First Breaks Picking: a Key Element to Improve Seismic Imaging through Full-Waveform Inversion on Land Data

Abstract

Seismic imaging on foothills domain has always been challenging. Data are generally noisy, often sparse and limited to 2D lines. Velocity model building for PSDM is extremely complex and often done with a geological model driven approach rather than a data driven one. Progress made on imaging techniques, Full-Waveform Inversion (FWI) on our case, allows us to apply fully data driven approach for seismic imaging on complex land dataset. The main objective is to “let the data speaks” and not to rely on an a priori geological model. The main pitfall in a FWI imaging workflow is poor initial velocity model that leads to cycle skipping. With land dataset, cycle skipping can easily occurs since we are working with rather noisy and often only 2D data. When we first try applying classic FWI on our data, the model update was not consistent with the expected geology and introduced velocity artifacts on the near surface. This model update degraded the stack image, clearly showing that the FWI did not properly converged due to cycle skipping. To avoid it, we used Laplace-Fourier FWI implementation. With this approach, first breaks picks are used to localize in time the inversion to the early arrivals which tend to be better predicted by the acoustic approximation. This has also the effect of boosting the low frequency content of the data, hence reducing the chance of cycle skipping. It is important to note that the Laplace-Fourier uses both the first breaks and the full wave information around the first breaks making it more robust than diving wave tomography. The data is localized in time with help of a damping term which is nothing but an exponential decaying function. On our process, we performed several iterations of model update through FWI, decreasing progressively the damping on the seismic data. This way, we updated first the background velocity trend and then more details were added in the model when reducing the damping on the seismic data. The resulting model shows a geologically consistent update and provide a clear imaging improvement. This method was applied on real 2D land seismic datasets from onshore Africa and Asia-Pacific. Standard FWI (without the use of first breaks) fails at correctly updating the model on both cases. Whereas the updated model using Laplace-Fourier method provided significant imaging improvements. Applying this workflow on land dataset to enhance seismic imaging could unlock the potential of underexplored onshore basins.

AAPG Datapages/Search and Discovery Article #90332 © 2018 AAPG International Conference and Exhibition, Cape Town, South Africa, November 4-11, 2018