Data Integration Methods for Environmental Exposures with Applications to Air Pollution and Asthma Morbidity

项目总结 准确和可靠的暴露估计对任何环境健康研究的成功都至关重要。这个 该项目的总体目标是开发和应用统计方法来改进暴露评估。 以及时空环境污染fi的暴露不确定性量化fi。这就完成了 通过将观测数据与其他数据源(包括最先进的计算机模型)进行统计整合 模拟和卫星图像。我们将根据目前在空气中的三个研究重点来开发方法 污染流行病学：a)确定暴露于空气污染风险最大的易受污染的亚人群； 评估气候变化下空气污染对健康的影响；及(C)了解空气污染的来源 制定控制策略。在目标1中，我们将开发多分辨率和多变量数据集成方法 环境空气污染浓度。我们将对来自监测网络的稀疏观测数据进行补充 来自化学品输送模式和多个卫星反演参数的模拟。建议的方法 将利用污染物之间的相关性和每个污染物内部的时空自相关性来 更好的预测。在目标2中，我们将开发用于气候模式模拟的多变量偏差校正方法 使用历史观察。目标是对多个变量执行联合偏差校正，以便 观察到的依赖关系将保留在未来的预测中。在目标3中，我们将开发集合信源分解 fi-Ne颗粒物污染(PM2.5)方法。这些方法将估计排放源的贡献。 通过组合几种算法的结果，这些算法结合了不同类型的外部信息和假设- 特兹。我们将进一步利用计算机模型模拟将源信息空间内插到位置 没有监视器。根据目标1、2和3开发的方法将用于创建每日(1)的国家数据库 PM2.5标准污染物和主要成分的浓度估计；(2)臭氧水平预测 由于未来不同排放情景下的气候变化，以及(3)来自 PM2.5的多种来源，包括燃煤、道路柴油和汽油燃烧、生物质燃烧、 并重新悬浮土壤/灰尘。我们还将提供不确定性估计、详细文档和R包 以确保这些方法和估计可用于其他环境健康研究。在目标4中，我们将 获取2005-2014年间25个城市的个人急诊科就诊数据。这个 数据集成产品将用于评估哮喘、急诊科就诊和多次 空气污染物和污染源。拟议的健康研究fi通过同时考虑老年人和 非老年易感人群支持有针对性、有效地降低风险和预防- 祭祀活动。虽然空气污染是这个项目的激励应用，但所提出的方法是 高度适用于其他环境暴露。

PROJECT SUMMARY Accurate and reliable exposure estimates are crucial to the success of any environmental health study. The overarching goal of this project is to develop and apply statistical methods to improve exposure assessment and exposure uncertainty quantification for spatio-temporal environmental pollution fields. This is accomplished by statistically integrating observations with additional data sources, including state-of-the-art computer model simulations and satellite imagery. We will develop methods motivated by three current research priorities in air pollution epidemiology: a) identifying susceptible sub-populations most at risk to air pollution exposures; (b) quan- tifying health impacts of air pollution under a changing climate; and (c) understanding sources of air pollution to develop control strategies. In Aim 1, we will develop multi-resolutional and multivariate data integration methods for ambient air pollution concentrations. We will supplement sparse observations from monitoring networks with simulations from a chemical transport model and multiple satellite retrieval parameters. The proposed methods will exploit the between-pollutant dependence and the spatio-temporal autocorrelation within each pollutant for better predictions. In Aim 2, we will develop multivariate bias-correction methods for climate model simulations using historical observations. The goal is to perform joint bias-correction across multiple variables such that the observed dependence is retained in future projections. In Aim 3, we will develop ensemble source apportionment methods for fine particulate matter pollution (PM2.5). The methods will estimate emission source contributions by combining results from several algorithms that incorporate different types of external information and assump- tions. We will further utilize computer model simulations to spatially interpolate source information to locations without monitors. Methods developed from Aims 1, 2, and 3 will be used to create national databases of (1) daily concentration estimates for criteria pollutants and major constituents of PM2.5, (2) projections of ozone levels due to climate change under different future emission scenarios, and (3) daily estimates of contributions from multiple PM2.5 sources, including coal combustion, on-road diesel and gasoline combustion, biomass burning, and resuspended soil/dust. We will also provide uncertainty estimates, detailed documentation, and R packages to ensure these methods and estimates can be used in other environmental health studies. In Aim 4, we will acquire individual-level emergency department (ED) visit data from 25 cities during the period 2005-2014. The data integration products will be used to estimate short-term associations between asthma ED visits and multiple air pollutants and pollutant sources. The proposed health study fills a major gap by considering both elderly and non-elderly susceptible populations to support the development of targeted, effective risk reduction and preven- tion activities. While air pollution serves as the motivating application in this project, the methods proposed are highly applicable to other environmental exposures.