Preparing dispersion model surface meteorological inputs using High-resolution Rapid Refresh (HRRR) data

The quality of meteorological input data is essential for air pollution dispersion modeling. Traditionally, dispersion models have relied upon observational meteorological data collected from weather stations. However, the sparse national distribution of weather stations limits model ability in capturing fine-scale meteorological variability, particularly in areas far from any stations. We aimed to replace traditional preparation methods for meteorological inputs with a novel framework using the 3-km resolution High-Resolution Rapid Refresh (HRRR) dataset. We developed several test scenarios to convert the HRRR variables to a format compatible with the Research LINE (R-LINE) source dispersion model. We then predicted traffic-related nitrogen dioxide (NO2) concentrations at 452 Air Quality System monitoring sites across the United States (U.S.) in the year of 2019 using those scenarios. To compare our approach with the traditional method, we also ran R-LINE using observational meteorological data. NO2concentrations predicted by R-LINE with different meteorological inputs (three HRRR scenarios versus observational weather station data) were compared against NO2 measurements using simple linear regression coefficients of determination (R2) and Indices of Agreement (IOA). The HRRR-derived predictions had slightly higher R2 values (averaging 0.29) than that derived from observational meteorological data (0.26). Stratified by distance between NO2sites and weather stations, HRRR-based predictions generally maintained slightly higher R2 values than those based on observational meteorology. Site-specific IOA analyses further showed that HRRR inputs performed better than observational meteorological data across most of the continental U.S. but not as well in urban areas. Because many meteorological parameters required by R-LINE are directly available from HRRR, the use of HRRR can reduce reliance on additional data sources during meteorological input preparation. This framework represents a practical and effective alternative for preparing input data for dispersion modeling, particularly in regions lacking nearby weather stations.Implications: This study presents a novel framework for using high-resolution meteorological data to drive regulatory dispersion models, which has predominately relied on observational meteorology data. It is also the first study to apply 3-km High-Resolution Rapid Refresh (HRRR) meteorological data in dispersion modeling and conduct a nationwide evaluation, providing guidance on where HRRR outperforms observational data for model predictions and what factors could have contributed to the outperformance. This framework advances exposure assessment in environmental health by using spatially resolved meteorological data and promotes methodological innovation in dispersion modeling.