3.6 Semidi section
The Semidi section (Figure 4) extends ~250 km from approximately the southwestern edge of the 1964 CE rupture near Sitkinak Island to ~70 km east of the Shumagin Islands (Nishenko & Jacob, 1990). The Semidi section is differentiated from the neighboring Kodiak and Shumagin sections based on historical and paleoseismic earthquake history: this portion of the AASZ hosted great historical ruptures in 1938 (Mw 8.3) and 2021 (Mw 8.2) (Elliott et al., 2022; Freymueller et al., 2021), although these appear to be much smaller than the 1788 rupture from paleoseismic records (Briggs et al., 2014; Nelson et al., 2015). We draw the boundary between the Semidi and the adjacent Shumagin segment based on the presumed western edge of the 2021 Mw 8.2 Chignik earthquake (Elliott et al., 2022), which also corresponded to a segment boundary in the interseismic model of Drooff and Freymueller (2021). von Huene et al. (1999) infer that the Patton-Murray hot spot swell may influence rupture character of the Semidi section, which exhibits much higher interseismic coupling than the Shumagin section to the southwest; alternatively, von Huene et al. (2012) also argue that the subduction of the head of the Zodiac fan may influence rupture behavior in this section. The Semidi section was recognized as a potential source for a Pacific basin-wide tsunami with risk implications for the western coast of the United States (Ross et al., 2013).
The geologic record of subduction earthquakes for the Semidi section is derived from geologic studies on Chirikof and Sitkinak Islands. On Chirikof Island, stratigraphic evidence of 13 paleotsunamis since ~3.5 ka, including the major historical1788 rupture, corresponds to a paleotsunami recurrence of 180-270 years (Nelson et al., 2015). Sitkinak is at the westernmost edge of the neighboring Kodiak section, and so is not strictly in the Semidi section; however, marshes at Sitkinak record a series of land level changes that we infer record ruptures of the Semidi section. Five land-level changes at Sitkinak (Briggs et al., 2014), between ~1050 BP and 1788 CE indicate a recurrence interval of ~222 years, in agreement with the paleotsunami record from Chirikof (Table 1).
Geodetic observations in the Semidi section consistently show a highly coupled region with a lower coupling toward the Shumagin islands to the west (Drooff & Freymueller, 2021; Li & Freymueller, 2018). Li & Freymueller (2018) estimated strong coupling in their ‘Semidi segment’ (~70%), with much lower coupling to the west in their ‘Shumagin segment’ (~40%). However, few data were used to constrain the location of the boundary, so the location was quite uncertain. Drooff and Freymueller (2021) revised the segment boundaries of Li and Freymueller (2018), incorporating additional data from Veniaminof volcano on the Alaska Peninsula, which had been excluded in the earlier study due to the presence of volcanic deformation. Drooff and Freymueller (2021) shifted the western boundary of the Semidi section to the east, leaving strong coupling in their segment 2, and broke the Shumagin region into two segments (their segments 3 and 4). Our location of the Semidi-Shumagin boundary corresponds to the boundary between the Drooff and Freymueller (2021) segments 2 and 3, which also corresponds closely to the southwestern edge of the 2021 rupture. For hazard estimates, we represent 70% coupling ~125 km from the deformation front, corresponding to a locking depth on the Slab2 interface of ~20 km (Figure 4). The slip deficit at shallow depth near the trench is highly uncertain due to poor model resolution, and depends strongly on the assumed model regularization (Xiao et al., 2021), but the total integrated moment accumulation rate does not vary much even where the appearance of the slip deficit distribution with depth varies a lot. Because we lack concrete information about whether the shallow part of the interface is locked or creeping, we adopt the estimates based on models that assume locking to the trench.