Space-time Modeling for Linking Climate Change, Pollutant Exposure, Built Environment
NIEHS Grant: R01ES014843
This is a joint collaborative effort between North Carolina State University, Duke University, and the University of North Carolina at Chapel Hill. The expertise at the three institutions complements each other and brings synergy. We will achieve the following objectives:
- The development of broad spatial-temporal statistical models to study how air pollution affects human health under climatic change conditions. We will improve upon existing methods by introducing Bayesian multivariate spatiotemporal statistical models that characterize simultaneously complex spatial and temporal dependence structures in environmental stressors, climatic variables, and health outcomes, while taking into account different sources of uncertainty in models and data. We will develop novel spatial quantile regression models for the climatic and pollution variables for better characterization of extremes, tail behavior, and complex dependences between weather and pollution.
- The development of Bayesian hierarchical shrinkage methods for assessing spatial associations between complex pollutant mixtures and health outcomes. We will improve upon existing approaches by simultaneously accounting for different pollutant types, such as ozone and particulate matter (PM) or speciated PM, characterizing the spatial temporal structure of the susceptible periods of fetal development (pregnancy outcomes) and the exposure lag (mortality outcome), while taking into account different sources of uncertainty in models and data.
- We will build neighborhood deprivation and environment indices for linkage to health outcomes. We will use the statistical frameworks above and data on birth weight and gestational age at delivery in the Pregnancy, Infection, and Nutrition (PIN) study, which examines neighborhood factors concerning the built and perceived physical environment in relation to pregnancy outcomes, to bring together GIS capabilities, deterministic models for air pollution, climate and weather, and novel spatial statistical modeling approaches for dimension reduction.
- We will combine the statistical models in aims 1-3 to study the impact of air pollution and extreme weather on human health under projected future climatic conditions. Health data to be examined include the following: U.S. daily mortality in 2001-2006 at the county level (and geocoded at the street level for the states of NC and NY). Birth weight (small-for-gestational age) and gestational age at delivery (preterm birth) in a sample of infants born in 10 U.S. states who participated as controls in the National Birth Defects Prevention Study (NBDPS), for whom geocoded latitude and longitude at delivery are available. Individual-level cardiovascular birth defects geocoded data are available, as well as individual-level geocoded cardiovascular birth defects data for 15,000 cases and controls in NBDPS. We will make this new methodology broadly applicable and disseminated by developing free-access software and conducting extensive validation and diagnostics of our approaches, as well as presenting measures of goodness-of-fit.