On the Reconstruction of Paleosalinities
1999
Methods potentially useful for paleosalinity reconstructions are summarized and applied to surface sediments and two cores from the Atlantic. The first approach is based on the oxygen isotope ratio in calcite tests of planktonic foraminifera in conjunction with an independent sea-surface temperature estimate. A two-step procedure from foraminiferal δ18O to the isotopic composition of sea-water and from there to an estimate of paleosalinity is proposed. The quality of the estimation of δ18O for sea-water depends heavily on the reliability of the independent temperature method. The final salinities are obtained through an assumed δ18O-salinity relationship for sea-water. Propagation of errors yields large deviations, especially in the tropics where slopes of the δ18O-salinity relationship are low. An uncertainty in temperature of ±1°C leads to errors in the salinity reconstruction of ±0.5 to ± 1.2 %o. Downcore application at a site in the western equatorial Atlantic indicates salinity increases of roughly 2.5 %o in the glacials, but also demonstrates sensitivity of the results to the temperature estimates and the δ18O-salinity relationship. Two additional methods of paleosalinity estimation were investigated both of which employ foraminiferal abundance data. An Artificial Neural Network was used for the first time to reconstruct paleosalinities. The obtained results were then compared to results from a Modern Analog Technique. Application to surface sediments yields comparable results for both methods, with standard deviations between 0.49 and 0.63 %o. Salinity calculations performed on downcore data in the tropical and subtropical South Atlantic indicate that the results are controlled by today’s temperature-salinity relationship. This leads to the conclusion that paleosalinity reconstructions from species composition of foraminiferal assemblages are unrealistic, because of a predominant response of the fauna to temperature.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
100
References
26
Citations
NaN
KQI