Description This repository contains the source code, model architectures, and analysis workflows used in the study: “Interpretable Deep Learning Reveals Hydro-Geomorphic Controls on Flood Susceptibility and Exposure in Data-Scarce Regions” The study develops a geomorphology-driven flood susceptibility (FS) framework using 40 digital elevation model (DEM)-derived geomorphic flood descriptors (GFDs). We implement and evaluate a suite of machine learning (ML) and deep learning (DL) models under a river-basin-based spatial cross-validation framework to ensure robust generalisation across hydrologically independent regions. The repository includes implementations of classical ML models (e.g., Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost) and deep learning architectures, including Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), Graph Neural Networks (GraphSAGE), and a hybrid CNN–GNN ensemble model that integrates local terrain morphology with basin-scale spatial connectivity. Model performance is evaluated using multiple metrics, including ROC–AUC, accuracy, precision, recall, and F1-score. Spatial cross-validation is explicitly implemented to mitigate spatial data leakage. Model interpretability is achieved using SHAP (SHapley Additive exPlanations), enabling identification of dominant geomorphic controls on flood susceptibility. Repository ContentsData preprocessing and geomorphic descriptor generation workflowsSpatial cross-validation framework (river-basin-based partitioning)Machine learning and deep learning model training scriptsCNN, GraphSAGE, and hybrid CNN–GNN model implementationsModel evaluation and benchmarking scriptsSHAP-based model interpretation and analysisFlood susceptibility map generation workflows The workflow follows three main stages: DEM preprocessing and derivation of geomorphic flood descriptors (GFDs)Model training and evaluation under spatial cross-validationModel interpretation and generation of flood susceptibility maps
Sabirul Sk, Arya Mehta, Subimal Ghosh et al.· Zenodo (CERN European Organi...· 0 citations
Description This repository contains the source code, model architectures, and analysis workflows used in the study: “Interpretable Deep Learning Reveals Hydro-Geomorphic Controls on Flood Susceptibility and Exposure in Data-Scarce Regions” The study develops a geomorphology-driven flood susceptibility (FS) framework using 40 digital elevation model (DEM)-derived geomorphic flood descriptors (GFDs). We implement and evaluate a suite of machine learning (ML) and deep learning (DL) models under a river-basin-based spatial cross-validation framework to ensure robust generalisation across hydrologically independent regions. The repository includes implementations of classical ML models (e.g., Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost) and deep learning architectures, including Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), Graph Neural Networks (GraphSAGE), and a hybrid CNN–GNN ensemble model that integrates local terrain morphology with basin-scale spatial connectivity. Model performance is evaluated using multiple metrics, including ROC–AUC, accuracy, precision, recall, and F1-score. Spatial cross-validation is explicitly implemented to mitigate spatial data leakage. Model interpretability is achieved using SHAP (SHapley Additive exPlanations), enabling identification of dominant geomorphic controls on flood susceptibility. Repository ContentsData preprocessing and geomorphic descriptor generation workflowsSpatial cross-validation framework (river-basin-based partitioning)Machine learning and deep learning model training scriptsCNN, GraphSAGE, and hybrid CNN–GNN model implementationsModel evaluation and benchmarking scriptsSHAP-based model interpretation and analysisFlood susceptibility map generation workflows The workflow follows three main stages: DEM preprocessing and derivation of geomorphic flood descriptors (GFDs)Model training and evaluation under spatial cross-validationModel interpretation and generation of flood susceptibility maps
Sabirul Sk, Arya Mehta, Subimal Ghosh et al.· Zenodo (CERN European Organi...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
