2D Travel-time tomography of downward continued streamer multichannel seismic data followed by a band-limited full waveform inversion. Application to the Alboran basin (SE Iberia)

Abstract

High-resolution velocity models can be retrieved by applying adjoint-state full-waveform inversion (FWI) tocontrolled source data. However the strong non-linearity of the problem makes the solution strongly dependent onthe initial model chosen and on the low frequency content of the seismic source. Besides, typical relatively-shortoffset multi-channel seismic (MCS) data lacks first -refracted- arrivals that are commonly used to obtain a suitablestarting model for FWI.Here we show that this problem can be solved by combining a joint refraction and reflection Travel TimeTomography (TTT) of a re-datumed version of the same data set to obtain an appropriate reference model with thecorrect low wavenumber on it that is subsequently refined by FWI. The proposed workflow is first described andthen applied to MCS data acquired with 6 km-long streamer during the TOPOMED-2011 experiment in the Gulfof Cadiz (SE Iberia).The applied strategy includes as a first step a wave equation-based downward continuation (DC) or redatumingof the MCS data to simulate a sea bottom acquisition geometry, followed by a joint travel-time tomographicinversion of first arrivals identified in the DC data set together with the top of the basement (TOB) reflection fromthe MCS common mid point gathers to finally perform the multi-scale FWI of the original streamer data usingthe model obtained by TTT as initial model. The robustness of the Vp and TOB geometry model obtained byjoint refraction and reflection TTT is assessed by comparing the results obtained using three independent datasets (different groups of shotgathers). The three models coincide within parameter uncertainty bounds, and thetwo-way-time transformed TOB geometry is also coincident with the time migrated image. We conclude thereforethat the velocity and reflector depth model obtained is robust. The joint DC refraction and reflection travel-timeinversion scheme helps to reduce the inherent existing velocity-depth trade-off. As a third step, we apply FWIusing the TTT model as initial model. We show that in this case FWI can be done starting at realistic frequencies,because the starting model has the low wavenumber information needed to avoid cycle-skypping problems. Themain limitation is the depth of the inverted model (∼2 km) severely limited by the short acquisition offset