Abstract:
Correlations between mean annual temperature (MAT) and the weighted average oxygen isotope composition of yearly precipitation (δ18Opt) are well-known, but the utility of modern relations to make reliable estimates of temperature change over geological time is uncertain. This question has been addressed by using seasonal subsets of the global data base of temperature and isotopic measurements to represent two different climate modes. A comparison of middle- to high-latitude δ18Opt/temperature relations for each climate mode reveals (1) a significant offset between them, and (2) a difference in the strength of their correlations. The offset in relations is due to differences in temperature and water vapor budget in the tropics, and can lead to serious underestimates of temperature change. Differences in the strength of correlations arise from the influence of climate mode-specific, non-temperature factors on δ18Opt. The overall result is that no single relation can be used in all cases to make unambiguous temperature estimates using a temporal record of δ18Opt values. One way to overcome these problems is to reconstruct δ18Opt/temperature relations for the time periods being investigated. If an appropriate proxy for δ18Opt is available, it may also be possible to estimate temperature without relying on δ18Opt/temperature relations. A promising alternative to these options is to use records of δ18Opt to test predictions of global climate models, an approach that may allow a reliable and more complete reconstruction to be made of climate change over geologic time.