A new GPU-accelerated flood forecasting system developed at the University of Illinois' Discovery Partners Institute (DPI) Climate Hub could soon give city managers, emergency responders and regional authorities the real-time flood intelligence they have long lacked, compressing simulations that once took hours into a matter of minutes.
The breakthrough lies in harnessing graphics processing units (GPUs), processors capable of running thousands of calculations simultaneously, to power physics-based flood models across complex urban landscapes and drainage networks. The result: near real-time forecasts at city scale, a capability that has so far eluded conventional computing approaches.
This is not just about predicting floods; we are forecasting when, how quickly, and who will be impacted, way ahead of time
Critically, the tool was shaped from the outset in close collaboration with the City of Chicago, the Metropolitan Water Reclamation District of Greater Chicago (MWRD) and state emergency management agencies — partners whose involvement reflects a widening mismatch between ageing urban infrastructure and the intensity of today's storms. MWRD President Kari K. Steele highlighted the system's potential to help authorities assess urban flooding, hyperlocal storm, and guide infrastructure investment decisions, and improve the protection of communities and ecosystems.
"Traditional flood models are often too slow for real-time decision-making," said DPI postdoctoral researcher Abhinav Wadhwa, who led the study. "What we've shown is that by leveraging GPU acceleration, you can compress what used to take hours into minutes without losing any physical processes, and with increased accuracy."
The system does not simply speed up existing approaches; it shifts flood forecasting from static risk maps to dynamic, evolving intelligence. "This is not just about predicting floods; we are forecasting when, how quickly, and who will be impacted, way ahead of time," Wadhwa added.
DPI Climate Hub lead Ashish Sharma stressed that today's storms are already outpacing the design limits of urban infrastructure, and that decision-makers urgently need tools that can match that scale and speed. The flood prediction capability will feed into AerisIQ, the team's high-resolution operational forecasting platform, with future integration of AI, machine learning, and large language models planned.
