Wednesday, October 16, 2019, 6:30 PM - 10:00 PM
Two Fountain Plaza, Buffalo, NY 14202
Biography: Margarete Jadamec is an assistant professor in geodynamics, with a joint appointment with the Department of Geology and the Computational and Data-Enabled Science and Engineering Program at the University at Buffalo, SUNY (UB). Dr. Jadamec leads the Geodynamics Research and Visualization Group at UB where her team integrates high-fidelity data-driven model design, high performance computing, and 3D virtual reality to solve complex problems in Earth Science. Dr. Jadamec received her PhD from the University of California, Davis (2009). She was Postdoctoral Fellow jointly in the School of Mathematical Sciences and the School of Geosciences at Monash University in Australia, where she co-founded the 3DALIVE visualization facility, co-sponsored with the Monash e-Research Centre and CSIRO Australia. She was awarded a National Science Foundation Postdoctoral Fellowship at Brown University (2011-2013) and received both Best Science Paper and Best Conference Paper at XSEDE12. Her innovative research has pushed the technological envelope in geodynamics, building geographically referenced simulations of the Earth's natural plate tectonic boundaries at unprecedented scales.
Abstract: The influx in digital information, improved data coverages, and access to large-scale computational resources and infrastructures has enabled a next generation of simulations in solid Earth science, in terms of the overall scope and level of detail that can be represented in a given simulation or set of simulations. At the same time, this poses challenges in terms of accurately representing the information within the model domain and formulating identifiable features with physical properties capable of driving the system dynamics, whilst maintaining solvable solutions and feasible simulation times. Moreover, as new levels of complexity are incorporated into the model design and problem size increases, it becomes harder to parse and conceptualize simulation input as well as output, requiring new analysis tools to accompany the new scale in computation. To this end, I will present results from high-fidelity simulations of plate tectonic deformation on Earth and show how the combination of data-driven simulation design, optimization of solvers for complex systems with large viscosity variations, and 3D virtual reality for simulation exploration, has led to a new kind of modeling in solid Earth science. In particular, the combined ability to represent complex features, simulate non-linear systems, and literally walk inside a virtual finite element and query neighboring flow fields, led to the discovery that the Earth's upper mantle, located just 100 km beneath the Earth's surface, can flow an order of magnitude faster than what was expected from surface plate motions. Furthermore, the inclusion of more detailed structure into model design and framing the problem three-dimensionally led to the first models to explain the enigmatic location of the tallest mountain in North America, Denali. Looking forward, new approaches to visualizing and rendering complex information will be critical to communicating scientific concepts and ensuring reproducibility.