.png){kind=icon}
.png){kind=icon}
Return
Flood susceptibility mapping using optimized deep learning models: a non-structural framework
.png){kind=icon}
OA
AI
DOI：10.1007/s13201-025-02548-5.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Floods are among the most destructive natural hazards, demanding accurate and efficient predictive tools for non-structural risk management. This study introduces a novel framework that integrates deep learning models—long short-term memory (LSTM) and recurrent neural network (RNN)—with two metaheuristic optimization algorithms, genetic algorithm (GA) and crow search algorithm (CSA), for flood susceptibility mapping (FSM). The innovation lies in hybridizing deep learning with metaheuristic optimization to enhance predictive accuracy. Using remote sensing and 12 key flood-conditioning factors, we produced high-resolution FSMs for Estahban, Iran. Five hundred and nine historical flood locations were used for model training and validation. The models were designed to predict continuous flood susceptibility values, enabling detailed spatial risk assessment using six developed models. Our findings reveal that optimized models significantly outperformed standalone models in predicting flood-prone areas. The RNN-GA model achieved the highest performance (area under the curve (AUC = 93.2%)), followed closely by LSTM-GA (AUC = 93.1%), RNN-CSA (AUC = 93%), and LSTM-CSA (AUC = 92.9%). Standalone models demonstrated comparatively lower accuracy, with RNN (AUC = 92.7%) and LSTM (AUC = 90%). This research contributes to developing a more effective and sustainable approach to flood management that complements existing structural measures.
Keywords:
Flood management
Flood susceptibility map
Deep learning algorithms
Remote sensing imagery
Metaheuristic algorithms
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
A
IF:
5.7
Papers:
2.2K
Citations:
1.2W