Modelling shoreline evolution over multiple time-scales

Abstract

In recent decades, research efforts to understand and predict beach evolution have increased since it is becoming clear that coastal erosion is likely to be exacerbated as a result of the intensification of storms and increased rates of sea level rise. The social and economic implications of changes along the beach are vast, hence the importance of developing predictive models. Despite the development of a variety of models based on different approaches to address shoreline evolution, the predictive capability of these models is still limited by an incomplete understanding of interactions between drivers and responses, and the different spatial and temporal scales at which they act. In general, traditional shoreline models (based on the equilibrium concept), have shown good performance at predicting shoreline changes from seasonal to multi-annual time-scales but still struggle to predict faster changes in the shoreline position. Therefore, a modi cation of one of the most popular shoreline evolution models (Yates et al., 2009) is introduced, showing good performance when predicting time-scales longer than seasonal but also the faster shoreline changes. The model was presented in a competition where 19 numerical models (a mix of established shoreline models and machine learning techniques) were tested at Tairua beach, New Zealand. The results showed that although traditional models and machine learning techniques had a good performance at reproducing the shoreline evolution during the 15 years of calibration, skill decreased during the 3 years of forecast prediction (unseen data). A model ensemble results in better performance than any individual model, accounting for uncertainties in model architecture. The study gave evidence of the difficulty in achieving reliable predictions over both short and long-term shoreline time-scales. For this reason, a new model approach based on the Complete Ensemble Empirical Mode Decom- position method is introduced and tested at two study sites (Narrabeen, Australia and Tairua, New Zealand). The new approach estimates the characteristic oscillations in the shoreline and drivers, allowing to predict the shoreline changes at individual time-scales, identifying the drivers with the largest contribution to shoreline change. Then the total shoreline position is predicted as the sum of all the significant time-scales. The approach is novel also because it uses as model drivers, sea level pressure fields and gradients, in addition to the more traditional bulk wave information. The new model displays better performance when compared to an established shoreline model. This approach bridges the short-term shoreline change driven by waves with longer-term changes driven by large-scale climate oscillations (e.g. El Ni~no Southern Oscillation). Finally, the new model approach was applied to a beach with an entirely different setting, Vougot beach, France. This beach is unique in many aspects: the large tidal range, the presence of o shore rocks, and the frequently observed dune erosion during storm events followed by resilience phases in between stormy winters, making the modelling extremely challenging. The dune/beach interactions were analysed throughout a centroid analysis in which the dominant beach change modes were identified The analysis allows to identify how the sediment contribution resulting from the dune erosion events `resets' the shoreline behavior. As a result, the shoreline oscillations at time-scales related to dune erosion and recovery events account for a large part of the explained variance. This methodology allows improving understanding of beach-dune interactions, and even more generally, prediction horizons at beaches where many processes operate and traditional approaches fail. Overall, the research provides useful insights to understand that in addition to the expected seasonal-annual shoreline changes caused by incident wave variability or long-term changes associated with longshore sediment transport, many other time-scales of change may co-exist and have significant impacts on shoreline evolution and its prediction.