On a data-model assimilation method to inverse wave-dominated beach bathymetry using heterogeneous video-derived observations

Abstract

A data-model assimilation method is presented that can combine heterogeneous remotely sensed video observations to inverse wave-dominated beach bathymetry. The ability of our method is assessed using synthetic cases from a laboratory experiment. For relatively flat beach profiles, good bathymetry inversion can be obtained with a small number (O(1)) of observations and a simple Gaussian-type background error matrix. For more complex barred-beach profiles, localisation of the background error matrix is crucial to depth inversion and the required number of observations is increased (O(10)). In the latter situation, both relevant location of observations and the corresponding localisation length scales can be properly defined without the knowledge of the local bathymetry, resulting in accurate bathymetry inversion, which is a major asset for practical applications. Our method also allows shifting between several sources of bathymetry proxy as well as overlapping which provides flexibility in data management. Multi-1D reconstruction can also be performed to estimate complex 3D beach morphologies. Our study therefore suggests that high-performing time-efficient nearshore bathymetry inversion can be achieved using a limited set of heterogeneous video-derived observations.