An explainable and transferable deep learning framework for spatiotemporal urban flood prediction by integrating Vision Transformer and U-Net

Key Points

• Accurate urban flood prediction provides critical insights for planning and response efforts, enhancing resilience to climate impacts.
• Results indicate a significant increase in predictive performance over traditional models, crucial for timely interventions.
• Framework integrates deep learning with advanced architectures—specifically Vision Transformer and U-Net—for robust analysis.
• The approach supports adaptability, yet further validation in diverse urban settings is needed for broader applicability.