LAND-LEVEL CHANGES DURING SIX MEGATHRUST EARTHQUAKES OVER THE LAST 3500 YEARS AT THE CASCADIA SUBDUCTION ZONE USING A DIATOM-BASED BAYESIAN TRANSFER FUNCTION (Invited Presentation)

The accurate assessment of Cascadia’s regional earthquake hazards requires well-constrained geophysical models of upper-plate deformation (coseismic) during megathrust earthquakes. Land-level change estimates using quantitative methods applied to microfossils (diatoms, foraminifera) found in stratigraphic records of mud-over-peat or -soil contacts from tidal wetlands offer the most precise means to reconstruct coseismic subsidence along the Cascadia subduction zone (CSZ). However, existing quantitative estimates of coseismic subsidence along the CSZ are temporally limited, with few estimates extending past the great 1700 CE earthquake. Thus, estimates of coseismic subsidence during multiple prehistoric megathrust earthquakes will yield important information on Cascadia’s past rupture behavior.

We expand on our current understanding of earthquake-related, land-level change by applying the first diatom Bayesian transfer function (DBTF) to fossil diatom assemblages from a stratigraphic sequence at the Redtail locality of Willapa Bay, Washington, USA where mud-over-soil contacts have been inferred to record subsidence during six megathrust earthquakes over the past 3500 years. The modern training set consists of 320 diatom species from 104 samples covering the subtidal to forested uplands environments of Willapa Bay. The DBTF overcomes the modern analog problem that has previously hampered the accuracy of diatom reconstructions by using grouped species response curves and integrating lithology prior information about each fossil sample. A 10-fold cross-validation provides an assessment of the predictive performance of the DBTF showing a good relationship (RMSE = 43) between observed and predicted elevations. We applied a modern analog technique (squared-chord distance) with a 20^% dissimilarity threshold derived from the modern training data on the fossil assemblages to confirm that the grouping of species improved the proportion of good analog samples from 51 to 77%. Estimates of coseismic subsidence (with 1 s uncertainty) across the mud-over-soil contacts of buried soils Y (1.41 ± 0.78), U (1.61 ± 0.75), S (1.94 ± 0.69), N (1.87 ± 0.74), L (2.12 ± 0.83), and J (1.45 ± 0.79) vary. Our estimates suggest heterogeneity of coseismic subsidence during past megathrust earthquakes along the CSZ.