A Hybrid Approach to Develop Runup Parameterisations for Reef-Lined Coasts

In recent years coastal ecosystems have been degraded, which in combination with an increased likelihood of more intense storms and higher sea levels associated with climate change, will lead to greater vulnerability to coastal flooding along reef-lined coasts. As a result, extreme water levels and storm impacts need to be accurately predicted in order to identify vulnerable areas. A commonly used method to assess coastal vulnerability is through runup parameterisations in combination with information on dune morphology. The present study uses numerical and physical modelling, as well as a machine learning approach, to develop runup parameterisations that take into account the geometry of complex bathymetries and incident wave conditions. A nonlinear non-hydrostatic numerical model (SWASH) is validated with laboratory data corresponding to an idealized reef geometry. Subsequently, the numerical model is employed to simulate wave runup for different combinations of water levels, wave conditions and reef geometries. Genetic programming is used to identify the most suitable predictors for wave runup in these environments. The numerical runup results suggest that the incorporation of crest elevation, lagoon width and depth, and forereef slope improve runup parameterisations for reef environments. Forereef slope and crest elevation were the geometric features that had the greatest influence on runup and should be accounted for when looking at storm impacts and coastal flooding. Reef roughness is not taken into account in the development of the new parameterisations, however its effects on wave runup are studied using the physical model. The results of this study could help improve the rapid assessment of coastal vulnerability along reef-lined coasts. Funding for this study was obtained through a postdoctoral scholarship from the Programa de Becas Posdoctorales UNAM-DGAPA, the Instituto de Ingeniería of the UNAM, and from CONACYT (Cátedras Program; Project 1146). We acknowledge Camilo Rendón Valdez, Juan Alberto Gómez Liera, Gonzalo Uriel Martín Ruiz and José López González for technical support.