The saga about measurement of temperature, in the United States, at least, is growing in reach and noise, and there are now dozens of websites with thousands of posts about it. One of the main temperature datasets, that of the Goddard Institute for Space Science (GISS), a part of NASA, is now in the news for what critics call ‘dancing data’. The GISS dataset is the one that the orthodox prefer to use, because it shows consistently warmer trends than all the others.
One Canadian sceptic, Walter Dnes, has been downloading GISS data every month for several years, and has downloaded lots more earlier data from the wonderfully named ‘Wayback Machine‘, which archives what appears on the web at the time it was placed there. Mr Dnes has a particular interest in temperature data, and the work he has done, not just for this post, is impressive indeed.
Mr Dnes then published his findings on WUWT. It turns out that the data do dance about, so that trend lines change their size and their sign for no obvious reason. For example, ‘In 7 years (December 2005 to December 2012), the rate of temperature rise for 1910-2005 has been adjusted up from +0.6 to +0.8 degree per century, an increase of approximately 30%.’ The essay contains lots of worrying statistics of this kind. We are in the land of ‘adjustments’ and ‘homogenisation’ again. And to make the point again, these are old data, past data, and they seem to change over time, again and again. In this particular case the data don’t all become cooler in the past and warmer in the present, the worry in my last post on this matter, but that is not the point. They just keep changing.
All this worries me, as a data-cruncher from way back, because anyone who works with data needs to be confident that the data are valid and reliable. When past data change you wonder what has gone on, and how solid your conclusions really are. Conspiracy theorists will remind you of Winston Smith, in Orwell’s 1984, whose job in the Ministry of Truth was to rewrite historical documents as the party line changed. It’s a neat analogy, but I distrust the rush to conspiracies.
The Commenters on Mr Dnes’s paper included someone who pointed out that the people at GISS don’t adjust data — that is done for the Global Historical Climatology Network (GHCN) by other people at NOAA/NCDC (sorry for all these abbreviations). Further discussion suggested to me that what we see in these data is a mixture of real temperature data, estimations and modelling. I doubt that it is now possible to disentangle the ‘data’, and doubt also that they can mean much.
The discussions led to a slanging match between the defenders of thermometers (the land-based and sea-surface temperature measurements) and those who prefer the measurements gained by satellites. The thermometer-lovers pointed out that the satellites don’t measure temperature at all, but ‘radiance’. I thought I knew about that, but decided to check. Dr Roy Spencer, who has been associated with the UAH temperature dataset from the beginning, describes it as follows:
Since 1979, NOAA satellites have been carrying instruments which measure the natural microwave thermal emissions from oxygen in the atmosphere. The intensity of the signals these microwave radiometers measure at different microwave frequencies is directly proportional to the temperature of different, deep layers of the atmosphere… [and the outcome is] global temperature datasets … that represent the piecing together of the temperature data from a total of fourteen instruments flying on different satellites over the years.
Wikipedia adds The satellite series is not fully homogeneous — it is constructed from a series of satellites with similar but not identical instrumentation. The sensors deteriorate over time, and corrections are necessary for satellite drift and orbital decay. Particularly large differences between reconstructed temperature series occur at the few times when there is little temporal overlap between successive satellites, making intercalibration difficult.
Does that inspire confidence? I prefer the satellite datasets (RSS and UAH) because of their near-global coverage, but I accept that that the datasets contain error. Fourteen instruments on several different satellites over 35 years? It would be extraordinary if there were not error.
Of course, the thermometers don’t measure temperature either. There are a great many types of thermometer, and each of them uses a change in something else to infer a change in temperature. All of them have problems. None of them is perfect. All temperature measurements have error. It seems to me that, for the most part, the likely error could be at least as large as the the change in temperature over time.
The longer these angry arguments drag on the more I am persuaded by some wise words of Tony Brown, a British climate historian who is interested in human records of climate experience over the past thousand years (and longer):
Sometimes those producing important data really need to use the words ‘ very approximately’ and ‘roughly’ and ‘there are numerous caveats’ or even ‘we don’t really know.’
It is absurd that a global policy is being decided by our governments on the basis that they think we know to a considerable degree of accuracy the global temperature of land and ocean over the last 150 years.