| Particulate matter (PM) suspended in the air is a major public health concern. Particulate air pollution has been found to be associated with health problems, such as asthma and increases in daily mortality. Because of the importance of PM to human health, PM is routinely monitored by environmental and health agencies of the government. The most commonly used model for estimating source contributions in PM is the effective variance (EV) chemical mass balance (CMB) model. The current EV CMB model has drawbacks, including negative estimated source contributions, problems with multi-collinearity of source profiles and biased estimates due to measurement errors. This dissertation introduces the air quality model and proposes two parallel modifications to improve EV CMB: a constrained CMB model and a Bayesian CMB model.; The constrained CMB model uses the effective variance to account for measurement errors of source profiles, but it also imposes inequality constraints on source apportionments.; The Bayesian CMB model is constructed by using a prior distribution for the source contributions that can only take nonnegative values. The proposed Bayesian CMB model also accommodates model selection between different potential sources. |
