Estimation of Rupture Scenarios along the Cascadia Megathrust from Interseismic Locking Models

In the West of Northern America, the Cascadia subduction zone that extends over one thousand kilometers has well-documented geological records of megathrust earthquakes. The most recent one occurred in 1700 AD with a moment magnitude of 9. Hence, it has been more than 300 years since the last earthquake, suggesting that Southern Cascadia is mature for the next large earthquake. Estimating future rupture scenarios is therefore crucial for earthquake hazard assessment in the region. Multiple interseismic locking distributions have been proposed for Cascadia. Since each locking model differs from another, it remains unclear how to estimate future rupture extents from interseismic locking distributions. Here, we use 3-D dynamic rupture simulations to investigate the potential rupture segmentation in Cascadia and test the dependency of rupture propagation on hypocenter, especially for the Southern Cascadia. We process the slip deficit distributions from locking models by interpolation and smoothening with a gaussian filter. We then calculate the corresponding stress changes with the assumption that all slip deficits would be released during a coseismic event and derive different initial stress distributions by prescribing constant dynamic stress. For the northern segment, the stress-shadowing (Lindsey et al. 2021) and the viscoelastic (Li et al. 2018) interseismic locking models based respectively on elastic and visco-elastic deformation have similar stress levels, lower than those derived from the Gamma model (Schmalzle et al. 2014). In addition, the Gamma model displays a distinct low-stress gap in the central segment but the stress-shadowing and viscoelastic models show smooth transition stress changes. Since the stress-shadowing and the viscoelastic locking models bear a resemblance, dynamic simulations are then developed based on the initial stress conditions derived from the viscoelastic and the Gamma models by prescribing artificial nucleation zones on the fault plane with varied hypocentre locations. Preliminary results demonstrate three major rupture scenario types - self-arrested, segmented, and full-margin ruptures for both stress models. Given the same conditions, both models indicate that Southern Cascadia with a shorter recurrence interval has a lower potential of growing into a margin-wide rupture compared to the central segment. The southern segment mainly hosts self-arrested and segmented ruptures with Mw ranging from ~6.7 to >7.3. Another finding is strong along-strike variations in stress distribution flavor segmented ruptures while homogeneous stress field promotes margin-wide ruptures. For ruptures initiating from the central segment, several segmented ruptures with Mw 8.14 to larger than 8.25 are observed from the Gamma model but such features are absent in the viscoelastic model. Apart from segmented ruptures, the margin-wide ruptures have amplitudes of ground surface vertical displacement comparable to the subsidence record in the A.D. 1700 megathrust earthquake, particularly for the along-strike variation in the Gamma model.