Concurrent 3D modelling and AI-based forecasting of storm surge and urban flooding - a case study of Super Typhoon Ragasa (2518)

The Guangdong-Hong Kong-Macao Greater Bay Area (GBA) is home to a cluster of inter-connected coastal cities along the Pearl River Estuary (PRE). The GBA also faces serious challenges of climate change, mean sea level rise, and more frequent extreme weather. The 2025 typhoon season set a record in Macao with two Signal No. 10 typhoon warnings issued in the same year. Super Typhoon Ragasa (2518) is one of the strongest typhoons in recent decades and has caused widespread coastal flooding in the PRE. An analysis of the coastal impact of Super Typhoon Ragasa (2518) in Macao is reported. The contributions include: (i) storm surge prediction using a 3D FVCOM model incorporating tide-surge-wave interaction; computed surge levels are in excellent agreement with observations. The contribution of wave setup on the total water level is quantified; (ii) concurrent real time forecasts of storm surge using a physics-informed Long Short-Term Memory (LSTM) neural network model; and (iii) integration of the storm surge models and a 2D InfoWorks model to predict flood inundation maps in the Macao Inner Harbour area, which is well-validated by a post-typhoon field survey and monitoring data. The comparative merits of 3D mechanistic versus data driven LSTM models for coastal flooding predictions in Macao and GBA are clearly elucidated.