Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection Palaeoclimate proxy indicators

A "proxy" climate indicator is a local record that is interpreted using physical or biophysical principles to represent some combination of climate-related variations back in time. Palaeoclimate proxy indicators have the potential to provide evidence for large-scale climatic changes prior to the existence of widespread instrumental or historical documentary records. Typically, the interpretation of a proxy climate record is complicated by the presence of "noise" in which climate information is immersed, and a variety of possible distortions of the underlying climate information (e.g., Bradley, 1999; Ren, 1999a,b). Careful calibration and cross-validation procedures are necessary to establish a reliable relationship between a proxy indicator and the climatic variable or variables it is assumed to represent, providing a "transfer" function through which past climatic conditions can be estimated. High-resolution proxy climate indicators, including tree rings, corals, ice cores, and laminated lake/ocean sediments, can be used to provide detailed information on annual or near-annual climate variations back in time. Certain coarser resolution proxy information (from e.g., boreholes, glacial moraines, and non-laminated ocean sediment records) can usefully supplement this high-resolution information. Important recent advances in the development and interpretation of proxy climate indicators are described below.

Tree rings
Tree-ring records of past climate are precisely dated, annually resolved, and can be well calibrated and verified (Fritts, 1976). They typically extend from the present to several centuries or more into the past, and so are useful for documenting climate change in terrestrial regions of the globe. Many recent studies have sought to reconstruct warm-season and annual temperatures several centuries or more ago from either the width or the density of annual growth rings (Briffa et al., 1995; D'Arrigo et al., 1996; Jacoby et al., 1996; D'Arrigo et al., 1998; Wiles et al., 1998; Hughes et al., 1999; Cook et al., 2000). Recently, there has been a concerted effort to develop spatial reconstructions of past temperature variations (e.g., Briffa et al., 1996) and estimates of hemispheric and global temperature change (e.g., Briffa et al., 1998b; Briffa, 2000). Tree-ring networks are also now being used to reconstruct important indices of climate variability over several centuries such as the Southern Oscillation Index (Stahle et al., 1998), the North Atlantic Oscillation (Cook et al., 1998; Cullen et al., 2001) and the Antarctic Oscillation Index (Villalba et al., 1997) (see also Section 2.6), as well as patterns of pre-instrumental precipitation and drought (Section

Several important caveats must be borne in mind when using tree-ring data for palaeoclimate reconstructions. Not least is the intrinsic sampling bias. Tree-ring information is available only in terrestrial regions, so is not available over substantial regions of the globe, and the climate signals contained in tree-ring density or width data reflect a complex biological response to climate forcing. Non-climatic growth trends must be removed from the tree-ring chronology, making it difficult to resolve time-scales longer than the lengths of the constituent chronologies (Briffa, 2000). Furthermore, the biological response to climate forcing may change over time. There is evidence, for example, that high latitude tree-ring density variations have changed in their response to temperature in recent decades, associated with possible non-climatic factors (Briffa et al., 1998a). By contrast, Vaganov et al. (1999) have presented evidence that such changes may actually be climatic and result from the effects of increasing winter precipitation on the starting date of the growing season (see Section Carbon dioxide fertilization may also have an influence, particularly on high-elevation drought-sensitive tree species, although attempts have been made to correct for this effect where appropriate (Mann et al., 1999). Thus climate reconstructions based entirely on tree-ring data are susceptible to several sources of contamination or non-stationarity of response. For these reasons, investigators have increasingly found tree-ring data most useful when supplemented by other types of proxy information in "multi-proxy" estimates of past temperature change (Overpeck et al., 1997; Jones et al., 1998; Mann et al., 1998; 1999; 2000a; 2000b; Crowley and Lowery, 2000).

Palaeoclimate reconstructions from corals provide insights into the past variability of the tropical and sub-tropical oceans and atmosphere, prior to the instrumental period, at annual or seasonal resolutions, making them a key addition to terrestrial information. Because of their potential to sample climate variations in ENSO-sensitive regions, a modest network of high-quality coral site records can resolve key large-scale patterns of climate variability (Evans et al., 1998). The corals used for palaeoclimate reconstruction grow throughout the tropics in relatively shallow waters, often living for several centuries. Accurate annual age estimates are possible for most sites using a combination of annual variations in skeletal density and geochemical parameters. Palaeoclimate reconstructions from corals generally rely on geochemical characteristics of the coral skeleton such as temporal variations in trace elements or stable isotopes or, less frequently, on density or variations in fluorescence. Dunbar and Cole (1999) review the use of coral records for palaeoclimatic reconstruction.

Ice cores
Ice cores from polar regions of northern Greenland, Canada and the islands of the North Atlantic and Arctic Oceans, Antarctica, and alpine, tropical and sub-tropical locations (e.g., Thompson, 1996) can provide several climate-related indicators. These indicators include stable isotopes (e.g., 18O), the fraction of melting ice, the rate of accumulation of precipitation, concentrations of various salts and acids, the implied atmospheric loading of dust pollen, and trace gases such as CH4 and CO2.

Recently, there has been increased activity in creating high-resolution Antarctic ice core series e.g., for the past millennium (Peel et al., 1996; Mayewski and Goodwin, 1997; Morgan and van Ommen, 1997). In certain regions, isotope information from ice cores shows the late 20th century temperatures as the warmest few decades in the last 1,000 years (Thompson et al., 2000a). Key strengths of ice core information are their high resolution (annual or even seasonal where accumulations rates are particularly high - see van Ommen and Morgan, 1996, 1997), availability in polar and high-elevation regions where other types of proxy climate information like tree-ring data are not available, and their provision of multiple climate- and atmosphere-related variables from the same reasonably well dated physical location (e.g., the GISP2 core; White et al., 1998a). A weakness of ice core data is regional sampling bias (high elevation or high latitude) and melt water and precipitation accumulation data are not easy to date accurately.

The best dated series are based on sub-annual sampling of cores and the counting of seasonal ice layers. Such series may have absolute dating errors as small as a few years in a millennium (Fisher et al., 1996). Dating is sometimes performed using volcanic acid layers with assumed dates (e.g., Clausen et al., 1995) but uncertainties in the volcanic dates can result in dating uncertainties throughout the core (Fisher et al., 1998).

Lake and ocean sediments
Annually laminated (varved) lake sediments offer considerable potential as high-resolution archives of palaeo-environmental conditions where other high-resolution proxy indicators are not available (e.g., arid terrestrial regions), and latitudes poleward of the treeline (Lamoureux and Bradley, 1996; Wohlfarth et al., 1998; Hughen et al., 2000). When annual deposition of the varves can be independently confirmed (e.g., through radiometric dating), they provide seasonal to interannual resolution over centuries to millennia. Varved sediments can be formed from biological processes or from the deposition of inorganic sediments, both of which are often influenced by climate variations. Three primary climate variables may influence lake varves: (a) summer temperature, serving as an index of the energy available to melt the seasonal snowpack, or snow and ice on glaciers; (b) winter snowfall, which governs the volume of discharge capable of mobilising sediments when melting; and (c) rainfall. Laminated lake sediments dominated by (a) can be used for inferences about past high latitude summer temperature changes (e.g., Overpeck et al., 1997), while sediments dominated by the latter two influences can be used to estimate past drought and precipitation patterns (Section

Ocean sediments may also be useful for high-resolution climate reconstructions. In rare examples, annually laminated sediments can be found (e.g., Hughen et al., 1996; Black et al., 1999) and it is possible to incorporate isotope and other information in climate reconstructions, much as varved lake sediments are used. Otherwise, sedimentation rates may sometimes still be sufficiently high that century-scale variability is resolvable (e.g., the Bermuda rise ocean sediment oxygen isotope record of Keigwin, 1996). Dating in such cases, however, must rely on radiometric methods with relatively poor age control.

Borehole measurements
Borehole measurements attempt to relate profiles of temperature with depth to the history of temperature change at the ground surface. The present global database of more than 600 borehole temperature-depth profiles has the densest geographic coverage in North America and Europe, but sparser data are available in other regions (e.g., Australia, Asia, Africa and South America). The depths of the temperature profiles range from about 200 to greater than 1,000 m, allowing palaeo-temperature reconstructions back several hundred to a thousand years. Although large-scale temperature reconstructions have been made to more than a millennium ago (Huang et al., 1997), they show substantial sensitivity to assumptions that are needed to convert the temperature profiles to ground surface temperature changes. Borehole data are probably most useful for climate reconstructions over the last five centuries (Pollack et al., 1998).

Figure 2.19: Reconstructed global ground temperature estimate from borehole data over the past five centuries, relative to present day. Shaded areas represent ± two standard errors about the mean history (Pollack et al., 1998). Superimposed is a smoothed (five-year running average) of the global surface air temperature instrumental record since 1860 (Jones and Briffa, 1992).

Figure 2.19 shows a reconstructed global ground surface temperature history (Pollack et al., 1998; see also Huang et al., 2000) from an average of the 358 individual sites, most located in North America and Eurasia, but some located in Africa, South America and Australia (similar results are obtained by Huang et al., 2000, using an updated network of 616 sites). Superimposed is an instrumental estimate of global surface air temperature (Jones and Briffa, 1992). The ensemble of reconstructions shows that the average ground temperature of the Earth has increased by about 0.5°C during the 20th century, and that this was the warmest of the past five centuries. About 80% of the sites experienced a net warming over this period. The estimated mean cumulative ground surface temperature change since 1500 is close to 1.0 ± 0.3°C. Uncertainties due to spatial sampling (see Pollack et al., 1998 and Huang et al., 2000) are also shown. It should be noted that the temporal resolution of the borehole estimates decreases sharply back in time, making it perilous to compare the shape of the trend shown in Figure 2.19 with better-resolved trends determined from higher-resolution climate proxy data discussed below.

While borehole data provide a direct estimate of ground surface temperatures under certain simplifying assumptions about the geothermal properties of the earth near the borehole, a number of factors complicate their interpretation. Non-temperature-related factors such as land-use changes, natural land cover variations, long-term variations in winter snow cover and soil moisture change the sub-surface thermal properties and weaken the interpretation of the reconstructions as estimates of surface air temperature change. In central England, where seasonal snow cover is not significant, and major land-use changes occurred many centuries ago, borehole ground surface temperature trends do tend to be similar to those in long instrumental records (Jones, 1999). In contrast, Skinner and Majorowicz (1999) show that borehole estimates of ground surface temperature warming during the 20th century in north-western North America are 1 to 2°C greater than in corresponding instrumental estimates of surface air temperature. They suggest that this discrepancy may be due to land-use changes that can enhance warming of the ground surface relative to that of the overlying atmospheric boundary layer (see also Lewis, 1998). Such factors need to be better understood before borehole temperature measurements can be confidently interpreted.

Continues on next page

Other reports in this collection