Xing's Research Uses AI to Improve Weather, Air Quality Forecasting

Jia Xing, a research associate professor in the Department of Civil and Environmental Engineering at the University of Tennessee, Knoxville, has nearly two decades of experience in atmospheric chemistry and physics modeling. His focus is on understanding the sources and distributions of atmospheric chemicals and their impacts on human health, weather, and ecosystems.

Xing developed the machine-learning-based reduced-form atmospheric chemistry model, DeepRSM, which has been adopted and promoted by the US Environmental Protection Agency for global applications in air quality forecasting and management.

Xing's latest research involves advancing the ability to predict weather and air quality by using AI and geoscience to create a new generation of forecasting tools that are faster, more accurate, and more responsive than current numerical model-based systems.

Xing has received National Science Foundation (NSF) funding for a study that aims to develop a machine learning-based surrogate for chemical transport models (CTMs), which are widely used in atmospheric research but often too computationally expensive for real-time forecasting.

He is also the lead principal investigator on a grant from the National Oceanic and Atmospheric Administration (NOAA) to integrate DeepCTM with NOAA's operational National Air Quality Forecast Capability (NAQFC). The integration has the potential to enhance the forecasting of atmospheric chemical concentrations to improve the efficiency, accuracy, and resolution in air quality predictions nationally and globally.

"My advantage is that I not only know the weather, but I also know the chemistry," said Xing, whose expertise developed while studying air pollution in China. "I know the aerosol and also some gasses they might influence. They have given feedback from the weather system. We're trying to implement the chemistry with the weather to improve the performance."

AI-enhanced Weather Models

he proposed AI model for the NSF project will be trained on geoscientific big data, including meteorological inputs, satellite remote sensing, and ground-based pollutant measurements, and will learn to replicate how CTMs change and move over time and space. This will enable real-time forecasting of air pollution and weather patterns across wide regions at a high resolution.

Xing conducted a preliminary study with help from the AI Tennessee program, and discovered there was major potential for his new model.

"Unlike most existing AI models, this framework explicitly incorporates meteorology-chemistry interactions, such as the feedback between air pollutants and weather conditions," Xing said. "This allows the model to provide more realistic and scientifically grounded forecasts."

Being able to more accurately predict the weather isn't just helpful to know how to dress each day or when to plan outdoor activities. With wildfires, heat waves, and floods, the volatility of the weather across the world can have devastating impacts socially and economically.

"It is very important for human health and emergency response systems," Xing said. "Traditionally, the method used is a model with simulation from the very beginning. But they have a very, very big problem with error propagation. That can be very turbulent and hard to predict over a very long time. It's very challenging. AI has potential to help us better predict things."

Hoping to 'Improve Lives'

For the NOAA study, Xing is collaborating with Youhua Tang, a senior researcher at George Mason University. Though the results will directly benefit NOAA and air quality forecasters, the study will also produce a series of scientific documents and publications for sharing with the scientific community to potentially improve atmospheric chemistry forecasting across the world.

"We definitely need other people to help, because it's a very big problem," Xing said. "The atmosphere system is a very huge system. What I've been addressing is always the chemistry part and linking that with the weather. I know there are a lot of experts who are working on weather through their own angle. I really hope to collaborate with other professors in the future to expand the knowledge."

Since he began his work in the field of using AI for air quality and weather prediction, Xing has harbored one overarching goal.

"I really hope it can improve our lives and help more people. That's my dream," he said. "Even though we're still facing the climate change challenge, we can do our best to protect human health and protect our properties, along with advancing our knowledge about science."

Provided by University of Tennessee at Knoxville