20-29. Machine learning/artificial intelligence for seafloor geodesy

CLOSED

----------

The frontier nature of the science and technology of seafloor geodesy makes it ideally suited to the application of new, state-of-the-art, physics-based, Machine learning (ML) and artificial intelligence (AI) algorithms. This Research Opportunity seeks a Fellow to explore applications of novel ML/AI algorithms to enhance the capabilities of seafloor geodesy for detection and characterization of tectonic deformation that occurs beneath the oceans. Seafloor geodesy is particularly important to the USGS’s mission in subduction zones, where the world’s largest faults exist mostly offshore, and that accommodate plate motions via slip not only seismically in earthquakes, but also aseismically in transient slow slip events (SSEs). Evidence for these comes from a variety of seafloor deformation measurements.  Just a few of the many key questions that confront seafloor geodesy and may be addressed by project proposals include:

(1) How can network design be optimized, considering the extreme technical demands and costs of seafloor instrumentation and operations, and the diversity of scientific   goals and signal types sought in today’s offshore deployments?

(2) Denoising, particularly discriminating ocean circulation-generated signals from much smaller tectonic deformation signals, commonly uses measured proxies for the circulation-generated signals, but which proxies may be most effective for this purpose? How can underlying physical, but indirect, nonlinear linkages between different circulation processes be identified to expand the number and efficacy of measurable proxies?

(3) To what extent can models of both ocean circulation (noise) and tectonic deformation be used to address questions in (1) and (2)?

(4) Can ML/AI algorithms reveal causal connections between ocean circulation pressures and tectonic defor mation, noting that an increasing number of studies show that stress changes comparable to those caused by ocean circulation on the seafloor may trigger earthquakes, SSEs and probably other slip phenomena (e.g., submarine landslides)?

Example data well-suited for addressing the above and other questions include those from the Alaska Amphibious Community Seismic Experiment and the Cascadia Initiative. These few-year network deployments included seafloor seismometers, pressure and temperature sensors, and hydrophones. In Cascadia, two cabled networks provide seafloor sensor data from a few isolated locations, but for much longer durations.  Both Cascadia and Alaska also benefit from the availability of other data (e.g., satellite sea surface height and temperature, bathymetric, and seismic imagery data). State-of-the-art regionally-specific model predictions of tectonic deformation and ocean circulation also exist, serving as idealized learning and algorithm testing tools.    Projects may use other datasets, and may focus on other regions and/or simulated data.

Although proposals need not specify particular ML/AI algorithms, the Fellow would be expected to collaborate with Mentor Kutz and his team, who have pioneered many algorithms and workflows that are ideally suited to questions relevant to seafloor geodesy. These include new strategies that overcome limitations of most ML/AI methods (Brunton, et al., 2016), ranging from network optimization, discovery of the underlying governing physics when abundant data are available, modeling of simultaneous processes with multiple time scales, to building of models that evolve in complexity as dictated by the available data and needs (e.g., as in rapid forecasting).

The probable duty station for this opportunity is the USGS field station at the University of Washington (UW) because we expect that the Fellow work most closely with Mentors Gomberg and Kutz, although regular interactions with all the Mentors also are anticipated.  USGS Mentors include earthquake seismologist Gomberg (also an affiliate UW professor), and marine geophysicists Brooks, Ericksen and Watt.  The other Mentors are UW faculty; Kutz is an applied mathematician, and the other Mentors have worked together previously on multiple studies. Hautala and Herrmann are physical oceanographers, Johnson is a marine geophysicist, and Wilcock straddles marine seismology and geodesy.  The Fellow will benefit from interactions with UW’s post-docs, graduate students, and undergraduates in the Department of Earth and Space Sciences and School of Oceanography, as well as in Kutz’s research group. In addition, the postdoctoral fellow will benefit from cohort-building activities that comprise the Mendenhall Fellowship Program and scientists in the USGS Earthquake Science, Geologic Hazards, and Pacific Coastal and Marine Science Centers. The Fellow will be encouraged to participate and present at professional meetings, complete professional journal publications, and ready ML/AI tools developed for dissemination to the scientific and broader communities.

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

References:  

Brunton, S.L., J.L. Proctor, and J.N. Kutz (2016). Discovering governing equations by sparse identification of nonlinear dynamical systems, Proc. Natl. Acad. Sci., 113 (15), 3932-3937, https://doi.org/10.1073/pnas.1517384113

Proposed Duty Station: Seattle, Washington

Areas of PhD:  Geodesy, seismology, physical oceanography, geo-informatix, marine 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 one of the following: Research Computer Scientist, Research Geodesist, Research Geologist, Research Geophysicist, Research Mathematician, Research Oceanographer, Research Physical Scientist

Human Resources Office Contact: Paj Shua Cha, 650-439-2455, pcha@usgs.gov