This post is a belated update to the lies, damn lies, and statistics post from the end of last year. The below figure has been updated with the temperature anomaly for 2010, and we can see that it sits just where we expect it to be if global warming was continuing on as before. So, no surprise there then.
All of the commonly used data sets tell the same story. However, in order to clearly demonstrate that we need to use a “trick!” The three most commonly used surface temperature data sets are those developed by the Climatic Research Unit at UEA, GISS TEMP from NASA and the NCDC from the US Department of Commerce. In fact these are the data sets that the World Meteorological Organization uses to calculate the global average temperature (2010 is in a three way tie for the warmest year on record). As before I will plot annual means from 1975 to 2010, this time for each of the above data sets.
We can see that the GISS and NCDC data sets agree pretty well with each other, but the HADCRUT data set is consistently lower than the others. This is due to the way the temperature anomalies are calculated for the different data sets. Each of the above groups calculates the temperature anomaly from a different base line. For HADCRUT the base line is from the average 1961-1990 values, for GISS 1951-1980 and for NCDC 1901 – 2000. These different baselines create the offsets seen in the above figure. Once we set all the baselines to the same value we can better compare the different data sets. (I have chosen to set all of the baselines to 1961-1990 as this is the baseline used in the previous post.)
We can now see that the three different data sets agree much more readily with each other. We have to note that altering the baseline does not change the temperature trends. In fact anomalies are used precisely so that it is possible to easily look at trends in the data. This is because they represent changes in temperature and not absolute temperatures. We could imagine a scenario where we wanted to look at the temperature history of a couple of different sites, one with a higher average temperature than the other. In order to make meaningful statements about the relative temperature changes of each site in comparison to each other we would need to normalise the temperatures in some way. This is what baseline averaging to produce anomalies does and it enables us therefore to make comparisons between different locations and estimate trends accurately.