Time-lapse full waveform inversion plus extended Kalman filter for high-resolution seismic models and uncertainty estimation

Time-lapse full waveform inversion (TLFWI) is a potential tool to revolutionize seismic monitoring by providing detailed spatiotemporal images of subsurface models. Due to the nature of seismic inversion, the quantification of the TLFWI uncertainty remains challenging. In this paper, we present a novel TLFWI by formulating full waveform inversion in the framework of extended Kalman filter. The proposed method utilizes full waveform inversion to predict velocity perturbation from the current time lapse to the next time lapse and uses the extend Kalman filter to assess the posterior covariance matrix. Numerical test is performed on Frio-II CO2 velocity models. The inverted high resolution results clearly reveal a continuous velocity reduction due to the injection of CO2. Meanwhile, the posterior variance images for each inverted model provide the quantitative evaluation of models. Therefore, the proposed TLFWI provides a reliable way to monitor the distribution of the injected CO2 and detect potential CO2 leaking hazards in the future CO2 sequestration experiments.