Developing a real-time data and modeling framework for operational flood inundation forecasting for Australian coastlines

AUSTRALIAN FLOODS

In Australia, the impact of tropical cyclones, extratropical storms, and extensive inland flooding have shown the vulnerability of low-lying areas in floodplains and along the coast to flooding. Particularly when flooding from rainfall, rivers, and the sea converge, it has caused significant damage to infrastructure and loss of life. To help mitigate the impacts, emergency services benefit from accurate flood inundation forecasts, real-time inundation analysis, and post-event flood maps to support decision-making before, during, and after events. This requires a sophisticated early warning system capable of integrating numerous real-time data, fast, large-scale compound modeling tools and flexible dissemination protocols to reach the various end-users.

THE NEED FOR A DATA AND MODELING FRAMEWORK

The challenges for developing an effective early warning system are found in the efficient integration of large meteorological and hydrological data sets, specialized modules to process the data, and open interfaces to allow easy integration of new and existing modeling and dissemination capabilities. We have therefore developed a cloud-based national Australian flood inundation forecasting system with the Delft-FEWS framework at its core.  This system, currently in the Proof of Concept (PoC) phase, handles large amounts of forecast data efficiently, orchestrates massive computations and disseminates the real time information in various ways to the end users. 

THE NEED FOR FAST COMPOUND FLOOD MODELS

Since flooding can occur by different drivers (marine, pluvial, riverine), all these processes need to be included dynamically to model compound flooding events with enough detail to produce accurate and relevant flood maps for the emergency authorities. Therefore, the reduced-physics solver SFINCS (Leijnse et al. 2021) was used, combining all relevant processes to model compound flooding events, with strongly reduced computation times to model large scales. We demonstrate in this work that we have developed 13 SFINCS models along the coastline of the states of New South Wales, Queensland and Northern Territories covering a total of 7000+ kilometers. Additionally, corresponding gridded hydrological WFLOW models have been set up that provide upstream river boundary conditions for the SFINCS models. The models were validated against historical events with good results before using them for simulating the 2022 event in real-time.

CONCLUSION

During the extensive floods in February 2022, the PoC demonstrated that a data and modeling framework solution can be built that is scalable, interoperable, fit-for-purpose, and adaptable to the real-world challenges of emergency management. It showcases the technologies to be used in development, and how the software solution can be adopted and applied by its intended users. The products of the system can easily and freely be shared with others and the method of creating those products is transparent.