Elastic near-surface model estimation from full waveforms by deep learning

• DOI: 10.1190/segam2020-w13-06.1
• Publication Date: 2020-10-01T01:28:46Z
• Authors: Vladimir Kazei; Oleg Ovcharenko; Pavel Plotnitskii; Daniel Peter; Tariq Alkhalifah; Ilya Silvestrov; Andrey Bakulin; Paul Zwartjes
• Keywords: Feature (linguistics); Data set; Deep Neural Networks

Strong near-surface heterogeneity poses a major challenge for seismic imaging of deep targets in arid environments. Inspired by the initial success learning applications to inverse problems, we investigate possibility building nearsurface models directly from raw elastic data including surface and body waves conditions. Namely, train convolutional neural network map into model supervised way on part SEAM Arid synthetic dataset evaluate its performance different same dataset. The main feature our approach is that estimate as set 1D vertical velocity profiles, utilizing relevant subsets input neighboring locations. This effectively reduces label spaces more practical application.