Why climate models undercount cloud-forming particles in polluted air (Preprint)

Climate models underestimate cloud condensation nuclei (CCN) measured near the surface, especially in polluted regions. Improved emissions and chemistry have not closed this gap. We test a structural cause. Climate models mix each emission at once through a grid box tens of kilometres wide. They therefore miss the weak near-surface wind that keeps fresh emissions concentrated. We used 3D-ICAM, a detailed three-dimensional aerosol model coupled to resolved winds. We ran it at 10 m resolution over 1 km tiles in Delhi, Beijing, London and Hyytiälä. We compared 3D-ICAM with a single 1 km grid box that mixes like a climate model but keeps the detailed aerosol. In stagnant winter air in Delhi and Beijing, the grid box had 40–69% less CCN at 10 m than 3D-ICAM. Yet their column means differed by 2–8%. 3D-ICAM reproduced 52–90% of observed CCN; the grid box 30–54%. Both share the same chemistry, so the gap comes from mixing. In ventilated London and the boreal forest, the models agreed. In stagnant polluted air, surface stations therefore sample a near-surface excess that a grid box cannot hold. Comparing grid boxes with these stations can mistake this mixing error for errors in emissions or chemistry. This is a preprint and has not been peer reviewed. For submission to PNAS.