Nelvinebi/Urban-Flood-Susceptibility-Mapping-in-Port-Harcourt-and-Warri-Using-GIS-and-ML

This project applies machine learning and GIS techniques to map urban flood susceptibility in Port Harcourt and Warri using realistic synthetic data, supporting flood risk assessment, spatial planning, and environmental research in data-scarce urban regions. Urban Flood Susceptibility Mapping in Port Harcourt and Warri Using GIS and Machine Learning (Synthetic Data) 📌 Overview This project models urban flood susceptibility in Port Harcourt and Warri, Nigeria, using GIS-based spatial factors and machine learning. Realistic synthetic data are used to demonstrate flood risk prediction and mapping workflows. 🎯 Objectives Simulate urban flood conditioning factors Train an ML model for flood susceptibility classification Produce GIS-ready flood risk outputs Support urban planning and flood risk assessment research 🗂️ Project Structure Urban-Flood-Susceptibility-Mapping/ │ ├── data/ │ └── urban_flood_susceptibility_dataset.xlsx │ ├── scripts/ │ └── urban_flood_susceptibility_portharcourt_warri_ml.py │ ├── outputs/ │ ├── flood_susceptibility_map.tif │ └── flood_zones.shp │ ├── README.md └── requirements.txt 📊 Dataset Description Synthetic dataset includes: Rainfall (mm) Elevation (m) Slope (degrees) Drainage density Impervious surface (%) Distance to river (m) Flood risk class (Low / High) 🤖 Methodology Generate realistic synthetic GIS variables Train a Random Forest classifier Predict flood susceptibility Export results as GeoTIFF and Shapefiles 🛠️ Technologies Used Python NumPy, Pandas Scikit-learn Rasterio, GeoPandas Matplotlib 🚀 How to Run pip install -r requirements.txt python scripts/urban_flood_susceptibility_portharcourt_warri_ml.py 🗺️ Outputs Flood susceptibility raster map (GeoTIFF) Flood risk zones (Shapefile) ⚠️ Disclaimer This project uses synthetic data for academic and demonstration purposes only. Results should not be used for real-world flood management decisions. 📄 License MIT License 👤 Author AGBOZU EBINGIYE NELVIN LinkedIn: *linkedin.com