STFS-urban: Spatio-temporal flood simulation model for urban areas

Area of Science:

• Environmental Science
• Hydrology
• Urban Planning

Background:

• Urbanization and extreme weather events necessitate improved urban flood resilience.
• Existing flood models often lack multi-spatiotemporal scale integration and fine-grained simulation capabilities.

Purpose of the Study:

• To introduce an innovative spatiotemporal urban flood simulation model.
• To assess the model's efficacy in integrating multi-scale data and refining flood simulations.
• To enhance urban flood resilience strategies through advanced modeling.

Main Methods:

• Developed a model comprising multi-scale radar rainfall inversion (MRI), fine-grained coupled flood simulation (FGCFS), and transformer-CNN flood prediction (TCFP) modules.
• Integrated diverse data at spatiotemporal scales from 7.5 to 60 minutes and 100 to 2000 meters.
• Validated the model using the Nanjing urban area as a case study.

Main Results:

• Radar rainfall inversion demonstrated strong performance (correlation coefficient > 0.8, RMSE < 5.2 mm).
• FGCFS achieved optimal simulation at 7.5 min × 500 m resolution (Nash efficiency > 0.69).
• TCFP showed efficient learning, with errors converging to 0.002 m, suitable for resolutions below 4 m.

Conclusions:

• The proposed model effectively harmonizes multi-scale data integration, refinement coupling, and dynamic allocation for urban flood simulation.
• The study provides a sophisticated framework and practical solutions for enhancing urban flood resilience.
• This work advances flood simulation modeling, crucial for environmental research and management.