Charlotte-Mecklenburg Storm Water Services has contracted engineering firm Woolpert to develop a flood forecast system using machine learning. The algorithms will use precipitation predictions to estimate how high county creeks will rise during storm events, allowing emergency managers to get ahead of the floodwaters instead of chasing after them.
Matthew Hornack, who oversees the project at Charlotte-Mecklenburg Storm Water Services, said the program could give them flood predictions 12 hours in advance.
“It’s going to look at our past history of data, and, at any snapshot in time, say ‘ah, when stuff like this has happened in the past this is what we’ve experienced,’” Hornack said.
Storm Water Services already maps inundation to determine where flooding is likely happening. More advanced notice would help the county emergency management office deploy resources, such as fire trucks and barricades, before that inundation occurs. By overlaying inundation mapping with maps of the county’s transportation infrastructure, such as bridges and culverts, Storm Water Services could alert emergency responders when roads and bridges were close to flooding.
“That swift-moving water can easily wash a car or vehicle off the roadway,” said David Kroenig, who manages Mecklenburg County’s flood mitigation and monitoring programs.
During Tropical Storm Debby, Storm Water Services sent about 25 warnings that led to some road closures. For example, emergency responders barricaded West Morehead Street at the Stewart Creek stream crossing in time to prevent cars from crossing dangerous floodwaters.
Kroenig said that the county’s ability to respond during and after flooding has improved, thanks in part to a $1 million grant from the U.S. Department of Homeland Security. Storm Water Services purchased and installed water monitors to supplement existing USGS stream gauges that measure water elevation. Then, the county mapped the space in between this network of gauges using topographic data to determine where flooding is likely happening, which removes some of the guesswork.
In the past, “we relied more heavily on people calling us, saying, ‘I’m surrounded by floodwater or my vehicle’s flooded,” Kroenig said.
During storms like Debby, Storm Water Services teams up with the Mecklenburg County Emergency Management Office in the county’s Emergency Operations Center — picture NASA’s Houston command center, but with about 50 local government employees gathered at the Charlotte Fire Department headquarters, monitoring storms instead of a spaceship.
This collaboration with Storm Water Services helps conserve county resources. Instead of deploying 10 emergency response units to 10 different calls in a similar area, Storm Water Services can identify the same flooded area on their map and tell the Emergency Management Office to dispatch a single unit to that neighborhood. These maps also help in the aftermath — cutting down the time it takes Storm Water Services to follow up with homeowners from a couple of weeks to 24 hours.
When Tropical Storm Debby abated, Storm Services visited about 80 sites, mostly single-family homes. The department wanted to verify that their mapping worked and check on people who had called during the storm. The mapping, they learned, had been accurate, even as the contours of flooding changed.
“We had flooding in areas that we hadn’t really seen it before,” Kroenig said.
Typically, Storm Services expects more flooding in southeast Mecklenburg, around Mint Hill and Matthews. This go-round, Kroenig said the team observed more floods in the northwest, which the maps correctly identified.
The inundation mapping feeds the Charlotte-Mecklenburg Flood Information and Notification System, or FINS — the application that allows anyone to monitor camera feeds from stream gauges, 24-hour rainfall totals and stream elevations.
The new flood prediction algorithms will use rainfall predictions to estimate how high county creeks will rise during storm events.
“It potentially allows you to get your valuables, maybe your vehicles, out of harm’s way,” Kroenig said. “But the home’s still going to flood.”
Kroenig said it would likely be years before the public had access to the predictions via FINS, since Storm Water Services needs to first develop the program, then test its reliability. There are some limitations to the program outside of the model’s control; the predictions are only as good as the rainfall forecasts, which as anyone who has ever tried to plan their day around the weather knows, can be fickle.
Kroenig said that they would like to avoid being the Storm Water Service who cried wolf … or, in this case, flood.
“Even with Debby, the forecast was much more onerous than what we actually saw,” Kroenig said.
