trends because CO2 in the atmosphere exhibits a seasonal cycle.

The next question is then what the projection should be for CO2 going forward. It should be the level that gives a 2 degree rise over 100 years.

It actually makes the predictions even worse, because the log (co2) part front loads the warming to the lower concentrations. A one percent rise in CO2 when concentrations are low gives a much larger rise than a one percent rise when CO2 concentrations are higher, if you believe the models

Nick

13 Tom Sullivan { 10.13.09 at 12:29 pm }

anthill – that’s spot at RealClimate not a debunk, it’s a comment, a blurb. And not convincing in the least.

14 Interglacial John { 10.13.09 at 1:13 pm }

anthill – how do you figure there was any “debunking” provided by Gavin? All I see is more cherry-picked data. Sea ice is simply a reaction to climate, it goes up and it goes down, no big deal.
The USS Skate surfaced at the North Pole March 17, 1959 to virtually no ice. That’s right skippy, MARCH 17th, as in WINTER AT THE NORTH POLE, TO VIRTUALLY NO ICE!!!! Guess it’s back to the drawing board for Gavin! Actually knowing history is a very powerful thing.

15 Noblesse Oblige { 10.13.09 at 1:27 pm }

Just for fun, there is a good chance that Richard Lindzen is correct in his statement that there has been no statistically significant warming since 1995. The reason is that annual data do a reasonable job of avoiding the redundancy inherent in monthly data. Monthly data display large month-to-month correlations of the AR(1) type which need to be eliminated in any calculation of statistical uncertainty in the slope. The net effect is to reduce the number of degrees of freedom in the data and therefore to increase the uncertainty of the slope. Taking annual data is not rigorous but gives a better estimate of the uncertainty than monthly data.

16 Climate Research News » A Cherry-Picker’s Guide to Temperature Trends { 10.13.09 at 2:33 pm }

[...] Read the entire article: ‘A Cherry-Picker’s Guide to Temperature Trends (down, flat–even up)’ [...]

17 John M { 10.13.09 at 5:47 pm }

anthill

Here’s a head start on the response (assuming you’re interested in more than just sniping).

http://rankexploits.com/musings/2009/adding-apples-and-oranges-to-cherry-picking/

18 Plocka körsbär | The Climate Scam { 10.14.09 at 1:31 am }

[...] Ett alldeles utmärkt område för cherry-picking är den globala temperaturens utveckling. Beroende på vilket startår man väljer och om man ser på månadsavläsningar eller årsdata kan man visst visa i princip vad man vill. Läs det alldeles utmärkta inlägget som Jan tipsar om – A Cherry-Picker’s Guide to Tem… [...]

19 Anna { 10.14.09 at 9:56 am }

A very interesting article!

What I think you have left out is the risks with using short-term data series on a dynamic system with many different factors acting on different time-scales.

If someone else already have written this I am sorry for copying, I have not had time to read through the comments.

20 John { 10.14.09 at 4:15 pm }

The above article suggests Fig. 2 depicts up to the present, however, Fig. 2 only shows through 2004.

21 Anders L. { 10.14.09 at 4:36 pm }

What do the “trend” datapoints actually say about temperature? Nothing, really, if you think about it for a little while.

22 Ken Maize { 10.14.09 at 5:31 pm }

Interesting and enlightening. I posted a blog and link to Chip’s analysis at http://www.powermag.com. Scroll down the blog section.

23 Jim Brown { 10.14.09 at 6:10 pm }

Anders L. said: ‘What do the “trend” datapoints actually say about temperature? Nothing, really, if you think about it for a little while.’

He is absolutely right. To expand a little bit: The burden of proof is upon he who asserts the positive. The AGW alarmists claim that carbon dioxide is dangerous and offer their evidence, but the evidence, and the “skeptics” offer evidence to the contrary. Neither is conclusive, so we must conclude that there is insufficient evidence of dangerous warming. To put it another way: If carbon dioxide were on trial by a fair jury, it would be declared “innocent.”

24 Andrew { 10.14.09 at 7:33 pm }

John { 10.14.09 at 4:15 pm } Figure two is label with the start year of the trends, and shorter than 5 years is pretty meaningless.

25 cknappenberger { 10.14.09 at 8:57 pm }

Thanks for the comments everyone.

Rob (#7) got it right. I posted this and headed for the hills–Durango, CO, to be exact, for a little R&R!

And it turns out, Lucia (at The Blackboard did an excellent job in filling in the details (I encourage everyone who has not done so to read her post on the topic as I think she provided answers to most everyone’s questions).

Gavin (at RealClimate) was correct that I erred on the less than conservative side in not dealing with auto-correlation when determining my significance values (but Lucia shows that it turns out to have mattered little), while Lucia was correct that I erred on the conservative side when broadly assigning a 0.2C/decade value to all model trends.

I always like to have a little something in my back pocket to deal with potential criticisms (and Lucia sleuthed this one out).

-Chip

26 Chris { 10.15.09 at 1:17 pm }

Chip,

I’m curious why you did not go back 30 yrs (i.e., beginning of satellite data)? From memory, the satellite-derived trend over that time span is about 0.14 C per decade, or 0.014 C per year. Also, instead of trend, why not plot actual delta T divided by number of intervening years? For example, if the delta anomaly (on annual basis) was 0.4 C between 1979 and 1999 (20 years), then the calculation would yield 0.02 C/yr. This would be more robust, I believe, than the linear least squares analysis (since no trend is assumed). However, temp anomaly data is rarely presented this way. I guess global warming alarmists believe there is a trend, when in reality, there may be none.

27 Recent Temperature Trends | Detached Ideas { 10.15.09 at 4:18 pm }

[...] overviews of controversial topics are sometimes difficult to find. Chip Knappenberger’s A Cherry-Picker’s Guide to Temper­ature Trends provides just such an overview of recent temperature trends. Knappenberger first charts all five [...]

28 cknappenberger { 10.15.09 at 4:49 pm }

Chris (#26),

Over at The Blackboard, Lucia takes the trends back to 1980 (and in some cases even earlier than that). So, if you are interested, you should check out her analysis.

And as to your second question, in general a statistical fit to all the data points is a more robust reflection of the behavior than is the difference between just the end points.

-Chip

29 Chris { 10.16.09 at 1:04 pm }

Yes, but Lucia only used Giss and Had data. Regarding the second point, who’s to say that linear least squares is the proper statistical fit? What if the data has a hump in it? Yes, this criticism is not new. Yes, I can do the analysis myself (both points) regading the satellite data. Politically, don’t you think it would be interesting to tell people that the global temp (satellite data) next Jan (for example) was the same as Jan, 1980, or 30 years ago? I think it would make a strong point that most people don’t even think is possible. A similar graph like you produced above would put that datapoint in context.

30 Why Reconstructions Matter Part II « the Air Vent { 10.25.09 at 5:30 pm }

[...] models. They have to evaluate the resulting skill in reproducing the observed mean climate. By http://masterresource.org/?p=5240at Lucia’s http://rankexploits.com/musings/2009/adding-apples-and-oranges-to-cherry-picking/ one [...]

31 Comments On AP Story “Statistics Experts Reject Global Cooling Claims” « The EIW Network { 10.27.09 at 2:14 pm }

[...] note: Also as Chip Knappenberger showed in his A Cherry-Picker’s Guide to Temperature Trends, all the data sets have shown cooling back to 2001. Picking years before that can yeild different [...]

32 Berry { 10.28.09 at 4:45 pm }

A nice analysis. However, it raises one question. As it is based on figure 1, I cannot recognize the GISS data pattern shown in figure 1 when compared to data presented on the NASA GISS website: http://data.giss.nasa.gov/gistemp/

For example: last 5 years of yearly average NASA GISS data still show a positive trend; only the 4-year trend starting 1998 shows a negative trend.

33 cknappenberger { 10.28.09 at 5:00 pm }

Berry,

The GISS data in Figure 1 should be the same as the red data points shown in this graphic from GISS:

http://data.giss.nasa.gov/gistemp/graphs/Fig.C.lrg.gif

-Chip

34 Berry { 10.29.09 at 12:56 pm }

Chip,
Thanks for pointing out. I overlooked the graph, which NASA GISS presents only at the end of its page.
This raises another question (which I should probably address at NASA rather than you) that is how can the monthly mean global temperatures show clearly that there is little or no increase in temperature (and indeed even negative trend in about the last 10 years), whereas other graphs showing yearly indexes etc. all seem to show an increasing temperature trend in the last 10 – 20 years.

35 cknappenberger { 10.29.09 at 1:04 pm }

Berry,

I think that if you work with the numbers yourself, you’ll see that the monthly data indeed produces the annual values that you are familiar with. As to what they appear to show, it is all a matter of your perception (and of course, the period that you are looking at!).

-Chip

36 Bob { 11.02.09 at 1:39 pm }

So, the longer the time interval in the data set, the more likely it is to show warming, with no cooling noted with longer than 8 years worth of data.

I think you prove your point that you can cherry pick data to show cooling, but when you use all the data, it shows warming.

So do a simple trendline for the total data set and what does it show?

37 John E Pearson { 12.26.09 at 2:53 pm }

I am wondering why anyone would ask you anything at all about climate? I searched for you on google scholar and come up with a bunch of uncited work published on cato institute and heritage foundation web sites. Do you have any serious publications in climatology?

38 Robert Bradley Jr. { 12.26.09 at 5:08 pm }

We normally do not approve comments that personally slam the author, and I will let Chip share some of his peer-reviewed work, but post-Climategate, the peer review process is tainted in climatology as Ken Green notes here: http://www.masterresource.org/2009/12/countering-kerrys-catastrophic-climate-claims-part-1-of-2-2/.

39 Chip Knappenberger { 12.27.09 at 1:24 pm }

Mr. Pearson (re: # 37),

I think that in my article I refer adequately to the information I relied on to base my analysis. None of it depends on my contribution to the peer-reviewed scientific literature. So I don’t see your comment as germane to the subject of my post.

If you are interested in further investigating my research, try google scholar using my given name “Paul C. Knappenberger” rather than my nickname “Chip.”

I hope this helps!

-Chip

40 Comment en finir avec la cueillette des cerises « Le Mythe climatique { 01.08.10 at 7:56 am }

[...] [...]

41 Margaret Heft { 01.19.10 at 3:24 pm }

Paul,
I am perplexed about your calculations as I, too, have researched global warming and came to a completely different conclusion, using both Google Scholar and Metasearch Systems. I see your references but what were your calculations to determine that global warming is happening at a much slower pace than previously thought? Actually, even a slower pace would still indicate the necessity of acting immediately to forestall a faster pace.
Global warming is more than yearly average temperatures; what are your variables?

42 On Weather Stations and Climate Trends - Dot Earth Blog - NYTimes.com { 01.28.10 at 10:51 am }

[...] findings are robust enough that a frequent critic of climate overstatement, Chip Knappenberger, has provisionally endorsed the findings and thrown cold water on the idea that bad weather [...]

43 Cherry Picking to Generalize ~ NASA Global Temperature Trends « mind of a Markov chain { 04.06.10 at 8:49 pm }

[...] Knappenberger of MasterResource calculates recent linear trends and compares among [...]

44 Al Black { 05.15.10 at 4:51 pm }

I love your Global Temperature Trend Graph – it compresses a lot of data into understandable information. Has anyone published a similar graph going up to 2010?
I’d really like to see what that shows.
Thanks for the public service you have performed.

45 Marcus Quintilian { 08.10.13 at 9:57 am }

Monthly data are bad because they are correlated but yearly data are not? A lagged cross-panel model analysis shows that the temperatures in May are not causing a rise in the temperatures in April.

Leave a Comment