Constructing meaningful FWI gradients for data from shaped DAS fibres

Matthew Eaid, Scott Keating, Kristopher A. H. Innanen

Distributed acoustic sensing (DAS) has garnered significant interest as a seismic acquisition technology, especially in borehole deployments. Using non-invasive optical fibres, DAS can be placed in producing wells, abandoned wells, and horizontal wells during hydraulic fracture treatments. This powerful property of DAS fibres allows for the dual purposing of wells without shutting in production and offers a unique opportunity to densely sample the long wavelength transmission portion of the wavefield. A lingering question is how to best utilize DAS data to estimate reservoir properties. Elastic full waveform inversion is a robust means of estimating elastic subsurface parameters, but it is conventionally formulated to incorporate particle velocity data supplied by geophones. In contrast to geophones in conventional surveys which provide measurements of three orthogonal components of particle velocity, DAS supplies a single measurement of tangential strain or strain rate along the fibre. We reformulate the conventional least-squares FWI objective function to incorporate the strain data supplied by distributed acoustic sensing fibres. The method developed in this paper can incorporate strain data from straight and shaped DAS fibres, and because it utilizes the conventional FWI formulation is capable of inverting displacement data from geophones and DAS fibre simultaneously. We explore the effect that shaping DAS fibres has on the quality of parameter estimations from a toy model by comparing inversion results from a straight DAS fibre in a horizontal well and three helical fibres with varying wind rates. A simultaneous inversion of reflection geophone data and DAS data from horizontal well is examined and provides enhanced parameter estimations over either dataset alone. The simultaneous inversion is then tested on a portion of the Marmousi 2 model.