On the design of ultra-sparse seismic surveys for monitoring subsurface CO2 storage sites using full waveform inversion

Abstract

Monitoring the dynamics of CO${}_2$ in subsurface reservoirs allows the conformance of carbon capture and storage (CCS) projects to be assessed. Full waveform inversion (FWI) of data from dense, time-lapse seismic surveys can provide high resolution images of dynamic changes. However, FWI solutions remain highly non-unique, so uncertainties must be accounted for to ensure that conformance verification is robust. Time-lapse seismic FWI is therefore expensive because, first, dense surveys are costly to acquire, and second, quantifying realistic uncertainties requires extreme computational power and memory. We first introduce a significantly less costly method to quantify Bayesian uncertainties in the maximum a posteriori (MAP – most likely) solutions of time-lapse seismic velocity changes. The method embodies strong prior information from the baseline survey to inform inversions of monitoring surveys. In contrast to comparable methods, these uncertainty estimates are shown to be of a reasonable magnitude to inform subsequent decision-making. This method also allows the quality of prospective survey designs to be assessed in terms of expected confidence in time-lapse imaging results, at reasonable computational cost. We therefore perform a time-lapse seismic survey design study to assess the quality of more economically attractive surveys. We demonstrate for the first time that even if extremely sparse acquisition geometries are deployed, potentially even involving only a single seismic source and recordings on a single fibre-optic cable, reasonable images of subsurface time-lapse velocity changes are produced, and uncertainties remain sufficiently low to enable robust decision-making.