 Novel optimized deep learning algorithms and explainable artificial intelligence for storm surge susceptibility modeling and management in a flood-prone islandMohammed J. Alshayeb (),
Hoang Thi Hang (),
Ahmed Ali A. Shohan () and
Ahmed Ali Bindajam ()
Natural Hazards: Journal of the International Society for the Prevention and Mitigation of Natural Hazards, 2024, vol. 120, issue 6, No 6, 5099-5128 Abstract: Abstract Sagar Island, located in the Indian Sundarbans Delta, is extremely vulnerable to storm surge flooding. Therefore, there is a need for a precise model to assess its susceptibility to storm surges, which is essential for efficient coastal management and reducing the risk of disasters. Traditional modeling methods often lack the capability to consider the intricate relationships between various influencing factors. This study aims to advance the field by developing robust deep learning (DL) models for storm surge susceptibility prediction, specifically incorporation of Bayesian optimization with DL models and implementing Explainable Artificial Intelligence (XAI) methods for model interpretation and data-driven management. Storm surge susceptibility in Sagar Island presents a critical problem requiring advanced predictive modeling. We employ Bayesian Optimization for hyperparameter tuning in Deep Neural Networks (DNN), 1D Convolutional Neural Networks (CNN), and LightGBM models, focusing on 11 variables with low multi-collinearity. Additionally, we applied Explainable AI techniques such as SHapley Additive exPlanations (SHAP) and permutation importance for model interpretability. DNN achieved the highest accuracy of 97.75%, with F1-score at 97.85%. LightGBM and CNN were close competitors with an accuracy of 97.5%. DNN’s true positive rate was 95%, compared to CNN’s 94% and LightGBM’s 93%. In the susceptibility mapping analysis, the CNN detected areas categorized as ‘Very Low’ and ‘Very High’ susceptibility, constituting 65.12% of the study area and covering 156.88 km² and 11.38% covering 11.38 km², respectively. Similarly, LightGBM exhibited a comparable pattern, but with a more pronounced representation of ‘High’ susceptibility zones, spanning 22.61 km², predominantly across the coastal and central regions of Sagar Island. Feature importance assessed through SHAP revealed “Rainfall” as 40% more impactful than “Cyclone Track.” While all three models demonstrated robust performance, DNN emerged as marginally superior in most evaluated metrics. Our study provides valuable insights for targeted storm surge management strategies in Sagar Island, combining high predictive accuracy with model interpretability. Keywords: Storm surge; Bayesian optimization; Deep learning; LightGBM; Explainable AI; Sagar Island (search for similar items in EconPapers) Downloads: (external link) Related works: Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text Persistent link: https://EconPapers.repec.org/RePEc:spr:nathaz:v:120:y:2024:i:6:d:10.1007_s11069-024-06414-6 Ordering information: This journal article can be ordered from DOI: 10.1007/s11069-024-06414-6 Access Statistics for this article Natural Hazards: Journal of the International Society for the Prevention and Mitigation of Natural Hazards is currently edited by Thomas Glade, Tad S. Murty and Vladimír Schenk More articles in Natural Hazards: Journal of the International Society for the Prevention and Mitigation of Natural Hazards from Springer, International Society for the Prevention and Mitigation of Natural Hazards |
Is your work missing from RePEc? Here is how to contribute.
Questions or problems? Check the EconPapers FAQ or send mail to .
EconPapers is hosted by the
School of Business at Örebro University.