A Deep Learning Approach to Quantitatively Characterising the Marine Near-Surface

Summary In this paper, we present a Deep Learning workflow developed specifically for inverting marine site investigation data, comparing and contrasting it against the results obtained using traditional stochastic inversion algorithms with both synthetic and field data. In particular, we assess its potential for rapidly deriving a range of subsurface parameterisations, including geotechnical engineering properties of direct interest for various site investigation problems.