Perdiou et al. (2016) analysed two data-series from a 60 m Campanian chalk interval in the Adda-3 well offshore Denmark with the objective of identifying orbitally forced cycle periods and thus time-calibrating this part of the Campanian. The data are (1) a series of 457 gamma-ray (GR) log values (vertical spacing 0.1524 m), and (2) a series of 844 X-ray fluorescence measurements of Fe concentration from core at an average resolution of 0.07 m (but with several gaps in the data). Perdiou et al. used Thomson's multi-taper method (MTM) to generate their periodograms. They then used the Matlab function Redconf (Husson et al. 2014) to model the background (random) noise and thus to estimate confidence levels (CLs) for discriminating candidate cycles from the noise model. The approach to the estimation of CLs used in Redconf is widely used in cyclostratigraphy, but Vaughan et al. (2011) showed that it generates false positive results; it identifies cycles that are not present in the data. By (1) reassessing one of Perdiou et al.’s datasets, and (2) using a synthetic random dataset, we show here that the same criticism applies to their spectral analyses, implying that their proposed Campanian time-calibration is not reliable.

Perdiou et al. (2016) based their Campanian time-calibration mainly on the Adda-3 GR data series, and our Figure 1 duplicates (in log–log presentation) the depth-domain spectral analysis shown in their figure 7 (top left). Spectral power exceeds the 99% CL at a number of points on the periodogram, suggesting statistically …