Skip to main content
U.S. flag

An official website of the United States government

22-36. Improving earthquake forecasting with machine learning

We seek applicants to develop a machine-learning model that can be used for earthquake forecasting. This model should provide probabilistic forecasts of the future earthquake rate, earthquake locations, and sizes.

Description of the Research Opportunity

Many features of earthquake sequences are not captured well by current earthquake forecasting models, including 1) large inter-sequence variations in aftershock productivity, 2) correlations between aftershock locations and background seismicity, stress changes, and physical properties of the crust, 3) changes in the size distribution of earthquakes (e.g., b-value), which may inform whether or not an earthquake is a foreshock to a larger event, 4) variability/correlations of Omori parameters describing aftershock decay, and 5) swarms, induced seismicity, and other transients. Capturing patterns within these signals requires new approaches, as does finding ways to use the substantial catalog information below the magnitude of completeness.  The confluence of newly available, modern high-resolution seismic catalogs – which contain orders of magnitude more data than conventional catalogs – and recent advances in supervised and unsupervised machine learning, particularly deep learning techniques, provide a unique opportunity to mine seismic catalogs for deeper levels of earthquake predictability than have previously been found. 

The goal of this opportunity is to develop machine-learning approaches to improve earthquake forecasting capabilities – namely, to better forecast the rate, locations, and sizes of earthquakes, and to supplement or potentially replace the empirical epidemic-type aftershock sequence (ETAS) models currently being used by the USGS for operational earthquake forecasting.  In their proposal, applicants should describe methods and architectures amenable to extracting relevant features from catalogs of earthquake times and locations, and to generating stochastic representations of future activity.  Furthermore, applicants may consider ways to use other types of geophysical data such as fault information and geodetic data in their model.  Applicants should also discuss their plans for model validation and testing, and comparisons to conventional forecasting methods (e.g., ETAS). 

A seismology background is not required for this opportunity. Interested applicants are strongly encouraged to contact the Research Advisor(s) early in the application process to discuss project ideas. 



Beroza, G.C., Segou, M. and Mostafa Mousavi, S., Machine learning and earthquake forecasting—next steps, Nat. Commun. 12, 4761 (2021), doi: 10.1038/s41467-021-24952-6 


Proposed Duty Station(s)



Areas of PhD

Computer science, geophysics, seismology, statistics, 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). 



Applicants must meet one of the following qualifications: Research Geophysicist, Research Computer Scientist, 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.)