We have been planning to add an opinion page to our site for some time but little did we think we would have anything as exciting as the CRU emails and files to discuss. (If you have found this page you probably know what we’re talking about. In case you don’t, CRU stands for the Climatic Research Unit at the University of East Anglia in the UK. It produces one of world’s main climate data sets (in conjunction with the British Met Office) and several prominent climate researchers work there. In late November 2009 more than 1000 emails to and from researchers at the centre were released on the Internet together with several files of data, programs and notes. It is not clear whether it was work of a hacker or a whistle-blower (whichever it was the emails and files were carefully chosen).
Initial reaction was highly polarised. Many sceptical bloggers reacted with glee; this was the smoking gun, which proved that climate change was a hoax/conspiracy. On the other hand realclimate.org, which represents the views of many climate modellers, took the line that the emails revealed nothing other than robust scientific discussion.
After a slight lull the news of the emails became a big story in the mainstream media, in part because they were released a few weeks before the Copenhagen conference to agree a successor to the Kyoto protocol. This was followed by responses from the CRU and IPCC. The professor who figured in many of the emails stood down while an enquiry instituted by the university was carried out. The IPCC also announced an enquiry. Many of the emails dealt with the temperature record produced by the CRU and the university announced that the data set would be reworked over the next three years and as much as possible of the data put into the public domain. As this decision is the most significant change to come about as the result of the emails it is the one we concentrate on here. There are 5 main temperature data sets. Three of them, the one from CRU and two from the USA, use observed temperature readings, and start in the mid-19th century. The other two use data from satellite data and start in 1979. All five are in broad agreement for their common period.
The use of observed data has two main problems. Firstly the records are not continuous; stations close or are moved and new stations open. A symptom of this is that the number of stations available for analysis varies over time reaching a peak between 1970 to 1990 and then declined. Secondly urban development in the vicinity of a met station can make its reading unrepresentative of the surrounding area. Other changes are subtler. From about 1870 onwards thermometers were housed in a standard Stephenson screen, which had whitewashed wooded slats. The whitewash came to be replaced by different paints, which changes the reflectivity of the station. More recent automatic instruments have different housings. Another change is that temperature measurements at sea went over from using water temperature based on a bucket dipped into the sea to using the temperature of the cooling water intake.
We have two concerns about the new data set to be created.
The first is the concept of a single temperature series. In combining data from many sources a number of assumptions, about how to treat missing data for example, have to be made. Often it is not possible to demonstrate that one set of assumptions is better than another. In the right-hand sidebar we show three temperature series we have calculated ourselves using 5º gridded data from the HadCRUt3v series (Figure below).
The first (red) data set using all the data is the same as the published HadCRU series. Temperature depends very heavily on latitude so for the second series we assumed that any grid filled represented temperature at the latitude (green line); this gives equal weight to each latitude band regardless of how many grids had data. The third alternative we examined was to use only grids, which had data for 90% or more of the time (blue). As can be seen there is not much difference between the first two series but third has some important differences. In particular the peak around 1940 is more pronounced as is the more recent temperature increase. We believe that several new series should be published with the assumptions behind each of clearly stated.
Secondly there is a worry that those developing the new series of measured data will ‘know what the answer should be’. For example the emails revealed there was some concern about the temperature peak in the mid 1940s; although not stated in the emails this was because climate models did not represent that peak. There might be a temptation to choose the data and the stations to fit the models.
For the new data to represent a real advance on previous data sets there is also need for some experimental work. One ‘experiment’ could be to set up a series of met stations with different types of Stephenson screens and different types of modern automatic sensors to see what differences, if any, there are. Another could be to set up a series of stations in urban and rural areas (or select carefully chosen stations currently in use) to develop a way of quantifying the urban heat island effect. Nighttime satellite images have already been used to identify urbanised areas; this could be taken a stage further to develop relationships between degree of urbanisation and temperature anomaly. A third experiment could be to measure sea temperature in different ways. Complicating all these experiments is a need to give good global coverage and to apply current findings to past climate records.