19-24. Rapid, robust characterization of earthquake uncertainties for advanced monitoring and research applications

The rapid, accurate, and fully automated estimation of earthquake source parameters (particularly hypocentral locations and magnitudes) is the holy grail of earthquake monitoring science.  While tools to estimate these source parameters are long standing, their associated uncertainties are not standardized and are usually derived only from misfit within a particular assumed model (also known as “model errors”). This characterization completely ignores other significant sources of uncertainty and bias (such as from errors in the velocity model) and often dramatically underestimates the total uncertainty.  This shortcoming significantly impairs downstream research activities ranging from tectonic interpretations to seismic hazard assessments, and it can have a significant impact on the interpretation of these source estimates by the public.  Perhaps less obviously, improved uncertainty analysis will also be required to facilitate next-generation monitoring itself, as part of the decision tree governing the release of fully automated earthquake solutions. 

The current paradigm of earthquake monitoring relies heavily on expert human judgement, both for identification and timing of seismic phases, and just as critically, for deciding when solution quality is sufficiently high to warrant public release of an event.  Although this approach has been successful within its scope, it does not readily scale to rapidly monitor earthquakes to lower magnitudes recorded with increasingly dense instrumentation, thus necessitating a new approach.  Recent developments in machine learning are poised to revolutionize aspects of seismic processing, such as seismic phase picking, but a key piece of an automated system remains underdeveloped: how do we determine overall solution quality?   

Methods for rapid release of automatic earthquake locations and magnitudes have been implemented on local scales within dense networks. Examples include automatic solutions published within minutes by several ANSS regional seismic networks and alerts distributed within seconds from the ShakeAlert Earthquake Early Warning (EEW) system.  At regional and global scales uncertainty analysis becomes more difficult because of lower signal-to-noise ratios and sparse station coverage. Simple metrics such as numbers of phases recorded and associated azimuthal gaps provide rudimentary checks sufficient for dense local networks, but they are inadequate in more complex cases encountered in regional and global earthquake monitoring.   

Systematic, intelligent, and accurate characterization of location and magnitude errors will provide a framework for release of automatic solutions, a benchmark for solution quality, and the means to interpret seismicity patterns using robust statistical analysis. The successful approach will likely utilize probabilistic methods from Bayesian inference and information theory. Bayesian methods allow both the parameter value estimation uncertainty and model design uncertainty to be rigorously treated. Concepts from information theory allow us to quantify how much information is gained or lost through different analysis methods. 

The focus of this Mendenhall Research Opportunity is to provide a new framework for estimating uncertainties in earthquake source parameters.  Addressing this problem requires innovative research into new approaches to characterizing location and magnitude uncertainties.  

We invite proposals from candidates with research experience in observational seismology, computational seismology, geophysics, and/or computer science.  Research should be targeted to approaches that can eventually be integrated into processing workflows at the National Earthquake Information Center (NEIC) and other monitoring networks.  This requires that the analysis can be performed rapidly, so that it can be implemented in real-time earthquake processing to facilitate release of high-quality fully automated earthquake solutions.  The approach should also be robust, such that it can accurately characterize uncertainties across scales (local to global) with widely varying network coverage.  The research will start with a review of existing techniques for source parameter uncertainty estimation used by the Advance National Seismic System (ANSS) and by EEW, then will investigate potential unifying approaches possibly incorporating Bayesian, empirical, or bootstrap techniques. Novel combinations of these approaches should be considered, perhaps through the use of machine-learning tools leveraging large datasets.  

Testing the accuracy of these tools by comparing the performance to standard methodologies is essential to ensuring their applicability to real-time operations. Candidates will be encouraged to propose not only methods for how to improve real-time automated source-parameter estimation but also to quantify the performance of different methods including current operational methods in order to produce rigorous criteria for accepting the use of new automated methods. 

In addition to developing tools to facilitate real-time monitoring operations, we encourage the application of these tools to advance our understanding of seismotectonics and the earthquake process. Improved uncertainty characterization may have fundamental implications for interpretations of seismic catalogs. As just one example with important scientific consequences, the newly developed tools could be used to assess whether or not earthquake locations near a given fault can be explained by activity on a single fault (that is, their scatter can be explained fully by location uncertainty), or if the scatter instead implies activation of multiple faults.  Improved magnitude uncertainty estimates could be examined for implications for seismic hazard assessment. 

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

Proposed Duty Station: Golden, CO 

Areas of PhD: Geophysics, seismology, computer science, 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 Civil Engineer, Research Computer Engineer, Research Computer Scientist, Research Engineer (General), Research Environmental Engineer, Research Geodesist, Research Geologist, Research Geophysicist, Research Mathematician, Research Mechanical Engineer, Research Physical Scientist, Research Physicist, Research Statistician.

(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: Joseline Martinez Lopez, 303-236-9559, jmartinezlopez@usgs.gov 

Apply Here