Funded by the Australian Research Council (ARC)
Discovery Projects DP210103704

Project Start Year: 2021

The rapid expansion and growing density of our cities, together with climate change, have dramatically increased the risks of urban flooding. Structural engineering solutions typically used for urban flood protection (e.g., detention basins, pipe upgrades) occupy substantial valuable urban space and are expensive to build. It is therefore important to advance non-structural and distributed solutions, in particular, small-scale retention systems that are part of popular water sensitive urban design (WSUD), such as rain tanks, raingardens, constructed wetlands, etc. Achieving desired flood protection targets using distributed WSUD systems requires computationally efficient flood models and planning tools to explore potential options.

Existing flood models are based on numerical methods that are accurate but computationally slow. The project aims to develop a new generation of computationally fast models taking advantage of Machine Learning (ML) approaches. We will test whether existing numerical methods for solving the full Shallow Water Equations could be advanced or replaced using cutting-edge ML techniques e.g., data-driven curve fitting, deep neural networks, and Physics-Informed Neural Networks (PINNs). The developed model will be integrated with the UrbanBEATS planning tool to develop the flood risk mitigation module.

The Team

Ana Deletic – Project Lead
Professor, Executive Dean
Queensland University of Technology (QUT)
Behzad Jamali
Research Associate
Water Research Centre (UNSW)
João P. Leitão
Senior Scientist
Urban Water Management (Eawag)
Peter M. Bach
Research Scientist
Urban Water Management (Eawag)