Flood susceptibility evaluation through deep learning optimizer ensembles and GIS techniques.

Area of Science:

• Environmental Science
• Geospatial Analysis
• Machine Learning

Background:

• Floods pose significant destructive risks globally, necessitating improved predictive modeling.
• Accurate flood susceptibility estimation is crucial for effective risk management and land-use planning.

Purpose of the Study:

• To evaluate the predictive performance of three distinct ensemble algorithms for flood susceptibility estimation.
• To compare the accuracy of Iterative Classifier Optimizer - Alternating Decision Tree - Frequency Ratio (ICO-ADT-FR), Iterative Classifier Optimizer - Deep Learning Neural Network - Frequency Ratio (ICO-DLNN-FR), and Iterative Classifier Optimizer - Multilayer Perceptron - Frequency Ratio (ICO-MLP-FR).

Main Methods:

• A geodatabase was compiled, including 14 flood predictors and 132 known flood locations.
• Correlation-based Feature Selection (CFS) was employed to assess predictor relevance.
• Three ensemble models (ICO-ADT-FR, ICO-DLNN-FR, ICO-MLP-FR) were trained and validated using statistical metrics and ROC curves.

Main Results:

• All evaluated ensemble models demonstrated high performance, with Area Under the Curve (AUC) values exceeding 0.89.
• The ICO-DLNN-FR model achieved the highest accuracy, with an AUC of 0.959.
• Feature selection identified key predictors influencing flood susceptibility.

Conclusions:

• Ensemble machine learning models offer powerful capabilities for flood susceptibility prediction.
• The ICO-DLNN-FR model is recommended for its superior accuracy in flood risk assessment.
• Findings support informed flood risk management and sustainable land-use planning.