Thursday, January 19, 2012

NASA PR Sorta Spins 2011 Temperature, Says It's 9th-Warmest

Line plots of global temperature anomaly 1880-2011NASA public relations says 2011 was the 9th-warmest year in their records, but that's not really true.

The scientists at NASA GISS have summarized 2011 here, and it's much more nuanced presentation than from NASA HQ. It reports they finds an average global temperature of

2011:   +0.51 ± 0.05 °C 

compared to the 1951-1980 baseline. (The uncertainty is 2 standard deviations.) That's statistically cooler than last year:

2010:   +0.63 ± 0.05 °C

which is the warmest year in their records.


the numbers depend on a lot of factors, which the GISS technical page rightly discusses, even if HQ doesn't (and most journalists probably won't either).
  1. The resolution of the underlying virtual network -- that is, as I understand it, GISS (and the Hadley center) take into account the fact that temperature stations aren't evenly distributed around the globe (or evenly distributed in time). There are regions with a higher density of stations than others. You don't want to just average all the stations across the world, because that would over-represent regions with many stations (like, say, the northeastern US) and under-represent regions with few stations (like, say, the Arctic, or undeveloped countries, or places were few people live, like the Sahara Desert). So they construct a virtual, even-spaced grid, and then average all the stations within a grid, and then average all the grids. That way all regions get an equal weighting. So the final number is going to depend on the chosen grid size; as they write, the 2011 average global surface temperature anomaly was +0.51 °C for a resolution of 1200 km, but +0.44 °C for a 250 km resolution (both are ± 0.05 °C). Which is better? It's a judgement call -- do you ignore regions where there's no data, or try to estimate them by interpolating across them? GISS believes the first is preferable, and give their reasoning in a footnote.
  2. Each year has a statistical uncertainty, which GISS calculates to be 2σ = 0.05 °C. But being good scientists they say comparing years depends on how close the years are, and go on to say
    "The size of this uncertainty and the small temperature differences among different years (Fig. 2) is one reason that alternative analyses yield different rankings for the warmest years. However, the magnitude of global temperature change of the past century is in good agreement among the GISS, NCDC (NOAA National Climatic Data Center), and HadCRUT (UK Met Office Hadley Centre and the University of East Anglia Climatic Research Unit) temperature analyses."
  3. Statistically speaking, you can't say one year is warmer than another if their difference is not statistically significant. It's like asking if Romney is ahead of Obama if a poll shows their support levels at 51% - 49%. It depends. If the uncertainty of the numbers if ± 3%, you can't make a statistically significant about them (i.e. one that holds to a certainty of p% -- usually in climate science 95), but you can make statements about the probability that Romney is ahead of Obama (you'd have to calculate the overlap of the Bell curves that surround each number, and it depends on the objectivity of the questions, the randomness of the poll participants, yadda yadda. So 2011 was "statistical significantly" cooler than 2010 by the statistical standards of climate science. That doesn't mean that you can rank it as "warmer" or "cooler" than an earlier year if the difference in their anomalies is less than 0.10 °C.
  4. They also, to their credit (and perhaps anticipating criticisms), elaborate on the nuances of whether the rate of warming has changed in the last several years, etc:
    The 12-month running mean (Fig. 3a) provides a useful alternative measure of temperature change on the annual time scale, and 60-month (5-year) and 132-month (11-year) running means (Fig. 3b) reduce the variability caused by the Southern Oscillation (El Niño-La Niña cycle) and the solar cycle. The current status of these running means (Fig. 3) adds some weak evidence for the frequent assertion that the rate of global warming has been less in the 21st century than in the last two decades of the 20th century. However, that impression is dependent on the end point, which is heavily influenced by the strong La Niñas in the past three years. If an El Niño occurs in the next few years, which is likely as we discuss below, the mean warming rate will probably exhibit no slowdown on the decadal time scale.
They then go on to summarize by saying, basically, Yes, warming has been slowing down lately, due to a strong La Nina (the one in 2010-2011, not the current one) and a solar minimum, but
We conclude that the slowdown of warming is likely to prove illusory, with more rapid warming appearing over the next few years.
  1. the NASA HQ press release is simplistic
  2. NASA GISS did a great job of presenting all the nuances
  3. there's certainly no reason to think global warming is "over" 
  4. you can be sure contrarian bloggers will try to say that it is, and
  5. later, when/if it does get warmer, these same bloggers will revert back to claims that the data is junk because someone, somewhere near a temperature station, once grilled a piece of chicken.

No comments: