Detection thresholds for large to great paleoseismic subduction earthquakes in south-central Alaska from coastal marsh records.

Abstract

Coastal paleoseismic evidence currently provide records of the recurrence of great earthquakes in the Prince William Sound section, Alaska, with widespread evidence of vertical surface displacement ≥ 1 m during seven great earthquakes in the past 4000 years. The established criteria for detecting land level changes ≥ 1 m relies heavily on identifying and tracing abrupt changes in sediment stratigraphy, peat-silt couplets which indicate ≥ Mw 8 earthquake produced subsidence. While methods are well developed for identifying displacement of ≥ 1 m, very little is known about the lower detection limit of vertical land displacement. Briggs and Barnhart (2017) suggest that subduction earthquakes < Mw 8 displace the coast by < 0.3 m and likely leave no geologic evidence in coastal marshes. However, recent studies suggest that a detection limit of ~0.1 to 0.2 m is possible within vegetated marsh environments.
Consequently, this thesis aims to identify the lower detection limits of co-seismic vertical deformation in coastal marshes of Prince William Sound to reveal large (Mw 7-7.9) to great (≥ Mw 8) earthquakes, which may have occurred between the 1964 CE and the preceding earthquake (~771 cal yr BP). The sampling design of previous studies looking at ≥ 1 m subsidence focuses on marsh-front locations, which may not detect smaller earthquakes. Thus, fieldwork was undertaken in Alaska to test a new methodology which focuses on high marsh environments; peat forming communities where the most precise quantitative estimates of deformation can be reconstructed.
The stratigraphic contact of the 771 ± 10 cal yr BP earthquake was traced inland where it is expressed as a peat-to-peat contact. AMS dating, shifts in diatom assemblages, and a pronounced fall in PMSE at the contact provide evidence of the ~771 cal yr BP earthquake. However, I argue that the magnitude of this subsidence is dependent on training set selection. Furthermore, I suggest that using a regional peat only training set provides the lowest detection limit (< 0.1 m) for the ~771 cal yr BP earthquake whilst still providing a suitable number of analogues. No additional earthquakes were identified in the peat at Girdwood, raising questions about the suitability of diatom transfer functions for detecting smaller seismic events in high marsh environments.