FloodSAT is an AI-powered flood detection and analysis system developed by ConCodes team during the NASA Space Challenge hackathon.
FloodSAT leverages satellite imagery to identify flooded areas in near real-time. It utilizes a deep learning semantic segmentation model fine-tuned on limited labeled Harmonized Landsat Sentinel-2 (HLS) data. FloodSAT provides an automated flood mapping and monitoring system to empower emergency managers with near real-time flood intelligence for improved situational awareness and response. It also enables new insights into flood patterns to inform community planning and infrastructure design.
- Emergency response: FloodSAT can provide near real-time flood intelligence to emergency managers, enabling them to make faster and more informed decisions.
- Flood monitoring: FloodSAT can be used to monitor flood extent and patterns over time. This information can be used to identify areas at risk of flooding and to develop flood mitigation strategies.
- Infrastructure planning and design: FloodSAT can be used to inform community planning and infrastructure design by identifying areas that are frequently flooded and by modeling the potential impact of future floods.
The ConCodes team is planning to continue developing the system in several ways:
- Improve the accuracy of the flood detection model: The team is working to improve the accuracy of the flood detection model by training it on a larger dataset of labeled HLS imagery.
- Add support for additional satellite data sources: The team is planning to add support for additional satellite data sources, such as Radar and SAR data, to improve the performance of FloodSAT in challenging environmental conditions.
- Develop new analytics capabilities: The team is planning to develop new analytics capabilities, such as flood damage assessment and flood prediction, to enhance the utility of FloodSAT for emergency managers and other users.