Uncertainty Quantification in Waveform-based Imaging Methods - a Sleipner CO2 Monitoring Study

Abstract

Full Waveform Inversion (FWI) and Controlled Source Electro-Magnetic (CSEM) inversion can provide quantitative estimates of selected subsurface properties. In the context of CO2 monitoring, FWI has the capability to locate and image a CO2 plume, seen as velocity changes caused by the injected CO2, while CSEM can be useful to determine the CO2 saturation. Quantifying the uncertainty of the reconstructed parameters is however challenging. In this study, preliminary work has been performed to assess the reliability of the results obtained using CSEM and FWI when monitoring CO2 storage at the Sleipner field offshore Norway. This paper begins with a brief review of the imaging techniques used. Then the uncertainty quantification methodology is described. Finally, preliminary results of uncertainty quantification for two CO2 monitoring examples, one synthetic case using CSEM and one real data case using FWI, are presented. For the synthetic CSEM case, the inversion of the EM data has a very positive effect on the uncertainty, clearly reducing the covariance of the probability distribution for possible conductivity models as compared to the prior knowledge of these models. For the real, rather large-scale FWI case, computationally efficient strategies had to be used. The covariance in this case is not reduced to the same extent as for the CSEM case, but an improvement is observed. In both cases, the quantification approach works as expected providing a means to assess the quality of obtained subsurface models. For future studies, a more formal way of providing prior model knowledge and a stricter formulation of the Bayesian inverse problem will be employed.