MuSIKAL: Multiphysics Simulations and Knowledge Discovery through AI/ML Technologies

Summary

The MuSIKAL project developed a coastal digital twin (DT) platform, leveraging multiphysics simulations and AI/ML technologies, to improve predictions and mitigation of coastal flooding along the U.S. Gulf Coast. The initiative, led by the University of Texas at Austin and supported by DOE Award DE-SC0022320, focused on integrating diverse data and multiscale models for storm surge and heavy rainfall events. Key achievements include the successful application of the SFINCS model to historical hurricanes like Ike (2008), Harvey (2017), and Beryl (2024), demonstrating hourly resolution flood dynamics. A hybrid AI/ML approach achieved near-real-time flood mapping with a 0.7 cm MAE and computational savings of approximately 21 seconds. The project also produced the UT-GraphCast Hindcast Dataset (1979–2024), offering daily 15-day global forecasts at a 0.25° grid, with future operational integration of its sister technology, STOFS-2D-Global, planned for 2026.

Key Facts

Source Document

https://example-government.gov/policy-document-link