Regime-based aerosol–cloud interactions from CALIPSO-MODIS and the Energy Exascale Earth System Model version 2 (E3SMv2) over the Eastern North Atlantic

Abstract. This study investigates aerosol-cloud interactions in marine boundary layer (MBL) clouds using an advanced deep-learning-driven synoptic-regime-based framework, combining satellite data (CALIPSO vertically resolved aerosol extinction and MODIS cloud properties) with 1° nudged Energy Exascale Earth System Model version 2 (E3SMv2) simulation over the Eastern North Atlantic (ENA; ∼10°×10°, 2006–2014). The E3SMv2 captures observed seasonal variations in cloud droplet number concentrations (Nd) and liquid water path (LWP), though it systematically underestimates Nd. We then partition ENA meteorology into four synoptic regimes (Pre-Trough, Post-Trough, Ridge, Trough) via a deep-learning clustering of ERA5 reanalysis fields, enabling regime-dependent aerosol-cloud interactions analyses. Both satellite and E3SMv2 exhibit an inverted-V LWP-Nd relationship. In Post-Trough and Ridge regimes, the satellite shows stronger negative LWP-Nd sensitivities than in Pre-Trough regime. The Trough regime displays a muted satellite LWP response. In comparison, the model predicts more exaggerated LWP responses across regimes, with LWP increasing too quickly at low Nd and decreasing more sharply at high Nd, especially in Pre-Trough and Trough regimes. These exaggerated model LWP sensitivities may stem from uncertainties in representing drizzle processes, entrainment, and turbulent mixing. As for Nd susceptibility to aerosols, Nd increases with MBL aerosol extinction in both datasets, but the simulated aerosol-cloud interactions appear oversensitive to meteorological conditions. Overall, E3SMv2 better captures aerosol effects under regimes that favor stratiform clouds (Post-Trough, Ridge), but performance deteriorates for regimes with deeper, dynamically complex clouds (Trough), highlighting the need for improved representations of those cloud processes in climate models.