Proof-of-Concept of a Neural Network Constrained by the Exner Equation for Temporal Interpolation of Flood–time Bed Evolution

Key Points

• This research aims to create a new analytical method for accurately estimating riverbed changes during floods using a neural network integrated with the Exner equation.
• Developed a neural network constrained by the Exner equation for sediment transport analysis.
• Utilized numerical simulation results from a one-dimensional riverbed model as training data.
• Evaluated the neural network's performance under unsupervised conditions for sediment transport estimation.
• Successfully estimated spatial derivatives of sediment transport with high accuracy.
• Demonstrated temporal extrapolation capabilities of the model for flood conditions.
• Confirmed the model's effectiveness for understanding flood-induced riverbed change structures.