18-21. Machine learning for the detection, identification, and location of volcanic seismicity

Approximately 10% of the Earth’s population lives in close proximity to one of the ~1500 active volcanoes on our planet and roughly half of the Nation's 161 active or potentially active volcanoes are considered dangerous because of the manner in which they erupt and the communities within their reach1. Seismology is the primary tool used for monitoring volcanic unrest because volcanic processes involving magmatic and hydrothermal fluids and their interaction with the solid Earth generate a broad range of observable seismic energy. Typical event types include brittle failure earthquakes in solid rock induced by stress changes due to fluid transfer, pressure oscillations due to the dynamics of liquid and gas motion in conduits and cracks, and magma fracturing and fragmentation. The oscillatory behavior of magmatic and hydrothermal systems produces the bulk of volcanic seismicity and the detection and location of these sources, the quantification of their source mechanisms, and the interpretation of the complex and nonlinear characteristics of this behavior forms the basis of modern volcano seismology2. 

The USGS volcano observatories (Alaska Volcano Observatory, California Volcano Observatory, Cascades Volcano Observatory, Hawaiian Volcano Observatory, and Yellowstone Volcano Observatory) are charged with monitoring the Nation’s volcanic systems. At present, large amounts of seismic data are continuously collected and archived at many U.S. volcanoes but much of this data remains underutilized. Seismic analysts typically focus on brittle-failure earthquakes which constitute a small subset of the seismicity occurring during volcanic unrest. Most oscillatory fluid-induced signals such as volcanic tremor typically exhibit low amplitudes, are emergent in character, and are generally difficult to detect and locate. Thus, seismologists are left to manually detect the prevalence of different event types, often in an ad hoc fashion. With this approach, it is difficult for subtle differences in the relative abundance of different event types to be identified and early recognition of changes within a volcano may go undetected.

Over the past decade, high-performance computing resources have become widely available within the USGS, and machine learning and advanced signal processing algorithms have emerged as valuable tools in seismic monitoring3-5. While much of the existing work on volcanoes using machine learning and deep learning algorithms has focused on a single volcano, often using only a single station, the development of algorithms that can leverage the entire available dataset at any volcano presents a significant and meaningful opportunity to support the monitoring efforts of the USGS volcano observatories.  As the Volcano Hazards Program (VHP) implements the National Volcano Early Warning System (NVEWS) the number of available seismic stations on our Nation’s volcanoes will impose new challenges in the analysis of large volumes of streamed seismic data and machine learning is expected to be the primary tool to accomplish the robust detection, identification, and location of volcanic seismicity.

The USGS volcano observatories collectively seek a postdoctoral fellow to develop algorithms to efficiently detect all types of volcanic seismicity in real time, investigate the best approaches to classify and locate this seismicity, and to implement these algorithms in the routine processing of seismic data at the USGS volcano observatories. The development of these tools, leading an effort to implement and standardize the routine analysis of continuous seismic data streams across the volcano observatories, and to look back in time at archived data streams to enhance our understanding of past volcanic behavior provides an excellent opportunity to develop a valuable research program.

Interested applicants are strongly encouraged to contact the Research Advisor(s) early in the application process to discuss project ideas.

References: 

1Ewert, J. W., Diefenbach, A. K., and Ramsey, D. W. (2018). 2018 Update to the U.S. Geological Survey National Volcanic Threat Assessment. USGS Scientific Investigations Report, 2018-5140. https://pubs.usgs.gov/sir/2018/5140/sir20185140.pdf

2Chouet, B. A., and Matoza, R. S. (2013). A multi-decadal view of seismic methods for detecting precursors of magma movement and eruption. Journal of Volcanology and Geothermal Research, 252, 108-175. doi:  https://doi.org/10.1016/j.jvolgeores.2012.11.013

3Kong, Q., Trugman, D. T., Ross, Z. E., Bianco, M. J., Meade, B. J., & Gerstoft, P. (2018). Machine learning in seismology: Turning data into insights. Seismological Research Letters, 90(1), 3-14. doi: https://doi.org/10.1785/0220180259 

4Bergen, K. J., Johnson, P. A., de Hoop, M. V., and Beroza, G. C. (2019). Machine learning for data-driven discovery in solid Earth geoscience. Science, 363(6433). doi: https://doi.org/10.1126/science.aau0323

5Ross, Z. E., Yue, Y., Meier, M. A., Hauksson, E., & Heaton, T. H. (2019). PhaseLink: A Deep Learning Approach to Seismic Phase Association. Journal of Geophysical Research: Solid Earth, 124(1), 856–869. https://doi.org/10.1029/2018JB016674

Proposed Duty Station: Moffett Field, CA

Areas of PhD: Computer science, seismology, geophysics, applied mathematics, or related fields (candidates holding a Ph.D. in other disciplines, but with extensive knowledge and skills relevant to the Research Opportunity may be considered).

Qualifications: Applicants must meet the qualifications for:  Research Computer Scientist, Research Geophysicist, Research Mathematician

(This type of research is performed by those who have backgrounds for the occupations stated above.  However, other titles may be applicable depending on the applicant's background, education, and research proposal. The final classification of the position will be made by the Human Resources specialist.)

Human Resources Office Contact: Audrey Tsujita, 916-278-9395, atsujita@usgs.gov