Experience of Fwi on Marine Seismic Data Using a Robust Optimization Procedure

ll Waveform Inversion (FWI) represents an important tool to obtain high resolution model of subsurface from active seismic data (Tarantola, 1986; Virieux, etal. 2009). In the last years many aspects about FWI have been studied concerning the implementation of efficient modelling algorithm (Chaljubetal, 2007; Moczo et al., 2007) and the formulation of inversion procedures (Fichtner, 2010). However, its application to real data requires specific operations on the seismograms to obtain an observed data that can be reproduced by a modelling algorithm. Besides, the use of an iterative gradient method requires the estimation of a starting model that must be as close as possible to the valley of the global minimum of the misfit function (Beydoun et al., 1988).In this work, we present an application of acoustic FWI on a marine seismic data. Specific processing operations are applied on both predicted and observed data to increase the robustness of the inversion procedure against the cycle skipping phenomenon, thus improving the reliabilityof the final model estimation. The predicted data are obtained by solving the 2D acoustic wave equation, whereas in the local optimization procedure the steepest descend algorithm is employed, using the L1 norm difference between the predicted and observed data to compute the misfit function. As the starting model, we use a model obbtained in a previous work (Tognarelli etal., 2015; Mazzotti etal., 2017) by a global optimization method based on genetic algorithms. To validate the final model, we pre-stack depth migrate the data using the final estimated velocity field, and we check the improvements of the flattening of the events in the common-image-gathers (CIGs).