This project aims to reprocess and analyze fiber-optic Distributed Acoustic Sensing (DAS) seismic recordings to reconstruct undersampled datasets with high fidelity. The primary focus is to apply novel anisotropic regularization techniques that minimize the loss of critical amplitude and travel-time information. These efforts are directed toward formulating Fresnel-zone and compressive-sampling criteria for anisotropic media, enabling the design of rapid, cost-effective acquisition surveys that remain highly sensitive to CO₂ plume anomalies.

The project integrates theoretical advancements in linearized Born and Kirchhoff inversion to optimize imaging in complex media. Specifically, inversion kernels will be developed and benchmarked to explicitly account for the directional wave speed variations observed in DAS data, ensuring accurate subsurface characterization.

Machine learning will be explored as a complementary tool to enhance data quality, building on collaborative research with experienced peers in the group. These efforts aim to train models capable of suppressing fiber noise and migration artifacts while enhancing the contrast of CO₂-induced anomalies. Additionally, the project will investigate distinct cost functions to minimize data misfit and effectively attenuate high-amplitude noise.

The work is further strengthened by a focus on fiber-optic signal calibration and anisotropic parameter estimation. The ultimate objective is to deliver robust techniques that translate seamlessly from algorithm development to field-ready workflows, bridging the gap between theoretical geophysics and practical monitoring solutions.