Answers are grounded in this paper and cite their sources.
Study constructs explainable machine learning models to assess flood susceptibility in urban areas, indicating effective risk mapping.
This study constructs ML models for flood susceptibility modeling in Kyrgyzstan's three primary urban hubs: Jalal-Abad, Osh, and Tokmok. By developing a robust and explainable modeling approach, this study provides valuable insights into flood susceptibility assessment in these flood-prone metropolitan areas. A detailed flood inventory dataset was created using Sentinel-1 satellite imagery to delineate flood and non-flood areas and to integrate values for flood-contributing factors. Machine Learning (ML) algorithms: Random Forest (RF), eXtreme Gradient Boosting (XGB), and AdaBoost were deployed to generate flood susceptibility maps. Subsequently, model performance was rigorously evaluated using various statistical metrics, including accuracy, kappa coefficient, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). All three models gave success and prediction rate curve values above 84%. The averages of the very high susceptibility class for the cities of Jalal-Abad, Osh, and Tokmok were 8.44%, 12.01%, and 37.8%, respectively. Additionally, SHAP (SHapley Additive exPlanations) analysis was employed to interpret and explain the importance of flood-contributing factors in the predictive models, enhancing model interpretability and providing insights into flood susceptibility mechanisms.
No takes yet. Share an insight, caveat, or question.
Linh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: