PREDICTION OF EQUILIBRIUM SUBAQUEOUS DUNE CHARACTERISTICS USING GENETIC ALGORITHMS

Sand dunes are ubiquitous in natural subaqueous environments and may pose a significant risk in coastal setups for many domains such as the offshore industry or marine renewable energies (Vantorre et al., 2013 ; Barrie and Conway, 2014). Observations of bedform development show that an initially flatbed evolves through different phases: an incipient bedform phase, a growing phase and a stabilizing phase leading to a fully developed dune field. Finite-amplitude dune equilibrium is essentially controlled by the flow depth, the bed shear stress, and the inertia length of the sediment transport in suspension and for bedload at smaller depths (Dore et al., 2023). Existing predictors fail to describe dune dimensions at equilibrium (height and wavelength) (Bradley and Venditti, 2017). Data availability and the poor description of complex morphodynamic phenomena has led to an increasing use of Machine learning in coastal science (Goldstein et al., 2019). In the present work we use Genetic algorithms (GAs) to perform regression analysis on dune laboratory data available in the literature. Our results showed that GAs outperform usual predictors against performance metrics.