The ARTIFACT project includes the following case studies to demonstrate and validate AI-based models and techniques for flood management:
1. AI-Based Urban Floods Models
- Goal: Develop flood prediction models to assess hazardous events, focusing on increasing speed and accuracy for early-warning systems.
- Implementation: Apply models to urban environments in Serbia, Germany, and the Netherlands. The focus will be on using data-centric approaches to optimize small datasets and generate synthetic data to enhance model training .
2. Nowcasting and Short-Term Forecasting:
- Focus: AI-driven nowcasting for rapid flood assessment using advanced methods like RNNs, GNNs, and GANs.
- Approach: Employ hybrid models that combine AI with traditional flood models to improve prediction speed and generalization across different case study sites .
3. Nature-Based Solutions (NBS) for Flood Proofing:
- Objective: Develop hybrid models coupling NBS (e.g., rain gardens, permeable pavements) with traditional infrastructure.
- Exploratory Modeling: AI-based agent modeling to assess the effectiveness of NBS at reducing flood volumes, peak heights, and damages.
- Examples: Apply these hybrid models to city-scale systems, testing configurations for urban flood mitigation .
- Specific focus on local challenges in Belgrade and Novi Sad, targeting populations of 100,000 and 30,000, respectively.
- Benefits include enhanced flood preparedness and implementation of early-warning systems tailored to regional needs.
- The open-access repository created by the project will host data from European and global case studies, supporting broader application of findings and promoting international collaboration .
