Probabilistic, high-resolution tsunami predictions in northern Cascadia by exploiting sequential design for efficient emulation

The potential of a full-margin rupture along the Cascadia subduction zone poses a significant threat over a populous region of North America. Previous probabilistic tsunami hazard assessment studies produced hazard curves based on simulated predictions of tsunami waves, either at low resolution or at high resolution for a local area or under limited ranges of scenarios or at a high computational cost to generate hundreds of scenarios at high resolution. We use the graphics processing unit (GPU)-accelerated tsunami simulator VOLNA-OP2 with a detailed representation of topographic and bathymetric features. We replace the simulator by a Gaussian process emulator at each output location to overcome the large computational burden. The emulators are statistical approximations of the simulator's behaviour. We train the emulators on a set of input–output pairs and use them to generate approximate output values over a six-dimensional scenario parameter space, e.g. uplift/subsidence ratio and maximum uplift, that represent the seabed deformation. We implement an advanced sequential design algorithm for the optimal selection of only 60 simulations. The low cost of emulation provides for additional flexibility in the shape of the deformation, which we illustrate here considering two families – buried rupture and splay-faulting – of 2000 potential scenarios. This approach allows for the first emulation-accelerated computation of probabilistic tsunami hazard in the region of the city of Victoria, British Columbia.

An Interdisciplinary Agent-based Evacuation Model: Integrating Natural Environment, Built environment, and Social System for Community Preparedness and Resilience

Previous tsunami evacuation simulations have mostly been based on arbitrary assumptions or inputs adapted from non-emergency situations, but a few studies have used empirical behavior data. This study bridges this gap by integrating empirical decision data from local evacuation expectations surveys and evacuation drills into an agent-based model of evacuation behavior for a Cascadia Subduction Zone community. The model also considers the impacts of liquefaction and landslides from the earthquake on tsunami evacuation. Furthermore, we integrate the slope-speed component from Least-cost-distance to build the simulation model that better represents the complex nature of evacuations. The simulation results indicate that milling time and evacuation participation rate have significant non-linear impacts on tsunami mortality estimates. When people walk faster than 1 m/s, evacuation by foot is more effective because it avoids traffic congestion when driving. We also find that evacuation results are more sensitive to walking speed, milling time, evacuation participation, and choosing the closest safe location than to other behavioral variables. Minimum tsunami mortality results from maximizing the evacuation participation rate, minimizing milling time, and choosing the closest safe destination outside of the inundation zone. This study's comparison of the agent-based model and BtW model finds consistency between the two models' results. By integrating the natural system, built environment, and social system, this interdisciplinary model incorporates substantial aspects of the real world into the multi-hazard agent-based platform. This model provides a unique opportunity for local authorities to prioritize their resources for hazard education, community disaster preparedness, and resilience plans.

Deep Coseismic Slip in the Cascadia Megathrust can be Consistent with Coastal Subsidence

At subduction zones, the down-dip limit of slip represents how deep an earthquake can rupture. For hazards it is important - it controls the intensity of shaking and the pattern of coseismic uplift and subsidence. In the Cascadia Subduction Zone, because no large magnitude events have been observed in instrumental times, the limit is inferred from geological estimates of coastal subsidence during previous earthquakes; it is typically assumed to coincide approximately with the coastline. This is at odds with geodetic coupling models, it leaves residual slip deficits unaccommodated on a large swath of the megathrust. Here we will show that ruptures can penetrate deeper into the megathrust and still produce coastal subsidence provided slip decreases with depth. We will discuss the impacts of this on expected shaking intensities

Relationship between the high-amplitude magnetic anomalies and serpentinized fore-arc mantle in the Cascadia subduction zone

A zone of significant high-amplitude magnetic anomalies is observed without a comparable gravity high along the Cascadia margin and is spatially correlated with the low-velocity fore-arc mantle wedge, which is understood to be serpentinized fore-arc mantle and is further considered to be the main source of the high-amplitude magnetic anomalies. To test this hypothesis, the magnetization-density ratio (MDR) is estimated along the Cascadia margin to highlight the physical characteristics of serpentinization (reduced density and increased magnetization). Interestingly, high MDR values are found only in central Oregon, where slab dehydration and fore-arc mantle serpentinization (50%-60% serpentinization) are inferred in conjunction with sparse seismicity. This result may indicate either poorly serpentinized fore-arc mantle (low degree of serpentinization) or that the fore-arc mantle is deeper than the Curie temperature isotherm for magnetite in northern and southern Cascadia. I thus propose that serpentinized fore-arc mantle may not be the major contributor to the high-amplitude magnetic anomalies in these segments. This finding means that magnetic anomaly highs and serpentinized fore-arc mantle may not be completely positively related in subduction zones. On the other hand, the MDR pattern reveals the segmentation of the Cascadia subduction zone, which is consistent with several previous observations.

Imaging the next Cascadia earthquake: Optimal design for a seafloor GNSS-A network

Summary The Cascadia subduction zone in the Pacific Northwest of the United States of America is capable of producing magnitude ∼9 earthquakes, likely often accompanied by tsunamis. An outstanding question in this region, as in most subduction zones, is the degree and spatial extent of strain accumulation, which will eventually release as an earthquake, on the subduction megathrust. Geodetic observations, including those from Global Navigation Satellite Systems (GNSS), may be used to image the strain actively accumulating on a fault before an earthquake ultimately occurs. Technology combining GNSS and underwater acoustic ranging (GNSS-A) is now capable of making centimeter-level horizontal geodetic observations on the seafloor. GNSS-A enables previously inaccessible observations to better image seismogenic portions of the Cascadia subduction zone. Because seafloor geodetic instruments, and the time and logistics associated with observations, can be cost-prohibitive, it is important to identify where deploying seafloor geodetic instruments will provide information that cannot be obtained through a similar investment in onshore geodetic networks. Here we leverage the concept of information entropy to 1) quantify the relative information provided by expanding GNSS observation networks offshore Oregon and Washington and 2) identify optimal locations for a network of seafloor geodetic instruments. The information gained by new observations, and their optimal locations, depends on the expected uncertainties on the seafloor velocity observations, modeling assumptions, and the modeling objectives.

Swipe Left on the “Big One”: Better Dates for Cascadia Quakes

Improving our understanding of hazards posed by future large earthquakes on the Cascadia Subduction Zone requires advancements in the methods and sampling used to date and characterize past events.