The new and improved global temperature database, Hadcrut4, has been updated at last. Previously it had been complete to 2010. Hadcrut4 is a combination of Crutem4 and Hadsst3 – land and ocean data – that includes better sampling of the Arctic – the most rapidly warming region on Earth. When Hadcrut4 first came out it ended in 2010 that was the warmest year in the dataset. This was used in trend analysis to imply a recent temperature increase.
When I first looked at Hadcrut4 I was aware that when the 2011 data was included it would almost certainly show a reduction in global temperature, and hence alter the tone of the implications that were inferred after its debut. Looking at the differences between Hadcrut4 and Hadcrut3, whose data was up to date, I estimated that Hadcrut4 for 2011 would be 0.400. This attracted some criticism. It is pleasing to note therefore that in the updated Hadcrut4 dataset 2011 has a temperature anomaly of 0.399!
The inclusion of the official data for 2011 does not change the statistics of the Hadcrut4 database as I determined – in the last decade or so it is warmer and flatter than the previous database, Hadcrut3. The recent temperature standstill is very evident. There is no statistical case to be made for a global temperature increase in the past 15 years.
Some still say this is cherry picking the data. But it is not cherry picking to look at the temperature of the recent warm spell, post-1980, and note the features evident in the data, such as El Nino, La Nina, volcanic dips and the post-1997 standstill. Perhaps the best way to deduce the length of the recent standstill is to start at the latest data and go back, year-by-year, until the hypothesis of constant temperature is violated. This leads one to 1997.
I anticipate this trend to continue when the annual data for 2012 is complete making the global temperature standstill 16-years long.
The Decline of 1998
A close inspection of the latest Hadcrut4 data shows some differences from the dataset that was first released in March of this year.
In the initial data set 1998 was the warmest year (though not statistically so) along with 2010 (0.53) closely followed by 2005 with 0.52. Now however, 2010 has been increased to 0.540, 2005 to 0.534 and 1998 reduced to 0.523. In fact every year in the top ten warmest years has been adjusted to some degree or other. Compared to Hadcrut3 2010 has increased by 0.07, 2005 by 0.06 and 1998 decreased by 0.006.
None of these adjustments are, considering the errors of measurement, statistically significant, but they do affect the ranking of years, which is important if the associated errors are not considered, as is often the case in the media.
The overall conclusion is that global temperature datasets are fluid and change from month to month, and this must be taken into account in any analysis. It would be nice to have explanations for such changes.