A practical approach to flood hazard, vulnerability, and risk assessing and mapping for Quang Binh province, Vietnam

Flood damage is often severe and directly affects housing, transport infrastructure, industrial, service, commercial, and land use. A flood risk assessment based on vulnerability indicators can provide valuable information to support decision-making and operational strategies to reduce disaster damage. The main objective of this study is to propose a framework for assessing flood risk based on flood hazard factor and its relationship with flood vulnerability indicators. We applied an integrated machine learning (ML) and analytic hierarchy process (AHP) framework for a case study of Quang Binh province, Vietnam. Several state-of-the-art ML models of AdaBoost, logistic regression, and AdaBoost-Logistic were applied to build a flood hazard map. AHP was employed to integrate vulnerability criteria for the assessment. We used 671 flooding locations of historical flooding events in 2007, 2010, 2016, and 2020 in Quang Binh province; and 14 flood conditioning factors relating to geo-environment to generate and verify the flood susceptibility models. Statistical indexes were applied to verify the used models. The validated result showed that the AdaBoost-Logistic ensemble model has the best performance of AUC = 0.996. The flood hazard map was combined with flood vulnerability maps to generate a valuable flood risk assessment map for Quang Binh province. The result of this study shows that 330,579 ha (40.99%) is in very low-risk zones, 349,511 ha (43.33%) in low-risk zones, 50,628 ha (6.28%) in medium risk zones, 48,688 ha (6.04%) in high-risk zones, and 27,121 ha (3.36%) in extremely high-risk zones. This proposed methodology and flood risk map result can be beneficial for selecting priority measures for flood risk reduction and management.