| dc.description.abstract | There is a well-documented association between exposure to fine particulate matter (PM2.5) and numerous health outcomes, with some evidence suggesting PM2.5 originating from coal combustion may have different health impacts. These studies typically estimate exposure to coal-derived PM2.5 based on the presence of certain chemical tracers measured in the air near exposed populations. Interpreting such user-defined source profiles requires a certain degree of subjectivity and approximation, and such approaches do not consider the contributions of individual coal power plants. This limits their relevance for informing air quality management interventions that must, ultimately, be implemented at individual sources (e.g., through scrubber installation, closing inefficient plants, etc.). Existing literature that does focus on specific point sources uses computationally expensive models for pollution transport, thus limiting their applicability to only a few power plants or groups of power plants.
In chapter one, we employ a recently-developed, reduced-complexity air quality model to provide the first national study of the association between long-term exposure to emissions from individual coal power plants and Ischemic Heart Disease hospitalization. The study provides a novel combination of observed data, statistical methods, and tools from environmental engineering. Rooting the approach to causal inference methods to isolate the coal emissions/health relationship represents an important step towards establishing the causal links between emissions and health necessary to drive policy changes.
In chapter two, we provide the first investigation of whether a purely statistical, data-driven approach to source-receptor mapping can reproduce knowledge typically produced by complex chemical transport models. The ability to do so would provide a more computationally nimble approach to estimate S-R relationships in a wider variety of settings. Specifically, we consider daily sulfur dioxide (SO2) emissions from 385 coal-fired power plants operating in the U.S. in 2005, and estimate a source-receptor mapping to 732 EPA Air Quality System (AQS) monitor locations measuring daily fine particulate matter (PM2.5). Results were framed as an ``emissions network'' -- power plants and monitors are nodes and significant associations between their daily time-series define edges in the network -- representing an annual pattern in coal emissions transport for 2005. The results of the proposed approach were shown to hold some promise in capturing general patterns of pollution transport and source-specific exposures, but was limited in its ability to recover individual source-receptor links relative to a recently proposed reduced complexity CTM. Our investigation uncovered several statistical challenges for which we provide initial progress towards addressing, with future refinements holding promise for improving the fidelity of the purely statistical approach.
In chapter three, we explore the value of the statistical, data-driven approach to source-receptor mapping to evaluate how source-receptor relationships vary over time. Specifically, we use the statistical methods to explore seasonal variability (winter, spring, summer, and fall) in coal emissions transport using daily SO2 emissions from coal-fired power plants operating in the United States from 2005-2010 and daily PM2.5 concentrations at air quality monitors. We fit four emissions networks per year (one each season) from 2005-2010 and compared them across seasons and years at various levels of granularity. Our results point to important short-term variability in source-receptor mappings that may not be captured in annual models. | |
