Computer & Information Sciences & Engineering associate professor Zhe Jiang, Ph.D., appeared on FOX Weather in September to help viewers understand a debate in weather forecasting — whether or not a disturbance in the Atlantic would become the season’s first hurricane.
The issue? Newer AI-based weather models and traditional physics-based models don’t agree.
Dr. Jiang, whose research focuses on AI and data science for coastal hazard forecasting, explained that the two kinds of models reason in fundamentally different ways. Physics-based models compute forward from the equations describing the atmosphere, while AI models learn patterns from decades of historical weather data. Neither can determine which individual storms will spin up a hurricane, and each fills that gap differently.
“We’ve had decades to learn the habits of physics-based models. With AI, we’re still learning,” Dr. Jiang said.
Although both kinds of models have biases, AI takes far less time and energy, allowing researchers to run hundreds or thousands of scenarios for uncertainty quantification.
“For hurricanes, knowing the range of what could happen matters as much as the single most likely answer,” Dr. Jiang said. “That’s a big part of why my group at UF works on AI for coastal flooding and storm surge forecasting.”
This was not Dr. Jiang’s first appearance. He was also on FOX Weather in June explaining the future of hurricane forecasting.