Spotlight on AI: UMD Researchers Help Communities Prepare for and Prevent Disasters

When disaster strikes, increased lead time, better detection and smarter tools can mean the difference between reacting to an emergency as it unfolds and anticipating or even preventing one before it happens. 

University of Maryland researchers are using artificial intelligence in bold and creative ways to help protect people, infrastructure and communities from a variety of threats including wildfires, unexploded bombs and massive ocean waves.  

Each Friday this October, we are highlighting how UMD is shaping the future of AI and transforming sectors of society through our research, education and service. This week, we’re focusing on disaster and emergency readiness projects spanning diverse domains—from land to sea, and even inside war zones

• Computer science Professor Heng Huang and his team secured a $1.86 million grant from the National Science Foundation to build advanced AI tools to better detect and forecast wildland fires. These tools combine data from satellites, atmospheric records, vegetation/fuel information and historical fire data to predict when and where fires might spread. The project also plans to release an open‑source dataset to help other researchers, and collaborate with agencies such as the U.S. Forest Service, NASA and the National Park Service so the tools are aligned with real operational needs.
• In the Maryland Robotics Center, UMD engineers and students are working on an ambitious system called RoboScout that combines autonomous drones, robot “dogs” and AI‑powered base stations to survey disaster zones and support first‐responders in real time. The system prioritizes which victims need help first, potentially saving lives and reducing risk to human responders in dangerous scenarios when every second counts.
• Unpredictable, ship-sinking, massive waves—also known as “rogue waves”—can form at sea and threaten vessels and offshore platforms with little to no warning. Researchers from the Department of Mechanical Engineering trained an AI system on billions of ocean wave measurements collected from buoys across the U.S. and Pacific that was able to correctly predict approximately 75% of impending rogue waves one minute ahead, and 73% five minutes ahead. The researchers say incorporating other data such as water depths, wind speeds and buoy locations can improve the accuracy and advance warning time of forecasts and future research could enable the system to predict heights of upcoming rogue waves or times when they’re likely to emerge.
• A team led by Postdoctoral Researcher Zihui “Helen” Ma in the Department of Civil and Environmental Engineering developed a machine‑learning system that analyzes social media posts (for example, tweets about “smoke” or “fire”) in real time to track unfolding wildfires. Researchers collected over 150,000 posts in a one‑month period in wildfire‑prone states and used AI to categorize them into topics such as health impact, damage, evacuation and monitoring. This approach offers a ground‑level “sensor” perspective that can complement satellite and official data—helping emergency responders get earlier hints of trouble, potentially speeding evacuations or targeting resources more effectively.
• Maria Molina, an assistant professor of atmospheric and oceanic science, and Chris Metzler, an assistant professor of computer science, are combining their expertise in AI and the study of the Earth’s systems, known as geoscience, to improve the lives of populations that are disproportionately vulnerable to extreme weather events. Both have joint appointments in the University of Maryland Institute for Advanced Computer Studies. Their work, supported by an award from the National Science Foundation (NSF), will help experts and non-experts alike better grasp uncertainty and inequity in predicting extreme weather and threatening climate conditions.
• Geographical sciences researchers in the College of Behavioral and Social Sciences developed an AI system to identify and map millions of artillery craters in Ukraine to pinpoint hot spots for dangerous unexploded ordnance that could kill or maim civilians for decades if left in place. Their system identified roughly 2.5 million strike craters across a 500‑mile arc of southern and eastern Ukraine, helping demining authorities quickly assess deadly risks across vast regions.
• As part of the XPRIZE Wildfire Competition, students and faculty with xFoundry@UMD developed a drone system that can detect wildfires with AI and then drop a water‑balloon to suppress them—within minutes of ignition. The system uses thermal and optical cameras, AI to detect the fire locations, GPS‑linked systems to drop suppression payloads and student‑engineered mechanisms for balloon release. The aim: locate and suppress a wildfire within 10 minutes, before it grows dangerously large.

To learn more about how the University of Maryland is shaping the future of AI, visit ai.umd.edu.