Abstract
This dissertation presents a comprehensive investigation into two interconnected aspects of wildland fire impacts: fire behavior quantification and prediction, and the emission and transport of bacteria and fungi in smoke. Using drone-based technology, artificial intelligence, and statistical modeling the research develops and demonstrates new methodological approaches to address critical knowledge gaps in assessing past and future fire behavior and understanding the emission and dispersion of biological aerosols from fires. This work increases fundamental knowledge, advances analytical methods for studying bioaerosol emissions and dispersal, and has direct applications for fire management, human health, ecosystem processes, and climate dynamics. The work establishes integrated frameworks for near-real-time monitoring and prediction of fire behavior while advancing the emerging field of pyroaerobiology—the aerosolization and transport of microbial life by wildland fire.The research design leveraged uncrewed aircraft systems (UAS) for three distinct but complementary investigations. In the first study, UAS equipped with thermal infrared (TIR) sensors were employed to develop a novel workflow for quantifying fire behavior metrics—fire radiative power, fireline intensity, and rate of spread—from repeat-pass imagery. Machine learning (a subset of artificial intelligence) models, specifically artificial neural networks and random forest, were trained on these high-resolution fire behavior data to predict rate of spread with high accuracy. This approach provides fire managers with computationally efficient, data-driven tools for proactive decision-making while enabling researchers to link fire behavior to ecological effects at realistic spatial scales.
The second study presents the first quantification of fungal spore concentrations and emission factors in wildland fire smoke, addressing a fundamental knowledge gap in pyroaerobiology and, more generally, air pollutant research. Using UAS equipped with specialized bioaerosol and carbon sampling payloads deployed across contrasting ecosystems—Kansas tallgrass prairie and Utah subalpine forest—this work revealed that fungal spores are 4–5 times higher in wildland fire smoke than background air. Results also corroborated elevated levels of bacterial cells in fire smoke that have been reported in a small number of previous studies and developed novel statistical frameworks using generalized linear mixed-effects models and Monte Carlo simulations to derive bacteria and fungi emission factors with associated uncertainties. The research demonstrated that fungal spores comprised 99% of estimated bioaerosol mass emissions from a subalpine forest fire and that near-fire concentrations exceed levels known to impede human respiratory functions, with important implications for public health, especially that of wildland firefighters working in smokey conditions.
The third study investigates the spatiotemporal dynamics of smoke bioaerosol diversity during smoke plume transport through simultaneously deploying three UAS to create three-dimensional transects across tallgrass prairie fire smoke plumes. This work revealed that bacterial diversity exhibits a pronounced increase (eightfold for effective number of taxa) in smoke relative to background air followed by decline with transport distance—patterns that closely mirror the behavior of particulate matter concentrations during plume advection. Air quality index (AQI)—a measure of smoke PM2.5 concentrations used for air pollution communications with the public—and relative humidity were both significantly associated with bioaerosol diversity. This study demonstrates novel advanced analysis techniques applied to bioaerosols for the first time. These findings provide crucial insights into the spatial extent of fire’s biological impacts and the mechanisms controlling bioaerosol dispersal.
Collectively, this research relies on UAS-based data acquisition and state-of-the-art analytical methodologies as transformative tools for wildland fire science, enabling integrated investigation of wildland fire smoke, its living inhabitants, and the fire behavior that produces it. The work provides foundational datasets and analytical frameworks that advance our ability to assess wildland fire impacts comprehensively, from immediate fire management needs to long-term ecological and health consequences. The integration of advanced remote sensing, machine learning, and statistical modeling approaches demonstrates the potential for real-time, data-driven wildland fire assessment systems that can inform both operational management decisions and scientific understanding of fire’s complex ecological and atmospheric effects. These contributions are particularly timely given the increasing complexity of wildland fire management in the context of climate change, expanding wildland-urban interfaces, and growing recognition of fire’s role in atmospheric biological processes.