USING EXPLAINABLE AI TO IDENTIFY BEACH VOLUME CHANGES UNDER MET-OCEAN CONDITIONS AT HASAKI COAST, JAPAN

Beaches play a significant role by providing and supporting various ecosystem services and protecting people from severe storms. However, they are vulnerable to met-ocean conditions. This study uses a 23-year morphological and met-oceanic dataset collected at the Hasaki coast, Japan to investigate the impact of natural conditions on the volumetric evolution of the beach. For this, we have considered a study area composed of 400 m long-shore and 65 m cross-shore sections on the foreshore of the coast. The area was divided by the high water level to separate the foreshore and backshore regions. In this study, we considered Seaside Volume (VS) for the medium-term changes. Power spectral analysis was utilized to investigate the characteristics of morphological changes and met-ocean conditions with a frequency component of more than 1024 days to define the medium-term components. Later, the XGBoost model combined with the SHapley Additive exPlanation (SHAP) and intercorrelation matrix was used to investigate the independent variables and avoid multicollinearity. The results showed that VS is dominantly influenced by cross-shore wave energy flux (Efx), longshore wave energy flux (Efy), wind velocity (W), and currents (v). Especially, it was found that Efx is the most influential parameter affecting VS, while Efy was found to influence erosion and accretion. Additionally, this study highlights that the direction and velocity of the v affect the volume changes in VS, with accretion occurring during the northward flow and erosion during the southward flow. Finally, while W had the least impact, it was still shown to play a role in influencing VS.