Sunday, January 05, 2014

RSS AMSU 2013: 10th warmest year on record

Roy Spencer et al. claim that 2013 was the fourth warmest year of their 35-year-long satellite era. Their satellite friends/competitors at RSS AMSU which has been my primary source of temperature data since the beginning of this blog seem to disagree.

RSS, 1979-2013, graphics by Mathematica.

According to the RSS AMSU data, 2013 was actually the tenth warmest year – quite a difference for two datasets using "almost" the same methodology.

The ranking is summarized by the table below.

01: {1998, 0.55},
02: {2010, 0.472},
03: {2005, 0.33},
04: {2003, 0.32},
05: {2002, 0.315},
06: {2007, 0.256},
07: {2001, 0.246},
08: {2006, 0.231},
09: {2009, 0.222},
10: {2013, 0.218},
11: {2004, 0.202},
12: {2012, 0.187},
13: {1995, 0.159},
14: {2011, 0.143},
15: {1999, 0.104},
16: {1997, 0.102},
17: {1987, 0.099},
18: {2000, 0.092},
19: {1991, 0.081},
20: {1990, 0.074},
21: {1988, 0.066},
22: {1983, 0.066},
23: {1996, 0.047},
24: {2008, 0.046},
25: {1994, 0.028},
26: {1981, 0.022},
27: {1980, 0.015},
28: {1979, -0.094},
29: {1993, -0.118},
30: {1989, -0.12},
31: {1986, -0.139},
32: {1982, -0.172},
33: {1992, -0.179},
34: {1984, -0.224},
35: {1985, -0.261}.

Apologies for the primitive formatting. I don't want to spend an hour with that. A message is that the ranking isn't terribly important or accurately known; it's enough to use a "slightly different dataset" for the 4th place to become the 10th one. If you're allowed to look at many different datasets, it's much easier to discover a record breaker if that is your goal.

See also e.g. the blog post 1 year ago: 2012 was the 11th warmest year (RSS AMSU).


  1. If we are on the top of a sinusoidal curve, one could fit one within some errors, then one could estimate an error by fitting the two different distributions, compare it with the reported errors and maybe discover where the discrepancy comes from.

    Anyway, all one can say is "what goes up will come down".

  2. Hey Luboš, what is your opinion with regards to the time series graphs the NOAA has posted up at ?

    Also, despite the fact that you post pretty prolifically on the matter, I still haven't been able to deduce any of your quantitative opinions with regards to global warming. Are you of the opinion that A) There is no evidence that there has been an average temperature increase of ~.8 degrees in the past 70 years, or B) A temperature fluctuation of 1 degree over 100 years is indistinguishable from statistical noise?

  3. Dear Dv, I don't know why you think that there's something interesting about the NOAA page and I believe that

    A) the global mean temperature averaged over the globe and over a few years has increased relatively to 1900 by 0.7+-0.15 where 0.15 is the std deviation. So there is a small but nonzero chance that the temperature change was actually negative but it was most likely comparable to the positive figure above.

    B) it is physically naive to talk about one type of "statistical noise". There are many colors and types of noise - and more generally, temperature data shouldn't be described by "statistical noises" but by physical theories. Realistic theories of the global mean temperature involve some kinds of noise but they also include factors/contributions that are largely understood.

    The right statement is that the 20th century temperature change doesn't statistically significantly differ by its magnitude from temperature changes in many previous centuries. The human activity has surely affected the temperature change, but so did the ocean cycles, volcanoes, solar activity, cosmic rays, and other things.

    What is more important is that it is utterly irrational for non-specialists to be worried about these mundane variations. Climatology in this sense is a boring esoteric field of classical physics that O(100) people in the world should be interested in as researchers and they should get O($100 million) for that annually. This field of science has no significant implications for the other branches of science and surely not for the ordinary life of the human society. The idea that millions of people should be worried about various claimed climatic "threats" is a postmodern religious cult with no scientific backing.

  4. It's easier to just review the methodologies of the two groups. The most important correction to raw satellite data has been the "diurnal correction" which comes from the fact that the satellites, as their orbits drift, will sample different parts of the Earth at different times of day than they did when they first launched: this introduces spurious trends in the data. One needs to estimate the diurnal cycle in temperature and then correct for it. UAH uses an empirical method derived from the data itself, when there was a brief period of three satellites observing the Earth simultaneously. RSS uses a climate model to calculate the diurnal cycle that would be expected in atmospheric temperatures.

    So now that we know the difference, we might be tempted to immediately prefer the empirical method for this correction, but another alternative would be to examine independent tests. A number have been performed, numerous papers by John Christy have been written on the subject, but also a paper by Ben Herman at the University of Arizona; it seems that the RSS diurnal corrections are excessive. But the most critical test of all came when AQUA went up: AQUA has stationkeeping thrusters, and requires no diurnal correction. UAH and RSS took different approaches to using this data: UAH relied on it to provide a stable "backbone" for the data during the period, RSS simply treated it the same as other satellites after diurnal corrections were applied to them: those satellites were drifting warm, so RSS used it's existing corrections to cool them and then combined them with AQUA. If RSS's diurnal corrections are too large, then we should have seen RSS cool relative to UAH, and indeed we did see that. This basically definitively proved that RSS's diurnal correction was excessive. It continues to be so, even now that AQUA is no longer the main satellite being relied on by either group (unfortunately, satellites don't last forever and AQUA began to experience some problems).

  5. Thanks for this insightful lecture, Werdna. The climate models may be and are imperfect or flawed and I believe you are right that these are flawed, too, but I still think that the RSS-like climate-model approach is the superior one.

  6. Hm, well I suppose so. In theory if there is no bias in the model's estimate of the diurnal cycle, the method should work. But the observational evidence suggests that in practice, it didn't work, and therefore there is probably a bias in the diurnal cycle estimate of the model they did in fact use.

    I personally prefer UAH's method because it appears to work better in practice. But I suppose there could still be a better theoretical approach.

  7. Exactly right.