The Automation of Numerical Models of Coseismic Tsunamis

• Main Author: Wiersma, Codi Allen
• Other Authors: Geosciences
• Format: Others
• Published: Virginia Tech 2019
• Subjects: Numerical Modeling; Tsunami; GeoClaw; Finite Volume
• Online Access: http://hdl.handle.net/10919/93273

The use of tsunami models for applications of 'now-casting', which is the prediction of the present and near future behavior, has limited exploration, and could potentially be of significant usefulness. Tsunamis are most often caused by earthquakes in subduction zones, which generates coupled uplift and subsidence, and displaces the water column. The behavior of the fault failure is difficult to describe in the short term, often requiring seismic waveform inversion, which takes a length of time on the order of weeks to months to properly model, and is much too late for any use in a now-casting sense. To expedite this length of time, a series of source models are created with variable fault geometry behaviors, using fault parameters from Northern Oceanic and Atmospheric Administration's Short-term Inundation and Forecasting of Tsunamis (SIFT) database, in order to model a series of potential tsunami behaviors using the numerical modelling package, GeoClaw. The implementation of modeling could identify areas of interest for further study that are sensitive to fault failure geometry. Initial results show that by varying the geometry of sub-faults of a given earthquake, the resulting tsunami models behave fairly differently with different wave dispersion behavior, both in pattern and magnitude. While there are shortcomings of the potential geometries the code created in this study, and there are significant improvements that can be made, this study provides a good starting point into now-casting of tsunami models, with future iterations likely involving statistical probability in the fault failure geometries. === Master of Science === Short term modeling of tsunamis generated by earthquakes is poorly explored. If an earthquake causes movement in a fault located underwater, and this movement will then cause the water column above it to be displaced. Tsunami models are sensitive to how the fault moves, and an accurate representation of this movement often takes much more time that the duration of a tsunami. This lengthy process is ineffective for short term modeling. This study instead estimates several possible scenarios of how the fault will behave, and model each of them. This will show how different locations of interest are sensitive to different geometries of fault failure. Initial results show that by varying this geometry, the tsunami wave behaves very differently, and will cause different amounts of run-up in the same location depending on which particular geometry is modeled. The automation of distinctly different earthquake sources serve as a good starting point for future work to be conducted to generate more accurate models.