Abstract
In this study, we illustrate the application of a 3-D reflection oriented workflow for full waveform inversion (FWI) to the offset data from the Sleipner field. The data set is having maximum offset of less than 2000 m, and has been pre-processed with a low-cut filter below 6 Hz, imposing strong challenges for the FWI application. To tackle these challenges, a reflection oriented FWI is applied to the data set, which utilize joint full waveform inversion (JFWI) to constrain the low wavenumber updates at the deepest part of the reconstructed model. It consists of two steps, an impedance model building serving as a prior reflector information followed by a velocity model building. In this case, JFWI workflow is taking advantage of the pseudo-time formulation to honour the zero offset traveltime, fast and robust asymptotic pre-conditioner for impedance model building, and graph space optimal transport misfit function to mitigate cycle skipping. To show the effects of limited offset, conventional FWI is performed. In this case, it is clear that the meaningful updates coming from the diving waves are restricted to the shallow part no deeper than 500 m of depth, while no meaningful perturbation are observed beyond the diving waves penetration. Taking advantage of the meaningful shallow updates, diving wave only inversion is performed prior to the impedance model building, and then followed by the JFWI workflow. The results of the field data application show that JFWI is able to produce meaningful velocity updates both in shallow part and the deeper part. The result is supported by satisfactory fit of the calculated data based on the JFWI model compared to the observed data. In addition, the velocity model fits the low wavenumber trend of the well log data. A subsequent run of conventional FWI is performed starting from JFWI model, in order to improve the resolution of the velocity model. The results is able to introduce higher wavenumber content to the velocity model, producing satisfactory fit with the observed data, and matching the well log data.