high-altitude locations, and few of the ice-core reconstructions of temperature have been explicitly calibrated against instrumental time series. The differences between ice-core records from various sites also make it clear that there is a pressing need to identify and differentiate the relative influences of local mesoscale and hemispheric atmospheric factors.
In her paper in this chapter, Street-Perrott (1995) discusses how fluctuations in the water level and surface area of relatively undisturbed lakes (i.e, there has been no human-induced change in the water budget) can provide a measure of climate variability on time scales of months to millions of years. The paper provides examples of the response of tropical lakes to variations in ocean-atmosphere interactions over the Pacific (Lake Pátzcuaro), the Indian Ocean (Lake Victoria), and the Atlantic (Lake Chad, Lake Malawi), and points to paleolimnological evidence for century-scale droughts in southern Africa and the tropical Americas.
Finally, Reifsnyder (1995, in this chapter) analyzes observational and paleoclimate records of temperature in an attempt to determine the realism of models' predicted global-warming rates. Some models predict a rate of warming that is 10 to 40 times faster than the natural warming that followed the last ice age. On the basis of his analysis, however, Reifsnyder argues that the climate change between now and the end of the next century, for a "business as usual" scenario, should take place at no more than twice the maximum rate estimated for changes since the last ice age. He further argues that, because models are known to overpredict by a factor of two the rate of change over the past century in response to the CO2 increase, the rate of global warming over the next century is in fact unlikely to be higher than any estimated post-ice-age maximum rate— a position which certainly could generate a lot of debate, given the other forcings likely to be operating.
Soil and varved sediments, and historical documents, are two sources of atmospheric proxy data that are not covered in this volume. Sedimentary and soil cores can reveal significant paleoclimate information. For example, in a recently published paper in Science, Weiss et al. (1993) attributed the demise of the Akkadian culture (fl. 3000 B.C.) of southern Mesopotamia (present-day northeast Syria) in 2200 B.C. to an abrupt change in climate. The combined archaeological and soil-stratigraphic data for the area point to a shift in climate toward the arid, and a dry period persisting for about 300 years.
A variety of historical sources, such as ancient inscriptions, personal diaries and correspondence, scientific and quasi-scientific writings, government records, and public and private chronicles and annals record climate events that were seen as having some significance at the time. The information may include observations of weather phenomena, such as the occurrence of extreme rain- and snowfalls, droughts, floods, or lake and river freeze-ups and breakups. Historical records, unfortunately, fail to give a complete picture of former climate conditions. They are often discontinuous observations, biased towards extreme events. The important long-term trends tend to go unremarked. A good source of discussions of documentary records of past climate is the recently published book by Bradley and Jones (1992).
While proxy indicators are extremely important, even crucial, to climate reconstruction, their limitations should be kept in mind. The most significant are:
It is not always clear whether the signal they record reflects only local conditions, or is representative of regional or global conditions.
Their accuracy is often unknown or untested.
They often reflect more than one variable, making interpretation difficult.
The absolute, and even the relative, timing of the record is often not certain.
Bradley and Eddy (1989) stressed the importance of the last of these limitations, and noted that without accurate dating it is impossible to determine whether certain events occurred synchronously or not. In the Vostok ice core of central East Antarctica, for example, variations in air temperature (Jouzel et al., 1987) and atmospheric CO2 concentration (Barnola et al., 1987) have been reconstructed for the last 160,000 years. Within the dating uncertainties, these records show striking correspondence between high CO2 concentrations and warm temperatures. However, whether the increases in CO2 concentration precede or follow the temperature increases cannot be assessed accurately from these records.
Accurate dating is also a problem in attempting to overcome the first limitation listed above. That is, in order to reconstruct a global picture of climate at a given time, it is necessary to reconstruct climate variables from a number of different locations for that time. The accurate dating needed to establish the relationships of the various records is not always available. This problem plagues most of the ocean records, because distinct annual varves are rarely present. Dating is mostly accomplished through other techniques, again of limited accuracy, which constrains the degree to which comparisons can be made. Sowers et al. (1993) attempted to relate the variations in the Vostok ice record to those in the ocean by correlating supposedly globally synchronous changes in oxygen isotope concentrations. This technique has met with some success, but the range of error precludes comparisons over time scales less than millennial.
Other limitations arise because recording is often discontinuous. For instance, the recording mechanism may be disturbed during or after the recording, or the recovery may