About the SAFE-Project
The SAFE-Project focuses on developing the first scalable, dynamic, and transferable AI-based global flood hazard mapping framework. It addresses a major gap in flood risk science: many regions worldwide still lack reliable flood hazard maps because conventional hydrodynamic models are expensive, time-intensive, data-heavy, and difficult to transfer to ungauged basins.
Project Vision
SAFE combines machine learning, with remote sensing, hydrological knowledge, and global datasets to produce interpretable and adaptable flood hazard functions. The goal is to enable rapid and scientifically grounded hazard mapping across diverse hydroclimatic conditions without extensive recalibration.
- Scalable mapping from basin to national and global levels
- Transferable AI-based functions for ungauged and data-scarce regions
- Dynamic updates under climate variability and future scenarios
- Open and practical support for research, planning, and disaster risk reduction
Project PI and Founder
Prof. Mohamed Saber (Associate Professor)
Founder and Principal Investigator of the SAFE-Project and the AI-Based Global Flood Hazard Mapping Platform. The project is developed as a research-driven platform to support global disaster resilience, scientific innovation, and practical flood hazard assessment using AI-integrated hydrological intelligence.
Why this platform matters
Flooding remains one of the most devastating hazards worldwide, while many vulnerable regions still lack usable, updatable hazard maps.
What makes SAFE different
It emphasizes interpretable AI functions, transferability across regions, dynamic updates, and a web-based mapping platform.
Who can use it
Researchers, disaster risk reduction agencies, planners, international organizations, consultants, and academic users.
Flood Hazard Maps
The interactive map platform to support visualization of the flood hazard mapping interface.
Datasets and Modeling Inputs
The SAFE-Project uses multi-source geospatial, hydrological, and observational datasets to train, test, validate, and transfer AI-based flood hazard functions across different basins and hydroclimatic settings.
1. Flood Observation and Target Variables
- Historical flood extent datasets
- Observed flood depth datasets
- Post-flood field surveys from affected basins
- Flood event records used for model training and validation
2. Topographic and Terrain Data
- Digital Elevation Models (DEM)
- Slope and terrain derivatives
- Flow direction and flow accumulation layers
- Landform and elevation-based conditioning factors
3. Hydrological and Surface Characteristics
- River networks and basin boundaries
- Drainage and hydrological descriptors
- Catchment-scale influencing factors
- Hydrological simulation outputs for benchmarking and comparison
4. Land Use and Environmental Variables
- Land use and land cover datasets
- Surface condition variables affecting runoff and inundation
- Environmental layers relevant to flood propagation and susceptibility
5. Climate and Rainfall Information
- Historical climate datasets
- Rainfall event information and extreme rainfall conditions
- Future climate scenario datasets
- Return-period-based hazard forcing for dynamic map generation
6. Modeling and AI Processing Framework
- Feature selection and multicollinearity testing
- Data fusion to harmonize diverse case studies
- Handling covariate and label shifts across regions
- Transfer learning strategies, including domain adaptation approaches
Main Modeling Workflow
- Collect and preprocess global flood and environmental datasets
- Optimize predictors using feature selection
- Train and test Machine Learning models
- Benchmark results against physically based hydrological models
- Apply trained functions to ungauged basins and future scenarios
- Disseminate results through an open web-based AI-GIS platform
Expected Outputs
- Flood hazard maps for multiple regions and return periods
- Transferable AI-based flood hazard functions
- Dynamic and updatable hazard mapping products
- Research-ready geospatial outputs for visualization and decision support
Contact Us
For scientific collaboration, data exchange, project discussion, or platform-related inquiries, please use the contact information below.
Project Contact
AI-Based Global Flood Hazard Mapping Platform
SAFE-Project research platform for scalable, dynamic, and transferable flood hazard mapping.
Project PI and Founder: Prof. Mohamed Saber
Institution: Water Resources Research Center, DPRI, Kyoto University
Address: Gokasho, Uji City, Kyoto 611-0011, Japan
E-mail: mohamedmd.saber.3u(at)kyoto-u.ac.jp
Website Policy & Data Use
- This website and all its content (data, maps, visuals, and materials) are the property of Kyoto University and are protected by copyright laws
- The platform is intended for research and educational use only
- Unauthorized copying, redistribution, modification, or commercial use of any content is strictly prohibited
- Data access, reuse, or collaboration requires prior permission from the project team
- Proper citation of the project and Kyoto University is required in any use
- The project team reserves the right to update or restrict access at any time