Computationally efficient modelling of wave driven flooding in Atoll Islands: Investigation on the use of a reduced-physics model solver SFINCS

Wave driven flooding threat posed to small island developing states (SIDS) with atoll islands underscore the urgency and critical need to assess flood risk. Recent studies suggest that the timeline for adaptation tipping points on these islands may be passed in the order of decades (Storlazzi et al., 2018). Assessment of climate change scenarios for these islands is critical and require the assessment of a large number of conditions through complex physics models which incurs a huge computational expense. This thesis aims to investigate the use of a reduced-physics model SFINCS is driven by wave generating boundaries from a complex model. This is assessed through a large number of runs based on synthetic database of Pearson (2016) preceded by sensitivity analysis for model parameters. Moreover, this thesis proposes and evaluates methodology to provide synthetic time series to drive the reduced physics model without the use of a complex physics model. Spectra for time series are provided through JONSWAP based spectra and machine learning methods including regression trees and artificial neural networks. In conjunction with spectra, phase models for the time series including random phase and “skewed” phase were tested for different reef locations. Results indicate that the model underestimates the high frequency wave transformation and subsequently wave runup when driven further from the shoreline. The reduced physics model was also found to be an order of a magnitude faster than XBeach with increases in computational speed observed when a smaller domain is resolved. Spectral methods including TMA+GAUSS, MULTIPLE JONSWAP and GAUSSIAN JONSWAP and machine learning methods including regression trees and neural networks indicate the ability to capture the spectral density at coral reefs. It was observed that the use of random phase leads to the underestimation of runup extremes when forced close to the shoreline. The mid reef was noted to be an optimal location to balance the development of wave shape, offer computational efficiency and the underestimation of high frequency waves. Results combining the spectra and phase suggest that the simplest method of representing spectra TMA+GAUSS and random phase are the ideal combination. This thesis briefly outlines how the reduced physics model can be used to setup site specific early warning systems on atoll islands and carry out climate change assessments. The reduced physics model offers the potential to determine flooding on atoll islands with reasonable accuracy and speed to be applied for disaster risk reduction measures.