UC Berkeley

Airborne particulate matter smaller than 2.5 µm in diameter (PM2.5) is among the biggest determinants of disease world-wide. In 2013, exposure to PM2.5 caused an estimated 7 million deaths and 189 million Disability-Adjusted Life Years (DALYs). About half of this calculated burden arises from the indoor residential use of solid fuels – like wood, coal, dung, and crop waste – for cooking, a practice common to about 2.8 billion people. The rest is attributable to ambient concentrations produced by combustion sources like power plants, heating utilities and appliances, and motor vehicles. Despite the ubiquity of the problem, many monitoring, research, and policy efforts employ proxies of total exposure like outdoor ambient concentrations. These proxies are inadequate for the quantification of actual exposures and can prove misleading when used to estimate health effects. More work is needed to progress the use of total exposure – which is often a complex function of a person’s interaction with numerous environments with varying PM2.5 concentrations – as a proper PM2.5 risk metric.

The first research chapter of this dissertation, Chapter 2, estimates changes in total PM2.5 exposure from indoor concentrations (including contributions from second hand smoke), outdoor concentrations, and time-activity in 2014 and in 2024 under alternative emissions policy pathways in one of the most polluted capital cities on Earth: Ulaanbaatar, Mongolia. Ulaanbaatar’s air pollution crisis is seasonal; heavy use of residential coal heating during its harsh and lengthy winters produces some of the worst air pollution in the world, but many cities are more polluted on an annual basis. With this in mind, seasonal exposure patterns are considered separately to produce estimates of annual exposures. These values are combined with projected background disease and population levels and some of the latest available exposure-response functions to project PM2.5-related health impacts. Policy pathways are estimated for business as usual; moderate reductions in heating, power plant, and motor vehicle emissions; and major reductions in the same sectors. The analysis estimates a 2014 population-wide annual average exposure of 59 µg/m3, which increases to 60 µg/m3 in 2024 under business as usual but falls to 32 µg/m3 and 12 µg/m3 under moderate and major emissions control policies, respectively. Annual PM2.5-related deaths and DALYs are estimated at about 1,400 and 40,000, respectively, in 2014. Under business as usual, about 18,000 deaths and 530,000 DALYs are accrued through 2024. Exposure reductions resulting from the moderate control policy pathway avert an estimated 110,000 DALYs and 4,000 deaths from the business as usual pathway between 2014-2024. An estimated 240,000 DALYs and 8,000 deaths are averted under major reduction policies. In all, Chapter 2 highlights the need for aggressive action, especially related to residential heating and tobacco smoking, to avert a growing pollution crisis in Ulaanbaatar.

Chapter 3 presents some of the first personal PM2.5 exposure measurements conducted in rural Lao women cooking primarily with wood. Measurements were taken during a stove intervention program in which traditional open fire and bucket stoves were ostensibly replaced with an ACE-1 fan stove. Average 48-hour concentrations before and after the intervention are reported at 123 µg/m3 and 81 µg/m3, respectively. Measurements of kitchen concentrations, ambient concentrations, and other environmental data are combined with an extensive set of survey responses to reliably model mean 48-hour average PM2.5 exposures before and after the intervention using machine learning, ensemble, and cross-validation techniques (for the full model: r2 = 0.26, predicted mean before intervention = 120 µmg/m3, predicted mean after intervention = 88 µg/m3).

Chapter 4 proposes the use of a household appliance, the smart smoke detector, as a tool for cost-effectively monitoring indoor PM2.5 concentrations, which are often overlooked by regulatory monitoring networks and health effects research. A particularly popular smart smoke detector, the Nest Protect, is reverse engineered. Its onboard optical sensor is co-opted and characterized for the real-time measurement of PM2.5 mass concentrations. Very good agreement is observed between processed Nest Protect signal and output from a co-located research grade monitor, the DustTrak II (r2 > 0.99).

The final chapter, Chapter 5, reiterates the thread common among Chapters 3-5 – advancing PM2.5 risk science through better estimation of total exposures – and discusses key areas for future research.