High Resolution Anisotropic Earth Model Building on Conventional Seismic Data Using Full Waveform Inversion: A Case Study Offshore Australia

Abstract

Abstract We present a case study from the North West Shelf of Australia where the complexity of the overburden consists of several thin multi-level channel systems filled with a combination of anomalously high or low velocity sediments. Not accounting for these strong velocity variations accurately, can lead to subtle image distortions affecting the underlying section down to and including the reservoir level. This can have significant impact on the volumetric estimates of reserves in place. To resolve these complexities in the overburden, full waveform inversion (FWI) was utilized to generate an updated earth model exploiting both early arrivals and reflection events. One caveat to using full waveform inversion is the need for low frequencies to be present in the seismic data, or, the initial starting velocity model must contain the correct low wavenumber components. However, conventional seismic data acquired at shallow tow depths are usually band limited particularly at the very low frequencies. Our case study will discuss these issues along with other limitations that this "conventional data" presented along with the workflows and quality control methods adapted to this data in order to converge to a plausible, high resolution earth model. Introduction An accurate earth model is fundamental to any depth imaging project. Full waveform inversion is an advanced model building technique incorporating the full two way wave equation. Full waveform inversion produces an accurate high resolution earth model by simultaneously using the information of travel time, amplitude and phase contained in the full recorded seismic wavefield. One pre-requisite to full waveform inversion is an initial starting model. In this case study, the initial starting model was derived from a smooth version of a reflection travel time tomography velocity field derived from a depth migration workflow. The full waveform inversion process utilizes this model and a two-way wave equation finite difference acoustic wavefield propagator to generate modelled seismic data. These modelled shots are then compared to the acquired (observed) recorded seismic shots. The residual differences are backward propagated from time to depth domain, into velocity gradients and velocity changes required to obtain an updated model (see Figure 1). As with solving any non-linear inversion problem, it is an iterative process and is repeated as required until the residuals between the modelled shots and the actual observed seismic data are minimized. Iterations start at low frequencies and progress to higher frequencies.