Sensitivities of Near‐field Tsunami Forecasts to Megathrust Deformation Predictions

Abstract

This study reveals how modeling configurations of forward and inverse analyses of coseismic deformation data influence the estimations of seismic and tsunami sources. We illuminate how the predictions of near‐field tsunami change when (1) a heterogeneous (HET) distribution of crustal material is introduced to the elastic dislocation model, and (2) the near‐trench rupture is either encouraged or suppressed to invert spontaneous coseismic displacements. Hypothetical scenarios of megathrust earthquakes are studied with synthetic Global Positioning System displacements in Cascadia. Finite‐element models are designed to mimic the subsurface heterogeneity across the curved subduction margin. The HET lithospheric domain modifies the seafloor displacement field and alters tsunami predictions from those of a homogeneous (HOM) crust. Uncertainties persist as the inverse analyses of geodetic data produce nonrealistic slip artifacts over the HOM domain, which propagates into the prediction errors of subsequent tsunami arrival and amplitudes. A stochastic analysis further shows that the uncertainties of seismic tomography models do not degrade the solution accuracy of HET over HOM. Whether the source ruptures near the trench also controls the details of the seafloor disturbance. Deeper subsurface slips induce more seafloor uplift near the coast and cause an earlier arrival of tsunami waves than surface‐slipping events. We suggest using the solutions of zero‐updip‐slip and zero‐updip‐slip‐gradient rupture boundary conditions as end‐members to constrain the tsunami behavior for forecasting purposes. The findings are important for the near‐field tsunami warning that primarily relies on the near‐real‐time geodetic or seismic data for source calibration before megawaves hit the nearest shore upon tsunamigenic events.

Rapid Geodetic Analysis of Subduction Zone Earthquakes Leveraging a 3‐D Elastic Green's Function Library

“…The most time‐consuming DPIA module is tasked with data downsampling (2.9 min), resolving data covariance (2.4 min), extracting GF matrix (3.9 min), and inverse analysis (8.3 min; Figure b) so that majority of time (~50%) is spent on matrix inversion. Though material heterogeneity (e.g., CRUST2.0) and compatible slab geometry (e.g., Slab1.0) are essential for satisfactory model resolution (Figures r and t–v; Hearn & Bürgmann, ; Kyriakopoulos et al, ; Masterlark et al, ; Tung & Masterlark, ), elastic models accounting for these complexities require longer time to be built than the conventional analytical solutions (Okada, ). To tackle this, the GFLED could be computed beforehand to bypass the lengthy FEM‐model‐build process for real‐time seismic analyses, since the information needed to develop such library could be obtained independently before seismic events.…”

“…During the PE, an earthquake early warning system in Central Mexico successfully delivered prompt notifications to the public regarding the arrivals of vigorous strong motions (Jacobson & Stein, 2018). The success of this warning demonstrates the potentials for expanded warning systems to rapidly characterize slip distributions, aftershock, and tsunami hazards (Minson et al, 2014;Newman et al, 2011;Tung & Masterlark, 2018a, 2018cTung, Masterlark, and Dovovan, 2018). In the past, the real-time inversion for finite-fault models has been carried out with high-rate GPS data (Allen & Ziv, 2011;Colombelli et al, 2013;Crowell et al, 2012;Grapenthin et al, 2014;Li et al, 2013;Minson et al, 2014) and seismic wave data (Hayes, 2011;Hayes et al, 2015;Hsieh et al, 2016;Newman et al, 2011;Ross & Ben-Zion, 2016).…”

“…Finite element models (FEMs) are well suited to simulate both of the above tectonic complexities (Hughes et al, ; Kyriakopoulos et al, ; Figure ) and assimilate existing crustal information such as CRUST2.0 that cannot be well accommodated by rectangular fault patches in an analytical half‐space (Okada, ). The advantages of using FEMs over customary half‐space solutions have been exemplified in recent earthquakes and volcanic eruptions (Hughes et al, ; Kyriakopoulos et al, ; Masterlark, ; Masterlark et al, ; Masterlark et al, ; Masterlark & Hughes, ; Trasatti et al, ; Tung & Masterlark, , , ; Williams & Wallace, ), contributing to significantly more accurate source information. Yet, the lengthy processes of building FEMs and the corresponding Green's function (GF) matrix are the bottlenecks of real‐time analyses (c.f.…”

“…The reliability of interpreting earthquake displacements largely depends on how well the elastic models simulate the lithospheric environments near the epicenter (Hearn & Bürgmann, 2005;Masterlark & Hughes, 2008;Moreno et al, 2009;Okada, 1985;Trasatti et al, 2011). Tung and Masterlark (2018c) discover significantly better recovery of coseismic displacements, earthquake sources, and tsunami behaviors when the elastic models account for the distinctive material contrast between the weaker continental crust and stiffer oceanic slab along the Cascadia subduction zone. This solution sensitivity toward material heterogeneity is also shown in other earthquake settings (e.g., Hikurangi margin; Hearn & Bürgmann, 2005;Williams & Wallace, 2015).…”

Finite Element Models of Elastic Earthquake Deformation

“…Matsuyama et al [65] underlines the importance of including non-uniform topography and bathometry in fault deformation model to assess the tsunami hazard and coastal impact upon tsunamigenic events. Subjected to the ongoing tectonic movements and irregular structural settings, seismogenic/tsunamigenic zones usually attain a variable topography or bathymetry, which can be well accommodated by our FEMs for better accuracy of source characterization and tsunami wave predictions ( Figure 1) [3,66].…”