Evaluating surface fluxes in WRF using eddy-covariance flux measurements in the Western and Eastern U.S.

Authors: WU, FAN - Pennsylvania State University; DAVIS, KENNETH - Pennsylvania State University; Anderson, Raymond; HORNE, JASON - Pennsylvania State University; Goslee, Sarah; MUNGER, WILLIAM - Harvard University; CAI, CHENXIA - California Air Resources Board; CUI, YUYAN - California Air Resources Board; ZHAO, ZHAN - California Air Resources Board; ZHONG, MIN - Pennsylvania Department Of Environmental Protection

Submitted to: Agricultural and Forest Meteorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/17/2026
Publication Date: 2/3/2026
Citation: Wu, F., Davis, K., Anderson, R.G., Horne, J., Goslee, S.C., Munger, W., Cai, C., Cui, Y., Zhao, Z., Zhong, M. 2026. Evaluating surface fluxes in WRF using eddy-covariance flux measurements in the Western and Eastern U.S. Agricultural and Forest Meteorology. https://doi.org/10.1016/j.agrformet.2026.111029.
DOI: https://doi.org/10.1016/j.agrformet.2026.111029

Interpretive Summary: Meteorological models are commonly used by state environmental agencies to forecast potential environmental events of concern, including smog formation and pollutant fluxes into the atmosphere. Understanding how well these models perform at assessing energy and mass exchanges between the land and atmosphere is critical for improving their performance. In this study, a commonly used model, the Weather Research and Forecasting (WRF) model, was evaluated against eddy covariance observations of land surface fluxes in major agroecosystems in the San Joaquin Valley of California and the Mid-Atlantic region of the Eastern United States. Results indicate that the WRF model shows significant biases in energy fluxes. WRF-eddy covariance differences were particularly notable in the San Joaquin Valley during the irrigation season in spring and summer. The results indicate that irrigation should be parameterized into WRF to improve performance, particularly in the San Joaquin Valley. The results are of interest to air quality districts and consultants who rely on WRF to forecast and model air quality.

Technical Abstract: The simulations of the atmospheric boundary layer (ABL) in weather models are tightly coupled with land surface fluxes, which are highly dependent on surface representation. Land use patterns in the San Joaquin Valley (SJV) of California and the Mid-Atlantic region feature a complex combination of agricultural, urbanized, and forested areas, which poses a challenge to flux simulation with models. In this study, we evaluate the surface flux simulations in the Weather Research and Forecasting (WRF) model that were configured with the physical modules commonly adopted by state regulatory and environmental agencies. Compared to the year-long Eddy-Covariance (EC) flux measurements collected from 16 sites with representative land use, the Pleim-Xiu land surface model (PX LSM) used in WRF shows substantial heat flux biases in both SJV and Mid-Atlantic regions. In the SJV, the model overestimates sensible heat (H) flux by 271 W m-2 (235%) and underestimates latent heat (LE) flux by 211 W m-2 (73%) at irrigated cropland and orchards during spring and summer days. In the Mid-Atlantic region, no sites are irrigated. The model overestimates both daytime H and LE fluxes by 58 W m-2 (44%) and 114 W m-2 (72%) in the spring and summer. Across all sites and times, biases for surface net radiation (Rnet) are small, at an average of -31 W m-2 (7%) during the day and 4 W m-2 (7%) during the night in the SJV and 77 W m-2 (25%) during the day and 5 W m-2 (14%) during the night in the Mid-Atlantic region. In both regions, the model tends to overestimate momentum flux during the day, but shows no clear consistency during the night. Irrigation is the primary source of discrepancies in the model’s simulations of summer, daytime turbulent heat flux simulations in the SJV. Enhancing the representation of irrigation in the WRF model, possibly through satellite remote sensing assimilation, could mitigate model biases, leading to more accurate simulations of surface fluxes and improved ABL development.