Australia (Commonwealth Union) – The realm of science, engineering, and decision-making, simulations and predictions play a pivotal role in understanding complex systems, mitigating risks, and making informed choices. Traditionally, these tasks required extensive computing power and often fall short in capturing real-world complexities. However, the advent of Artificial Intelligence (AI) has ushered in a new era, significantly transforming the landscape of simulations and predictions.
For weather forecasting, AI-driven models can process enormous volumes of meteorological data and historical patterns to provide more precise and timely predictions. AI’s speed is equally impressive.
University of Melbourne researchers have unveiled a groundbreaking simulation model, featured in the journal “Nature Water,” which possesses the remarkable ability to swiftly and precisely predict flooding in the midst of an ongoing disaster. This innovative model has the potential to revolutionize emergency responses by condensing the time required for flood forecasting from hours and days to a mere matter of seconds, thus enabling the rapid and accurate prediction of flood dynamics as emergencies unfold.
The creators of this model, including University of Melbourne PhD student Niels Fraehr, alongside Professor Q. J. Wang, Dr. Wenyan Wu, and Professor Rory Nathan from the Faculty of Engineering and Information Technology, have named it the “Low-Fidelity, Spatial Analysis and Gaussian Process Learning (LSG) model.” This LSG model stands out by producing predictions that are on par with the precision of our most advanced simulation models but does so at an astounding speed, performing computations 1000 times faster.
Professor Nathan, an expert with 40 years of experience in engineering and environmental hydrology, emphasized the enormous potential of this development as a critical tool for emergency responses. He highlighted that while existing advanced flood models can accurately simulate flood behavior, they are constrained by their sluggish computational speed and cannot be effectively employed during the unfolding of a flood event.
He said, “This new model provides results a thousand times more quickly than previous models, enabling highly accurate modelling to be used in real-time during an emergency. Being able to access up-to-date modelling during a disaster could help emergency services and communities receive much more accurate information about flooding risks and respond accordingly. It’s a game-changer.”
During rigorous testing on two vastly dissimilar yet equally intricate river systems within Australia, the LSG model demonstrated its exceptional predictive capabilities. In Southern Australia’s Chowilla floodplain, it achieved a remarkable 99 percent accuracy in flood prediction within a mere 33 seconds, compared to the 11 hours required by currently employed advanced models. Similarly, for the Burnett River in Queensland, it provided accurate predictions in just 27 seconds, as opposed to the 36 hours taken by existing models.
What sets this new model apart is its velocity, which empowers responders to account for the substantial unpredictability inherent in weather forecasts. Conventional flood forecast models often concentrate on the most probable scenario to anticipate flood events. In stark contrast, the LSG model crafted by the research team has the unique ability to simulate how the inherent uncertainty in weather forecasts translates into real-time flood impacts as a flood event unfolds. Leveraging mathematical transformations and an advanced machine learning approach, this model rapidly harnesses vast datasets while running on readily available computing systems.
Professor Nathan, reflecting on the culmination of two years of development work, underscored the model’s multifaceted potential benefits, not only within Australia but also on a global scale. “This new model also has potential benefits in helping us design more resilient infrastructure. Being able to simulate thousands of different flooding scenarios, instead of just a handful, will help design infrastructure that holds up to more unpredictable or extreme weather events,” Professor Nathan said. He added, “As our climate becomes more extreme, it’s models like these that will help us all be better prepared to weather the storm.”
