temperature (Rutherford et al. 2005). One reconstruction does not use tree ring networks at all for century-scale and longer changes, but instead relies on a combination of geochemical and sedimentary proxies (Moberg et al. 2005b).
Temperature records from about A.D. 1600 to the present derived from large-scale surface temperature reconstructions are consistent with other sources of temperature information for the period, including borehole temperatures and glacier length records.
Prior to about 1600, information is sparser and the pattern of change is not necessarily synchronous, but periods of medieval warmth are seen in a number of diverse records, including historical information from Europe and Asia; cave deposits; marine and lake sediments; and ice cores from Greenland, Ellesmere Island, Tibet, and the equatorial Andes.
Many challenges remain as research progresses to use large-scale surface temperature reconstructions to learn about climate history (Hughes 2002, Rutherford et al. 2005, D’Arrigo et al. 2006). There are two major structural challenges. First, the amount of high-quality proxy data available for analysis decreases markedly as one moves back in time. The great richness of tree ring network data available for 1700, for example, is largely depleted by A.D. 1000. Large-scale temperature reconstructions should always be viewed as having a “murky” early period and a later period of relative clarity. The boundary between murkiness and clarity is not precise but is nominally around A.D. 1600. Second, the finite length (about 150 years) of the instrumental temperature record available for calibration of large-scale temperature estimates places limits on efforts to demonstrate the accuracy of temperature reconstructions. Further research should be aimed at providing independent checks on reconstructions using borehole temperatures, glacier length records, and other proxies.
The role of statistical methods is not trivial. Each individual proxy provides a record of environmental change, but the process of combining these environmental signals into a large-scale spatially averaged temperature requires statistical evaluation. Even if a single proxy is a perfect recorder of the local environment, the question remains of whether the local environments are adequately or representatively sampling the large-scale temperature field. In addition, most proxy records lack the annual chronological precision found in tree ring data; the typical dating error might be 1–5 percent of the age of the sample for annually layered records such as lake varves and 5–10 percent for radiometrically dated records spanning the last 2,000 years.
The committee identified the following limitations of large-scale surface temperature reconstructions that would benefit from further research:
There are very few degrees of freedom in validations of the reconstructed temperature averaged over periods of decades and longer. The RE validation metric used by Mann et al. (1998, 1999) is a minimum requirement, but the committee questions whether any single statistic can provide a definitive indication of the uncertainty inherent in the reconstruction. Demonstrating performance for the higher-frequency component (e.g., by calculating the CE statistic) would increase confidence but still would not fully address the issue of evaluating the reconstruction’s ability to capture temperature variations on decadal-to-centennial timescales.
Using proxies sensitive to hydrologic variables (including moisture-sensitive trees and isotopes in tropical ice cores and speleothems) to take advantage of observed