This summer in Moscow the weather has been more than just unpleasant. It’s been dangerous. Not only has Moscow suffered through its worst heat wave ever, the heat has dried out forests and peat bogs in surrounding areas, leading to extensive wildfires which have filled the air with smoke. This has made the air in the city of 10,000,000 unhealthy to breathe; citizens are advised to stay indoors and keep their windows closed. Ordinarily, in such oppressive heat that suggestion would be a bad joke — but with air pollution at such dangerous levels it is, unfortunately, sound advice.
Many have suggested that the recent spate of extreme weather events, including the Russian heat wave, Pakistani floods, flooding and landslides in China, are exactly those events that become more likely due to man-made global warming. As Gavin Schmidt at the Goddard Institute of Space Studies has said, it’s better to think in terms of odds: warming might double the chances for a heat wave, for example.
It has also been suggested that the extremity and frequency of recent events is more than could be considered plausible, without humanity’s influence on climate. I decided to look at the temperature in Moscow, both in the past and at present, to estimate just how extreme the present heat wave really is. Daily data since just prior to 1950, and for a few scattered years prior to that, are available from both ECA and NCDC. Both sources supply daily high temperature data for Moscow, but the NCDC data are more up-to-date (they’re updated daily) so let’s see what they have to say.
July is Moscow’s hottest month, so it’s no surprise this year’s heat wave would occur in July (although the heat wave is still in progress). Here’s the monthly average of daily high temperature, for each July in the record which has reports for at least 20 days of the month (most of them have reports for all 31 days, and from 1960 onward no month is missing more than one day’s data):
Clearly, this July has been significantly hotter than previous years in the record. In fact the average daily high temperature for July 2010 is 3.6 standard deviations above the mean of all recorded July values. For a normally distributed random variable, the chance of being so extreme is only 0.0003 — less than 1 chance in 3000. Which agrees with statements from Russian meteorological officials that such a heat wave hasn’t been experienced in Moscow in at least 1,000 years.
And that means that the suggestion that this heat wave is just a natural variation, not due to global warming, is implausible. Or as we say here in Maine, t’aint likely.
UPDATE UPDATE UPDATE
Reader Richard C links to an article reporting that 17 countries have set new high-temperature records this year, while only one has set a new low-temperature record. And the year isn’t over yet.
Record temperatures in 2010:
Belarus, 7 August, 38.9C (102F) at Gomel
Ukraine, 1 August, 41.3C (106.3F), Lukhansk, Voznesensk
Cyprus, 1 August, 46.6C (115.9F), Lefconica
Finland, 29 July, 37.2C (99F), Joensuu
Qatar, 14 July, 50.4C (122.7F), Doha airport
Russia, 11 July, 44.0C (111.2F), Yashkul
Sudan, 25 June, 49.6C (121.3F), Dongola
Niger, 22 June, 47.1C (116.8F), Bilma
Saudi Arabia, 22 June, 52.0C (125.6F), Jeddah
Chad, 22 June, 47.6C (117.7F), Faya
Kuwait, 15 June, 52.6C (126.7F), Abdaly
Iraq, 14 June, 52.0C (125.6F), Basra
Pakistan, 26 May, 53.5C (128.3F), Mohenjo-daro
Burma, 12 May, 47C (116.6F), Myinmu
Ascension Island, 25 March, 34.9C (94.8F), Georgetown
Solomon Islands, 1 February, 36.1C (97F), Lata Nendo
Colombia, 24 January, 42.3C (108F), Puerto Salgar
Open Mind 
Tamino,
second image is broken in Red Hot
[Response: ??? There’s only one image.]
And when you consider the other highs worldwide, how much more unlikely does it being normal become?
http://www.guardian.co.uk/world/2010/aug/12/heatwave-record-temperatures-world
You may have a page formatting error. It looks like you are supposed to have a second graphic below the first, but for me it’s showing a blank box, suggesting a link to a graphic is broken.
(Otherwise, the subtitle and tabs are nice additions to your site)
jg
[Response: I don’t see that on my browser (Firefox). WordPress has made changes recently, perhaps there’s some bug somewhere?]
No, there are two, you’ve got an img tag linking to:
which doesn’t exist, immediately following the one to:
which does.
(feel free to delete this after cleaning up!)
[Response: I think I found the problem, and that it’s now fixed.]
hmmmph— just them commies setting fars to get the Christian nations to bankrupt thimselves fighting the UN myth of global warming
Hope you heard it here first—-MD
On second or maybe third thoughts. If the other events (heat records) are unrelated then we could examine the probability of the simultaneous heat records although it wouldn’t change the probability of the Moscow event…, but if they are related, have a common cause, then we need to examine the probability of the cause being in that condition? Probabilities always make my head hurt.
Genuinely curious: why would it not be correct to view this July temp as just an outlier? Just eyeballing it, the trend seems to be a fairly shallow upward angle…
[Response: Of course it’s an outlier. But to call it “just” an outlier risks implying that outliers can’t be meaningful or important, which just ain’t so. After all, if the high temperature were a million degrees, that would be an outlier too.]
5 seconds to find this “LAX got to only 67 degrees, breaking a record set in 1926, according to the National Weather Service.” And of course it will never post here.
[Response: Well let’s see … that was a new record for the given date for one location, not a new all-time record for Los Angeles, or for California, or for the United States of America. Such “for this date” records for one location are almost never statistically meaningful at all; they’re broken dozens, or even hundreds, of times every day.
But the records listed above are new ALL-TIME records, not for individual cities but for COUNTRIES. And the record heat in Moscow isn’t a once-in-a-lifetime event, it’s a once-in-three-thousand-years event.
At least it was …
Clearly you really don’t know the difference between notable events that are outside the bounds of normal variation and those that aren’t. Neither do most of your denialist buddies, like Anthony Watts who posts about meaningless cold spells all the time — it’s his bread-and-butter.]
It would be interesting to see an analysis where the the maximum temperature is fit to the General Extreme Value Distribution, and the return period of July 2010 max temperature calculated. I sort of expected that the Russian officials were basing their statements on such an analysis, but upon googling I found out that they were citing historic records.
Say the return period is O(1000) years and then this happens again in a decade. Do you conclude that GEV analysis is crock or that the PDF has changed?
risks implying that outliers can’t be meaningful or important
Understood, but absent a trend or a discrete precipitating event (e.g. a huge fireball from the sun), what is meaningful about an outlier?
Again: a genuine question, not a statement masquerading as one.
[Response: If there’s been no real climate change, just natural variations, then an outlier is just an extremely improbable event.
If there *has* been real climate change, then there can be changes in more than just the mean (i.e. the trend), there can also be changes in the extremes.
This may be one of the reasons that “climate change” is in many circumstances more descriptive than “global warming.” Absent man-made climate changes, the probability of such an extreme value is very low. The occurrence of an event this unlikely is evidence (not proof but evidence) that the underlying system has changed.]
It seems to me that the existence of an outlier event such as the Moscow temperatures can occur for two reasons:
1) Global warming changes the probability of the event because the baseline temperatures are higher: eg, a 14 degree F excursion from normal was a 1 in a 1000 year event, but given that the baseline has shifted up by, say, 5 degrees F, then it is only a 9 degree F excursion, which is a 1 in a 100 year event.
2) Some shift in weather systems has changed the probability of such excursions: eg, some climate shift has increased the probability that the Jet Stream or some Arctic high pressure event or whatever thing can cause high temperature excursions. Such a weather system shift can change either the baseline or the variability of the system, either of which can change the probabilities of extreme events.
Of course, 2b is that it is entirely possible (even likely) that large scale pattern shifts (if they exist) are themselves a result of anthropogenic climate changes, given that such anthropogenic climate changes are probably the largest outside forcing disturbance that the system is currently experiencing.
In addition, even if “2” is the case, global warming _still_ would have changed the baseline, so really, the excursion could then have been an effect of some combination of:
1) a temperature induced baseline shift
2) long term changes in the pattern weather variability
and 3) the right particular types of weather variability this past month to lead to high temperatures.
We can make decent estimates of 1 (see http://climate2008.igce.ru/v2008/pdf/resume_ob_eng.pdf for Russian temperature increases of about 2 degrees C) but for figuring out how to partition between 2 and 3 requires either good regional climate models or a couple more decades of observations…
-M
One paper that came out in 2008 argued that a new circulation pattern has emerged which plays an important role in the destruction of Arctic sea ice. Although the paper itself uses a different name, people typically refer to it by a much more common name: the Arctic Dipole.
Please see:
Unfortunately as of yet I have been unable to locate a copy of the paper itself.
In addition, higher temperatures speed up the hydrological cycle, with the rate at which water cycles through the system roughly doubling (as with the rate of evaporation) for every ten degrees Celsius. With this one expects greater variability in addition to the raising of the baseline with the shifting upward of average temperature. Masters (who is not a climatologist but meteorologist) has also suggested that during the transition from one climate quasi-equilibrium to another, the transition itself may result in greater variability — and that especially strong ridges of high pressure in the subarctic may play a role in this.
Found your paper!
On Asgeir’s CV.
Greater variability is one of the expected results of climate change according to the IPCC as reported by the World Meteorological Organization:
Hat-tip to Joe Romm:
NASA reports hottest January-July on record, says that 2010 is “likely” to be warmest year on record and July is “What Global Warming Looks Like”
August 12, 2010
http://climateprogress.org/2010/08/12/nasa-hottest-year-on-record-what-global-warming-looks-like/
T, you might find the work of Jan Gert van Oldenborgh of the KNMI (Dutch Met service) interesting. The Economist quotes him on the Russian heatwave, and I covered his analysis of last winter’s hot and cold extremes here.
I think M’s analysis above is on the money: lots of reasons why in my recent post, Fire and Rain — especially the Rob Carver/Stu Ostro stuff.
But even with global warming, it’s still a very improbable event. Looking only at your graph above, is it really any less improbable now than it was in the 1960s? Using only the information presented here, I think that’s a tough case to make.
PDA,
The term “outlier” is one of the most abused in statistics. To truly be an outlier, an event must not be “representative” of the distribution of interest. Of course, it could be part of the distribution, and if so, it is giving you much more information about the parent distribution than would an event near the mean. Discarding outliers is risky and foolish unless you deonstrate beyond doubt that they are not of interest.
semi off topic – i read that the wildfires in russia have destroyed forests that will take 15 years just to start regaining, and 300 years to reach their previous size.
i wonder how much CO2 is released in the process? probably very small compared to fossil fuel burning, but still part one of those scary feedbacks.
The way to deal with these facts is simply turn the graph upside down.
That way we reverse the trend(s).
See, problem solved.
Imagine seeing a line going from the lower left corner to the upper right corner. Now turn it 180 degrees so that the upper right corner will be in the lower left…
The is, of course, a mathematics of extreme events and there is another new book with that phrase as the subtitle. Reviews indicate it is thoroughly readable by the non-specialist.
AFAIK the whole “toss outlier” argument rests on the fact that human or instrumental error happens, and outliers are often garbage due to this fact.
What’s being documented in Russia is not due to drunken Russians misreading thermometers or whatever. It’s real. Russia’s burning. People are dying. Temps are being confirmed by a large number of sources.
Thank you, Paul. I had seen something there too but it wouldn’t load for me until now.
Now correct me if I’m wrong, but given that temperature is finite, doesn’t that mean that the extreme-value behavior will be Weibull? If that’s the case, then this event is even more remarkable.
What might be interesting on this subject is an article in this weeks New Scientist about how ‘Blocking Events’ have resulted in a freezing of the track of the northern JetStream, resulting in both the Russian Heatwave & the Pakistani Floods. And references to Mike Lockwood of University of Reading and his work on the history of links between Solar minima and the northern Jetstream and its impact on northern weather and variability. This has implications for how the so called MWP needs to be considered. Also what sort of mechanism could lead from low solar activity to impacts on the Jet Stream.
Although there is no serious suggestion that Solar Variability is responsible for Climate Change trends, there is some grounds for asking whether Solar Variability could be disproportionately implicated in Climate variability. Not just Svensmarks Cosmic Ray theories which are unlikely to be anymore than a minor contribution to that variability. But other factors such as the new data from SORCE suggesting that UV output from the Sun varies by larger proprtions over the Solar Cycle than overall TSI
One gets the feeling that Solar cycle factors, perhaps mediated through the upper atmosphere – Stratosphere & up – may have a bigger influence on variability than is conventionally believed.
And understanding far more about what drives the Noise gives us far more confidence in differentiating that from Signal when looking Climate Change.
Being able to predict the Noise is very important in winning public confidence in the science about Signal because Joe Public can’t differentiate between them.
Just a question from a thicko.
How were the error bars calculated, as they vary from year to year?
I have it on good authority which I’m not allowed to identify–which means I can’t prove it–that the fires have reached Chernobyl, but they’re not admitting it. The reactor complex is better shielded than it used to be, but the fires are releasing radioactive particles from the soil into the air.
Some in the media are reporting on it, even if the governments are not. http://www.nytimes.com/2010/08/11/world/europe/11russia.html?_r=1
So what was the all-time low?
The odds of a 1 in 3000 year event for a particular location occurring one or more times over 1000 years is 28%. You don’t need very many independent locations for the odds of such an event to occur to approach 100%.
Wow, Genius, so why is there only one record low, this summer?
The extremes of a distribution are hard to interpret, but when all the trends point in one direction, it ought to tell you something.
It’s happening over a huge area, not a single point where there might be, oh, an urban blast furnace or whatever.
Timothy, is this the cite to the Wang paper you are looking for:
Zhang, X., A. Sorteberg, J. Zhang, R. Gerdes, and J. C. Comiso (2008), Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system, Geophys. Res. Lett., 35, L22701, doi:10.1029/2008GL035607.
John McCormick
Surely someone must have or must be be working on analyzing the instrumental record in order to extract World-wide trends in the record highs and record lows. Plots of the percentage of stations recording one or the other would be informative. I know it has been done for both the US and Australia. But given that the data exists, why not do it globally?
BearCountry: you make the point well! 2003 – thousands died in France during the European heatwave. 2009 – Australian heatwave. 2010 – Russian heatwave.
What are the odds of three very rare “independent” events happening in the same decade?
Heatwaves are hotter and more frequent now. That’s what the data tells us. I don’t want to get bogged down in probability, because I’m useless with odds – and the main problem here is determining the baseline probability, which is difficult enough on its own. The 1 in 3000 is for a single month outlier temperature at a single (specific) location. Before we can go any further, we need a way to calculate the severity of a heatwave with respect to duration, geographic area and sustained temperature anomaly. It’s tempting to use fatalities as a proxy, but that doesn’t work – you would have to factor in population density, then you need to know the area….
So: do the chances of some weather station somewhere in the world recording a 1 in 3000 monthly anomaly approach 100%? Duh. Of course. Hundreds of record highs are set every single day. Does this have anything to do with heatwaves? Only tangentially.
Don’t criticise Tamino for failing to do a complete analysis. The heatwave isn’t even over, the data isn’t all available, and this subject is ultimately going to generate a good few scientific papers – the topic is just too big for a single blog post.
Carrot,
difficult to tell from Tamino’s fig. just by eyeballing, but it is quite possible that the Moscow distribution has shifted enough between 1960 and 2010 that the return period of this July’s record T has decreased by a factor of 5-10. See:
Click to access raisanen_ruokolainen_071008.pdf
Short version:
Click to access MistraSWECIA_NL1_raisanen4_6.pdf
Looking at USHCN data for ND, max temps in July 1936 were quite similar statistically.
50deg plus is shocking.
‘Specially in Burma.
I’m waiting for a ~50 deg C day here in Melbourne. We got close last year on Black Friday, and I reckon it’s not far away. That’s going to shock a few people.
Wonder what Andrew Bolt will have to say then? “UHI, wtf!!”
I see the bogs that are burning were drained 90 years ago. They used the peat for burning and creating electricity. 4,200 acres aere burning.
RP Jr. tries his hand at looking at this, and while I’m not a statistician, his math looks very fishy to me.
http://rogerpielkejr.blogspot.com/2010/08/why-rare-events-are-certainty.html
Er, this is the prosecutor’s fallacy.
http://en.wikipedia.org/wiki/Prosecutor%27s_fallacy
What you’ve calculated (1 in 3000) is the probability of the observed temperatures given only natural variation at work. But the probability you want is the probability of only natural variation at work given that those temperatures were observed.
Now, with climate change, these types of events should be occurring more frequently. They probably are, and your update suggests that this is the case. But that’s how this should be looked at (more Moscows in our future), instead of trying to attribute blame to individual extreme weather events since it can’t be done (at least, certainly not as easy as a quick probability calculation like you’ve done).
And every time one of us tries to blame an individual event on global warming, a denialist will point to a cold snap 6 months later to declare global warming to be false. Stick to the high ground, no overselling our case, is all I’m saying.