Tracing biological, anthropogenic, and inorganic sources of coarse aerosols via single-particle fluorescence and optical morphology

This is the authors' abstract. We don't add key points for this paper.

Coarse-mode aerosol particles influence the environment, climate, and human health in diverse ways depending on their type. While mineral dust and sea spray aerosol (SSA) dominate this size range, rarer biological particles can have outsized impacts, such as initiating hydrometeor freezing at relatively warm temperatures. Accurate type-specific characterization of coarse-mode aerosol is therefore essential for investigating their roles in climate and the environment. We provide a new reference dataset for fluorescence spectra and morphology characteristics of coarse-mode aerosols from common sources, including pollen, dust, bacteria, and microplastics, measured in controlled experiments with a Multiparameter Bioaerosol Spectrometer (MBS). Comparisons with published datasets revealed consistent source-dependent fluorescence features, but also highlighted similarities between biological and non-biological particles that can bias fluorescence-based classifications. To explore solutions for these confounding similarities, we developed a supervised machine learning classification algorithm integrating fluorescence and morphology information, and evaluated it using MBS and comprehensive chemical tracer observations from Zeppelin Observatory, Svalbard. This evaluation illustrates challenges for inductive methods in distinguishing biomass burning from biological particles, and dust from SSA, suggesting that important particle classes may be missing and/or laboratory-generated aerosols significantly differ from real-world counterparts. We show that domain adaptation using complementary observations can help address these difficulties. Compared to a fluorescence-only approach, the domain-adapted algorithm reproduces the previously published annual bioaerosol cycle while yielding higher summertime concentrations matching those reported from offline analyses. This open-source algorithm provides a basis for quantifying bioaerosols across diverse environments and can be refined with future field and laboratory efforts.