UPDATE: Corrected the typo in Figure 3. 1988 now correctly reads 1989.
# # #
There’s lots of blogosphere chatter about the warm temperatures in Russia in November 2013. In their global State of the Climate Report this month, NOAA stated:
According to Roshydromet, Russia observed its warmest November since national records began in 1891. Some areas of the Urals, Siberia, south of the Far East region, and on the Arctic islands in the Kara Sea had temperatures that were more than 8°C (14°F) higher than the monthly average.
NOAA even discussed the record warm temperatures on their global map here.
It might be true that Russian land surface air temperatures were at record levels for the month of November, but NOAA failed to present something that’s blatantly obvious in the data. In 1988, surface air temperature anomalies for much of Russia shifted upwards by more than 1 deg C.
The Russian “hotspot” stands out very clearly in the NOAA map presented in Figure 1. Based on it, I’ve used the coordinates of 50N-70N, 30E-140E for the NOAA NCDC data, and the climate model outputs, presented in the following graphs. That region covers a major portion of Russia.
Figure 1
Figure 2 presents the NCDC land surface air temperature anomalies for the Russian “hotspot”, for the period of January 1920 to November 2013. I’ve highlighted about when the shift occurred. Before that shift, surface temperatures there warmed very little, if at all. And after it, surface temperatures appear to have warmed, but not at an excessing rate. We’ll confirm that later.
Figure 2
The shift is much easier to see if we smooth the data with a 13-month filter, minimizing the visual impact of the monthly variations. In fact, with the aid of period average temperatures (the horizontal lines) and with some color-coding, the shift in 1988 becomes obvious. See Figure 3. Based on the period-average temperatures before and after 1988, that climate shift raised Russian “hotspot” surface temperatures by about 1.1 deg C.
Figure 3
MODEL-DATA COMPARISON BEFORE AND AFTER THE 1988 SHIFT
Figure 4 is a model-data comparison graph for the surface air temperature anomalies of the Russian “hotspot” for the period of January 1920 through December 1987. Both the NCDC surface temperature data and the climate model outputs have been smoothed w/ 13-month running average filters. The climate models are the multi-model ensemble mean of the models stored in the CMIP5 archive, using the historic and RCP6.0 scenarios. The CMIP5 archive, as you’ll recall, was used by the IPCC for their 5th Assessment Report. And we discussed why we use the model mean in the post here.
Figure 4
NOTE: The trends in Figures 4 and 5 are based on the “raw” data and model outputs, not the smoothed versions.
The models did a reasonable job of simulating the warming rate from 1920 to 1987. In more than 65 years, they only overestimated the warming by about 0.23 Deg C. But the models perform quite poorly for the period from January 1989 to November 2013. See Figure 5. During this much-shorter 25-year period, the models overestimated the warming by more than 1.1 deg C.
Figure 5
Let’s state that again: the models overestimated the warming by more than 1.1 deg C over the most recent 25-year period.
Climate model failings at the regional levels are not unusual. We discussed those failings in numerous posts over the past year and in my book Climate Models Fail.
WHAT CAUSED THE SHIFT?
The timing of the shift in the Russian surface temperatures is similar to the shift in Scandinavian surface air temperatures. See the post here. There we discussed that the shift in surface temperature was possibly a response to a shift in the sea level pressure and interrelated wind patterns associated with the Arctic Oscillation.
Additionally, see de Laat and Crok (2013) A Late 20th Century European Climate Shift: Fingerprint of Regional Brightening? The authors argue that a shift in the North Atlantic Oscillation (similar to the Arctic Oscillation) in the late 1980s caused more sunlight to warm European surface temperatures in an apparent shift. I would suspect that something similar occurred over Russia at that time as well.
CLOSING
Like other regions, a climate shift, not the long-term effects of manmade greenhouse gases, is responsible for a major portion of the warming that occurred over much of Russia.
And, of course, climate models performed poorly when attempting to simulate the warming that occurred there since the 1988 shift, overestimating the warming by a large amount. So what else is new?
SOURCE
The NCDC surface temperature data and the CMIP5-archived climate model outputs are available through the KNMI Climate Explorer.
Robert Brooke says: @ur momisugly December 20, 2013 at 5:29 am
Isn’t 1988/89 also when the ‘great dying of thermometers’ took place? Large numbers of Soviet military bases with weather stations, many in cold remote locations, closing post Glasnost.
>>>>>>>>>>>>>>>>>>>>
EM Smith looked into that a while ago.
Thermometer Years by Latitude Warm Globe: As the Thermometers March South, We Find Warmth
http://chiefio.wordpress.com/2009/08/17/thermometer-years-by-latitude-warm-globe/
Also: http://chiefio.wordpress.com/2009/08/05/agw-is-a-thermometer-count-artifact/
And: http://chiefio.wordpress.com/2010/03/08/assume-a-spherical-cow-therefore-all-steaks-are-round/
Then there is Verity Jones (Digging in the clay) who has also done a lot of work and a series of posts on The ‘Station drop out’ problem
http://diggingintheclay.wordpress.com/2010/01/21/the-station-drop-out-problem/
She says of her graphs of Asia
Blow up of the graphs:
http://www.climateapplications.com/GHCN/images/Raw_Asia_Stacked.png
http://www.climateapplications.com/GHCN/images/Raw_Asia_Unstacked.png
….
Frank Lansner over at Jo Nova’s also addressed the Russian station issue.
“Russia” has complained to CRU refers to this 2009 report:
And finally WUWT: http://wattsupwiththat.com/2010/03/08/on-the-march-of-the-thermometers/
OOPS forgot the Jo Nova link for Frank’s comment: http://joannenova.com.au/2012/01/does-the-pdo-drive-global-temps-and-is-there-a-siberian-connection/#comment-924371
@- Steve from Rockwood
“Seems hard to believe you could have a one-time shift in almost 100 years of temperature data caused by an effect that is defined as an “oscillation”.”
Good point.
There is no historical data that shows any similar temperature shifts from this “oscillation”.
@-” A shift up should be followed by and preceded by a shift down. Where is the shift down?”
The AO shifted back down to the pre-eighties level some years ago, which makes attributing this November’s warmth to the same cause as the shift in 1988 very unconvincing.
Joe Born says:
December 20, 2013 at 6:45 am
Forgive the tangent, but here’s something that seems to be apparent to everyone but me: Why the 13-month filter?
An odd numbered filter is used because it includes the “current month” and an integer number of months on both sides so one winds up with smoothed data centered around a point common to the unfiltered data. If an even numbered filter is applied one would have smoothed data referenced to midway between two of the original data points. If one were plotting such data in Excel or some such they’d have to adjust the indices appropriately which is slightly less convenient.
Off topic but interesting news … especially, if you know where Ellesmere Island is!
A biologist has discovered 400-year-old moss in Nunavut. The moss was buried under a glacier on Ellesmere Island where it survived under the ice. Catherine Lafarge (sp) is a biologist with the University of Alberta. She was able to grow the moss in the lab. Lafarge says this discovery could help in space travel.
“Looking at is there any life on Mars or whatever. I do think something like Arctic organisms would be one of the first group of organisms that you would try to see whether they could survive in extraterrestrial systems.”
Lafarge plans to look at ice patches for ancient plant species on Baffin Island and in the Yukon next year. Some of the ice is up to 120,000 years old.
Well lets not forget Climategate:
http://blogs.telegraph.co.uk/news/jamesdelingpole/100020126/
Joe Born: On the 13-month filter, you need an odd number of samples to make the filter symmetrical about the month in question. 13 months gives you six months before and six months after the month.
Joe and John: The 13-month smoothing averaging is done on the anomalies, effectively the residuals that remain after averages for each month over many years are taken. So the idea that, say a January 13-month average is raised by including two July readings is not valid. It is common analytic practice to look for patterns in residuals of fits.
If you go to Fig 3 (Monthly November ice extent for 1978 to 2013) at http://nsidc.org/arcticseaicenews/ you will see an “obvious” break in the curve at 1988 that would correspond to your 1 degree jump. The reason no one has commented on it before is that it looks like noise.
Bob: What do the satellite measurements for the Siberia area report? Do they see a similar step up at the end of the 1980s?
A 13-month filter is the shortest possible filter for eliminating seasonal effects while remaining symmetric about a given month. The next available filter would be 25 points long. A 3 point filter would show the average seasonal effect, which should reveal itself as a sinusoid (all this assumes continuous monthly data – unshifted).
Thanks Bob, this is very interesting. I can see a scientific paper here.
The a strongly positive ARCTIC OSCILLATION does seem to correlate with warm tempertaures in Russia . For example , the AO INDEX was 3.106 and 3.279 in January and Februray 1989. . The winter temperature in Moscow during January and Februray of 1989 was only -2.3C and -.0.7C when the tyicals are vastly colder [ january/february average in the 1980’s was closer to -8C and as high as -17.7 in January 1987 and -14.1 in February 1987]. So like in November 2013 , there was a spike in the warm AO and the temperatures went up.[ daily high for AO in NOVEMBER 2013 was over 4]. What caused this short term spike in the positive AO? . During November 2013, there was a major spike also in solar activity when the solar flux shot up to 174 and the sunspot number close to 290. I don’t know if there is a connection but the extra solar activity could account for the sudden higher positive AO].
Thanks Bob, Very good post.
Again to show that ENSO causes permanent surface temperature shifts, not just oscillations.
For what it’s worth, if one splits this temperature record into a pair of before/after linear trends without the (hokey IMHO) jump discontinuity, the natural break point is circa 1970 and the CMIP5 model comparison would look a lot better than what Tisdale has shown.
http://i42.tinypic.com/27dr8.png
@- Curt
” What do the satellite measurements for the Siberia area report? Do they see a similar step up at the end of the 1980s?”
Yes, both the 1988 ‘shift’ and the present extreme are validated and confirmed by satellite data and other readings of temperature and ice extent from non-Russian sources.
The changes in weather monitoring stations in the USSR from the political collapse are a red herring.
Especially as the big collapse happened AFTER November 1988.
What Shift?
please note. The NCDC chart is based on Monthly data and of course when you use a subset of all the data you’ll increase the odds of finding odd stuff. A good analyst looks at all datasources. Plus, defining russia by lat lon is a really stupid approach when you can actually use the exact border to pull out the data.
http://berkeleyearth.lbl.gov/auto/Regional/TAVG/Figures/russia-TAVG-Trend.pdf
http://berkeleyearth.lbl.gov/auto/Regional/TMAX/Figures/russia-TMAX-Trend.pdf
http://berkeleyearth.lbl.gov/auto/Regional/TMIN/Figures/russia-TMIN-Trend.pdf
note also.
http://static.berkeleyearth.org/graphics/figure1.pdf
http://static.berkeleyearth.org/graphics/figure2.pdf
REPLY: I see the shift in your Tavg chart. If you had presented graphs that were at useful time scales to the discussion at hand, you’d like see it too. Remember from our conversation at AGU I see the “specks” that you do not. I have tool to test for this, I’ll run it tonight and do a new post. – Anthony
I’m not at all sure using a 13 month filter rather than 12 months is a good idea. True you can get linear phase (or zero phase with respect to the center) if you use 13, but at the expense of imperfect cancellation of the full-year periodicity. If you must use 13, re-weight the first and last months by 1/2.
That’s right folks: global warming caused the Soviet Union to fall.
Back in the day I was told that co2 positive feedback crazy warming would be gradual not in big shifts.
Whatever the source of the change or the motive for the change, it is once again evidence that the manmade part of AGW has a lot to do with how data is ‘handled’ and very little to do with actual temperatures.
herkimer says:
December 20, 2013 at 7:01 am
My previous post did not show the current weather in NORTH ASIA properly
http://www.findlocalweather.com/weather_maps/temperature_north_asia.html
What I find interesting is the two cold areas. If those two cold areas were warm in November, and no data was in the middle, GISS and HadCRUT4 would have very different anomalies. Unfortunately we are forced to have this discussion without knowing the HadCRUT4 November anomaly. However the satellite data, whatever their differences with GISS, do not give grounds for confidence here by ranking November 2013 as 9th for UAH and 16th for RSS.
I can`t recall when it was…but wasn’t there a year when the headlines all screamed record warm October and it was later realized that Russia used their September data for October?
Just a thought.
Slightly OT-
Wind patterns over Russia for Nov. 15, 2013. Cool app.
http://earth.nullschool.net/#2013/11/15/1500Z/wind/isobaric/1000hPa/orthographic=-267.35,35.26,279
John Eggert says:
December 20, 2013 at 7:25 am
———————————————-
John, It doesn’t make enough of a difference because you are talking about only one extra month divided by 13. While July is much warmer than January (where I live) it isn’t much different than adjacent months June and August. I tried your idea of weighting the first and last months by 0.5 (e.g. 0.5, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0.5 divided by 12) compared to equal weights and the one graph is hidden below the other virtually indistinguishable (I used the HadCrut 3 data from 1850 onward).
The shift in 1988 doesn’t seem to be sea ice related around Russia, where from 1987 to 1988, in the North of the country there was very little sea ice change in just one year.
Most sea ice extent in March 1987
http://www.ospo.noaa.gov/data/sst/mean_mon/March.87.monmean.gif
Most sea ice extent in March 1988
http://www.ospo.noaa.gov/data/sst/mean_mon/March.88.monmean.gif