Anyone who has seen the raw temperature output from a weather station must have wondered at the marvel of averages. The output is all over the place – large fluctuations in temperature from hour to hour and day and night. Yet from those measurements the result is just one number – the monthly average – that finds its way into climate data.
Picking meaningful information from the variable set that are weather stations often seems more art than science; truncated sequences, gaps, changes of equipment, changes of sites, changes in the local environment, to name but a few factors that have to be taken into consideration, or sometimes not taken into consideration.
A new analysis of some of the statistical methods used in getting something out of temperature readings from weather stations carried out by Steirou and Koutsoyiannis of the National Technical University of Athens has been gaining some publicity as its conclusions are startling. The researchers say that the statistical manipulation of the data to correct errors often introduces even greater errors, as well as exaggerating positive trends.
Such statistical pitfalls are everywhere when one manipulates data like this. Consider the recent case of Dr Joelle Gergis of the University of Melbourne whose paper on 1000 years of climate data in Australia has had to be withdrawn for rewriting when it was pointed out that the “hockey sticks” produced by the calculations were artifacts. Then there is also the original hockey stick, once the unquestioned (by some) emblem of global warming, which was also shown to be in its broad detail an artifact of data processing.
Considering the processes applied to temperature time series Steirou and Koutsoyiannis say: “It turns out that these methods are mainly statistical, not well justified by experiments and are rarely supported by metadata. In many of the cases studied the proposed corrections are not even statistically significant.”
“In total we analyzed 181 stations globally. For these stations we calculated the differences between the adjusted and non-adjusted linear 100-year trends. It was found that in the two thirds of the cases, the homogenization procedure increased the positive or decreased the negative temperature trends.”
The process of ‘correcting’ raw data inevitably involves changing one number into another number. Sometimes this is called ‘smoothing.’ It always renders subsequent data manipulations irreleveant and meaningless. It introduces errors and increases the ‘uncertainties’ (variances) in extracted data.
ALL of the work involving climate data seeking ‘trends’ has been done on ‘smoothed’ data rather than raw data as data can be smoothed to some extent (rounding, selecting, change of scale, etc.) before it is even recorded.
Discussing statistical methods is interesting, but not enlightening.
It’s getting warmer.
Actually, once you remove the statistical manipulations and “corrections” there is almost overwhelming evidence that the climate is in fact cooling ever so slightly across the past millenium. All the supposed warming produced by the “adjusted” data, is an artifact of the adjustment, and not a real trend. Then of course, there are the myriad problems with the data sets themselves. If Major League Baseball kept score the way MOB and NOAA keep temperatures, the Cubs would be perpetual World Series Champs.