Hydro-Morphological Characterization of Coral Reefs for Wave Runup Prediction

Abstract

Many coral reef-lined coasts are low-lying with elevations <4 m above mean sea level. Climate-change-driven sea-level rise, coral reef degradation, and changes in storm wave climate will lead to greater occurrence and impacts of wave-driven flooding. This poses a significant threat to their coastal communities. While greatly at risk, the complex hydrodynamics and bathymetry of reef-lined coasts make flood risk assessment and prediction costly and difficult. Here we use a large (>30,000) dataset of measured coral reef topobathymetric cross-shore profiles, statistics, machine learning, and numerical modeling to develop a set of representative cluster profiles (RCPs) that can be used to accurately represent the shoreline hydrodynamics of a large variety of coral reef-lined coasts around the globe. In two stages, the large dataset is reduced by clustering cross-shore profiles based on morphology and hydrodynamic response to typical wind and swell wave conditions. By representing a large variety of coral reef morphologies with a reduced number of RCPs, a computationally feasible number of numerical model simulations can be done to obtain wave runup estimates, including setup at the shoreline and swash separated into infragravity and sea-swell components, of the entire dataset. The predictive capability of the RCPs is tested against 5,000 profiles from the dataset. The wave runup is predicted with a mean error of 9.7-13.1%, depending on the number of cluster profiles used, ranging from 312 to 50. The RCPs identified here can be combined with probabilistic tools that can provide an enhanced prediction given a multivariate wave and water level climate and reef ecology state. Such a tool can be used for climate change impact assessments and studying the effectiveness of reef restoration projects, as well as for the provision of coastal flood predictions in a simplified (global) early warning system.

A Clustering Approach for Predicting Dune Morphodynamic Response to Storms Using Typological Coastal Profiles: A Case Study at the Dutch Coast

“…(1) A 1st clustering of profiles based on a combination of morphological and hydrodynamic parameters to obtain N number of initial TCPs (iTCPs). The number of iTCPs N should always be larger than the number of final TCPs M. Here, different values of N were tested, and a value of N = 300 iTCPs was chosen based on an elbow approach (Scott et al, 2020; see Supplementary Figure 4), ensuring a high inter-cluster similarity while having a small number of iTCPs.…”

“…This means that the test profile will be first matched to the iTCPs and then the TCP assignments of the iTCPs will be followed to finally match the test profile with the TCPs. Two matching methods were compared following Scott et al (2020). First, a direct matching approach was tested, which was simply done by matching the test profile with the most similar iTCP, defined by the smallest pairwise distance between the test profile and the iTCPs.…”

“…Input reduction in coastal applications has been commonly applied to reduce model forcing (Walstra et al, 2013;Chiri et al, 2019;de Queiroz et al, 2019) for long-term morphodynamic cases. Recently, Scott et al (2020) presented a methodology to reduce a large dataset (>30,000) of coral reef topo-bathymetric profiles to a representative subset (on the basis of wave runup) of 50-312 profiles, using machine learning, statistics and a numerical model. Using clustering algorithms to obtain the representative profiles and a probabilistic matching technique, they predicted wave runup for a test subset of profiles with a mean error of 9.7-13.1%, highlighting the potential of these techniques.…”

“…Using clustering algorithms to obtain the representative profiles and a probabilistic matching technique, they predicted wave runup for a test subset of profiles with a mean error of 9.7-13.1%, highlighting the potential of these techniques. Alternatives for capturing the variability in topo-bathymetric characteristics, such as the creation of schematic cross-shore profiles based on a set of parameters (Pearson et al, 2017;Cohn et al, 2019), lack the realistic representation of the actual coastal features and can be computationally infeasible, since adding parameters and/or increasing their resolution can increase the number of simulations exponentially (Scott et al, 2020).…”

“…Here we built upon Scott et al (2020), with a focus on sandy coasts and dune erosion. Our case study is the Dutch coast where dune erosion is of critical importance for coastal zone management, with the dune system being the first line of defense against coastal flooding (de Vries et al, 2012) of the low-lying hinterland which is below mean sea level at almost one quarter of the country (Schiermeier, 2010).…”