Previous posts have discussed climate variability in general, and modelled decadal trends in temperature specifically. However, I should have considered decadal trends in observations as well, especially as there is a long temperature record available for the UK.
The annual mean Central England Temperature (CET) timeseries, which runs from 1659-2010, is shown below in Figure 1. There are many periods of warming and cooling temperatures, with an overall upwards trend.
In Figure 2, the left panel shows a grey histogram of changes in CET from one year to the next; the middle panel shows the distribution of decadal trends in CET, and the right panel shows the same for 30-year trends. Using changes from year-to-year, there is no obvious difference between positive and negative changes, but for 30 year trends there are clearly more warming periods (59%) than cooling periods (41%).
The black lines show similar estimates from a single gridpoint in the UK from a 5000 year control simulation of the HadCM3 GCM, with no changes in radiative forcings. For single year changes and decadal trends, the two distributions are very similar, but there seems to be wider distribution of variability in the observations than the GCM for 30 year trends, for both positive and negative trends. This may be because the GCM underdoes the variability on these timescales, or because radiative forcings which influence the observations are inducing additional apparent variability. Additionally, periods of stronger warming than might be expected from the GCM (e.g. 1700-1730), are balanced by periods which cool more than might be expected.