Quantifying flood risk using machine learning and data integration

Mentors:
Lead: Emily Hector (Statistics, NC State)
Collaborator: Sankar Arumugam (Civil Engineering, NC State)

Intellectual merit and significance:
Climate change affects flood risk, and quantifying these changes can reduce adverse impacts by carefully designing civil engineering projects [1]. Understanding the evolution of flood risk in a changing climate is challenging because the climate is a global phenomenon, and flooding depends on local factors such as topology and water management strategies. Addressing this problem requires diverse data and expertise, including global and regional climate model simulations, fine-scale data on extreme flooding events, and extreme value analysis. Fusing these data sources and scientific knowledge requires advanced machine learning and close collaboration between data and physical scientists.

Objectives:
Develop new machine learning tools to anticipate changes in the likelihood and magnitude of extreme flooding events under a changing climate over the next thirty years across the continental US.

Outcomes:
Maps of flood risk across the US with uncertainty for a range of climate scenarios and statistical tests that the flood risk has changed from the pre-industrial baseline.

References:

1. Seon-Ho Kim, Jeongwoo Hwang, A Sankarasubramanian (2024). Understanding the variability of large-scale statistical downscaling methods under different climate regimes. J Hydrology, 641,131818.