Poor ambient air quality represents one of the largest environmental risks to public health. Epidemiological studies, risk assessment, and urban planning require the quantification of long-term personal exposures over a large population. This brings challenges in detailed spatiotemporal mapping and the change of support problem when linking the air pollution concentration to, for example, health outcomes.
The research fellow will firstly focus on the development of statistical and machine learning models for spatiotemporal mapping and the air pollutants NO2 and Ozone. The statistical modeling approach predicts at unknown target spatial locations and times by capturing relationships between response and covariates, modeling the spatial or spatiotemporal processes of the observations and covariates, and learning features from remote sensing images. Spatiotemporal scales are important considerations in air pollution modeling. Specific challenges to be addressed are assimilating data of different sources, modeling spatially heterogeneous response-covariate relationships, uncertainty assessment, and efficient computation. Then, the research fellow will investigate how to relate the predicted air pollution concentrations to personal exposures health outcomes, and socioeconomic variables. Specifically, if we could, and how to optimize a pollution concentration prediction model with different supports of the health outcomes or socioeconomic variables?