The blog world is jump starting discussion of 2008 annual temperatures. Yesterday at 1:56 pm Eastern, NASA employee Gavin Schmidt and climate modeler, purely in his “private” capacity, posted an article arguing that the results were consistent with climate models – an activity that lesser minds might think relates to his employment.
Lucia commented here, perhaps redundantly, that Schmidt’s comment was “tendentious twaddle”.
As a result of Ross McKitrick’s “T3″ concept (discussed at CA here), I’ve been following tropical [tropospheric] temperatures from time to time, as these are supposed to be “fingerprints” of AGW and decided that it would be timely to update tropical (20S-20N) estimates, using reported satellite data and collating tropical averages from gridded CRU, GISS and NOAA versions as these are not reported separately. I calculated results over the satellite period (1979-2008) for consistency (as I’ve been doing in the past.) In each case, I zero-ed the data over the 1979-1997 period – the choice doesn’t matter much – but I didn’t want to include the most recent period in the normalization period and I wanted a relatively long period. I also divided the satellite temperatures by 1.2 according to a factor sent to me by John Christy. This yielded the following:
We now have a 30-year period of satellite records. Within that period, 2008 ranked 26th out of 30 (5th coldest), 23rd for RSS, 16th for CRU and 15th for NOAA and GISS. Tropical temperatures in 2008 were lower in 4 of 5 indices than in 1988, the year of Hansen’s famous testimony.
Although climate scientists downplayed individual cold months early in the year, Schmidt seized on recent Oct and Nov data from GISS:
In GISTEMP both October and November came in quite warm (0.58ºC), the former edging up slightly on last month’s estimate as more data came in.
In the tropics, temperatures in October and November have rebounded from lows early in the year, but GISS is running a bit warmer than the others (about 0.2 deg C relative to the 1979-1997 average).
When I squint at the tropical data, it has two quite different aspects depending on your perspective. On the one hand, you can look at it and conclude that the average level of the cycles is moving up, with the 2008 low being higher than prior lows, with the overall package moving up in an disorderly way. On the other hand, you can look at the data and surmise that it has consistent with 1/f noise or 1/f noise with a very minor trend. It’s hard to believe that this is the sort of result that Gavin Schmidt “wants”.
For people that look at data, it is obvious that the data (in each incarnation) is highly autocorrelated; the degrees of freedom in autocorrelated data can be much lower than people think and, accordingly, the confidence intervals can be surprisingly wide (a point made in Santer et al 2008, though there are some defects in their analysis that we discussed before).
Wide confidence intervals are a two-edged sword. For people who attempt to argue that this data shows no trend, the wide confidence intervals permit a trend much higher than observations (and perhaps high enough to be consistent with the less aggressive models). For modelers who argue (a la Santer) that the confidence intervals are wide (and wide enough so that observations are not “inconsistent” in some sense with models), the problem is that such wide confidence intervals will probably include the null hypothesis of no trend or a minimal trend. It’s interesting to watch, but I don’t think that you can conclude very much either way purely on the statistics of the data, as people so often try to do.
The bottom line: In the GISTEMP, HadCRU and NCDC analyses D-N 2008 were at 0.43, 0.42 and 0.47ºC above the 1951-1980 baseline (respectively). In GISTEMP both October and November came in quite warm (0.58ºC), the former edging up slightly on last month’s estimate as more data came in. This puts 2008 at #9 (or #8) in the yearly rankings, but given the uncertainty in the estimates, the real ranking could be anywhere between #6 or #15