One year after hurricanes Helene and Milton caused severe flooding in the Tampa Bay area, a mobile application developed by researchers at the University of South Florida (USF) has become an important tool for monitoring flood conditions. The Community Resiliency Information System (CRIS) HAZARD app, created by a team led by USF St. Petersburg GIS and Remote Sensing Professor Barnali Dixon, uses crowdsourced photos and artificial intelligence to provide real-time updates on flooding.
Residents can upload photos of flooded areas through the app, allowing others in the community to stay informed about local conditions. Dixon described the app as similar to Waze but for floods: “The more accurately we can predict floods, the better we can prepare for them,” Dixon said. “The CRIS-HAZARD app helps both residents and decision-makers by providing up-to-date information.”
Over the past year, 23 static cameras have been installed across Pinellas County in locations known to be prone to flooding. These cameras take images every 15 minutes during flood events and send them directly to the CRIS-HAZARD system. The platform combines these camera images with geolocated photos submitted by users. Machine learning models then analyze these visuals to estimate water depth and measure how extensive the flooding is.
“We can determine the severity of flooding, whether it’s minor, moderate or major,” Dixon said. “By turning people’s real-life experiences into usable data, we can train our AI and machine learning models to make flood predications more accurate.”
The app is available on smartphones and desktops. Users are encouraged to submit pictures showing flood levels with reference points such as ankle or knee height, which helps improve accuracy during storms.
Alec Colarusso, a PhD candidate at USF’s School of Geoscience who assists Dixon, highlighted how important user contributions were during Hurricane Helene: “Users uploaded photos with a known flood height, and it not only helped others understand what was happening in real time, but it also helped validate our models and future predictions,” he said.
After launching on September 18, 2024, the CRIS-HAZARD app was quickly put into use when Hurricane Helene made landfall as a Category 4 storm north of Tampa Bay just eight days later. Less than two weeks after that event, Hurricane Milton struck south of Tampa Bay as a Category 3 hurricane. The back-to-back storms resulted in widespread flooding throughout the region.
Initially there were only eight cameras operating in St. Petersburg during those storms; now there are 31 positioned across Pinellas County communities including Clearwater, Belleair Beach, and Tarpon Springs.
“We worked with the flood plain managers and used historical flood data and damage reports to help determine the best locations for the cameras,” Dixon said.
The placement strategy aims for broad coverage across high-risk zones while helping verify crowdsourced data from residents. When users upload photos of flooding incidents, computer vision identifies nearby camera feeds to analyze water depth; this data is then compared against camera images for greater accuracy.
As new storms occur and more data is collected through community input and automated cameras, CRIS-HAZARD continues to develop its capabilities as a resource for emergency responders and residents alike. By combining artificial intelligence with public participation, it offers faster tracking of floods while supporting emergency responses amid changing weather patterns.
