Grant: Discovery Grants Program - Individual

Floods are the most frequent weather-related natural disasters, affecting the largest number of people globally, and with economic damages in excess of $82 billion in 2021 alone. Climate change and urbanisation have led to an increase in floods globally in recent decades and this trend is projected to continue. In Canada, recent research corroborates this projection, with an increase in the frequency and severity of floods across the country. Flood forecasting systems can help develop better flood protection, inform timely mitigation and adaptation strategies, and are ultimately key to saving lives and reducing the damages associated with floods. Traditional flood forecasting methods, such as those used across Canada, are hindered by a number of significant challenges, including the complexity and uncertainty associated with changing hydrological systems and the limitations of process-based models. Flood forecasting techniques that use Artificial Intelligence (AI) have the potential to not only address these limitations, but to also improve our understanding of hydrological systems, and therefore our ability to forecast floods. The long term objective of this research program is to improve our ability to predict and prepare for floods by harnessing the power of novel and emerging AI technologies. In the short term, the program will develop an AI powered, Canada-wide, operational flood forecasting system which includes scaling-up current methods to include large-scale spatial characteristics from a diverse range of watersheds and to address the perceived limitations of AI's "black box" nature. This program will propose new hybrid methods that combine traditional process-based and AI models to improve forecasts. Results of this research program will be used to create a set of best practices for AI flood forecasting. This framework will include a central data repository to aid model performance comparison and future AI model development. The research results will be deployed to a new web-based, open-access, operational, flood forecasting platform to allow end-users and researchers to view flood forecasts across Canada and will act as a hub for AI flood research. The proposed research is the first of its kind in Canada and will improve flood forecasting capability across the country. At the global level, the use of AI for large-scale operational flood forecasting is only just beginning to gain traction and is still firmly in the research realm. The proposed research strategically aims to address open questions on the development and use of AI-based flood forecasting that, once resolved, will increase user confidence in, and uptake of, these methods. It takes advantage of key opportunities to make advancements in areas of the field that are ripe for breakthroughs within the next 5 years: the emerging applications of interpretable AI and the development of hybrid AI.