Detection of Salinity Changes in the North Atlantic due to the Messinian Salinity Crisis

Abstract

The detection of astronomical climate forcing is an important tool in the study of Earth’s history. In order to quantify the reliability of these detections, spectral analysis and statistical testing with appropriate confidence levels is required. To gain insight into the appropriate significance levels, a numerical simulation of the sedimentary process was made based on the Piobbico core (Ripepè and Fischer, 1989). Results from these simulations were analyzed in the typical manner based on Mann and Lees [1996]. The confidence level with the lowest cumulative type I and II errors was found to depend on the signal strength of the cyclic component in the stratigraphic record. If the signal strength is in a range between 10% and 20% the ideal confidence level was in the range of 90% to 99%, thus for this signal strength range, the commonly applied 95% confidence limit is appropriate. However, when signal strength is greater than 20% the ideal confidence level falls into the 99 - 99.9% due to the frequency of type I errors. For lower signal strengths a confidence level of 90% was found to be better, but even at this level the combined rate of type I and II errors was high at 20%.