Text

theses

Climate change critically affects both the atmospheric processes involved in the dynamics of air pollution systems and biogenic emissions including tree and grass pollens and fungal spores. Synergistic action of allergenic pollen with air pollutants like ozone and particulate matter has been reported as potentially exacerbating the symptoms of allergies. This dissertation investigated the spatiotemporal distributions predicted for allergenic pollen and ground-level ozone across the contiguous United States (CONUS) in 2004 and 2047 reflecting the Representative Concentration Pathways (RCP) 8.5 scenario, and estimated human exposures to those pollutants. In addition, Machine Learning (ML) methods were evaluated and applied to local-scale prediction of airborne allergenic pollen concentration.
It was estimated that ragweed pollen season will start earlier and last longer in 2047 under the RCP 8.5 scenario across the CONUS, with increasing average pollen concentrations in most regions. The response of the oak pollen season varies across the nine climate regions of the CONUS, with the largest increase in pollen concentration occurring in the Northeast region. The oak pollen season length was estimated to shorten by 1-2 days for most regions, except for the Southeast and Southwest regions.
Analyses of observed ragweed pollen counts and ozone concentrations from 1990 to 2010 indicate that the ragweed pollen season started earlier at 76% of the monitoring stations, and the annual average number of co-occurrence of ragweed and ozone exceedances (daily maximum 8-hour average ozone > 70 ppb) ranged between 0 to 17 days. Co-occurrences of ragweed pollen and ozone exceedances under climate change were investigated based on simulated ragweed pollen and ozone concentrations. Although the co-occurrence of ragweed pollen and ozone exceedances is scattered across the CONUS, it influences a remarkable fraction of the population. Inhalation exposures to ragweed pollen are higher outdoors than indoors, with significant correlation with pollen concentration. Males tend to have higher inhalation exposures to ragweed pollen and ozone than females. The inhalation exposure to ragweed pollen and ozone per unit body weight decreases with age.
Prediction of ragweed pollen concentration at the local scale, based on meteorological factors and previous ragweed pollen observations, was conducted using ML models including Support Vector Machine (SVM), Random Forest, XGBoost, Neural Network, Decision Trees, and a Bayesian Generalized Linear Model. The model parameters were optimized and the final models were evaluated using a repeated 10-fold cross-validation. Random Forest and XGBoost models outperformed other models, and pollen concentration of the previous day is the most important predictor variable for both models.

ETD_10329

Ph.D.

Includes bibliographical references

doi:10.7282/t3-c6hd-ns81

ETD doctoral

The author owns the copyright to this work.