Date of Award
Summer 2021
Document Type
Thesis - Restricted
Degree Name
Master of Science (MS)
Department
Mechanical Engineering
First Advisor
Roy, Somesh
Second Advisor
McDonald, Walter
Third Advisor
Borg, John
Abstract
The dispersion of pollutants originating from common sources in an urban environment were modeled. Modeling analyses were developed using AERMOD to evaluate PM2.5 emissions from a fireworks event, PM10 and NH3 from urban traffic infrastructure and H2S from a local water reclamation facility under real world conditions. Modeled hourly concentrations were compared with observed concentrations to evaluate model performance.AERMOD is able to successfully capture major trends in hourly concentration profiles when the source contribution is large relative to the background concentration, as demonstrated in the fireworks case. However, in modeling hourly-averaged pollutant concentrations from sources whose hourly contribution is small relative to historical levels, historical pollution or ‘background concentrations’ were shown to be important for satisfactory model performance. In the fireworks case, modeled concentrations were adjusted to account for a measured background concentration which significantly improved model performance and reduced the normalized mean square error of the model to 0.03 sq. µg/m3. In the traffic case, background concentrations were estimated using a gradient boosting model that was trained to predict roadway contributions based on travel activity metrics (traffic speed and traffic volume). While this method improved on the quantitative evaluation metrics from the unadjusted model, the improvements were not sufficient to meet model reliability criteria. Further investigation into this method is recommended for future work. A sensitivity analysis conducted using traffic sources demonstrated that AERMOD exhibits small sensitivity to changes in relative humidity, but is more sensitive to wind direction, particularly with the use of area sources. AERMOD also showed, as expected, a sensitivity to source geometry and possibly to terrain complexity. This sensitivity analysis indicates that model predictions likely reflect the cumulative uncertainties in the model inputs.