Beach resilience to storm-induced erosion: a machine learning-driven approach to identify key morphometrics to recovery

Abstract

In the context of ongoing discussions about climate change, the focus on beach resilience has gained significant attention in contemporary studies. However, a comprehensive understanding of beach resilience, particularly in the short term, remains limited. This study utilizes a dataset of 104 storm events in Hasaki beach, located on the East coast of Japan, to investigate the 10-day beach recovery. The study considers four types of distinct beach profile patterns based on sandbar formations. Employing XGBoost and the SHAP explanation method, the influence of morphometric indicators on beach resilience were explored. Resilient beach profiles are anticipated to exhibit rapid recovery following erosional storm events. The results reveal that morphometrics play a crucial role in determining the short-term, 10-day, recovery of beaches, with specific morphometric features demonstrating pronounced effects based on profile patterns. The study contributes to the current knowledge of post-storm beach recovery and provides insights that could inform discussions on beach resilience.