Development of Satellite-Based Emission Inventories and Indoor Exposure Prediction Models for PM2.5
CitationTang, Chia-Hsi. 2016. Development of Satellite-Based Emission Inventories and Indoor Exposure Prediction Models for PM2.5. Doctoral dissertation, Harvard T.H. Chan School of Public Health.
AbstractEpidemiological studies have documented significant relationships between outdoor particle exposure and adverse health effects whilst indoor residential PM2.5 exposure is found to be the most influential on total PM2.5 exposure as people spend more than 80% of their time indoors. Accurate exposure assessments for ambient and indoor environments are therefore of equal importance. In this dissertation, we aim to develop methodologies that enhance our ability to quantify fine particulate matter (PM2.5) exposure in both macro and micro perspective.
With advanced remote sensing technologies becoming more prevalent and less expensive, there is great potential in employing satellite date to analyze and illustrate ambient air quality in real-time over large geographic areas. In Chapter 1, we introduced a top down approach to construct PM2.5 emission inventory through the integration of mass balance and satellite retrieved daily Aerosol Optical Depth (AOD) at 1km x 1km resolution. The satellite-based inventory provides spatially- and temporally-resolved emission estimates as opposed to the conventional source-oriented inventory that has time lag issues with limited spatial variability to the extent of its data source (usually ground monitoring network). Subsequently in Chapter 2, we quantified the temporal and spatial trends of PM2.5 emission in the North East U.S. using the satellite-based emission inventory. Satellite-based emission trends are in agreement with that of the source-oriented inventories released by the US EPA, showing major reductions achieved in urban areas as well as along important traffic corridors. The technique of this part of the study can be applied to nation-wide or global remote sensing data for estimating PM2.5 emissions and hence improving the quantification of fine particles effects on climate, air quality and human health.
While big data may be the game changer for resolving ambient air quality problems, we still face the challenge of data scarcity in microenvironments. In Chapter 3, prediction models that utilize few available samples to assess indoor PM2.5 exposure were developed. We estimated infiltration rate of ambient particles penetrating indoors using sulfur as the tracer element at 95 residences in the Greater Boston Area. Mixed effects model was employed in order to predict infiltration for individual residences. We then estimate indoor levels of PM2.5 and its black carbon (BC) content using outdoor measurements and the estimated infiltration factor. We cross validated the aforementioned models to evaluate their predictive power specifically at dates without indoor information. Cross validation results of the infiltration model (R2=0.89) indicates that mixed effects captured infiltration rates for individual households adequately. We also found strong predictability when sulfur infiltration surrogate and outdoor measurements of PM2.5 and BC were used in predicting indoor exposure levels (R2= 0.79 [PM2.5], 0.76 [BC]).
Altogether, the methodologies introduced in this dissertation may serve as frameworks to (1) quantify and illustrate ambient emission of PM2.5 or other pollutants in a macro perspective and (2) determine the relationships between outdoor and indoor air quality and to predict indoor air pollution which are critical information for developing solutions of micro-level air quality problems.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:32644532