SHAKY CORRELATIONS: INFERRING EARTHQUAKE RECORDS FROM TURBIDITES

Depositional records extend well beyond instrumental or historical archives and have been widely used along fault systems to better constrain earthquake histories, cyclicity, and variability through time and space. Turbidite paleoseismology, or the study of deposits left behind by seismically triggered turbidity currents, can be useful to infer earthquake histories deep in time and across vast spatial scales. Inference of seismic triggering relies on assuming synchronous triggering of sediment gravity flows across a rupture zone, by earthquake shaking. Flows leave behind extensive deposits that can be correlated based on age dating and/or sedimentological and stratigraphic character. However, depositional characteristics of turbidites commonly have great spatial variation, making correlation uncertain and subjective; and marine age constraints typically have uncertainties of decades or more, making events separated by sub-decadal intervals difficult to distinguish. In this work, we re-evaluate previously correlated turbidite facies in core photographs and X-Ray computed tomography images from the margin of the Cascadia Subduction Zone and present a new method to correlate turbidites that yields a more objective and repeatable stratigraphic framework to underpin earthquake recurrence. We use a dynamic time warping optimization algorithm to correlate pairs of Cascadia sediment core magnetic susceptibility logs, which provides measures of correlation coefficients and their significance. We compare these measures to those derived from distributions of randomly generated turbidite sequences and find that only a small number of core pairs can be more confidently correlated than randomly stacked turbidites. Results suggest that caution should be used when using low-confidence correlations for turbidite paleoseismology and this methodology promises a more robust correlation strategy for future stratigraphic studies.