Outdoor air pollution is estimated to contribute to more than 100,000 premature deaths in the United States each year, according to the National Weather Service. Accurate air quality forecasts - designed to protect public health, alerting communities to dangerous levels of pollutants linked to asthma attacks, heart disease and premature death - are critical for helping people limit exposure and for guiding regulatory action.
However, a new study led by Fan Wu, a doctoral student in Penn State's Department of Meteorology and Atmospheric Science, suggests some of the computer models that agencies rely on may not be getting it right. Wu and the multi-institutional team found that models used to predict air pollution can seriously misrepresent how heat and moisture move between farmland and the atmosphere, potentially skewing air quality forecasts used for policy decisions.
The study, published in the journal Agricultural and Forest Meteorology, evaluated how well the Weather Research and Forecasting (WRF) model - a widely used regulatory weather model - simulates surface fluxes, which are the exchanges of heat, moisture and momentum between the Earth's surface and the atmosphere. These fluxes directly influence the atmospheric boundary layer, the lowest part of the atmosphere that interacts with the surface and contains the air people breathe. The depth and mixing of that layer are crucial in determining how pollutants such as fine particles (PM 2.5) and ozone build up or disperse.
The team evaluated the Pleim-Xiu Land Surface Model (PX LSM), a land module within WRF used by air quality agencies in California and the Mid-Atlantic region, including Pennsylvania and Maryland. The researchers ran WRF simulations and compared the results to yearlong, real-world measurements collected from 16 flux towers that directly measure heat and moisture exchanges across California's San Joaquin Valley and the Mid-Atlantic.
The team found that the model performs very differently in the two regions.
In California's San Joaquin Valley, the model makes irrigated farm fields appear far too hot and dry. During summer daylight hours, it overestimates the heat flowing from the surface to the air - known as sensible heat flux - by about 260 watts per square meter, or 274%. It also underestimates the cooling effect from evaporation - latent heat flux - by about 200 watts per square meter, or 68%, especially during spring and summer daylight hours. The problem stems from the model's exclusion of irrigation, meaning it does not capture how added water cools and moistens the surface.
"These significant heat flux errors over irrigated fields can distort air quality forecasts," Wu said. "If the model puts too much heat into the atmosphere, it makes the atmospheric boundary layer too deep, giving pollutants in the model more room to dilute. That can lead to underestimates of pollution near the surface, where people breathe."
In the Mid-Atlantic, model errors were smaller and more balanced. The system tended to slightly overestimate both heating and evaporation, running too hot over cities and somewhat too wet over vegetated areas. However, overall, it captured surface-atmosphere exchanges more realistically than in California.
Across both regions, the researchers also found that the model overestimates how strongly the surface slows and stirs the wind during the daytime, with mixed performance at night. The findings point to broader challenges in how the model represents surface-atmosphere interactions. The researchers said the results also suggest that including a representation of irrigation, perhaps by integrating space-based observations of vegetation and soil moisture, could strengthen air quality forecasts in heavily farmed regions.
"If WRF better represented irrigation and land use details, we would expect more accurate simulations of daytime PM2.5 and ozone concentrations in state modeling systems, which could help agencies create more effective plans to reduce pollution," Wu said.
According to Ken Davis, professor of meteorology and atmospheric science and research team member, the next step is determining whether improving how models represent irrigation leads to better air quality forecasts - and whether those improvements are practical for states to adopt.
"We're testing whether tools like NASA's Land Information System or a simpler irrigation module can reduce the surface heat flux errors we identified," Davis said. "First, we need to show that these approaches improve the weather model. Then we need to determine whether states can realistically implement them. If they can, adoption should be straightforward."
Davis added that the team must make sure that improving the meteorology actually improves the air quality simulation.
"Sometimes these complex systems contain compensating errors," Davis said. "If better surface modeling improves both the weather and air quality simulations - and early signs in the San Joaquin Valley suggest it does - then we're headed in the right direction."
In addition to Wu and Davis, Penn State researchers on the project include Jason Horne, a doctoral student in the Department of Meteorology and Atmospheric Science. The team also included Li Zhang, Yu Yan Cui, Zhan Zhao and Chenxia Cai, California Resources Board; Ray Anderson and Sarah Goslee, U.S. Department of Agriculture (USDA); and Min Zhong, Pennsylvania Department of Environmental Protection.
Support for this project was provided by NASA, USDA, U.S. National Institute of Standards and Technology and Penn State.
