Data assimilated time-lapse viscoacoustic full waveform inversion for monitoring CO₂ geological storage

Monitoring dynamic evolution of CO₂ plume in subsurface, as well as detecting and quantifying any potential leakages are crucial for long-term geological storage of anthropogenic carbon dioxide. Dense time-lapse seismic monitoring has made it possible to image the nearly continuous subsurface changes of seismic properties due to CO₂ plume distribution. Time-lapse acoustic/elastic full waveform inversion (TLFWI) algorithms are able to recover high spatial resolution statics subsurface velocity models by retrieving waveform information from time-lapse seismic data, but the neglecting attenuation caused by the injected CO₂ blurs the final velocity models. Another question is the TLFWI methods barely have strong constraint on extracting temporal information from dense time-lapse data to reconstruct high resolution dynamic evolution subsurface models. Besides, to quantify these subsurface models derived by TLFWI with uncertainty analysis is also challenging. Here we propose a data assimilation based time-lapse viscoacoustic full waveform inversion (TLQFWI) method, using hierarchical matrix powered extended Kalman filter (HiEKF) to simultaneously predict dynamic evolution images of subsurface velocity and attenuation changes and quantify the corresponding uncertainties of the changes over time. We demonstrate the validity and applicability of the proposed TLQFWI-HiEKF with a realistic CO₂ monitoring models derived from Frio-II CO₂ injection sites.