Abstract

In the realm of aerosol classification, South America poses distinctive challenges due to data constraints and the varied methods used, often without consideration of their uncertainties or suitability for the specific problem at hand. This study delves into the impact of uncertainties in aerosol optical properties on widely employed classification methods in the continent. Employing partial derivatives, we propagated uncertainties within 2-D threshold-based classification schemes across 30 sites within the AERONET network. The analysis examines the interplay between total and aerosol-type-specific potential misclassification rates. Our findings underscore pronounced uncertainties in complex schemes, and those reliant on the Angstrom exponent. Additionally, differentiation between maritime and continental aerosols poses significant challenges across all methods, with a potential uncertainty of 63%. The finest performance is particularly evident in tropical dry climates and rainforest environments, where top-performing methods achieve an average uncertainty rate of 21%. In contrast, regions characterized by the mountain grasslands & scrublands biome and coastal locations present formidable challenges, resulting in 76% averaged misclassification. Notably, the fine mode fraction-based scheme excels in coastal and island environments, while aerosol relative optical depth emerges as a valuable metric for classification in mountainous, rainforest, rural, and urban areas. The implications of this study are significant for aerosol classification applications, offering a replicable methodology capable of reducing uncertainties. We recommend further measurements on the continent, as well as the use of our insights to select the most suitable classification method depending on the study site and aerosol conditions.