Research Highlights

Aerosol-cloud interactions remain one of the largest uncertainties in physical process understanding and long-term projections. A key challenge is disentangling causality in observed aerosol-cloud relationships, as both variables can be independently influenced by large-scale meteorology. To address this, we apply an explainable machine learning (ML) framework to isolate and examine the individual effects of aerosols and meteorological factors (MFs) on cloud liquid water path (LWP), using recent observations from the Eastern Pacific Cloud Aerosol Precipitation Experiment (EPCAPE) field campaign conducted by the Atmospheric Radiation Measurement (ARM) User Facility.

Atmospheric particles act as nuclei for the formation of cloud droplets. To study these nucleation processes, pumped counterflow virtual impactors are often used because they can separate cloud droplets or ice crystals from inactivated particles. In this study, we compared the performances of three commercial units in a laboratory setting using dry aerosols and cloud droplets. We quantified how the transmission efficiency varies with flow rates and particle types, and identified the operating conditions that optimize the collection of activated particles.

While a non-monotonic (“inverted-V”) cloud response to aerosol perturbation—initial cloud thickening via precipitation suppression followed by enhanced evaporative dissipation—has been previously reported, its meteorological conditioning remains unresolved. Using a deep-learning framework to objectively classify Eastern North Atlantic synoptic regimes, we show that the cloud response is strongly regime-dependent and that the U.S. Department of Energy's earth system model (E3SMv2) systematically overestimates liquid water loss, particularly in dynamically complex, precipitating environments with strong vertical motion. These biases are linked to uncertainties in models representing drizzle, entrainment, and turbulent processes.