On reading the Research commentary "It was the best of times, it was the worst of times" by P. Jones (Science's Compass, 24 Apr., p.544), the reader might see more disagreement than actually exists between that piece and a recent paper by Mann et al. (1). We take this opportunity to clarify some possible misunderstandings. The point expressed in the piece by Jones regarding the need for extensive and independent cross-validation of proxy-based reconstructions is indeed one that is wholly embraced by Mann et al. (1). The Northern Hemisphere mean temperature series shown in (1) is based on the calibrations which exhibited the greatest skill, that is, the fraction of instrumental variance described in both calibration and cross-validation or "verification." This reconstruction was based on all available data, which included proxy data, and the few long instrumental and historical records. However, a variety of additional independent calibration-verification experiments, although not shown, were clearly referred to in (1) and are described in detail on Nature's supplementary information Web site (2), referred to by Mann et al. (1). In several of these experiments, only true "proxies," that is, natural archives, were used in the temperature pattern reconstructions. The long historical and instrumental records dating back several centuries in Europe and North America were withheld from the calibration experiment. These data were then used for just the kind of independent long-term verification advocated by Jones for all paleoclimate reconstruction studies. Those tests demonstrated that pure proxy-based reconstructions of global surface temperatures were able to reproduce quite faithfully the actual instrumental temperature records that are available several centuries back in time [a dozen in Europe and western Asia, and one in North America--see (2)]. We all advocate strongly the independent verification of proxy-based climate reconstructions by the use of long instrumental and historical data, withheld from the calibration, for independent long-term cross-validation.
The comparison shown by Jones between Mann et al.'s Northern Hemisphere temperature reconstruction (1) and two other recent estimates is useful in several ways. For example, it demonstrates the robustness of the conclusion that the 20th-century warming is unusual in the context of the past several centuries, on the basis of largely independent estimates. However, the comparison might be misleading to those readers unfamiliar with the details of seasonal and spatial sampling contributing to the different estimates. Apparent discrepancies between these different estimates are largely associated with these sampling differences. The Mann et al. series represents estimated annual calendar-mean conditions and a spatial average over the entire--tropical and extratropical--Northern Hemisphere. The other two series shown in (1) represent estimated warm-season half-year conditions with a more extratropical Northern Hemisphere emphasis in the data used. As mentioned (1), certain proxies, like tree-ring density data, are highly effective indicators of warm-season temperatures. This makes warm-season temperatures the logical quantity to reconstruct with such data when used alone. In contrast, the analysis described by Mann et al. (1) attempted to exploit the complementary seasonal information in a diverse set of proxy, instrumental, and historical indicators. The statistical calibration-verification experiments performed in that study indicated that annual mean conditions could be more accurately represented than warm- or cold-season half-year conditions. More work is needed to resolve the seasonal details of climatic variability in past centuries. A key aim of future efforts must be to further improve multiproxy networks to both extend reconstructions of global climate further back in time, and to reduce uncertainties in existing estimates. As we decrease present uncertainties [represented, for example, by the substantial error bars shown for the Mann et al. reconstruction (1), or the differences between independent warm-season temperature estimates shown in the piece by Jones], it is to be hoped that we will soon be able to better constrain patterns of climate variability in past centuries. Such improved constraint will aid us in verifying important aspects of the numerical climate models now used to describe possible future climate scenarios.
Michael E. Mann
Raymond S. Bradley
Department of Geosciences,
University of Massachusetts,
Amherst, MA 01003-5820, USA
Malcolm K. Hughes
Laboratory of Tree Ring Research,
University of Arizona,
Tucson, AZ 85721, USA
Philip D. Jones
Climatic Research Unit,
School of Environmental Sciences,
East Anglia, Norwich, NR4 7TJ,
United Kingdom