Detecting cycles in stratigraphic data: Spectral analysis in the presence of red noise

摘要: [1] We discuss the detection of cyclic signals in stratigraphic ‘time series’ using spectral methods. The dominant source variance record is red noise, which greatly complicates process searching for weak periodic signals. We highlight two issues that are more significant than generally appreciated. first lack a correction ‘multiple tests’ – many independent frequencies examined periods but significance test appropriate examination single frequency. second problem poor choice null hypothesis used to model spectrum non-periodic variations. Stratigraphers commonly assume noise first-order autoregressive AR(1) practice often gives very match real data; fact goes largely unnoticed because checking rarely performed. These problems have effect raising number false positives far above expected rate, extent literature on spatial cycles dominated by positives. In turn these will distort construction astronomically calibrated timescales, lead inflated estimates physical deterministic forcing climate and depositional processes pre-Neogene, may even bias models solar system dynamics long timescales. make suggestions controlling positive emphasize value Monte Carlo simulations validate calibrate analysis