Date of Award


Document Type


Degree Name


Organizational Unit

College of Natual Science and Mathematics, Chemistry and Biochemistry

First Advisor

John A. Huffman, Ph.D.

Second Advisor

Mark Siemens, Ph.D.


Atmospheric, Bioaerosols, Clustering, Instrumentation, Laser-induced fluorescence, Pollen


Atmospheric aerosols are ubiquitous throughout the Earth’s atmosphere and can be important with respect to environmental systems and human health. Pollen particles are a class of primary biological aerosol particles (PBAPs) that cost the United States billions of dollars a year in loss of productivity and healthcare costs due to allergy and respiratory effects. Traditional methods of pollen detection rely on collection and subsequent identification by visual microscopy, yet few measurement stations exist in the United States. As such, current pollen forecasting models have relatively high prediction uncertainty, especially in regions without sampling stations. Recently, laser-induced fluorescence instrumentation has been applied as one method to bridge gaps in bioaerosol detection and classification, though this instrumentation suffers from prohibitively high cost or analysis barriers.

This thesis describes the development, characterization, and preliminary application of a new single-particle fluorescence spectrometer geared towards bioaerosol, particularly pollen, analysis. A sequence of four laser or LED sources are used to excite the particles, which emit fluorescent light that is magnified then diffracted through a transmission grating into a simple digital camera. This instrument operates similar to a traditional spectroscope, though is able to collect spectral light from several small particles simultaneously. This process allows for spectroscopic analysis of many particles at the same time. The instrument went through several phases of both development and characterization. Development included the addition of several new excitation sources (two light-emitting diodes and one laser) to expand the number of fluorophores probed. A monochrome camera was also added to the system to circumvent issues caused by inexpensive point-and-shoot cameras. Methods to size the particles, as well as calibrations for camera parameters and systemic defects were also implemented. For defects in the optical surface and differences in source intensity, a spatial interpolation map was developed that reduces the error of identical particles depending on their location on the CCD from 17% to 3%.

Utilizing these techniques, four clustering and classification methods were examined with 8 species of commercial pollen in Chapter 4. The random forest (RF) and gradient boosting algorithms performed exceptionally well, both classifying above 95% accuracy. The RF technique was examined further due to computational advantages. Testing on source reduction revealed that the 405 and 450 nm sources were less important in classification models, with the latter having particularly low (3%) importance.

The classification techniques were utilized on freshly collected pollen standards in Chapter 5. 34 types of pollen were collected and classified to 90% accuracy at the species level. Pollen was also classified by species, allergenicity, as well as by plant type depending on their collection months, with one scenario being classified at 98% accuracy. A proof-of-concept was also provided for the prediction of new, ambient pollen samples to a developed random forest classification model from standard collections, in which several particles collected in a central location of the Botanic Gardens were classified as a type of tree that was seen to be pollinating on the same day.

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Rights Holder

Benjamin E Swanson


Received from ProQuest

File Format




File Size

226 p.



Included in

Chemistry Commons