be incomplete. This introduces gaps or perturbations that disguise or break the general patterns. An excellent example of the type of inconsistencies that may occur from the different recording properties of different climate proxies is provided by Grootes's (1995) paper in this chapter. However, the promising news in this area is that more researchers are reporting long-term reconstructions from different proxy sources and different geographic locations. The latest is the study by Lara and Villalba (1993) of a 3620-year reconstructed temperature record from alerce trees (Fitzroya cupressoides) from southern Chile. Their work, which has shown the alerce tree as being the second longest-living tree after the bristlecone pine, expands the availability of long-term climate records for the Southern Hemisphere.
In Chapter 2 Diaz and Bradley (1995) provide a useful discussion of a number of potential sources of uncertainties and biases associated with climate variable reconstruction, even for the relatively well-understood tree rings. They warn that changes over time in the composition of the tree-ring network used for reconstruction are likely to affect the high-frequency variance and, to some extent, the low-frequency variance as well. Considerable effort has been made to reduce the uncertainties associated with the long-term reconstruction of climate variables from the composition of tree-ring samples. For instance, an index value (temperature, for instance) for a given site is obtained from samples of trees that may vary considerably in age or even be dead. Attention must therefore be given during the reconstruction analysis to getting the chronology right for every tree; removing the biological growth trend; identifying the nature of the climate signal and its strength; replication; and the way in which the analysis method deals with changes in site conditions over the recorded interval. As a result of this careful scrutiny, tree-ring records exhibit exceptionally high fidelity, with comparatively minor interpretational problems. Similar sets of considerations must be addressed for other proxy indicators, no doubt including some not yet identified.
It is clear that accurate dating is required to assess the rate at which past climate changes have occurred and to reconstruct globally synchronous records of climate, particularly for changes on time scales of less than a century. The duration of high-frequency, short-term events may be less than the normal error associated with most methods used in dating the proxy records available. The issue of dating thus merits continued attention. Similarly, a sustained effort must be made to identify problems and uncertainties in reconstructions based on proxy indicators and to assess the associated errors; this should come naturally with increased experience in these relatively novel techniques. Last. combining records that have been drawn from different areas and that use different types of indicators into a consistent picture will be crucial for the study and reconstruction of global climate variations.
Finally, as with climate modeling, the reconstruction of past climates requires the employment of a full range of different proxy indicators. These must be developed more fully so that they can be used to intercalibrate the reconstructed climates and cross-check the reconstructions; to provide multiple images of climate through different climate indicators; and to reduce the influence of noise through averaging. In this early stage, we are seeing only the tip of the iceberg; the results of the use of proxy indicators only hint at the climate insights yet to be won by means of these invaluable resources.