3.2 Changes in Surface Climate: Temperature

3.2.1 Background

Improvements have been made to both land surface air temperature and sea surface temperature (SST) databases during the six years since the TAR was published. Jones and Moberg (2003) revised and updated the Climatic Research Unit (CRU) monthly land-surface air temperature record, improving coverage particularly in the Southern Hemisphere (SH) in the late 19th century. Further revisions by Brohan et al. (2006) include a comprehensive reassessment of errors together with an extension back to 1850. Under the auspices of the World Meteorological Organization (WMO) and the Global Climate Observing System (GCOS), daily temperature (together with precipitation and pressure) data for an increasing number of land stations have also become available, allowing more detailed assessment of extremes (see Section 3.8), as well as potential urban influences on both large-scale temperature averages and microclimate. A new gridded data set of monthly maximum and minimum temperatures has updated earlier work (Vose et al., 2005a). For the oceans, the International Comprehensive Ocean-Atmosphere Data Set (ICOADS) has been extended by blending the former COADS with the UK’s Marine Data Bank and newly digitised data, including the US Maury Collection and Japan’s Kobe Collection. As a result, coverage has been improved substantially before 1920, especially over the Pacific, with further modest improvements up to 1950 (Worley et al., 2005; Rayner et al., 2006). Improvements have also been made in the bias reduction of satellite-based infrared (Reynolds et al., 2002) and microwave (Reynolds et al., 2004; Chelton and Wentz, 2005) retrievals of SST for the 1980s onwards. These data represent ocean skin temperature (Section 3.2.2.3), not air temperature or SST, and so must be adjusted to match the latter. Satellite infrared and microwave imagery can now also be used to monitor land surface temperature (Peterson et al., 2000; Jin and Dickinson, 2002; Kwok and Comiso, 2002b). Microwave imagery must allow for variations in surface emissivity and cannot act as a surrogate for air temperature over either snow-covered (Peterson et al., 2000) or sea-ice areas. As satellite-based records are still short in duration, all regional and hemispheric temperature series shown in this section are based on conventional surface-based data sets, except where stated.

Despite these improvements, substantial gaps in data coverage remain, especially in the tropics and the SH, particularly Antarctica. These gaps are largest in the 19th century and during the two world wars. Accordingly, advanced interpolation and averaging techniques have been applied when creating global data sets and hemispheric and global averages (Smith and Reynolds, 2005), and advanced techniques have also been used in the estimation of errors (Brohan et al., 2006), both locally and on a global basis (see Appendix 3.B.1). These errors, as well as the influence of decadal and multi-decadal variability in the climate, have been taken into account when estimating linear trends and their uncertainties (see Appendix 3.A). Estimates of surface temperature from ERA-40 reanalyses have been shown to be of climate quality (i.e., without major time-varying biases) at large scales from 1979 (Simmons et al., 2004). Improvements in ERA-40 over NRA arose from both improved data sources and better assimilation techniques (Uppala et al., 2005). The performance of ERA-40 was degraded prior to the availability of satellite data in the mid-1970s (see Appendix 3.B.5).