Our confidence in the validity of large-scale surface temperature reconstructions is based, in part, on the fact that the individual proxy data series used to create these reconstructions generally exhibit strong correlations with local environmental conditions. In most cases, there is a physical, chemical, or physiological reason why the proxy reflects local temperature variations. Our confidence is stronger when multiple independent lines of evidence point to the same result, as in the case of the Little Ice Age cooling and of the 20th century warming.

Although the reconstructions based on borehole temperature composites and glacier length records in Figure O-5 do not extend back far enough to provide an independent check on the tree-ring- and multiproxy-based reconstructions for periods prior to the 16th century, there is additional evidence pointing toward the unique nature of recent warmth in the context of the last one or two millennia. This evidence includes the recent melting on the summits of ice caps on Ellesmere Island and Quelccaya and other Andean mountains, the widespread retreat of glaciers in mountain ranges around the world (which in some places has exposed decomposing organic matter that dates to well before A.D. 1000), the recent disintegration of the Larsen B ice shelf in Antarctica, and the fact that ice cores from both Greenland and coastal Antarctica show evidence of 20th century warming (whereas only Greenland shows warming during medieval times). Ice cores from the Andes and Tibetan plateau and the recession of the ice caps on mountains in equatorial Africa, which reflect both temperature and hydrologic processes, also suggest that the 20th century climate is unusual in the context of the last few thousand years.

What are the limitations and strengths of large-scale surface temperature reconstructions?

The main reason that our confidence in large-scale surface temperature reconstructions is lower for periods before about A.D. 1600 is the relative scarcity of precisely dated proxy evidence. Other factors limiting our confidence in these reconstructions include:

• The relatively short length of the instrumental record (about 150 years) only provides a few pieces of independent information to both calibrate and validate surface temperature reconstructions over large spatial scales and multidecade time periods. Instrumental records used for calibration and validation of proxy data have also been collected during a period when both global mean temperatures and human impacts on the environment have increased substantially.

• Although care is taken when selecting, analyzing, and interpreting proxy data, there is always the possibility that the relationship between the proxy and local surface temperature may have varied over time. Most proxies are sensitive to temperature only during certain times of year, and the proxy may reflect temperature variations on timescales longer than the calibration period.

• In the absence of a consensus as to which methods or statistical formulas are most appropriate for calibrating and validating these reconstructions, different choices made by different investigators and research groups also contribute to the differences between them. In some cases the choice of whether or not to include one or more proxy records in a reconstruction has also been a factor.