Person:

Proistosescu, Cristian

We address the problem of detecting quasi-periodic variability at orbital frequencies within pre-Pleistocene climate records using depth-derived and orbitally tuned chronologies. Many studies describing orbital variability in pre-Pleistocene sediment hosted isotope records employ climatic records that are set on orbitally tuned chronologies, without accounting for the bias in spectral power estimates introduced by orbital tuning. In this study we develop a method to quantify the effects of tuning upon spectral estimates and, in particular, to more properly determine the statistical significance of spectral peaks associated with orbital frequencies. We apply this method to two marine sediment δ18O records spanning the Oligo-Miocene, from ODP cores 1090 and 1218. We find that using linear age–depth relationships reveals statistically significant spectral peaks matching eccentricity in core 1090, and obliquity and precession in core 1218, where the last appears most significant. Tuning the chronologies to the orbital solutions of Laskar et al. (2004) increases the statistical significance of the precession peak, whereas the obliquity and eccentricity peaks become indistinguishable from those expected from tuning noise. This result can be understood in that tuning records with high signal to noise ratios tends to lead to more significant spectral peaks, whereas a linear age–depth relationship is better suited for detecting peaks when signal to noise ratios are low. We also demonstrate this concept using synthetic records.