Thread: Machine Learning for Global Earthquake Monitoring - USGS Mendenhall Postdoc Research Opportunity

Started: 2019-10-07 15:36:24
Last activity: 2019-10-07 15:36:24
Institution: USGS NEIC
Open Until: 2020-01-06

Mendenhall Research Fellowship Program
Opportunity 18-10: Leveraging machine learning for next-generation global earthquake monitoring

The U.S. Geological Survey (USGS) National Earthquake Information Center (NEIC) detects, locates, and characterizes tens of thousands of earthquakes globally each year with a large emphasis placed on both the timeliness and accuracy of event characterization. The rapid analysis of fundamental earthquake properties, including location, depth, magnitude, and other source characteristics, is critical to the NEIC’s rapid estimation of the impact and seismotectonic context of an earthquake. Improving the speed and accuracy of earthquake detection, association, location, and magnitude estimation algorithms is fundamental to improving the NEIC’s capabilities.

Recent research has demonstrated that machine-learning-based event-processing methodologies generalize well, have unprecedented accuracy, are fast, and have the potential to revolutionize seismic monitoring systems. The application of machine learning to NEIC’s operational systems has enormous potential to improve global earthquake monitoring and advance seismological research. Recent machine learning research has focused largely on local and regional datasets, which generally have a smaller variety of source types and higher quality signals as compared to the global case. Therefore, it is necessary to further develop these techniques with a specific focus on global monitoring needs. While the first wave of deep-learning algorithms has focused on methods that improve detection capabilities, there is a wide range of other machine learning applications that could enhance the rapid characterization of an earthquake’s location, magnitude, depth, mechanism, and seismic source type (e.g., explosion vs. earthquake).

The focus of this Mendenhall Research Opportunity is to develop a scientific framework leveraging machine learning that will ultimately provide the rapid and more complete characterization of earthquake properties. This work can be targeted at improving the NEIC’s ability to detect and associate seismic signals, or at estimating other seismic characteristics (e.g., location, magnitude, mechanism, and source-type). While the monitoring component of this project has direct implications for the rapid characterization and understanding of large high-impact earthquakes and the improved detection of smaller events, the research should also improve our understanding of earthquake sources and seismotectonics.

Candidates are encouraged to explore machine-learning-based methods to aid in global earthquake seismology and to consider how their research could be implemented into NEIC real-time operations and response activities. With the development of these technologies, research into the potential pitfalls of these algorithms, and rigorous comparison with standard monitoring techniques will be required.

In addition to the core development of machine-learning-based algorithms, research should apply these technologies to aid our understanding of earthquake processes. Research may include studies aimed at improving the characterization of specific earthquake sequences or event types and subsequent analysis of the scientific implications of these results. Proposals that aim to improve our understanding of earthquake sequence seismotectonics, induced seismicity, subduction zone processes, or other areas of large seismic hazard are also invited.

We invite proposals from candidates with experience in geophysics, computational seismology, machine learning and/or observational seismology and seismic monitoring.

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

Further details and how to apply can be found at:

Other Mendenhall Opportunities can be found at:
02:36:05 v.22510d55