3.2 Prince William Sound section
The Prince William Sound section (Figure 3) extends ~320 km between Bering Glacier and Seward and encompasses the region of maximum slip in 1964 (Ichinose et al., 2007; Suito & Freymueller, 2009). The plate interface is shallow here (~6°) (Hayes et al., 2018; Worthington et al., 2012) reflecting subduction of the relatively buoyant Yakutat microplate with the Pacific Plate (Eberhart-Phillips et al., 2006) and leading to a wide seismogenic zone. Deep (60-80 km) nonvolcanic tremor located downdip from the 1964 rupture implies show, persistent slip along the interface between the Yakutat microplate and North America (Wech, 2016).
The geologic record of subduction earthquakes along the Prince William Sound section is well-established relative to the rest of the AASZ after many decades of paleoseismic studies (Carver & Plafker, 2008; Hamilton & Shennan, 2005a, 2005b; George Plafker et al., 1992). Marsh stratigraphy and geomorphologic studies show evidence for between seven (Shennan et al., 2014) and nine (Carver & Plafker, 2008) subduction ruptures along the Prince William Sound section since ~4 to ~5 ka. Recurrence intervals between the seven youngest ruptures range from ~420 to 880 years, with the mean recurrence interval between the most recent six earthquakes of 594 -15/+18 years reported by Shennan et al. (2014) (Table 1).
Geodetic observations in the Prince William Sound section document a major slip patch in 1964 and a complex mix of subsequent postseismic and interseismic motions (Cohen & Freymueller, 1997; Li et al., 2016; Suito & Freymueller, 2009; Elliott & Freymueller, 2020). Strong coupling is attributed to shallow dip and geometric complexity at the easternmost end of the AASZ (Christensen & Beck, 1994). We simplify the available geodetic models and assume 100% interseismic coupling extending over 300 km inland from the trench, corresponding to a downdip limit of approximately 30 km depth (Li et al., 2016) (Figure 3). The depth and lateral extent of the segment are defined primarily from Li et al. (2016), which identified the presence of multi-year slow slip events in the Cook Inlet region. The Li et al. (2016) study distinguished between regions that have been persistently locked over the entire time span of geodetic observations, and regions of the interface that have accumulated slip deficit over certain time intervals and then released it in multi-year slow slip events. Other studies, such as Elliott & Freymueller (2020), used a single set of velocities and thus represent an average between the slow slip and non-slow slip intervals. The Elliott & Freymueller (2020) model estimated a more detailed upper plate block model than that assumed in Li et al. (2016), but this model also excluded some data from western Prince William Sound that was hard to fit. Some of the sites in southwest Prince William Sound move at nearly Pacific plate velocity, and all models persistently underestimate these observations unless the slip deficit rate is allowed to exceed the plate convergence rate. For example, Savage et al. (1998) used a plate convergence rate of 65 mm/yr, almost 20% too high, to model the observed velocities of a profile in western Prince William Sound (that study also did not account for postseismic deformation). In part due to the exclusion of some of the data from this profile, the Elliott & Freymueller (2020) model estimates a lower average slip deficit rate within parts of the Prince William Sound section than other studies did, but these variations all lie within the region of the massive Prince William Sound asperity as defined by coseismic slip models (e.g., Ichinose et al., 2007; Suito & Freymueller, 2009). Rather than subdivide the section further, we average the slip deficit over the whole polygon.
There is a discrepancy between the geodetic slip deficit rate and the observed geological recurrence rate for great earthquakes in Prince William Sound, as noted and discussed by Freymueller et al. (2008). The geologic recurrence interval (Table 1) is estimated to be 594±~20 years (Carver & Plafker, 2008; Shennan et al., 2014), but given the observed plate convergence rate and 100% coupling coefficient required to fit the interseismic geodetic velocities, the geodetic estimate for Mw 8.85 to 9.05 earthquakes is only ~200-300 years, depending on assumptions. In short, a fully locked plate interface, which is clearly needed to fit the interseismic geodetic observations, would result in even more frequent great earthquakes than observed; Reducing the coupling coefficient to ~50%, to match the geologic recurrence rate, would produce an enormous misfit to the geodetic data. The Elliott & Freymueller (2020) model includes significant permanent shortening of the upper plate, with the crustal block in Prince William Sound moving rapidly northward, and implying significant permanent contraction within the Chugach Mountains. However, that study excluded some of the data from SW Prince William Sound that were difficult to fit with any model (as noted above, those data were fit in earlier studies by allowing a slip deficit rate that exceeded the rate of plate motion). Some combination of permanent northward motion of the crustal block(s) in Prince William Sound (Elliott & Freymueller, 2020), a reduced incoming plate rate if the subducting crust is Yakutat Block rather than Pacific plate (Freymueller et al., 2008), or perhaps the occurrence of slow slip events to shallower depth than yet observed, or additional slip in M<~8.5 earthquakes that would be invisible in the geologic record would be required to explain the discrepancy. Future work would be necessary to fully explain the apparent mismatch between the geodetic and geologic record along the Prince William Sound section.