Missouri S&T researchers use tornado simulator to analyze effects of wind and pressure on buildings

Missouri recorded the third most tornado warnings in the United States in 2025. Researchers at the Missouri S&T Wind Hazard Mitigation Lab are now analyzing damage from the state's most prolific storms to better understand the impact of tornadic winds on buildings.

The 2025 tornado season began on March 14, a night when wildfires and damaging tornadoes occurred simultaneously in Mid-Missouri. One EF-2 tornado struck Rolla, damaging businesses and three schools. One school required evacuation for the rest of the academic year, with total repair costs for school buildings estimated at $14 million.

Dr. Grace Yan, who leads the Wind Hazard Mitigation Lab, was among the first on the scene to survey the damage in Rolla. Following an EF-3 tornado that struck St. Louis in May 2025, Yan's team used drones to fly over the storm path more than 50 times to collect images. This research involves a detailed classification of structural failures to understand how different buildings withstand various levels of force.

"We trained the entire team to rate the damage to each building based on damage, whether it was complete destruction, severe damage, heavy damage from negligible to slight, so everyone knows the damage level," Yan said. Because the manual survey process is time consuming, computer science major Michael Hotra started developing a machine learning program called "You Only Look Once," or "YOLO" to automate the rating process. The software learns specific parameters from storm footage to detect buildings and assign damage ratings.

The team also uses a large tornado simulator to study how wind behaves across different landscapes. Yan noted that the shape of the land significantly changes the intensity of a storm. "Our finding is that compared to the flat region, the presence of terrain will increase the wind speed and decrease the pressure at the center," Yan said.

Students in the lab are also measuring the load placed on buildings by testing how pressure changes as the simulator passes over different roof types. They compare gabled roofs, commonly found on houses, to flat roofs used on warehouses and box stores. Small tubes on the models feed air data directly into a digital acquisition system to determine the worst loading scenarios for structures.

Yan hopes the numerical simulations can eventually help the National Weather Service and the National Oceanic and Atmospheric Administration with historical records and storm ratings.

"The EF-scale only represents the observed damage, it's not as related to wind speed or wind pressure," Yan said. "By doing this type of numerical simulation we hope we can provide this important information for NOAA."