Across South Africa and much of the Global South, informal settlements are often located in environmentally vulnerable areas such as floodplains, riverbanks, and poorly drained land. These communities face disproportionate exposure to climate-related hazards, compounded by limited access to reliable data, early warning systems, and formal planning frameworks.

The talk presents a practical, scalable GeoAI-driven approach to flood risk mapping and early warning using entirely open-source tools and openly available datasets. The workflow integrates satellite imagery from Sentinel-2, digital elevation models (DEMs), rainfall data, and socio-economic indicators within platforms such as Google Earth Engine, QGIS, and Python-based libraries. Spectral indices such as NDVI and NDWI are derived to characterize vegetation and surface water dynamics, while terrain variables and hydrological features are incorporated to improve flood susceptibility modelling.

A key component of the methodology is the application of machine learning techniques to classify flood-prone areas and generate risk surfaces. Importantly, the model goes beyond physical hazard mapping by integrating social vulnerability indicators, enabling a more holistic understanding of risk that reflects both environmental exposure and community sensitivity. The result is a multi-dimensional flood risk map that can support targeted interventions and resource allocation.

Using case studies from Gauteng Province, the presentation demonstrates how this open-source workflow can identify high-risk zones within informal settlements, validate model outputs, and produce actionable insights for disaster management practitioners. The talk will also highlight practical challenges encountered, including data quality issues, model transferability, and implementation constraints in resource-limited settings.

By sharing lessons learned and reproducible methods, this presentation aims to contribute to the broader FOSS4G community by showcasing how open geospatial technologies can be applied to real-world problems. It emphasizes the role of open data, collaborative tools, and accessible platforms in bridging information gaps and supporting climate resilience. Ultimately, the session seeks to inspire practitioners, researchers, and policymakers to adopt and adapt open-source geospatial solutions for disaster risk reduction and sustainable urban development.