Pollen classification using a single particle fluorescence spectroscopy technique

Abstract

Most methods of classifying airborne pollen are labor-intensive and prohibitively expensive, which limits the scales at which predictions about ambient allergy conditions can be provided. A spectroscopic technique was developed toward inexpensively providing information about important pollen groups. The technique uses a single-particle fluorescence spectrometer to analyze super-micron atmospheric bioparticles, e.g., pollen and fungal spores, collected onto a substrate. Thirty-four pollen species were sampled from plants at the Denver Botanic Gardens over one year. Particle size and fluorescence emission spectra following excitation from each of four optical sources were collected for each of 916 particles as a first application to freshly collected pollen. Subsequently, a random forest (RF) algorithm was used to classify individual particles, with three key observations. First, categorizing pollen samples into broader categories (sampling month, allergenicity level, or plant type) increased the relative classification accuracy (F value) somewhat, compared to the classification to pollen species. Second, excitation wavelengths most important for classification were identified. Third, comparison of fresh and commercially available pollen of the same species suggests that differences introduced by chemical processing can alter spectroscopic properties and could impact classification.

The scope of analysis and particle statistics was limited, because sample collection and spectral analysis were each conducted manually. While particle numbers were too low to extrapolate results toward continuous analysis, results suggest that rough separation of fresh pollen samples is possible using only fluorescence and particle size. The comparison of results may thus be useful to others developing spectroscopic techniques to analyze bioparticles.

Copyright © 2022 American Association for Aerosol Research

Graphical Abstract

EDITOR:

• Hans Moosmüller

Acknowledgements

The authors acknowledge Prof. Emeritus Donald R. Huffman, University of Arizona, for discussions and technical support; Dr. Cathy Durso, University of Denver, for help with classification and statistical analysis; University of Denver through a PROF grant, Phillipson Graduate Fellowship (BES), and College of Natural Science and Mathematics (JAH) for financial support; Denver Botanic Gardens for help collecting pollen samples, especially Associate Director of Horticulture Cindy Newlander.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 https://oteros.shinyapps.io/pollen_map/

2 https://pollensense.com/

3 https://www.pollenlibrary.com

Additional information

Funding

Financial support from the University of Denver’s PROF grant, the Phillipson Graduate Fellowship (BES), and the College of Natural Sciences and Mathematics (JAH) is acknowledged.