• Earthquake source mechanisms and three-dimensional wavefield simulations in Alaska

      Silwal, Vipul; Tape, Carl; Christensen, Douglas; West, Michael; Ruppert, Natalia; Freymueller, Jeffrey (2018-08)
      This thesis presents: (1) a set of earthquake source mechanism catalogs for Alaska and (2) a threedimensional seismic velocity model of Alaska. The improved earthquake sources are used within the velocity model for generating synthetic seismograms, which are then compared with recorded seismograms to assess the quality of the velocity model. An earthquake source mechanism can be modeled as a moment tensor, which is a 3 × 3 symmetric matrix. We estimate the moment tensor for earthquakes by comparing observed waveforms (body waves and surface waves) with synthetic waveforms computed in a layered model. The improved moment tensor solutions are obtained by utilizing both the body waves and surface waves at as many broadband stations as possible. Further improvement in the inversion technique is obtained by (1) implementation of L1 norm in waveform misfit function and (2) inclusion of first-motion polarity misfit in the misfit function. We also demonstrate a probabilistic approach for quantifying the uncertainty in a moment tensor solution. Moment tensors can be used for understanding the tectonics of a region. In the Cook Inlet and Susitna region, west of Anchorage, we determined moment tensor solutions for small-tointermediate magnitude (M ≥ 2.5) crustal earthquakes. Analyzing these small earthquakes required us to modify the misfit function to include first-motion polarity measurements, in addition to waveform differences. The study was complemented with the probabilistic hypocenter estimation of large historical earthquakes (Mw ≥ 5.8) to assess their likelihood of origin as crustal, intraslab, or subduction interface. The predominance of thrust faulting mechanisms for crustal earthquakes indicate a compressive regime within the crust of south-central Alaska. Wavefield simulations are performed in three regions of Alaska: the southern Alaska region of subduction, the eastern Alaska region with the accreting Yakutat microplate, and the interior Alaska region containing predominantly strike-slip faulting, including the Minto Flats fault zone. Our three-dimensional seismic velocity model of Alaska is an interpolated body-wave arrival time model from a previous study, embedded with major sedimentary basins (Cook Inlet, Susitna, Nenana), and with a minimum shear wave velocity threshold of 1000 m/s. Our comparisons between data and synthetics quantify the misfit that arises from different parts of each model. Furtherwork is needed to comprehensively document the regions within each model that give rise to the observed misfit. This would be a step toward performing an iterative adjoint tomographic inversion in Alaska.
    • Modeling the coseismic and postseismic deformation of the 2002 Mw7.9 Denali, AK earthquake

      Harper, Hugh; Freymueller, Jeffrey T.; Christensen, Douglas; Holtkamp, Stephen; Tape, Carl (2017-08)
      The 2002 Mw7.9 Denali fault earthquake was among the largest intraplate earthquakes on record, and the ongoing crustal deformation of the event is still observed today. Understanding the deformation patterns in the years following the earthquake can give insight into the viscoelastic properties of the crust and upper mantle. Additionally, an accurate and predictive model of this deformation is essential to developing and increasingly complete tectonic model of Alaska. Using primarily GPS measurements, deformation can be measured to millimeter-level precision. To develop a coseismic and postseismic model of the earthquake, 224 GPS coseismic displacement measurements (along with SAR and geologic measurements from past studies) are inverted for fault slip distribution. Coseismic slip and consequent stress changes drive the forward postseismic deformation model, which is constrained by 119 postseismic GPS time series. Both models use a 1D elastic structure. The preferred 1D coseismic model fits the coseismic data with a weighted residual sum of squares (WRSS) of 4.86e3 m², with more deep slip than a homogeneous model and a geodetic moment of 8.92e20 N m (Mw 7.97). The Maxwell viscoelastic parameters used for the first postseismic model run are 3e19 Pa s for the lower crust; 5e18 Pa s for the viscoelastic shear zone; and 10e19 and 10e20 south and north of the fault, respectively, for the asthenosphere. The respective Kelvin parameters are all an order of magnitude less. The deep coseismic slip (a product of the 1D elastic model) eliminates the need to add deep slip, which was done in past studies. Based on time series analysis, the decade-plus of data will certainly improve the model prediction relative to previous models, but future observations will be needed to verify this. No preferred postseismic model is developed, and more postseismic models will be run to better fit the observations.