Guest Post by Willis Eschenbach
Once again, Anthony has highlighted a paper claiming that volcanoes have great power over the global temperature. Indeed, they go so far as to say:
From the reconstruction it can be seen that large eruptions, such as Mount Tambora in 1815, or clusters of eruptions, may result in a hiatus of over 20 years, a finding supported by model results.
SOURCE: Determining the likelihood of pauses and surges in global warming, Andrew P. Schurer et al, hereinafter Schurer2015, paywalled here.
Here we go again, sez I, another excuse for the current temperature plateau … and right out of the box, I note that we are dealing with a “reconstruction”, with findings that are “verified by model results”. Be still, my beating heart …
So I thought I’d give their claim a looksee by comparing their recent (post-1800) eruption dates with the Berkeley Earth land temperature record. You know, first see what the observations have to say.
I’m using a variant of the method that they use in the paper, which is to “stack” the temperature records, aligning them at the year of the eruption. I have used the years since 1800 that they have identified in the ice cores as being the years of the big injections of sulfates into the stratosphere from volcanic eruptions. These include inter alia the eruptions of Pinatubo (1991), El Chichon (1982), Mt. Agung (1963), the big eruption of Krakatoa (1883), and the largest eruption during the period, Tambora (1815). Tambora was the volcano responsible for a few areas having a “Year Without A Summer” in 1816. Here are all of the years of big stratospheric eruptions:
Figure 1. Table S3 from Schurer2015.
Now, “Crowley and Untermann” is reconstruction of historical stratospheric eruptions based on sulfate levels in ice cores in Greenland and Antarctica. After a bit of searching I found the data here. Figure 2 shows their best estimate from the data of the tropical AOD, based on the ice core sulfates.
Figure 2. Aerosol Optical Depth (AOD), a measure of the amount of stratospheric aerosols, based on ice core records of the sulfate levels. The dotted horizontal line was the accept/reject line for consideration in Schurer2015. One hemisphere had to be above an AOD of 0.1 to be considered for the paper.
Now, in this graph we can see how much more sulfate aerosols were ejected into the stratosphere by Tambora in 1815 than by any other volcano in the last two centuries.
So I’ve taken those years shown by the dotted lines, and I’ve “stacked’ them in Figure 3, showing the span from five years before to five years after each eruption. In the background of Figure 3 are all of the 11-year spans in the dataset, shown in gray. This indicates the expected variation of the data. Over that, I’ve shown the nine eruptions used for this period in Schurer2015.
Figure 3. Stacked eruptions, aligned on the eruption date at Year 0. In gray in the background are all possible 11-year spans from 1800 to 2014.
There are several things of interest. First, as I’ve discussed before, even the mighty eruption of Tambora doesn’t change the global temperature much if at all. Globally the “Year Without A Summer” is the “Year Without A Difference”. Yes, I understand that in some locations the summer was cooler. But it seems the anecdotes have far outpaced the reality. In Missing The Missing Summer I showed eight datasets which cover the time period. None of them shows an unusual summer in 1816. And in Figure 3 you can see that according to the Berkeley Earth data, globally the “missing summer” of Tambora is scarcely a blip.
More to the point, the average of the stack shows that in general there’s no significant effect from the volcanoes. Nor is there any “dose-related” effect. Tambora, with the largest AOD, doesn’t have a significant temperature drop after the eruption. On the other hand, the largest post-eruption drop is from the eruption with the smallest AOD, that of 1903.
I know that this is difficult to accept after years of hearing how volcanoes strongly affect the climate, but here’s what I say is happening. When the stratospheric aerosols cut down the sun slightly, the tropical surface cools slightly. Because the tropical surface is cooler, the cumulus clouds form later in the day, allowing in more sun. So the weaker sun is compensated for by the longer time that it is striking the surface, and as a result there is very little change in the global average temperature.
In this manner, despite even the size of the 1815 Tambora eruption, and despite the fact that local areas showed cooling after Tambora, the stratospheric aerosols from this huge eruption had little effect on the global average temperature. More stratospheric aerosols are simply balanced out by less tropospheric clouds, and the beat goes on.
Conclusions? Well, my main conclusion is that Schurer2015 way overstates the effect of volcanoes on the global surface air temperature. For them to have a 20-year effect is very doubtful, particularly given the short lifespan of the stratospheric aerosols, as shown in Figure 2. In no case do they remain airborne for more than six years or so. How could their effect last another fourteen years?
In fact, as Figure 3 shows, on average there is no global effect at all visible in the first five years.
So I’ll say, as I’ve said before, that while volcanoes can certainly affect local areas, rumors of the power of volcanoes to affect global average temperatures have been greatly exaggerated.
Best to all,
w.
A Strong Request: If you disagree with something that has been said, please quote the exact words you disagree with. That lets us all be clear about a) the person you are addressing, and b) just what they said that you object to.
Further Reading: As I mentioned above, I’ve written extensively on how little the eruptions affect the temperature. Here is a list of my previous posts on the subject:
Today I thought I’d discuss my research into what is put forward as one of the key pieces of evidence that GCMs (global climate models) are able to accurately reproduce the climate. This is the claim that the GCMs are able to reproduce the effects of volcanoes on the climate.…
Prediction is hard, especially of the future.
[UPDATE]: I have added a discussion of the size of the model error at the end of this post. Over at Judith Curry’s climate blog, the NASA climate scientist Dr. Andrew Lacis has been providing some comments. He was asked: Please provide 5- 10 recent ‘proof points’ which you would…
The claim is often made that volcanoes support the theory that forcing rules temperature. The aerosols from the eruptions are injected into the stratosphere. This reflects additional sunlight, and cuts the amount of sunshine that strikes the surface. As a result of this reduction in forcing, the biggest volcanic eruptions…
Dronning Maud Meets the Little Ice Age
I have to learn to keep my blood pressure down … this new paper, “Abrupt onset of the Little Ice Age triggered by volcanism and sustained by sea-ice/ocean feedbacks“, hereinafter M2012, has me shaking my head. It has gotten favorable reports in the scientific blogs … I don’t see it at…
Since I was a kid I’ve been reading stories about “The Year Without A Summer”. This was the summer of 1816, one year after the great eruption of the Tambora volcano in Indonesia. The Tambora eruption, in April of 1815, was so huge it could be heard from 2,600 km…
New Data, Old Claims About Volcanoes
Richard Muller and the good folks over at the Berkeley Earth Surface Temperature (BEST) project have released their temperature analysis back to 1750, and are making their usual unsupportable claims. I don’t mean his risible statements that the temperature changes are due to CO2 because the curves look alike—that joke has…
BEST, Volcanoes and Climate Sensitivity
I’ve argued in a variety of posts that the usual canonical estimate of climate sensitivity, which is 3°C of warming for a doubling of CO2, is an order of magnitude too large. Today, at the urging of Steven Mosher in a thread on Lucia Liljegren’s excellent blog “The Blackboard”, I’ve…
Back in 2010, I wrote a post called “Prediction is hard, especially of the future“. It turned out to be the first of a series of posts that I ended up writing on the inability of climate models to successfully replicate the effects of volcanoes. It was an investigation occasioned…
Volcanoes: Active, Inactive, and Retroactive
Anthony put up a post titled “Why the new Otto et al climate sensitivity paper is important – it’s a sea change for some IPCC authors” The paper in question is “Energy budget constraints on climate response” (free registration required), supplementary online information (SOI) here, by Otto et alia, sixteen…
Stacked Volcanoes Falsify Models
Well, this has been a circuitous journey. I started out to research volcanoes. First I got distracted by the question of model sensitivity, as I described in Model Climate Sensitivity Calculated Directly From Model Results. Then I was diverted by the question of smoothing of the Otto data, as I reported…
The Eruption Over the IPCC AR5
In the leaked version of the upcoming United Nations Intergovernmental Panel on Climate Change (UN IPCC) Fifth Assessment Report (AR5) Chapter 1, we find the following claims regarding volcanoes. The forcing from stratospheric volcanic aerosols can have a large impact on the climate for some years after volcanic eruptions. Several…
I see that Susan Solomon and her climate police have rounded up the usual suspects, which in this case are volcanic eruptions, in their desperation to explain the so-called “pause” in global warming that’s stretching towards two decades now. Their problem is that for a long while the climate alarmists…
Eruptions and Ocean Heat Content
I was out trolling for science the other day at the AGW Observer site. It’s a great place, they list lots and lots of science including the good, the bad, and the ugly, like for example all the references from the UN IPCC AR5. The beauty part is that the…
Well, we haven’t had a game of “Spot The Volcano” in a while, so I thought I’d take a look at what is likely the earliest volcanic eruption for which we have actual temperature records. This was the eruption of the Icelandic volcano Laki in June of 1783. It is claimed to…
[also, see update at the end of the post] Anthony recently highlighted a couple of new papers claiming to explain the current plateau in global warming. This time, it’s volcanoes, but the claim this time is that it’s not the big volcanoes. It’s the small volcanoes. The studies both seem to…
Enjoy.
UPDATE: I don’t think Willis will mind if I include this excellent graph prepared by Joe D’Aleo of Weatherbell:
As you can see in the bottom panel, the aerosol loading during “the pause” is close to zero. So much for Schurer2015 and their claims – Anthony
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
http://icecap.us/images/uploads/HOW_VOLCANISM_AFFECTS_CLIMATE.pdf
Here is the data and the real story showing volcanic /climate connections. It is quite clear and has the data to back it up with.
To further argue this point is a waste of time.
Willis say below which I say you are 100% wrong and all data sources other then yours show that what you present is false.
I know that this is difficult to accept after years of hearing how volcanoes strongly affect the climate, but here’s what I say is happening. When the stratospheric aerosols cut down the sun slightly, the tropical surface cools slightly. Because the tropical surface is cooler, the cumulus clouds form later in the day, allowing in more sun. So the weaker sun is compensated for by the longer time that it is striking the surface, and as a result there is very little change in the global average temperature.
In this manner, despite even the size of the 1815 Tambora eruption, and despite the fact that local areas showed cooling after Tambora, the stratospheric aerosols from this huge eruption had little effect on the global average temperature. More stratospheric aerosols are simply balanced out by less tropospheric clouds, and the beat goes on
>>Ian Wilson July 22, 2015 at 5:29 am
Only one major problem with Willis’ methodology. What if major volcanic eruptions are a symptom of extended cooling periods. This would mean that each large volcanic eruption in the tropics (or cluster of eruptions) would be preceded by an extended period of global cooling.(say five, ten or even fifteen years).
What Willis should do is normalize his curves in figure 3 to temperatures five (ten or even fifteen) years prior to the eruption rather than the time of eruption.<<
No. Willis is showing there is no correlation between eruptions and lower temp. Barring the observation about reduced temp. ranges post eruption that would seem to be the case. Your observation suggests that dropping temps and volcanic activity are related to something else. Others have suggested that volcanic activity increases during periods of low solar activity. Can willis' data be plotted against the correspomding solar cycles?
The human mind is pre-set to look for patterns as a way to simplify the flood of data it encounters. Even if a perceived pattern leads to logical fallacies (‘post hoc ergo propter hoc’, the fallacy of accident, false cause, etc.), if it *seems* to show a correlation once, we tend to accept a coincidence as a connection and use it for future generalizations without subsequent re-examination.
When a coincidental relationship fails to repeat, we find it more convenient to add caveats, codicils, and conditions than to entertain the idea that we may have been wrong on the first guess.
This is how superstition, mythology, and the belief in miracles and magic are born.
I agree with you Willis! Good post!
Deep oceans and volcanoes have something new in common. They both have great potential for excuses and with limited data reliability for global application.
Page keeps returning to the ad forcing me to scroll back to where I was before. Unaccepable.
I use Chrome with Ad Block Plus extension. No ads, just the ‘about these ads’ link at the top.
http://www.techienews.co.uk/9736585/link-volcanic-eruptions-climate-change-unearthed-ice-cores/
The volcanic /climate connection may be short lived but it is real .
TABLE 3. Effects of Large Explosive Volcanic Eruptions on Weather and Climate
Effect Mechanism Begins Duration
Reduction of diurnal cycle blockage of shortwave and emission of longwave
radiation immediately 1–4 days
Reduced tropical precipitation blockage of shortwave radiation, reduced evaporation 1–3 months 3–6 months
Summer cooling of NH tropics and subtropics blockage of shortwave radiation 1–3 months 1–2 years
Stratospheric warming stratospheric absorption of shortwave and longwave
radiation 1–3 months 1–2 years
Winter warming of NH continents stratospheric absorption of shortwave and longwave
radiation, dynamics 1 year one or two winters
Global cooling blockage of shortwave radiation immediately 1–3 years
Global cooling from multiple eruptions blockage of shortwave radiation immediately 10–100 years
Ozone depletion, enhanced UV dilution, heterogeneous chemistry on aerosols 1 day 1–2 years
Willis – you are obviously aware of some fairy tales about volcanoes. The biggest fairy tale of all is that volcanic eruptions are followed by an imaginary “volcanic cooling.” It simply does not exist as I proved in my book “What Warming?” in 2010 (see pp. 17-21). Any and all alleged volcanic “coolings” identified on global temperature graphs are simply misidentified La Nina valleys. Comes from their neglect and ignorance of the fact that all global temperature curves are a concatenation of the two ENSO phases. Best known is the “cooling” that follows Pinatubo. Roy Spencer insist on marking this 1993/94 La Nina valley as “volcanic cooling” on his monthly El Nino chart despite my having explained to him that he is wrong.
“They cannot be identified, either visually or by any other method, in the global temperature records”
Right, with your eyes closed it probably is,
http://virakkraft.com/BEST-aerosols-1800-1850.png
All your data seems to show is that, at best, without historical knowledge of which years were ‘eruption heavy’ it would be impossible to identify them.
Also it seems that the climate already knew about 1809/15 and 32 sometime in advance and so did some pre-emptive cooling. Warming in the case of 1832!
Missing word there, sorry, “…it probably is difficult”
Like I said earlier, there are also ENSO, AO/NAO, PDO/NPI …
Here is adding an ENSO recon from KNMI. (Note the BEST coverage in 1800 is only eastern north america, europe and india as far as I can see, so the global impact was perhaps around the half of this)
http://virakkraft.com/BEST-aerosols-ENSO-1800-1850.png
That’s hilarious, LGL. From just after the date of the first eruption shown 1809 to the time after the Tambora eruption, biggest one during the last two centuries, the temperature is rising.
And you claim that this RISE in temperature from about 1810 to 1825 shows that two huge volcanoes erupted during this time?
Well spotted … not.
w.
No, the scaled ENSO minus the BEST record shows that huge drops in forcing occurred.
… and the obvious explanation is the volcanoes.
http://climate.envsci.rutgers.edu/pdf/ROG2000.pdf
Every major volcanic eruption has been documented with a degree of global cooling..
The nonsense about the lack of a major volcanic eruption not producing a global cooling effect to some degree is just that, nonsense. Table 3 from the above detailed volcanic study shows the clear effects a major volcanic eruption will have on the atmosphere and has overwhelming support other then the few that have posted their counter assertions here which are 100% false.
@SDP
Willis did not deny that volcano-injected particles decreased solar radiation to the surface, which is the subject of your reference.
He suggests that decreased solar radiation is compensated by fewer tropical clouds and thus lower cloud albedo.
And his conclusions are wrong in the short run as table 3 clearly shows from the study I sent in my above post at 3:10pm
As I geologist what I find interesting is that it is almost impossible to spot the KT boundary (with dinosaur extinction) without resorting to Iridium geochemistry. There appears to have been such a small perpetuation of the sedimentary record that you cant spot it – which implies that global climate was not as greatly affected as we are expected to believe. I don’t expect volcanoes would have a bigger effect.
@A Crooks
July 22, 2015 at 4:18 pm
“As I geologist what I find interesting is that it is almost impossible to spot the KT boundary (with dinosaur extinction) without resorting to Iridium geochemistry”.
I assume you’ve read Richard Muller’s Book “Nemesis – the death star”, in which the Alvarez project to prove the collision cause of the Dinosaur extinctions is described.
I’m frustrated by my inability to find any follow up on the Alvarezes’ tentative explanation of a dwarf twin star to the sun with a periodicity of 26 million years as the motor behind such collisions (after a projectile travel period of 4 million years to reach the inner planets).
Have other Iridium layers coinciding with major extinctions been found, and do any of them support this cycle?
Willis,
Large nuclear bomb tests made in the late-50s and 60s injected radioactive particles into the lower stratosphere that were wonderful tracers of how such injected particles evolved. This could be analogous to volcano-injected particles and aerosols. The NRC supported collection of air samples via high flying jets to study how such particles moved and dispersed. Injections in the Pacific at low to mid latitudes were found to circle eastward around the Earth a few times while gradually widening. It required some time for the particles to cross the equator.
This might indicate that after a large volcano, any cooling would be largest in a latitude band, and that a few years might be required for the effect to disperse worldwide. Of course, these bomb particles may have been removed from the troposphere by rain rather quickly compared to aerosols.
It might be interesting to examine temperature records in the same latitude band as these large volcano eruptions.
Consider when there is a full solar eclipse, the temperatures dramatically drop for a few minutes. Ash/Dust in the upper atmosphere from volcanoes does the same, but not for years and years. And there is really nothing we can do about it. If Mt.Vesuvius creates a Plinian eruption like in AD 79, this would cause a lot of dust but it was preceded by a bad earthquake. Same as the Thera eruption. (Santorini in 1300 BC) It was massive and records of it are recorded in Northern Europe and China, and in Eqypt. Bit difficult to put its affect in a time line. But at this time there was a swarm of earthquakes, long droughts and more volcanic eruptions. And of course the displaced sea peoples and the Jewish exodus from Egypt. Records in Egypt about a long drought, causing people to eat children, is terrible really. Thank goodness we have the Red Crescent and Red Cross nowadays.
lgl- your diagram makes our points. Look at table 3 if you get a chance it is telling and an excellent summary of volcanic effects and where they might be and duration of time.
Willis: Temperature data is often noisy enough to obscure the signal from volcanos. This problem gets worse the further back in time (fewer stations) you go. If you go to WoodForTrees.org, you can plot monthly data and averages of N months of data to smooth out noise. If you gradually increase the N, El Chichon and Pinatubo become very clear with N is 30. 30 Months appears to be enough to remove most of the noise from El Nino events. With a 30 month average, the 1997/8 El Nino event has disappeared, but the cooling from El Chicon and Pinatubo remain with an amplitude of almost 0.5 degC despite the averaging.
Can I detect Tambora using this approach? Not for sure. 1808 and 1815 (Tambora) are one cold period of about a decade. 1832 and 1835, possible. 1883, probably. 1903 probably, but stayed cold for rest of decade. 1915, biggest cooling signal in 20th century with no candidate volcano. (Novarupta, 1912 was followed by 3 warm years, not cooling.) Spotting the volcano is much easier this way.
Thanks, Frank, but I fear that’s not enough information to duplicate your research.
Next, I don’t trust the kind of boxcar averaging used by WFT. It is well known for creating spurious false trends and results.
Next, it WFT is using a centered average. This means that at any point you are averaging both information from the past and information from the future … a dubious procedure at best for this kind of an analysis.
Finally, you have an “N” there of 2, El Chichon and Pinatubo … drawing statistically based conclusions with an N of 2 is dubious even on a good day with a following wind …
w.
Willis , did you look at table 3? If not you should.
http://climate.envsci.rutgers.edu/pdf/ROG2000.pdf
Willis: Thanks for the reply. Sorry I wasn’t particularly coherent earlier. Perhaps this link to WFT will help:
http://woodfortrees.org/plot/best/from:1900/to:2010/mean:30
I’m working with BEST data; Box 1) From (time) = 1900; Box 2) To (time to) = 2010; Box 3) Mean (sample) = N = 31. If I understand correctly, Box 3) means that the monthly temperature for January 2000 is average with the 15 months before and the 15 months afterwards and reported above January 2000. The same procedure creates about 12*111 = 1332 individual datapoint that are plotted for this time period. If you want to view the raw monthly data, change to Mean (sample) = 1 or put nothing in Box 3. (Putting compress = 12 in Box 3 may produce the annual averages you used in your post. Add series can overlay raw and smoothed data.)
If one successively tries values of N=7, 13 (somewhat like annual data, but more data points), 19, 25, 31, and 37 in Box 3, a signal for Pinatubo first stands out from the noise with N = 19 and a cooling of 0.5 degC. Increasing N above 19 smooths out some of the cooling and removes some of the noise.
Now let’s look for El Chichon in 1982. It is difficult to see because the second strongest El Nino since 1950 occurred in early 1983. (See link to MEI index.) The peaks for the 1983 and 1998 El Ninos are still visible with N=19 and N=25, but not with N=31. At this point, I become convinced I can “spot the 1982 volcano” and I can’t see a clear signal for the 1983 and 1998 El Ninos.
http://www.esrl.noaa.gov/psd/enso/mei/
If you look at Paul_K’s Blackboard analysis of Pinatubo, he thinks that maximum cooling occurred about 18 months after the eruption. With N=31, the datapoint for 18 months after the eruption is the average of months 3-34 after the eruption. If Paul_K is correct, I have averaged all of the months of strongest cooling with some months of modest cooling, but no months when Pinatubo wasn’t having some effect.
http://rankexploits.com/musings/2012/pinatubo-climate-sensitivity-and-two-dogs-that-didnt-bark-in-the-night/
So empirically and theoretically, N=31 months appears to do a good job of preserving the signal from strong volcanos and removing some of the high-frequency noise. And we should expect to see maximum cooling with this much smoothing 1-2 years after the actual eruption date, not in the same year as the eruption. (There are certainly better ways to smooth over 31 months, but they require more effort on my part.)
I’ll leave it up to you to see if you can “Spot the Volcano” elsewhere in the BEST record with Box 3 = 31 (or whatever value that you think is most appropriate). I will note for the record that the strongest cooling (0.5 degC) with Box 3 = 31 was around 1915 and I can’t associate that cooling with a major volcano. It appears to be a false positive. The strong warming about 1913 is where one would expect to see cooling from Novarupta/Katmai in southern Alaska. (Although this was the biggest eruption since 1900, volcanos at high latitudes are supposed to be less efficient at cooling the planet and this one was very far from the Antarctic ice core you used to select volcanos.) You may conclude that most major volcanos have produced about 0.5 degC or more cooling (smeared out by averaging), but my method still gives some false positives and false negatives. IMO, you couldn’t “spot the volcano” without this type of smoothing simply because the noise is too strong.
Frank, thanks for your further information. I went to your WFT site and got the data. It is an excellent example of why just squinting at a graph and declaring that you see the volcanoes can be very misleading. Here is the WFT data including their average, overlaid with the most recent large eruptions. I’ve followed Schurer et al. and used the Crowley and Untermann data to pinpoint the exact month that the stratospheric aerosols jumped, identifying the exact start of the eruption affecting temperatures. Here is that graph …
I’m sure that you can see the problem … the temperatures started dropping in all three cases well before the eruption … which, unless we are into reverse causality, means that those temperature drops were not caused by the eruptions.
So I fear that you are looking at coincidence. If you take a look at the full Berkeley Earth dataset you’ll find plenty of drops of comparable size which were NOT caused by volcanoes … which is why, as I’ve said many times, you cannot pick the eruption years out of the temperature record without knowing them in advance. They simply make too little difference to the global temperature to rise above the other fluctuations in the record.
Finally, I say again, the WFT “averaging” sucks. They use a centered “boxcar” filter for their average, which causes all kinds of artifacts. In addition, it uses information from both the future and the past, so it is responding to events that haven’t happened yet …
Like I said, it’s about the worst choice for an average I can imagine. You cannot depend on anything you think that you have “found” using that filter.
w.
Willis: Thanks for a second reply and the plot. You wrote: “I’m sure that you can see the problem … the temperatures started dropping in all three cases well before the eruption … which, unless we are into reverse causality, means that those temperature drops were not caused by the eruptions.
When you average in 15 months from before the month of the eruption – which should be random noise – and 15 months after the month of the eruption – which should be cooling plus noise – I expect to see cooling on the date of the eruption. Above I said: “And we should expect to see maximum cooling with this much smoothing 1-2 years after the actual eruption date, not in the same year as the eruption.” So I would mark Pinatubo and El Chichon with a band 1-2 years after the eruption date and expect to see a minimum temperature caused by the volcano around that period. So I don’t think this plot demonstrates “reverse causality”. It looks about right to me.
The boxcar filter was the only tool available at WFT, so I didn’t have a choice. I don’t have any personal experience with smoothing. What filter would you recommend for this situation? Every filter introduces some distortion, but it is common for signals to be obscured by noise. A citizen scientist’s job to extract those signals when possible without introducing artifacts. (Some climate scientists seem to have other agendas.) Ideally, one would take Paul_K’s best-fit curve for temperature vs time and put that signal through the filter to establish how much distortion it causes and whether it introduces any artifacts. The filter I used will certainly widen the period of volcanic cooling and reduce its depth. If one doesn’t want this kind of distortion, can you remove high frequency noise after a Fourier transform and transform back into the time domain?
If you plot the raw monthly data and smoothed data on the same plot (or look at Paul_K’s modeled fit to monthly data), it is obvious that the noise is comparable to the signal we expect to observe. In that case, we have no possibility of “Spotting the Volcano” without processing the signal. So our inability to “Spot the Volcano” in noisy monthly data or yearly averages (which recompressing an event into 4-5 still noisy data points) doesn’t tell us that a cooling of about 0.5 degC doesn’t follow most eruptions.
Should the existence of a modest number of false positives and false negatives in a smoothed record be a concern? Not necessarily. All major volcanos are not created equal (especially in latitude) and we have other forced and unforced variability to contend with. If half of the cooling events with an amplitude of at lest 0.4 degK were false positives and a signal was detected within +/-1 year of the expected minimum for 80% of large volcanos, this would be a situation unlikely to occur by chance. It makes me uncomfortable to not see an unambiguous signal from the biggest volcano, Tambora, but it took place during a period with the poorest temperature data and seven years after another eruption. I suppose someone has modeled what was expected to happen during this period, but most modelers (unlike Paul_K) tend to use high climate sensitivity and high sensitivity to aerosols. So the inability to see their predictions of a 1-2 degK cooling following Tambora doesn’t prove anything to me without first seeing what reasonable alternative models (like Paul_K’s) hindcast.
FWIW, I suspect my crude smoothing is resolving a roughly 0.5 degK cooling event following Pinatubo, El Chichon, and other major volcanos. It is far short of publishable proof.
Frank July 24, 2015 at 2:37 am
Many thanks for your continued comments, Frank.
My general advice is, do not filter. Let me direct you to an actual statistician for information on this, William Briggs, “Statistician to the Stars”, and his post entitled Do not smooth times series, you hockey puck!. You might also enjoy my post on Data Smoothing and Spurious Correlation …
Let me give you a practical demonstration. You say that when you smooth the Berkeley Earth data, you can see clearly that the 1982 eruption of El Chichon caused a drop in temperature … and you are quite correct, as the smoothed data you presented shows.
However, here is the unsmoothed Berkeley Earth data for El Chichon …
As you can see, there is absolutely no visible temperature effect from El Chichon—nothing immediate, nothing delayed, nothing at all … and as a result, we can be sure that any simple smoothing or filtering that “reveals” that El Chichon had such an effect, like the smoothing you have done with a 30-year boxcar filter, is merely revealing an artifact of the smoothing.
And that is why I (and others, such as the paper discussed in the head post) use the stacking method to reveal any commonalities in volcanic effects … and as Figure 3 shows, no such commonalities are revealed. In fact, that analysis shows that on average, the temperature dropped more BEFORE the seven eruptions than after the eruptions …
All the best,
w.
Willis: Thanks for the reminder of Brigg’s comments on smoothing. His main objection is to using smoothed data as input into another analysis – for example, looking for correlation. I haven’t committed that sin. And I am somewhat aware of the distortion smoothing can introduce.
When I made my plots at WFT, I overlaid the raw and smoothed data. I certainly knew what before and after looked like.
I used EXCEL to create a model signal, add noise and then smooth it with my crude boxcar filter. The signal was a sine curve: sin(PI()*t/60). t runs from 1 to 1000 months. The lower half of the sine curve was a model for 8 volcanic cooling events lasting 60 months with a maximum cooling of 1 vertical unit (which I will call degC).
I used 2*Rand()-1 to create a series of random numbers from -1 to +1, which I multiplied by an adjustable constant (currently 1) to create noise. Then I created autocorrelated noise from the simple noise by adding noise for month t to an adjustable percentage (currently 90%) of the correlated noise for month t-1. Then I created a signal plus noise and a signal plus autocorrelated noise to display with and without smoothing.
With 60 monthly data points to characterize a trough, signal with almost any amount of uncorrelated noise wasn’t very challenging. (Annual data could be more challenging.) Smoothing mades things a little clearer and reduced the apparent amplitude of the sine curve by 25%.
Auto-correlated noise is far more challenging – the sine signal was essentially invisible. The smoothing pulled 8 minima from data with no apparent pattern. However, one maximum was missing and a second was ambiguous and could have been missed. The amplitude of the troughs varied from -0.5 to -1.5. One maximum was at 2.5! The smoothed signal was a very distorted version of the original sine curve function.
If you wanted to invest the time, you could probably convince yourself that smoothing (hopefully with a more sophisticated filter) can help you recover a known signal from autocorrelated noise – albeit imperfectly. The same technique might help pull a real – albeit imperfect – volcanic signal from the temperature record and possibly explain why noise might still render some signals weak or undetectable. Enough smoothing to suppress the 1998 and 1983 El Ninos is important. For some eruptions, the smoothed curve is still likely to be ambiguous.
Willis is consistent in one area he always approaches the climate as a single cause and effect when the reality is there are multiple forcings going on at the same time which obscures the climate /item exerting a given force on the climate at a given point of time.
What could one say accept that is not how it works.
Salvatore Del Prete July 24, 2015 at 6:51 am
As is far too often the case, Salvatore, you are making accusations about what I do and don’t do without quoting a single one of my words.
You may wonder why I so rarely reply to you. It is because you are consistently either too stupid or too deliberately underhanded to quote my words, despite my repeated requests that you do so. Instead, you just make slimy remarks intended to hurt my reputation … all that hurts is your reputation, Salvatore.
Or it would if your reputation could get any worse …
w.
” You may wonder why I so rarely reply to you. It is because you are consistently either too stupid or too deliberately underhanded to quote my words, despite my repeated requests that you do so.”
…
False dichotomy.
..
There might be a third option, namely that he’s getting under your skin for cause.
..
A wise man would ignore him, so I guess that option is off the table.
Joel D. Jackson July 24, 2015 at 11:38 am
Thanks for the thought, Joel, and that ‘s likely true, I’m working on becoming more Canadian. However, I am loath to let untrue, uncited, unquoted accusations about me stand unchallenged. I do care about my reputation, tattered though it may be … I just need to cut out the extras in my replies. It’s a work in process.
w.
Salvatore writes: “Willis is consistent in one area he always approaches the climate as a single cause and effect when the reality is there are multiple forcings going on at the same time which obscures the climate /item exerting a given force on the climate at a given point of time. What could one say [except] that is not how it works?”
So what? The IPCC has the same difficulty with respect to GHGs and climate change. Volcanos, at least, are the biggest known forcing, though difficult to understand due to their transient nature. To further their political objectives, the IPCC works only with climate models with high climate sensitivity to forcing and high sensitivity to aerosols. Those models out to produce cooling signals which ought to stand out more clearly from the noise in the temperature record.* Willis has good reasons to be skeptical, especially with all of the hype about the “Year without a Summer”. For Pinatubo and El Chichon, however, we aren’t dealing with an assumed forcing estimated from the amount of aerosol estimated to have been ejected into the stratosphere, we have measurements of the actual forcing (change in SWR). There is less reason for skepticism here.
http://www.esrl.noaa.gov/gmd/webdata/grad/mloapt/mlo_transmission.gif
And during Pinatubo, we have satellite measurements of incoming and outgoing radiation. After climate scientists failed to create a model (mixed layer size, flux into the deeper ocean, climate sensitivity) that was consistent with all of this data, Paul_K did so at the Blackboard. That model said that Pinatubo produced a maximum cooling of 0.6 degK 18 months after the eruption. However, changing the parameters of that model produces other outcomes that are still consistent with the noisy data. It would be very interesting to see if other major volcanos did produce cooling of about 0.5 degC. Willis keeps telling us that volcanos don’t produce any detectable cooling, but he hasn’t processed the data in a way that is capable of “spotting volcanic cooling” of this size. As with the Hockey Team and the MWP, the question is whether Willis will process the data in a way that optimally and fairly reflects the magnitude of temperature change given the noisy data. (My crude methodology isn’t completely convincing.)
*I made a quick search to see just how much cooling models actually predict after Tambora. Gavin’s name is on a modeling paper, but it didn’t disclose a maximum hindcast global cooling (:)). They used a Tambora that was only 2X and 3X the size of Pinatubo, elsewhere you can find 6X. Zorita has a very crude study whose output shows -0.75 degK after both 1808 and 1815. The CMIP5 project made hindcasts for the last millennium, but a review made no mention of Tambora – the biggest transient forcing during the millennium. Following the usual slimy methodology, that project allowed the modelers to choose their preferred forcing input, the biggest of which was only 2X Pinatubo. (The rest are overlaid and can’t be seen.) As usual, Willis has pick a weak spot of the consensus doesn’t appear to want to confront.
http://www.researchgate.net/profile/Eduardo_Zorita/publication/228794710_The_influence_of_volcanic_solar_and_CO2_forcing_on_the_temperatures_in_the_Dalton_Minimum_(17901830)_a_model_study/links/0c960526c327d67b92000000.pdf
http://www.clim-past.net/9/393/2013/cp-9-393-2013.pdf
Willis , the simple truth is your view on the climate and my view are the opposite.
I simply disagree with many of your conclusions on the climate 100% ,nothing more ,nothing less.
Time will show which one of us is more correct. take care
Thanks, Salvatore.
w.
Thanks. Not personal at all you are a brilliant writer and good communicator and I wish we did agree more.
FWIW, maybe not much, my opinion is that large, tropical volcanic eruptions cool the globe or a hemisphere for a few years at most, ie they affect the weather. Big eruptions at higher latitude might actually have a slight net warming effect. The mix of ash, SO2 and other ejecta matters. There are of course also more pronounced local to regional effects.
The rare supervolcano eruptions may have a more lasting effect, influencing climate, but within a fairly brief period, conditions return to whatever state normal functioning of the system dictates.
However, even flurries of regular but large volcanic eruptions (up to VEI-7) do not affect climate, as imagined by Mann, et al, to try to explain away the LIA, the better to get rid of the MWP. Nor can the two decade-long plateau in recent warming be hand waved away by volcanism not in evidence.
Sturgis ,look at table 3 of this report. I think they have it about right.
http://climate.envsci.rutgers.edu/pdf/ROG2000.pdf
I neglected to mention the differing effect with season.
I’m dubious of an interdecadal, climatic effect of volcanic eruptions, even in the LIA. I’d put it the other way around. The LIA was already frosty, with many regions at the margin of agriculture. Thus the effect of Tambora, which occurred during the Dalton Minimum, the last severe blast of the LIA, was magnified relative to say, Krakatoa (adjusted for magnitude, of course).
Reblogged this on Climate Collections and commented:
Excellent discussion on volcanic aerosols with outstanding Weatherbell graphic added.