Have you ever noticed that discussions explaining how recent negative short term trends are somehow to be expected forget to mention volcanoes? Well, today, I’m going to show you a quote, and two figures. Two figures are required because the quote involves both avoiding mentioning volcanoes when mentioning them matters.
The Quote
The quote I will engage comes from Easterling and Wehrner. In their introduction, they observe that chatter at blogs and forums are suggesting the earth is cooling, using recent short term trends to support this claim. EW provide only one example of such a claim, which happens to be an editorial at Investors Business News Daily published in November of 2008. The claim seems to amount to this:
Due to a decline in solar activity and other factors, the Earth is cooling and has been since 1998.
If someone had posted that here on Nov. 2008, I would have simply observed that using data available on that date (i.e. ending in October 2008), the trend since 1998 was positive. Positive trends cannot provide evidence of cooling. No need for a peer-reviewed journal article, right?
Nevertheless, that sort of claim motivated a peer reviewed journal article which advances some rather bold claims about the internal variability of the climate. I’m not going to address the bolder claims in this post; instead, I’ll restrict myself to discussing the introduction which sets the stage for the article.
In the introduction, EW writes:
It is true that if we fit a linear trend line to the annual global land-ocean surface air temperature [Smith et al., 2005] shown in Figure 1 for the period 1998 to 2008 there is no real trend, even though global temperatures remain well above the long-term average. The unusually strong 1997 – 1998 El Nin˜ o contributed to unusual warmth in the global temperature for 1998 at the start of this period resulting in only a small, statistically insignificant positive trend. However, if we fit a trend line to the same annual global land-ocean temperatures for the 1977 – 1985 period or the 1981 – 1989 period we also get no trend, even though these periods are embedded in the 1975 – 2008 period showing a substantial overall warming. Furthermore, if we drop 1998 and fit the trend to the period 1999 – 2008 we indeed get a strong, statistically significant positive trend. It is easy to ‘‘cherry pick’’ a period to reinforce a point of view, but this notion begs the question, what would happen to the current concerns about climate change if we do have a sustained period where the climate appears to be cooling even when, in the end, the longer term trend is warming?
Ok… So, let’s unpack this.
The time period discussed by the unnamed author of the Investors Business News Daily article
The Investor’s business news daily article included a claim based on 116 months.
The figure below illustrates 116 month trends computed based on a multi-model mean computed from model data available from “The Climate Explorer” and also from observations of the earth’s surface temperature:
As you can see:
a) The end date for the 116 month trend specifically referenced in the IBD article is highlighted with a brown vertical line. At the time that editorial was written, the trend since 1998 fell below the multi-model mean projection. This discrepancy has contributed to many blog and forum discussions, many of which make no claims the earth has entered any sort of period of long term cooling. One of the reasons “many” make no claims this particular trend is proves the existence of long term cooling is simple: The trend is positive. Proof of cooling will require a negative trend, and this trend must be of sufficient duration to be statistically significant. (I haven’t seen any such trend yet.)
The argument that the trend since 1998 does not prove cooling is so simple, one wonders why it motivated any sort of detailed analysis or a submission to GRL.
Whatever the motive for writing the GRL article explaining that trend does not prove cooling, we bloggers get to discuss the other more difficult questions.
The more difficult questions for those making projections is this: Is this discrepancy inconsistent with the projections of much larger warming? One might suggest EW seem to engage that with their analysis about 20 year trends; I’ll discuss that further at a later point. For now, and let’s simply look at the peculiarly odd discussion in EW which involves explaining something about trends embedded in the periods from 1975-2008.
What do we see if we apply the ordinary eyeball test to the chart illustrating 116-month trends embedded in the the period 1975-2008? (The possible choice of trends are those with end dates to the right of the olive vertical dashed line. )
One of the things we see is the current trend falls below projections. We also see that for brief periods of time, the observed 116-month trends are negative. However, having super-imposed the model-mean trends, we also notice that the the multi-model mean also exhibits brief negative trends during the earlier periods. Moreover, periods when the negative trends are observed appear to be in rought agreement.
Why? Well, when AOGCMS are driven by volcanic aerosols, the models “predict” the earth’s surface will cool after eruptions similar to those of El Chicon or Pinatubo. These erupted in the 80s and 90s respectively. So, dips in the trend during the 80s and 90s can be explained: El Chicon and Pinatubo caused these dips.
The reason those chattering about the recent low short term trends are chattering about trends since 1998 is that, though perfectly aware negative trends can be embedded within long term positive trends, and they also know that climatologists explain that these often arise as a result of violent volcanic eruptions. They also know there have been no reports of climatologically significant eruptions since Pinatubo erupted in the early 90s.
The current weakly positive short term trends are different from the two periods highlighted in EW. Those were caused by volcanic eruptions; the current ones are not.
So, in short: The existence of nearly decade long negative trends in the 80s and 90s is almost irrelevant to the discussion of hypothetical nearly decade long current negative trends. The only way to make the dicussion seem relevant is to not mention volcanic eruptions..
Oddly, when placing the trend since 1998 into the context of historical variations, Easterling and Wehner did not consider the volcano connection.
This is unfortunate.
Their article appeared to have been aimed at silencing chatter in the blogosphere. But when climatologists at blogs or in the peer review literature try to explain away the recent short term slow warming by explaining that low trends were also observed during the 80s and 90s (but fail to mention that these excursions are explained by the eruptions of Pinatubo and El Chichon), the collective nose of the blogosphere is likely to snort. The collective mind of the blogosphere shares the tacit knowledge that volcanoes erupted in the 80s and 90s and that climatologists often attribute the temperature declines during those periods to the volcanic eruptions. To whatever extent the blogosphere has a collective mind, it can recognize when climatologists are feeding them apples and calling them oranges.
With regard to the introduction in EW, the blogosphere is likely to respond with “can you give examples of negative trends not caused by volcanic eruptions that are also embedded during periods of brisk warming”?
The answer to that question is: “Ehrmm… no”.
The time periods discussed by EW
I know I will get flack if I do not also include a graph showing trends over the second time period discussed in EW. EW chose to frame their discussion of the 116 month trend discussed in Investors Business Daily using slightly shorter. For those who wish to see graphs of the shorter trends, I’m also showing the 9 year trends:
Notice this graphs does not substantively change any points. volcanoes still erupted in the 80s and 90s. Nine year trends did decline with minimums occurring shortly after volcanoes erupted. One notable features: It’s a bit easier to find an actual negative trend using 9 year trends rather than 9 year 9 month trends.
OT, slightly, but how were Dessler et al able to get the following study into GRL based upon measurements from 2003 -2008. You can now measure trends over 5 years?
Bah, link here,
http://geotest.tamu.edu/userfiles/216/Dessler2008b.pdf
I think the window you are using is too short to see the negative trends. If you plot Hadley 20 year trends they go negative from about 1890 to 1910 and again from about 1950 through 1965 with a smaller period in the 70’s. Whether these are associated with volcanoes I don’t know.
BarryW–
EW’s discussion is to describe negative trends embedded in a longer period with briskwarming. If you go back far enough, there are no longer periods of brisk warming. So, of course you can see short term negative trends during periods that have no trend over the long term, and eve more during periods that have cooling trends over the longer term.
BarryW–
The period 2003-2008 is fairly free of volcanic aerosols.
OK, look at the ten year trends from 1950 through 1980 and you’ll see both large high and low trends. Volcanic?
Hmm, well, probably, if you compare with the volcanic aerosol data here:
http://data.giss.nasa.gov/modelforce/strataer/
(As a note, I’m told that pre-Pinatubo, the data is somewhat speculative-though I’m not sure what to make of that).
Lucia-
I read the EW paper and sent them both an email message, pointing out that the main argument they made (a typical climate model predicts ten-year periods of declining temperatures after 2000 with significant frequency, so any current cooling is quite “normal”) is flawed. The model they used does not include volcanic eruptions, but even without volcanoes, the model’s hind-casts of the 20th century are substantially more variable than the 20th century actually was. The 20th century should have yielded MORE variability than the (non-volcanic) model hind-cast, not less. This was evident in the tabular comparison EW made between the 20th century hind cast number of statistically significant 10 year periods of cooling and the actual number of these cooling periods that took place. If the model itself (without volcanoes) is more variable than reality, then predicted probabilities of future cooling periods are almost surely overstated.
The excessive model variability is also evident in EW’s comparison of late 20th early 21st century temperature history and the model’s predicted variability from 2000 onward. EW (cleverly?) used two separate graphs with different x and y axis scales, and the different scales had the net effect of magnifying the apparent variability on the late 20th early 21st century data. This makes the variability of the late 20th century appear to the eye comparable to the model’s predicted variability starting in 2000. When the two graphs are adjusted to match in both x and y scales (that is, no phony magnification), the greater variability of the model forecast is obvious by visual inspection.
EW did not reply to my message… no surprise there I guess.
I am currently traveling in South America, and do not have access to my desktop computer, but if you are at all interested, when I return early next week I can send to you what I sent to them, along with a few other comments. Maybe you could make better statistical argument than me.
SteveF– Do you mean to suggest the variability of ECHAMP relative to the earth looks like this:
Here is the temperature impact of all the different forcings in GISS Model E. Note that the volcanic impact is huge in the models and that the volcanic temp impact has effectively been Zero since 1999.
http://img509.imageshack.us/img509/8388/modeletempimpactst.png
The volcanic forcing estimates are really just made up by Hansen and his top pal, Dr. Makiko Sato. The satellite optical depth and Short Wave radiation data shows that the impact of Pinatubo was over by the end of 1993 – only 18 months – versus the 9 years they have built into the models.
Lucia-
The model graph I saw was a single model run starting in 2000 (through 2100 I thin). The second graph was the historical data (starting in 1980 I think) through 2008.
Bill and BarryW–
One difficult issue with attributing variability to volcanoes prior to 1970 is that the estimates of forcing differ in several references. Measurements of the dust veil or optical depth weren’t as good in earlier times. There is a strong dip in the temperature at the time that might be attributed to Fuego in the 70s, but it would be much smaller according to the multi-model mean. Was the dip Fuego? Or ENSO? I don’t have any idea.
The other difficult issue as you approach the ’40s ’50s is possible sudden changes in bias in sea surface temperatures measurements associated with changing from measuring with buckets to jet inlets, back again, back again etc.
SteveF–
The graph I inserted is 4 model runs during the 20th century and observations. That models is very, very variable even without volcanic eruptions.
Lucia – OT
How goes the tip jar?
SteveF– The tip jar goes well. I have to send out thank you notes and also create the thermometer!
Lucia-
So could the very high variability of this model data not be use to show that EW exaggerates the probability of 10 year (least square linear fit) trends after 2000?
BTW – I was really shocked that such a silly paper was not shot down by reviewers.
SteveF–
I think so, yes. Others may think othewise. We shall see what we shall see. . .
The article cited by EW nominated “a decline in solar activity” as a factor that affects the trend since 1998, but this factor doesn’t seem to have figured in the discussion so far on this thread. Is this not also relevant to the argument that the world has failed to warm in line with model predictions? (I’ve recently been confronted with argument that “[T]he last 10 years should have had a strong cooling trend based on natural influences (which would slow the effect of many feedbacks), so it would not be a big surprise if we saw another big increase once we move out of the low point of the solar cycle”).
Ian– According the models, the solar activity isn’t supposed to have much of an effect. But, it does have some. So I guess we’ll see.
Here is a plot of the 10-year HADCRUT3 trend as a function of time over the last 50 years. (Roughly the same as Lucia’s yellow curve but going back a bit further). You can see that there were several periods in the past when the 10-year trend went to roughly zero, including the 1980s as mentioned by E&W.
But each of these periods follows a major volcanic eruption: Agung, Fuego, Chichon, Pinatubo. There is a nice fit, with the trend starting to fall almost immediately after the eruption. But the current non-warming is not associated with any volcanic eruption, and therefore is unusual (if I was a climate scientist I might even describe it as ‘unprecedented’).
globalwarmingquestions.googlepages.com/trendvolc.jpg
Here’s the link to that image:
http://globalwarmingquestions.googlepages.com/trendvolc.jpg”
Lucia,
I am not so sure I agree with your answer to your qeustion:
While you are correct that it is difficult to find a negative trend not associated with a volcanic eruption in recent decades, the influence of volcanic eruptions occurs at random places in the dataset and perhaps masks other forms of natural variability. So, how about this question: Had the volcanic eruptions not occurred could you find negative trends embedded during periods of brisk warming?
While it may be impossible to answer this question (after all the volcanoes did occur), I think there is some evidence (based on observed ENSO variability) that the answer to that question isn’t such a slam dunk “no.”
-Chip
Chip–
It’s true that they might have happened if the volcanoes hadn’t erupted. The difficulty is the question of whether or not they would, conceivably have happened if the eruption had not occurred is different from whether or not they did occur.
The result is that if we correct empirical data for what we think is the plausible effect of volcanic activity, we are no longer looking at purely empirical data. Such comparisons are useful– but one must make it clear that we are actually making that comparison.
So, with regard to EW, we have the difficulty that they pose their discussion as showing these sorts of negative trends can happen even without volcanic intervention because they actually did happen. But, at best one might plausibly argue that they might have happened even if the volcano not erupted. The first argument would be very, very convincing evidence that such trends can occur and have occurred.
The second one is speculative, requires us to agree to the precise level of correction for the volcanic eruption and suggests we cannot exclude the possibility that they might occur. Would I say I can prove it’s impossible for such trends to occur? I can’t and I don’t. That would be overstating just as much as EW overstated.
We can estimate the probability of such occurances a variety of ways. But EW don’t do that. They substitute an argument that appears simpler, but which is wrong.
Lucia – Good point. We don’t know if there was non-volcanic cooling which happened during the 1980s and 1990s. Although we could take Easterling and Wehrner’s article as stating that those two events were not due to volcanic effects. 🙂
There is a similar situation with the 1945-1975 cooling. The experts claim it was due to man-made aerosols, but that claim obscures whether there were other causes of cooling. They’ve adjusted their models to ascribe that 30 years of cooling as being due to aerosols, and related research also doesn’t detect other cooling effects.
Scooter–
EW don’t state the cause of cooling during the 80s and 90s. They just mention it happened while advancing the case that 20 year periods with zero or cooling trends not caused by volcanic eruptions could be embedded during periods with brisk warming.
Readers who are not aware of the issue associated with volcanoes are likely to infer that the cooling during the 80s and 90s happened during periods unaffected by volcanic eruptions. Otherwise, the reader would need to stop and say “huh?” After all, unless writer (i.e. EW) were to advance a much more nuanced or sophisticated argument about correcting for volcanic aerosols etc. they have just introduced an irrelevant fact into their argument to suggest that absent volcanic eruptions 20 year long periods of no-warming can be embedded on longer term periods of warming.
I recognize that those who believe 20 year volcano-free periods without warming can be embedded in longer periods with brisk warming may be frustrated by lack of data. Maybe they believe believe empirical evidence for such things would exist if only a) the earth had experienced 100 year long periods of brisk warming, b) volcanoes conveniently did not erupt and c) we were there to measure such things. The fact is, we have no direct empirical evidence of any such thing.
Do they explain the drop around the end of the century (19th) the same way? If they do what happened in between?
Lucia,
I agree entirely!
-Chip
Lucia nice nuisanced critique of the Easterling & Wehrner paper. I made a similar comment on this paper some weeks ago on Skeptical Science (Examining the science of global warming skepticism). This is a very weak paper.
In relation to attempting to adjust the data as if the volcanoes hadn’t happened Steve Goddard attempted this on WUWT (Parts I & II)
http://wattsupwiththat.com/2009/01/13/how-did-the-el-chicon-and-pinatubo-volcanic-eruptions-affect-global-temperature/
http://wattsupwiththat.com/2009/01/14/how-did-the-el-chichon-and-pinatubo-volcanic-eruptions-affect-global-temperature-records-part-2/
Geoff–
I agree with Stephen that one of the reasons the computed trends since the mid 70s are as high as they are is that they begin during periods when temperature were depressed by volcanoes. Comparison of models driven by volcanic aerosols to those not driven by volcanic aerosols give the same result.
How large is the effect? Once again, I agree with Stephen that this is difficult to figure out. To estimate it, we need to first decide how we estimate much effect the individual volcanoes had. We could blank out–as Stephen did. We could do a multivariate analysis using the Dust Veil index (one measure of obscuration). We could do one using the optical depth (a second measure of the obscuration.) We could try to find information for time varying distributions of aerosols over the globe. (Like I believe those would be precise. Heh!)
Then, we do math or run model to compare trends with and without the volcanoes. So, in the end, the magnitude of the effect depends on how we estimate their impact.
But I think everyone agrees they did have an effect. My gripe with EW in this paragraph is that, when connected to what they claim to show, it obfuscates. They claim to be engaging discussions at blogs and forums. Climate blog-war addicts are aware that interpretations of recent short term cooling by “most” at blogs and forums all take into account the fact that the recent periods is not trendless or cooling due to the eruption of any volcano. Later in the paper, E&W make much of the “no volcanoes” issue with ECHAMP and model projections into the 21st century.
So, the organization of that paper is structured to cause less knowledgeable readers to be unaware that the example trend in the empirical record are due to volcanic eruptions and so irrelevant to the full argument. Whether E&W did this intentionally or unintentionally, it is unfortunate. Because clearly they will not achieve their apparent goal of convincing anyone in the blogosphere that there is any decent evidence that 20 year negative trends could be embedded in much longer (say 100 year) periods of brisk warming.
I know I am not at all convinced of E&W’s argument. It is very weak on the empirical front and equally weak on the interpretation based on models.