Data-driven investigations of broken wave behaviour in the surf and swash zones

This thesis investigates waves on sandy, wave-dominated, beaches using data-driven approaches. Although waves - particularly broken waves - are the main driver for sediment dynamics in the nearshore, little research has been dedicated to track and study individual waves. By applying machine learning techniques to collocated remotely sensed and in-situ data, this thesis answered questions regarding the fraction of broken waves (Q_b) in the surf zone, assessed the probability of a broken wave being captured by another broken wave (p(c)), and showed how changes in offshore conditions correlated to changes in swash zone motions. Q_b was shown to be highly variable inter- and intra-beach and was directly correlated to nearshore parameters. Tides drove variations in Q_b of up to 70% for a given water depth and increases in infragravity wave energy correlated with lower Q_b values at the surf-swash boundary. Bore-bore captures were found to be common in the surf and swash zones (p(c)≈40%) and a direct correlation between increasing infragravity wave energy at the shoreline and increases in the rate of occurrence of bore-bore capture events was observed. The observed bore-bore capture events did not drive extreme (horizontal)shoreline maxima events frequently (≈20%), but most of the observed extreme events (≥95%) were directly driven by bore-bore captures. Offshore conditions correlated to changes in the probability distribution function (PDF) of shoreline height timeseries on a gently sloping beach. Variations in offshore wave height and period correlated to variations in shape and modality of shoreline height PDF. Similar results showed that trough-to-peak swash height PDFs, both in the incident and infragravity frequency bands, were multimodal and correlated to changes in offshore and surf zone conditions. Results obtained here showed that purely theoretical approaches to nearshore phenomena unsatisfactorily predicted the observed data. Theoretical Q_b models showed errors of 40%, and no currently available theoretical approach could precisely describe the observed swash zone motion PDFs. The data-driven models developed here, on the contrary, showed high skill when predicting the observed data. These results demonstrate that the novel approaches developed in this thesis have potential to be further developed into coastal management tools.