Seasonal aerosol variations at the Land-Ocean boundary: Insights from a global AERONET network analysis

Abstract

Aerosols exert a significant influence on Earth’s climate system via radiative forcing, cloud formation, and air quality. Despite regional variability in their optical properties, coastal boundary aerosols remain poorly characterized, which limits the accuracy of climate assessments. In this study, we develop a hybrid classification framework that combines k-means clustering and a multilayer perceptron neural network to classify coastal aerosols. Using observations from 58 global sites in the Aerosol Robotic Network, we identify four representative coastal aerosol regimes: urban and industrial pollution aerosol, mineral dust aerosol, biomass-burning smoke aerosol, and marine aerosol dominated by sea salt. Our findings reveal strong seasonal dominance in coastal aerosol composition, with mineral dust accounting for up to 75% of the total aerosol burden in summer. Multiwavelength optical properties indicate that the wavelength gradient of aerosol optical depth may decrease from 0.3 to 0.12, highlighting regime-dependent spectral variability. Coarse-mode aerosol optical depth also increases substantially in winter, reaching levels approximately three times those observed in other seasons. Distinguishing coastal aerosol regimes across regions and seasons can improve climate-model evaluation and support evidence-based policies to protect vulnerable coastal ecosystems worldwide.