Excellence in Research: A Data-Driven Computational Framework for Seismic Detection, Modeling and Prediction

This project is co-funded by the Geophysics (PH) Program and the Historically Black Colleges and Universities - Excellence in Research (HBCU-EiR) Program, along with support from Integrative and Collaborative Education and Research (ICER) funds of the NSF Geosciences Directorate.

Embarking on innovative scientific approaches that fully exploit fast-growing datasets across different disciplines is necessary to achieve great scientific discovery. Such necessity stimulates urgent research initiatives across various scientific disciplines. Seismology, being a data-driven science with huge datasets recorded for more than a century, will definitely benefit from the developments of new scalable algorithms that can process such massive data volumes.

Having tremendous potential, geophysics and particularly seismology innovation using machine learning and big-data analytics based on multiple seismic datasets has so far been trailing behind. Broader impacts of this project include: (1) launching a new interdisciplinary research in the areas of machine learning, big-data analytics, computational techniques and geophysics in which undergraduate STEM students and graduate students of the research project will be cross-trained to transcend traditional disciplinary boundaries, (2) creation and distribution of big-data analytics machine learning techniques useful for detecting underground nuclear explosions, modeling crustal structure and predicting the spatial distribution of aftershocks following major earthquakes, (3) delivering analytical and computational techniques that have much to offer to the field of seismology and solid-Earth geophysics at large, and (4) imparting research-enriched learning experiences to STEM undergraduate and graduate students through educational activities at Morgan State University and summer research internships at national laboratories.

This project focuses on developing a data-driven computational framework for seismic detection, modeling and prediction. Having the training and expertise in geophysics/seismology, machine learning and big-data analytics to explore computational techniques, the investigators of this project will address the challenges in the development of underground nuclear explosion detection methods, predicting the spatial distribution of aftershocks following major earthquakes, and modeling crustal structure.

The specific objectives of this research are: (i) to investigate the development of automatic nuclear explosion detection methods utilizing approximate nearest neighbor methods to search large archives along with the integration of template matching and iterative seismic processing framework, (ii) to explore machine learning methods to predict the likelihood that aftershocks would occur in a particular location on a spatial grid based on modeling transfer of elastic energy between regional stress fields and a set of localized faults, and (iii) to examine data analytics algorithms for modeling crustal structure using multiple complex seismic datasets that provide research communities with open source big-data analytics tools.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.