Research

From the ocean floor off Puerto Rico to a billion-year-old fossil failed rift under Illinois-Indiana and pressure buildup beneath Delaware Basin, I use earthquakes and noise to read what the crust is trying to tell us.

My research integrates observational seismology, computational geophysics, and machine learning across three interconnected themes. The tools drive the science: improved velocity models and relocation schemes lead to more accurate earthquake locations; machine learning enables the detection of many more small earthquakes, and these enhanced, more complete catalogs, combined with source-mechanism analyses, reveal fault geometry and stress localization. Integrating source characterization with InSAR-derived surface deformation reveals how strain is partitioned between seismic and aseismic slip among interconnected fault networks.

Current active fieldwork includes a dense nodal deployment in southeast New Mexico (Summer 2026) targeting the crustal velocity structure beneath the most seismically active region in the state, and ongoing ensemble catalog development for the New Mexico using the Pick Aggregator.

Active Research Threads
Machine Learning · Seismic Phase Picking · High-Resolution Earthquake Catalogs

Making ML pickers work where they weren't trained

Most ML pretrained models struggle when applied to data distribution that is significantly different from what they are trained on. The Pick Aggregator runs different picker-model combinations in parallel and accepts a pick only when at least two independent combinations agree, producing catalogs that generalize across instrumentation and data types and reduce false positives. Application to Puerto Rico reveals a previously unresolved slab tear northeast of the island.

Pick Aggregator SeisBench . PhaseNet . EQTransformer Puerto Rico Subduction Zone OBS + Land Data In review manuscript, 2026
Read more →
Surface Wave Dispersion . Receiver Functions. High-Resolution Crustal Velocity Structure

Imaging the crust with its own background noise

Cross-correlating ambient seismic noise between station pairs reconstructs surface wave Green's functions, which encode the velocity structure of the crust along each path. Joint inversion with teleseismic receiver functions resolves both layer velocities and interface contrasts. Applied to the Wabash Valley, this approach revealed a billion-year-old rift pillow concentrating present-day intraplate seismicity. The same methodology now drives a 200-sensor nodal deployment in southeast New Mexico.

Ambient Noise Tomography . Receiver Functions · Joint Inversion Computer Programs in Seismology (Herrmann 2013) Wabash Valley Crustal Shear Wave Structure . Published 2019 Nodal Array Field Deployment 2026 . SE New Mexico
Read more →
Seismic Source Characterization · InSAR · Spatiotemporal Evolution of Induced Seismicity

Mapping a decade of fault reactivation in the Delaware Basin

Over 5,400 relocated earthquakes in southern Delaware Basin and satellite InSAR deformation spanning 2009–2022 tell a story of saltwater disposal driving pore pressure southeast along permeable faults, reactivating shallow normal faults at 1.5 km depth, and producing concurrent seismic and aseismic slip that the ground surface records as centimeter-scale linear subsidence.

Hypocentroidal Decomposition Relocation . Southern Delaware Basin Seismicity . InSAR . Injection . Production MLOC package (Bergman et al. 2023) Published 2024, 2025
Read more →
Stochastic Ground Motion Modeling · Seismic Attenuation · Seismic Hazard

Mapping spatial variations in seismic attenuation using weak-motion records

Over 10,000 weak motion records from the Mw 8.8 Maule and Mw 8.2 Iquique aftershock sequences reveal that seismic energy decays faster through the warm, fluid-saturated crust of central Chile than through the cold, dry crust of the north. Stochastic source–attenuation models constrain frequency-dependent Q(f) across three sub-regions, providing physical validation of non-ergodic ground motion models and a path toward better seismic hazard assessment along the Chilean trench.

Stochastic Method Chilean Subduction Zone: Iquique · Valparaíso · Maule SIMulation Package (Boore 2003) In review manuscript, 2026
Read more →