Project Research Application: Geologic Hydrogen

The algorithmic workhorse for creating a map of the subsurface from seismic surveys is full waveform inversion. Many fundamental questions remain wide open about it, including how to deal with the lack of convexity in optimization, how to make use of AI to speed it up or supersede it, and how to quantify the uncertainty inherent in its predictions. Among the various projects that the Imaging and Computing group has led over the years, one current topic of great interest is the convergence of AI and uncertainty quantification. We provide the first method for sampling the “epistemic” posterior of a neural network that is consistent with the Bayesian setting. The upshot is for the practitioner to have correct, not arbitrary, error bars on the seismic image.

Sponsored by: TotalEnergies

ERL Personnel: Laurent Demanet

While the potential for subsurface fluid injection and extraction to trigger earthquakes has long been recognized, the sharp increase in the extent and vigor of injection-induced seismicity calls for much deeper understanding than is currently available. One of the features of some highly destructive earthquakes is their supershear rupture propagation, with velocities faster than the shear wave speed that typically lead to large magnitude events. The intensity and the patterns of strong ground motion for supershear earthquakes have been shown to be inherently different from those of regular (sub-Rayleigh) ones, calling for a need to elucidate the controlling factors behind rupture speed to understand. Fluids fill the pore space of crustal rocks, and pore pressures are essential to understand the stability of geologic fault. However, poro-mechanical effects are often neglected in the analysis and interpretation of earthquake rupture speeds. Here we address the overarching question: what is the role of pore fluids in the generation of supershear earthquakes? We break this down into two fundamental questions: (1) How does pressurization rate from injection impact rupture speeds? (2) Can heterogeneity in frictional properties elicit a transition between rupture regimes (e.g., from slow-slip, to regular earthquake, to supershear earthquake)? We address these fundamental questions through a combination of high-resolution computational modeling and theoretical stability analyses that probe the regime transitions of the coupled fluid/porous solid system.  We anticipate that our results will shed new light in our understanding of supershear earthquakes, paving the way for improved forecasting of earthquake hazards.

Sponsored by: MIT-Spain UPM Seed Fund

ERL Personnel: Ruben Juanes (PI)

Collaborators: Luis Cueto-Felgueroso (co-PI, UPM Spain)