Bathymetry Inversion using a Deep-Learning-Based Surrogate for Shallow Water Equations Solvers

Abstract

A surrogate model for shallow water equations solvers using convolutional autoencoder has high fidelity.• Gradient-based optimization is used to perform inversion with surrogate's automatic differentiation.• Physically-based regularizations on bed elevation value and slope are necessary for usable inversion results.

Bed Topography Inference from Velocity Field Using Deep Learning

“…In recent years, deep learning has become one of the most powerful tools for overcoming some deterministic approaches' limitations [13,[16][17][18]. It can be used to image bed topography from surface flow data.…”

“…The training of neural networks for bathymetry estimation can be performed by solving the shallow water equations and using the resulting solutions. A typical example is provided by Liu et al [17], who employed shared encoders and separate decoders, where bathymetry's input image is encoded and then decoded to three outputs, namely the flow's longitudinal and transverse depth-averaged velocity components and the water surface elevation. Two-dimensional (2D) simulations using randomly generated input bathymetry data were used to generate the training data.…”