Reduced-complexity modelling of decadal-scale shoreline change

<p dir="ltr">Robust and reliable models are needed to understand how coastlines will evolve over the coming decades, driven by both natural variability and climate change. While many models of shoreline change have been developed for sandy coasts and are increasingly being used to generate forecasts to the end of the century, few studies have evaluated how these models perform over such multi-decadal timescales. This thesis evaluated the performance of six recently developed models against a high-quality 40-year dataset of monthly shoreline change at Narrabeen-Collaroy Beach in southeast Australia, one of very few such datasets globally. The six models selected were ‘reduced-complexity’ hybrid frameworks that each couple popular sub-models of cross-shore and longshore processes. This class of models has been identified by multiple recent reviews to be most promising for multi-decadal simulations. The complex cross-shore and longshore dynamics at Narrabeen-Collaroy Beach, varying both spatially alongshore and temporally at interannual timescales, provided a rigorous test for the coupled model frameworks.</p><p dir="ltr">Modifications to the set-up of five of the six models were required to achieve a stable output at this site. Some models were found to be excessively sensitive to decadal-scale shifts in the wave climate, while others (using the popular ‘one-line’ approach as a sub-model to simulate longshore sediment transport) were very sensitive to small biases in nearshore wave direction. Following modifications, the six models showed similar performance across the embayment on average but with large variability between transects. Their performance was promising considering their simplicity and low computational cost, but likely not yet sufficient for routine applications to inform coastal management. In particular, the models appeared to aggregate key processes at this timescale into parameter values rather than capturing them directly, suggesting time-varying parameters and/or changes to model structure may be necessary for decadal-scale simulations. The results also highlighted the importance of site-specific calibration, validating performance against unseen data, and testing models across morphodynamically different sites.</p><p dir="ltr">Subsequently, given the popularity of one-line models and the high sensitivity to wave direction bias observed in these results, a standalone one-line model was used to test three methods to correct for this issue. Two methods corrected the input wave conditions while the third corrected the output modelled shoreline positions. The methods provided different degrees of trade-off between computational expense and the resulting effectiveness of the correction. No solution was ‘optimal’, but different solutions may be best suited for different cases.</p><p dir="ltr">Finally, recent advances in extracting shoreline positions from satellite imagery with sub-pixel accuracy have resulted in a transformative step forward for the field, providing an unprecedented global 35-year dataset of shoreline change. A simple comparison between two models calibrated with shorelines derived from field survey data, satellite imagery, and aerial photogrammetry illustrated that the error of satellite-derived shorelines appears to be within the error already inherent in reduced-complexity models and does not substantially affect their performance for ‘best-case scenario’ sites such as Narrabeen-Collaroy Beach. It is argued that utilising this data to achieve site-specific calibrations for individual beaches, and to test the performance of shoreline change models over a much broader range of morphodynamically diverse sites than previously possible, will be key to further advancing our ability to model shoreline change over multi-decadal timescales.</p>