Public officials and key decision-makers will be able to make important public safety decisions in real time thanks to a graphics processing unit (GPU)–driven city-scale flood forecasting system developed by Discovery Partners Institute (DPI) Climate Hub.   

GPUs are specialized processors that perform many calculations simultaneously, enabling complex flood simulations to run significantly faster than traditional computing systems. 

The project introduces a breakthrough in urban flood prediction, using GPU-accelerated computing to deliver near real-time forecasts at a city scale rather than across broader regions. This research has the potential to operationalize preparedness for disaster response and infrastructure planning worldwide. 

Urban flooding is driven by extreme rainfall, rapid urbanization, and aging drainage systems. Today, nearly 1.8 billion people live in flood-prone areas globally, with economic losses reaching hundreds of billions annually. DPI’s approach could reduce flood simulation time by up to 80 percent, enabling emergency managers to anticipate impacts as storms unfold rather than after the fact. 

In addition to improving public safety and reducing economic losses, enhanced flood modeling supports environmental protection by reducing contamination from floodwaters, limiting ecosystem damage, and informing more sustainable infrastructure and water management decisions.

This city-scale urban flood forecasting system uses state-of-the art GPUs to process physics-based flood models across complex city landscapes and drainage networks at unprecedented speed.  

“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 study shifts flood forecasting from static risk maps to dynamic, real-time intelligence that enables faster, more informed decisions. “This is not just about predicting floods; we are forecasting when, how quickly, and who will be impacted, way ahead of time,” Wadhwa added.  

The project was shaped through collaboration with stakeholders across multiple levels, including city agencies such as the City of Chicago, regional partners like the Metropolitan Water Reclamation District of Greater Chicago (MWRD), and state agencies. These partnerships highlight the growing gap between legacy infrastructure and the scale of today’s extreme weather. 

Ashish Sharma, who leads the DPI Climate Hub and is the project lead, said, “Cities today are facing storms that exceed the design standard in their infrastructure, and we need operational capabilities that match the scale and speed of these risks, so leaders can act decisively to protect lives, secure infrastructure, and drive economic growth.”   

Sharma’s team is advancing on AerisIQ, a high-resolution weather forecasting technology platform that integrates real-time data with high-performance computing and ongoing advancements in AI/ML. “Our goal is to generate actionable science,” Sharma said. “This peer-reviewed flood prediction capability strengthens AerisIQ as an operational platform for real-time decision-making and serves as a foundation for our next generation of AI/ML-driven decision-support tools, including the integration of large language models (LLMs).” 

The video below, created by the research team to illustrate dynamic flood visualization outputs, demonstrates how next-generation flood forecasting moves beyond static maps to real-time, actionable intelligence.  

Caption: This video shows our GPU-accelerated flood forecasting system with dynamic real-time flood intelligence. DPI’s Climate Hub can predict where flooding will occur, when it will arrive, and when it will recede, 24 to 48 hours ahead of time, enabling faster and more targeted responses. 

“This advanced modeling will help cities and regional authorities like the MWRD assess urban flooding, hyperlocal storms and a growing need for infrastructure investments to contain and convey this volume of stormwater,” said MWRD President Kari K. Steele. “Improved situational awareness enables proactive planning, efficient resource allocation and better protection of communities, infrastructure and our environment. I thank our research partners and staff who contributed to this report and their dedication to building resilience for our region.” 

The study also highlights implications across the full disaster cycle, from preparedness to response to recovery and to mitigation. Clayton Kuetemeyer, Deputy Director of the Office of Emergency Management at the Illinois Emergency Management Agency and Office of Homeland Security, and a coauthor of the study focusing on policy and applications, noted that “the real opportunity is bridging the gap between science and action. When cities and regions have reliable, real-time flood intelligence, it dramatically improves the ability to alert communities and manage response and recovery operations. Integrating these tools can also better inform planning and decision-making for infrastructure investments that mitigate risk and improve community disaster resilience.”  

Overall, this study shows deep academic, civic, and utility partnerships, informed by the action-driven science of flood forecasting technology, providing city, regional, and state officials with situational awareness to prepare more effectively and respond swiftly, with future planned integration of AI and ML enabling more LLM-based targeted warnings and enhanced decision-making capabilities. 

Read the full study in npj Natural Hazards.