FIGURE 2-2 Annual mean and summer (June–August) mean temperature anomalies in degrees Celsius for Northern Hemisphere extratropical land areas from the HadCRUT2v surface temperature dataset. SOURCE: Jones et al. (2001).


Because proxy-based surface temperature reconstructions often depend on either local or large-scale average land air temperatures and/or SSTs, any errors in the instrumental temperature record will reduce the confidence in the reconstructed temperature record. Several factors influence the land air temperature measurements over time. As land use has changed (e.g., from forest to urban), many thermometers in the land air temperature record have responded to the changes in the thermal properties of their surroundings, yielding temperature changes that are real but not likely due to large-scale climatic causes. The geographic distribution of the land air temperature sites has also grown significantly since 1856, so data from regions that previously had no measurements now provide a more accurate large-scale average, giving rise to larger errors in the earlier part of the record. Similarly, compilations of SST measurements suffer from poor calibration and sampling in the earlier decades. These types of problems are estimated to introduce a potential error (95 percent confidence) of ~0.10°C for the earliest decades of the global and Northern Hemisphere average temperature values (Folland et al. 2001b). In the most recent decades, improved coverage and better knowledge of instrumental biases, such as the effects of urbanization, reduce the error range to ~0.04°C (Brohan et al. in press). These potential errors are relatively small

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement