At a global scale, observed temperatures are covered by five main data sets:
- The Climate Research Unit (CRU) at the University of East Anglia in conjunction with the Hadley Centre, British Meteorological Office.
- NASA’s Goddard Institute for Space Studies (GISS).
- The National Climatic Data Center (NCDC) of the National Oceanic and Atmospheric Administration.
- The University of Alabama, Huntsville (UAH)
- Remote Sensing Systems (RSS)
The first three use land based measurements and, given that there is a finite number of meteorological stations, there is some overlap in the data sources. The data are processed independently with the CRU preferring to use a sub-set of the data limited to stations with long records. All three data sets data sets also use information on sea surface temperatures. Whereas temperatures over land are based on air temperature, over sea they are based on sea temperature. In the case of the CRU data set these are provided by the British Meteorological Office. In the case of the GISS data set they also used the same sea temperature as the CRU data set until 1981 but since then they have been calculated independently. The number of data points varies both geographically, advanced industrial countries tend to have more data points, and temporally, the number of stations in the 19th century was very limited and political upheavals and wars have also affected the number of operating stations. The CRU data set has global temperature estimate from 1850 to the present. The GISS and NCDC have temperature from 1880 to the present. The percentage of the earth’s area covered by grid cells of the CRU data set has been around 80% since the 1950s but in the 1850s was only around 20%. The other two data sets are based on data from the same satellites but independently processed. Strictly speaking what they measure is lower troposphere temperature which is not the same as the near ground measurements used for the first three. They have data from 1979 to the present.
Satellites have the advantage that they provide uniform coverage of the whole earth except for small regions near to the poles. These areas represent about 3% of the earth’s surface. Both land based and satellite data need continual revision. In the case of land based stations, it is necessary to allow for the “urban heat island effect”. This refers to the fact that meteorological stations in urban areas record higher temperatures than those in nearby rural locations. Reasons for this include the proximity of heated buildings and exhaust from air conditioning units and vehicles. All the land based data sets compensate for this effect. For example the GISS data set adjusts the rate of rise of urban stations so that it is similar to nearby rural stations; if there is no nearby rural station then the data is not used. In the case of satellites, adjustments have to made for changes in the satellites’ orbits and the deterioration of sensors. It is also necessary to make adjustments when one satellite is replaced by another. Figure 2
shows annual temperatures for the whole earth from all five measurement sources. It is normal to express temperatures as the difference between a given value and the average for a fixed period, usually 30 years. In this case we have chosen the period 1979 to 2008. This enables satellite data to be plotted on the same graph as land-based data without need for further adjustment. In terms of quality of data, the graph shows that all temperature estimates are broadly consistent. The differences are slightly larger in the early years, when there was a limited number of stations. A second feature is that the satellite temperature records seem to have more variance; hot years are relatively hotter and cold years are relatively colder.
Using the average of the three long-term stations there were distinct periods in the last hundred years:
- 1909 to 1944: an increase of 0.61 ºC,
- 1944 to 1975: fluctuating temperatures with an overall fall of 0.35 ºC,
- 1975 to 2005: an increase of 0.75 ºC.