Benchmarking Multiple Methods for the Spectral Analysis of Seafloor Morphology

Abstract

The influence of glacial sea-level changes on the quasi-periodic spacing of abyssal hills has remained controversial for over a decade. Spectral analyses of seafloor morphology have yielded contrasting conclusions regarding whether periodicities associated with orbital Milankovitch forcing are recorded in abyssal hill spacing, raising questions about the sensitivity of results to methodological choices. An important component of spectral analysis is prewhitening, or methods by which to mitigate the effects of the strong red background spectrum, characteristic of seafloor bathymetry. Various methods can have first-order implications for the inferred significance of spectral peaks. Here we benchmark four approaches for analyzing bathymetric spectra to detect periodicities in seafloor morphology. Through synthetic testing, we confirm that spectral analysis is better performed using a multitaper approach relative to other standard approaches and that a leveled spectrum can be derived by either prewhitening with an autoregressive and moving-average model or by leveling with a von Kármán model. Application of the different methods to previously published and newly analyzed bathymetric data sets shows that methodological choices substantially influence the inferred significance of spectral peaks, even when similar spectral features are present. Our results provide practical guidance for selecting appropriate spectral methods and bathymetric data sets when testing hypotheses linking glacial/interglacial sea-level variations to seafloor morphology and establish a methodological framework for future investigations of periodic signals in oceanic crustal accretion.