Answers are grounded in this paper and cite their sources.
WRF commonly overestimates near-surface wind speed, which can distort pollutant transport and accumulation in coupled air-quality simulations. We developed an XGBoost model to correct 10 m wind components and tested three ways of introducing the corrected winds into WRF–CMAQ: replacement in WRF output, substitution during MCIP processing, and dynamic assimilation through observation nudging. The correction reduced wind-speed bias and improved the representation of coastal and vertical wind fields. Its effect on air quality, however, depended on the integration pathway. Static replacement produced little change, whereas observation nudging allowed the correction to modify the three-dimensional flow and boundary-layer structure. This improved O 3 and PM 2.5 simulations under strong-transport conditions but intensified existing accumulation biases during stagnant periods. These results show that accurate ML corrections do not necessarily improve air-quality simulations unless they are coupled to the model in a dynamically consistent manner.
No takes yet. Share an insight, caveat, or question.
Kim et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: