The instrumental surface temperature record (“instrumental record”) is derived from traditional thermometer readings and provides the basis for generating the large-scale (global mean or hemispheric mean) surface temperature estimates used in climate change studies. The global average temperature is produced as a combination of near-surface land air temperatures and temperatures of the sea water near the surface (or sea surface temperatures [SSTs]) for the oceans. Land air temperatures are measurements taken by thermometers mounted in shelters about 1.5 meters above the land surface, or higher in areas where snow cover may be substantial. About 2,000 stations report land air temperatures for the global compilations shown in this chapter. The stations are not spatially distributed to monitor all land areas with equal density; unpopulated and undeveloped areas have always tended to have poor coverage.
SSTs are measured by ships, buoys, bathythermograph profilers, and, since 1981, satellites. Ships generally take the water temperature in one of three ways: buckets (the oldest method), hull sensors, and water drawn in to cool the engines (injection temperatures). The depths of ship measurements vary from 1 to 15 meters. Buoys are more standardized and report temperatures generally at 1 meter as well as several other depths depending on the buoy type.
Very few land air temperature records begin prior to 1856, so estimates of large-scale (i.e., global and hemispheric) averages are uncertain before that time. The average SST for all oceans is less well known than land air temperature, especially during the middle to late 19th century, when large portions of the tropical and southern oceans were poorly sampled (and these areas remain comparatively undersampled). Differences in the types of measurement methods, the generally unknown calibration of instruments, and the sparse geographic and temporal sampling in many areas contribute to uncertainties in the estimates of large-scale averages. In addition, the proxy indicators discussed in Chapters 3–8 are generally not directly sensitive to 1.5 meter air temperature. For example, borehole temperature profiles are sensitive to the ground surface temperatures, and ice isotopic ratios are sensitive to cloud-level atmospheric temperatures. Significant systematic differences can exist between temperatures at such different elevations with respect to ground, and these differences represent one of the inherent uncertainties in performing surface temperature reconstructions.
Figure 2-1 shows three large-scale averages of annual mean surface temperature anomalies from the HadCRUT2v dataset (Jones et al. 2001), which is commonly used in both proxy reconstructions and more general global climate studies.1 The three estimates are for (1) global, (2) Northern Hemisphere, and (3) Northern Hemisphere extratropical land areas only (20°N–85°N). The Northern Hemisphere extratropical land area estimate has the largest variability of the three because the mid- and high-