No abstract

No abstract

Nha Trang Coast is located in the South Central Vietnam and the coastal erosion has been occurring rapidly in recent years. Hence it is crucial to accurately monitor the shoreline changes for better coastal management and reduction of risks for communities. In this paper, we explored, for the first time, a statistical forecasting model, Seasonal Auto-regressive Integrated Moving Average (SARIMA), and two Machine Learning (ML) models, Neural Network Auto-Regression (NNAR) and Long Short-Term Memory (LSTM), to predict the shoreline variations from surveillance camera images. Compared to the Empirical Orthogonal Function (EOF), the most common method used for predicting shoreline changes from cameras, we demonstrate that the SARIMA, NNAR and LSTM models outperform the EOF model significantly in terms of prediction accuracy. The forecasting performance of the SARIMA model, NNAR model and LSTM model is comparable in both long and short-term predictions. The results suggest that these models are highly effective in detecting shoreline changes from video cameras under extreme weather conditions.

Abstract:A field experiment was conducted on a sandy beach with a low tide terrace (Nha Trang, Vietnam) to investigate the swash zone hydro-and morphodynamics throughout different tide and wave conditions. A 2D Lidar was used to measure runup properties and bed level changes on the swash zone. An energetic monsoon wave event provided energetic conditions during the initial stage of this experiment while mild wave conditions were observed during the remaining days. Swash dynamics were clearly modulated by wave and tide conditions. Preliminary results indicate that wave climate is linked with extreme runup and beach erosion and recovery processes while tide level seems to affect swash spectral signature (dominated by infragravity band during low tide and incident band during high tide) and linked with asymmetrical morphological response of the swash.

No abstract