Tuesday, July 07, 2009

UAH: June 2009: anomaly near zero

Global mean temperature according to UAH MSU for the first 8.5 years i.e. 102 months of this century. Linear regression gives a cooling trend by a hefty -1.45 °C per century in this interval. So if someone tells you that the trend is "of course" positive as long as we omit the year 1998, you may be very certain that he or she is not telling you the truth.
UAH MSU has officially released their June 2009 data. This time, they're faster than RSS MSU. The anomaly was +0.01 °C, meaning that the global temperature was essentially equal to the average June temperature since 1979. June 2009 actually belonged to the cooler half of the Junes since 1979.

Global warming is supposed to exist and to be bad. Sometimes, we hear that global warming causes cooling. In this case, global warming causes global averageness. In all three cases, it is bad news. The three main enemies of environmentalism are warm weather, cool weather, and average weather.

It is not a coincidence that these enemies are very similar to the four main enemies of communism. The four main enemies that were spoiling the success of communism were Spring, Summer, Fall, and Winter. :-) See Anthony Watts' blog for additional discussion.

Bonus: trends over different intervals

You may have been intrigued by my comment that the centennial cooling trend during the last 8.5 years is -1.45 °C. What is the result if you choose the last "N" months and perform the linear regression?

You may see that the cooling trends are dominating for most intervals shorter than 110 months; the trend in the last 50 months is around -6 °C per century. The local minimum near "N=140 months" ago appears because of the 1998 El Nino of the century that makes the 1998-2009 trend cooling, too. Only when the period gets longer than 150 months i.e. 12.5 years (but less than 31 years), the trend becomes uniformly positive (as far as we can say), around 1.2 °C per century for the intervals whose length is close to 30 years.

Note that those 12.5 years - where you still get a vanishing trend - is from January 1997 to June 2009. If you consider the UAH mid troposphere data instead (relevant for the part of the atmosphere where the greenhouse warming should be most pronounced, according to both proper atmospheric science and the IPCC report, page 675), all the trends are shifted downwards:

You need to consider time periods longer than 180 months i.e. 15 years (at least from Summer 1994) - but shorter than 31 years - to see a uniformly positive warming trend. And the trend that you can calculate from those 30+ years is just 0.4 °C per century and chances are that this 30+-year trend will actually drop below zero again, in a few years. At any rate, the blue graph makes it clear that in the right context, the longer-term warming trend converges to zero at a very good accuracy.

According to the IPCC, the surface warming trend should be around 3 °C per century which should translate to a 4-5 °C warming per century in the mid troposphere where the greenhouse effect has the strongest muscles. You see that according to the last 30 years of the data, the IPCC overestimates the warming trend by one order of magnitude!

Because the mid troposphere is the dominant locus of the greenhouse "fingerprint", this is the most appropriate method to check the validity of the IPCC predictions. Their order-of-magnitude error is equivalent to the mistake of a biologist who confuses squirrels and elephants.

To be more specific about a detail, half of the Earth's surface is between 30°S and 30°N - because, as Sheldon Cooper said in TBBT, sine of 30 degrees is exactly 1/2. But the mid-troposphere warming (8 km above the surface) is faster than the surface at least between 40°S and 40°N, i.e. on the majority of the surface, so it is likely that even when you take the global averages of both quantities, the mid-troposphere should see a faster warming than the surface.

Someone may argue that those 30 years represent too short an interval and the trend will be higher in 100 years. But such a reasoning is a wishful thinking. Moreover, periods longer than 30 years don't really belong to the present generation. In 30 years, most of the population of the Earth won't remember the year 2009 - and they shouldn't be affected by stupid fads of those mostly dumb people from 2009.


  1. Hi Lubos... I read your article at WUWT and decided to come here for reading it from first hand. Congratulations!

    You know the first line of attack against your first-rate assessment will be the shortness of the period considered. However, your database, taken from UAH/AMSU, is considering the "standard" temperature for June 1979, which is scientifically correct.

    On the other hand, if someone considers eight and a half years a very short term, it would be enough to prolong the period to 11550 years and see that actually the Earth has been cooling, not warming. I consider this "small" peaks the same as wavelets riding on longer waves, i.e. a very small warmhouse (hot house) occurring on a very deep icehouse which could end anytime in the near future. Anyway, the current warmhouse cannot be compared to the Holocene Optimum neither with the climate in the geological timescale.

    On another note, the best evidence on a cooling of the Earth is the alternating phases of transgression-regression of the sea. In other eras, the transgression phase reached almost 50% of flooded continental area. If a transgression phase occurred in our times, the most it would reach would be 5 to 10%. Nothing to be worried about. ;)

  2. I mentioned your analysis on one of the Amazon blogs and the AGW people went on the attack. You might be interested in the critique.

    " Barton Paul Levenson says:
    Let's have a little fun with VV's (and Lubos Motl's) dishonestly cherry-picked regression.

    He takes every month from January 2001 to June 2009, which gives him 102 points, a very adequate sample size if monthly values were valid to use in this context. The regression line is:

    Anom = 0.29908 - 0.00122 P

    where P is the period in months, with January 2001 = 1, February 2001 = 2, etc. The t-statistic on the negative slope is -2.57, which is significant at the 98% level (except, of course, that the score is artificially inflated by using monthly data where annual averages would have been more appropriate).

    But wait! There's more! The amount of variance account for by the regression is about 6%. Oh my God, that's pretty crummy. Means we haven't accounted for 94% of the temperature variation.

    Now, let's increase the sample still further. We'll take ALL the monthly UAH temperature anomalies for the globe, dating back to the first one made, in December 1978. This is a period of more than 30 years, adequate to discerning a valid climate trend. The sample size is now N = 367, and the regression line is:

    Anom = -0.12239 + 0.001034 P

    where P = 1 is for December 1978, and so on. Note that the magnitude and sign of the coefficient have been almost reversed 180 degrees. And the t-statistic on the slope is now 11.7, which is significant at well beyond the 99.9% level. What's more, we're now accounting for 27% of the variance, more than four times as much as with the VV/Lubos cherry pick.

    So is the long-term trend (you know, climate, as opposed to weather) up or down? You tell me. "

    Any ideas on how to convince people who refuse to look at the data logically?

  3. Well, Vangel, I never wanted to claim that the temperature evolution is anything else than noise. I thought it was the other side who want to claim that there is a trend that they used to call "global warming".

    I agree with all the partial statements you listed.

    If he wanted to say that there's nothing statistically significant about any of these trends, I wholeheartedly agree with it and I have never claimed otherwise. If he repeats the very same analysis with the same standards and applies them to longer intervals, he will get the same qualitative answer: no trends are statistically significant.

    The other part of my article contained graphs showing the dependence of the slope on the timeframe we choose. Because I showed all possible timeframes, this completely eliminates any possible cherry-picking - except for cherry-picking of "now" which was done by Nature, not me (because that's where we are "now"). ;-)