Spatiotemporal prediction of fine particulate matter using high-resolution satellite images in the Southeastern US 2003-2011

Is model-estimated PM2.5 exposure equivalent to station-observed in mortality risk assessment? A literature review and meta-analysis

Model-estimated air pollution exposure assessments have been extensively employed in the evaluation of health risks associated with air pollution. However, few studies synthetically evaluate the reliability of model-estimated PM2.5 products in health risk assessment by comparing them with ground-based monitoring station air quality data. In response to this gap, we undertook a meticulously structured systematic review and meta-analysis. Our objective was to aggregate existing comparative studies to ascertain the disparity in mortality effect estimates derived from model-estimated ambient PM2.5 exposure versus those based on monitoring station-observed PM2.5 exposure. We conducted searches across multiple databases, namely PubMed, Scopus, and Web of Science, using predefined keywords. Ultimately, ten studies were included in the review. Of these, seven investigated long-term annual exposure, while the remaining three studies focused on short-term daily PM2.5 exposure. Despite variances in the estimated Exposure-Response (E-R) associations, most studies revealed positive associations between ambient PM2.5 exposure and all-cause and cardiovascular mortality, irrespective of the exposure being estimated through models or observed at monitoring stations. Our meta-analysis revealed that all-cause mortality risk associated with model-estimated PM2.5 exposure was in line with that derived from station-observed sources. The pooled Relative Risk (RR) was 1.083 (95% CI: 1.047, 1.119) for model-estimated exposure, and 1.089 (95% CI: 1.054, 1.125) for station-observed sources (p = 0.795). In conclusion, most model-estimated air pollution products have demonstrated consistency in estimating mortality risk compared to data from monitoring stations. However, only a limited number of studies have undertaken such comparative analyses, underscoring the necessity for more comprehensive investigations to validate the reliability of these model-estimated exposure in mortality risk assessment.

Prediction and model evaluation for space-time data

Evaluation metrics for prediction error, model selection and model averaging on space-time data are understudied and poorly understood. The absence of independent replication makes prediction ambiguous as a concept and renders evaluation procedures developed for independent data inappropriate for most space-time prediction problems. Motivated by air pollution data collected during California wildfires in 2008, this manuscript attempts a formalization of the true prediction error associated with spatial interpolation. We investigate a variety of cross-validation (CV) procedures employing both simulations and case studies to provide insight into the nature of the estimand targeted by alternative data partition strategies. Consistent with recent best practice, we find that location-based cross-validation is appropriate for estimating spatial interpolation error as in our analysis of the California wildfire data. Interestingly, commonly held notions of bias-variance trade-off of CV fold size do not trivially apply to dependent data, and we recommend leave-one-location-out (LOLO) CV as the preferred prediction error metric for spatial interpolation.

Methodological Challenges in Spatial and Contextual Exposome-Health Studies

The concept of the exposome encompasses the totality of exposures from a variety of external and internal sources across an individual's life course. The wealth of existing spatial and contextual data makes it appealing to characterize individuals' external exposome to advance our understanding of environmental determinants of health. However, the spatial and contextual exposome is very different from other exposome factors measured at the individual-level as spatial and contextual exposome data are more heterogenous with unique correlation structures and various spatiotemporal scales. These distinctive characteristics lead to multiple unique methodological challenges across different stages of a study. This article provides a review of the existing resources, methods, and tools in the new and developing field for spatial and contextual exposome-health studies focusing on four areas: (1) data engineering, (2) spatiotemporal data linkage, (3) statistical methods for exposome-health association studies, and (4) machine- and deep-learning methods to use spatial and contextual exposome data for disease prediction. A critical analysis of the methodological challenges involved in each of these areas is performed to identify knowledge gaps and address future research needs.

Using satellite data on remote transportation of air pollutants for PM2.5 prediction in northern Taiwan

Accurate PM2.5 prediction is part of the fight against air pollution that helps governments to manage environmental policy. Satellite Remote sensing aerosol optical depth (AOD) processed by The Multi-Angle Implementation of Atmospheric Correlation (MAIAC) algorithm allows us to observe the transportation of remote pollutants between regions. The paper proposes a composite neural network model, the Remote Transported Pollutants (RTP) model, for such long-range pollutant transportation that predicts more accurate local PM2.5 concentrations given such satellite data. The proposed RTP model integrates several deep learning components and learns from the heterogeneous features of various domains. We also detected remote transportation pollution events (RTPEs) at two reference sites from the AOD data. Extensive experiments using real-world data show that the proposed RTP model outperforms the base model that does not account for RTPEs by 17%-30%, 23%-26% and 18%-22% and state-of-the-art models that account for RTPEs by 12%-22%, 12%-14%, and 10%-11% at +4h to +24h, +28h to +48 hours, and +52h to +72h hours respectively.

A review on the application of the exposome paradigm to unveil the environmental determinants of age-related diseases

Age-related diseases account for almost half of all diseases among adults worldwide, and their incidence is substantially affected by the exposome, which is the sum of all exogenous and endogenous environmental exposures and the human body’s response to these exposures throughout the entire lifespan. Herein, we perform a comprehensive review of the epidemiological literature to determine the key elements of the exposome that affect the development of age-related diseases and the roles of aging hallmarks in this process. We find that most exposure assessments in previous aging studies have used a reductionist approach, whereby the effect of only a single environmental factor or a specific class of environmental factors on the development of age-related diseases has been examined. As such, there is a lack of a holistic and unbiased understanding of the effect of multiple environmental factors on the development of age-related diseases. To address this, we propose several research strategies based on an exposomic framework that could advance our understanding—in particular, from a mechanistic perspective—of how environmental factors affect the development of age-related diseases. We discuss the statistical methods and other methods that have been used in exposome-wide association studies, with a particular focus on multiomics technologies. We also address future challenges and opportunities in the realm of multidisciplinary approaches and genome–exposome epidemiology. Furthermore, we provide perspectives on precise public health services for vulnerable populations, public communications, the integration of risk exposure information, and the bench-to-bedside translation of research on age-related diseases.

Influence of Spatial Resolution on Satellite-Based PM2.5 Estimation: Implications for Health Assessment

Satellite-based PM2.5 estimation has been widely used to assess health impact associated with PM2.5 exposure and might be affected by spatial resolutions of satellite input data, e.g., aerosol optical depth (AOD). Here, based on Multi-Angle Implementation of Atmospheric Correction (MAIAC) AOD in 2020 over the Yangtze River Delta (YRD) and three PM2.5 retrieval models, i.e., the mixed effects model (ME), the land-use regression model (LUR) and the Random Forest model (RF), we compare these model performances at different spatial resolutions (1, 3, 5 and 10 km). The PM2.5 estimations are further used to investigate the impact of spatial resolution on health assessment. Our cross-validated results show that the model performance is not sensitive to spatial resolution change for the ME and LUR models. By contrast, the RF model can create a more accurate PM2.5 prediction with a finer AOD spatial resolution. Additionally, we find that annual population-weighted mean (PWM) PM2.5 concentration and attributable mortality strongly depend on spatial resolution, with larger values estimated from coarser resolution. Specifically, compared to PWM PM2.5 at 1 km resolution, the estimation at 10 km resolution increases by 7.8%, 22.9%, and 9.7% for ME, LUR, and RF models, respectively. The corresponding increases in mortality are 7.3%, 18.3%, and 8.4%. Our results also show that PWM PM2.5 at 10 km resolution from the three models fails to meet the national air quality standard, whereas the estimations at 1, 3 and 5 km resolutions generally meet the standard. These findings suggest that satellite-based health assessment should consider the spatial resolution effect.

New Homogeneous Spatial Areas Identified Using Case-Crossover Spatial Lag Grid Differences between Aerosol Optical Depth-PM2.5 and Respiratory-Cardiovascular Emergency Department Visits and Hospitalizations

Optimal use of Hierarchical Bayesian Model (HBM)-assembled aerosol optical depth (AOD)-PM2.5 fused surfaces in epidemiologic studies requires homogeneous temporal and spatial fused surfaces. No analytical method is available to evaluate spatial heterogeneity. The temporal case-crossover design was modified to assess the spatial association between four experimental AOD-PM2.5 fused surfaces and four respiratory–cardiovascular hospital events in 12 km2 grids. The maximum number of adjacent lag grids with significant odds ratios (ORs) identified homogeneous spatial areas (HOSAs). The largest HOSA included five grids (lag grids 04; 720 km2) and the smallest HOSA contained two grids (lag grids 01; 288 km2). Emergency department asthma and inpatient asthma, myocardial infarction, and heart failure ORs were significantly higher in rural grids without air monitors than in urban grids with air monitors at lag grids 0, 1, and 01. Rural grids had higher AOD-PM2.5 concentration levels, population density, and poverty percentages than urban grids. Warm season ORs were significantly higher than cold season ORs for all health outcomes at lag grids 0, 1, 01, and 04. The possibility of elevated fine and ultrafine PM and other demographic and environmental risk factors synergistically contributing to elevated respiratory–cardiovascular chronic diseases in persons residing in rural areas was discussed.

Mortality Burden due to Exposure to Outdoor Fine Particulate Matter in Hanoi, Vietnam: Health Impact Assessment

Objective: This study reports the mortality burden due to PM_2.5 exposure among adults (age >25) living in Hanoi in 2017.

Methods: We applied a health impact assessment methodology with the global exposure mortality model and a PM_2.5 map with 3 × 3 km resolution derived from multiple data sources.

Results: The annual average PM_2.5 concentration for each grid ranged from 22.1 to 37.2 µg/m³. The district average concentration values ranged from 26.9 to 37.2 µg/m³, which means that none of the 30 districts had annual average values below the Vietnam Ambient National Standard of 25 µg/m³. Using the Vietnam Ambient National Standard as the reference standard, we estimated that 2,696 deaths (95% CI: 2,225 to 3,158) per year were attributable to exposure to elevated PM_2.5 concentrations in Hanoi. Using the Interim Target 4 value of 10 µg/m³ as the reference standard, the number of excess deaths attributable to elevated PM_2.5 exposure was 4,760 (95% CI: 3,958–5,534).

Conclusion: A significant proportion of deaths in Hanoi could be avoided by reducing air pollution concentrations to a level consistent with the Vietnam Ambient National Standard.

Spatiotemporal high-resolution prediction and mapping: methodology and application to dengue disease

Dengue disease has become a major public health problem. Accurate and precise identification, prediction and mapping of high-risk areas are crucial elements of an effective and efficient early warning system in countering the spread of dengue disease. In this paper, we present the fusion area-cell spatiotemporal generalized geoadditive-Gaussian Markov random field (FGG-GMRF) framework for joint estimation of an area-cell model, involving temporally varying coefficients, spatially and temporally structured and unstructured random effects, and spatiotemporal interaction of the random effects. The spatiotemporal Gaussian field is applied to determine the unobserved relative risk at cell level. It is transformed to a Gaussian Markov random field using the finite element method and the linear stochastic partial differential equation approach to solve the "big n" problem. Sub-area relative risk estimates are obtained as block averages of the cell outcomes within each sub-area boundary. The FGG-GMRF model is estimated by applying Bayesian Integrated Nested Laplace Approximation. In the application to Bandung city, Indonesia, we combine low-resolution area level (district) spatiotemporal data on population at risk and incidence and high-resolution cell level data on weather variables to obtain predictions of relative risk at subdistrict level. The predicted dengue relative risk at subdistrict level suggests significant fine-scale heterogeneities which are not apparent when examining the area level. The relative risk varies considerably across subdistricts and time, with the latter showing an increase in the period January-July and a decrease in the period August-December.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10109-021-00368-0.

Estimating monthly PM2.5 concentrations from satellite remote sensing data, meteorological variables, and land use data using ensemble statistical modeling and a random forest approach

Fine particulate matter (PM_2.5) is associated with various adverse health outcomes and poses serious concerns for public health. However, ground monitoring stations for PM_2.5 measurements are mostly installed in population-dense or urban areas. Thus, satellite retrieved aerosol optical depth (AOD) data, which provide spatial and temporal surrogates of exposure, have become an important tool for PM_2.5 estimates in a study area. In this study, we used AOD estimates of surface PM_2.5 together with meteorological and land use variables to estimate monthly PM_2.5 concentrations at a spatial resolution of 3 km² over Taiwan Island from 2015 to 2019. An ensemble two-stage estimation procedure was proposed, with a generalized additive model (GAM) for temporal-trend removal in the first stage and a random forest model used to assess residual spatiotemporal variations in the second stage. We obtained a model-fitting R² of 0.98 with a root mean square error (RMSE) of 1.40 μg/m³. The leave-one-out cross-validation (LOOCV) R² with seasonal stratification was 0.82, and the RMSE was 3.85 μg/m³, whereas the R² and RMSE obtained by using the pure random forest approach produced R² and RMSE values of 0.74 and 4.60 μg/m³, respectively. The results indicated that the ensemble modeling approach had a higher predictive ability than the pure machine learning method and could provide reliable PM_2.5 estimates over the entire island, which has complex terrain in terms of land use and topography.

Encoder-Decoder Full Residual Deep Networks for Robust Regression and Spatiotemporal Estimation

Although increasing hidden layers can improve the ability of a neural network in modeling complex nonlinear relationships, deep layers may result in degradation of accuracy due to the problem of vanishing gradient. Accuracy degradation limits the applications of deep neural networks to predict continuous variables with a small sample size and/or weak or little invariance to translations. Inspired by residual convolutional neural network in computer vision, we developed an encoder-decoder full residual deep network to robustly regress and predict complex spatiotemporal variables. We embedded full shortcuts from each encoding layer to its corresponding decoding layer in a systematic encoder-decoder architecture for efficient residual mapping and error signal propagation. We demonstrated, theoretically and experimentally, that the proposed network structure with full residual connections can successfully boost the backpropagation of signals and improve learning outcomes. This novel method has been extensively evaluated and compared with four commonly used methods (i.e., plain neural network, cascaded residual autoencoder, generalized additive model, and XGBoost) across different testing cases for continuous variable predictions. For model evaluation, we focused on spatiotemporal imputation of satellite aerosol optical depth with massive nonrandomness missingness and spatiotemporal estimation of atmospheric fine particulate matter [Formula: see text] (PM2.5). Compared with the other approaches, our method achieved the state-of-the-art accuracy, had less bias in predicting extreme values, and generated more realistic spatial surfaces. This encoder-decoder full residual deep network can be an efficient and powerful tool in a variety of applications that involve complex nonlinear relationships of continuous variables, varying sample sizes, and spatiotemporal data with weak or little invariance to translation.

Particulate matter exposure predicts residence in high-risk areas for community acquired pneumonia among hospitalized children

Particulate matter exposure is a risk factor for lower respiratory tract infection in children. Here, we investigated the geospatial patterns of community-acquired pneumonia and the impact of PM_2.5 (particulate matter with an aerodynamic diameter ≤2.5 µm) on geospatial variability of pneumonia in children. We performed a retrospective analysis of prospectively collected population-based surveillance study data of community-acquired pneumonia hospitalizations among children <18 years residing in the Memphis metropolitan area, who were enrolled in the Centers for Disease Control and Prevention sponsored Etiology of Pneumonia in the Community (EPIC) study from January 2010 to June 2012. The outcome measure, residence in high- and low-risk areas for community-acquired pneumonia, was determined by calculating pneumonia incidence rates and performing cluster analysis to identify areas with higher/lower than expected rates of community-acquired pneumonia for the population at risk. High PM_2.5 was defined as exposure to PM_2.5 concentrations greater than the mean value (>10.75 μg/m³), and low PM_2.5 is defined as exposure to PM_2.5 concentrations less than or equal to the mean value (≤10.75 μg/m³). We also assessed the effects of age, sex, race/ethnicity, history of wheezing, insurance type, tobacco smoke exposure, bacterial etiology, and viral etiology of infection. Of 810 (96.1%) subjects with radiographic community-acquired pneumonia, who resided in the Memphis metropolitan area and had addresses which were successfully geocoded (Supplementary Figure F2), 220 (27.2%) patients were identified to be from high- (n = 126) or low-risk (n = 94) community-acquired pneumonia areas. Community-acquired pneumonia in Memphis metropolitan area had a non-homogenous geospatial pattern. PM_2.5 was associated with residence in high-risk areas for community-acquired pneumonia. In addition, children with private insurance and bacterial, as opposed to viral, etiology of infection had a decreased risk of residence in a high-risk area for community-acquired pneumonia. The results from this paper suggest that environmental exposures as well as social risk factors are associated with childhood pneumonia.

Satellite-based ground PM2.5 estimation using a gradient boosting decision tree

Fine particulate matter with an aerodynamic diameter less than 2.5 μm (PM_2.5) is one of the major air pollutants risks to human health worldwide. Satellite-based aerosol optical depth (AOD) products are an effective metric for acquiring PM_2.5 information, featuring broad coverage and high resolution, which compensate for the sparse and uneven distribution of existing monitoring stations. In this study, a gradient boosting decision tree (GBDT) model for estimating ground PM_2.5 concentration directly from AOD products across China in 2017, integrating human activities and various natural variables was proposed. The GBDT model performed well in estimating temporal variability and spatial contrasts in daily PM_2.5 concentrations, with relatively high fitted model (10-fold cross-validation) coefficients of determination of 0.98 (0.81), low root mean square errors of 3.82 (11.57) μg/m³, and mean absolute error of 1.44 (7.45) μg/m³. Seasonal examinations revealed that summer had the cleanest air with the highest estimation accuracies, whereas winter had the most polluted air with the lowest estimation accuracies. The model successfully captured the PM_2.5 distribution pattern across China in 2017, showing high levels in southwest Xinjiang, the North China Plain, and the Sichuan Basin, especially in winter. Compared with other models, the GBDT model showed the highest performance in the estimation of PM_2.5 with a 3-km resolution. This algorithm can be adopted to improve the accuracy of PM_2.5 estimation with higher spatial resolution, especially in summer. In general, this study provided a potential method of improving the accuracy of satellite-based ground PM_2.5 estimation.

Imputing Satellite-Derived Aerosol Optical Depth Using a Multi-Resolution Spatial Model and Random Forest for PM2.5 Prediction

A task for environmental health research is to produce complete pollution exposure maps despite limited monitoring data. Satellite-derived aerosol optical depth (AOD) is frequently used as a predictor in various models to improve PM2.5 estimation, despite significant gaps in coverage. We analyze PM2.5 and AOD from July 2011 in the contiguous United States. We examine two methods to aid in gap-filling AOD: (1) lattice kriging, a spatial statistical method adapted to handle large amounts data, and (2) random forest, a tree-based machine learning method. First, we evaluate each model’s performance in the spatial prediction of AOD, and we additionally consider ensemble methods for combining the predictors. In order to accurately assess the predictive performance of these methods, we construct spatially clustered holdouts to mimic the observed patterns of missing data. Finally, we assess whether gap-filling AOD through one of the proposed ensemble methods can improve prediction of PM2.5 in a random forest model. Our results suggest that ensemble methods of combining lattice kriging and random forest can improve AOD gap-filling. Based on summary metrics of performance, PM2.5 predictions based on random forest models were largely similar regardless of the inclusion of gap-filled AOD, but there was some variability in daily model predictions.

Ensemble-based deep learning for estimating PM2.5 over California with multisource big data including wildfire smoke

INTRODUCTION

Estimating PM2.5 concentrations and their prediction uncertainties at a high spatiotemporal resolution is important for air pollution health effect studies. This is particularly challenging for California, which has high variability in natural (e.g, wildfires, dust) and anthropogenic emissions, meteorology, topography (e.g. desert surfaces, mountains, snow cover) and land use.

METHODS

Using ensemble-based deep learning with big data fused from multiple sources we developed a PM2.5 prediction model with uncertainty estimates at a high spatial (1 km × 1 km) and temporal (weekly) resolution for a 10-year time span (2008-2017). We leveraged autoencoder-based full residual deep networks to model complex nonlinear interrelationships among PM2.5 emission, transport and dispersion factors and other influential features. These included remote sensing data (MAIAC aerosol optical depth (AOD), normalized difference vegetation index, impervious surface), MERRA-2 GMI Replay Simulation (M2GMI) output, wildfire smoke plume dispersion, meteorology, land cover, traffic, elevation, and spatiotemporal trends (geo-coordinates, temporal basis functions, time index). As one of the primary predictors of interest with substantial missing data in California related to bright surfaces, cloud cover and other known interferences, missing MAIAC AOD observations were imputed and adjusted for relative humidity and vertical distribution. Wildfire smoke contribution to PM2.5 was also calculated through HYSPLIT dispersion modeling of smoke emissions derived from MODIS fire radiative power using the Fire Energetics and Emissions Research version 1.0 model.

RESULTS

Ensemble deep learning to predict PM2.5 achieved an overall mean training RMSE of 1.54 μg/m3 (R2: 0.94) and test RMSE of 2.29 μg/m3 (R2: 0.87). The top predictors included M2GMI carbon monoxide mixing ratio in the bottom layer, temporal basis functions, spatial location, air temperature, MAIAC AOD, and PM2.5 sea salt mass concentration. In an independent test using three long-term AQS sites and one short-term non-AQS site, our model achieved a high correlation (>0.8) and a low RMSE (<3 μg/m3). Statewide predictions indicated that our model can capture the spatial distribution and temporal peaks in wildfire-related PM2.5. The coefficient of variation indicated highest uncertainty over deciduous and mixed forests and open water land covers.

CONCLUSION

Our method can be generalized to other regions, including those having a mix of major urban areas, deserts, intensive smoke events, snow cover and complex terrains, where PM2.5 has previously been challenging to predict. Prediction uncertainty estimates can also inform further model development and measurement error evaluations in exposure and health studies.

Novel Approaches to Air Pollution Exposure and Clinical Outcomes Assessment in Environmental Health Studies

An accurate assessment of pollutants’ exposure and precise evaluation of the clinical outcomes pose two major challenges to the contemporary environmental health research. The common methods for exposure assessment are based on residential addresses and are prone to many biases. Pollution levels are defined based on monitoring stations that are sparsely distributed and frequently distanced far from residential addresses. In addition, the degree of an association between outdoor and indoor air pollution levels is not fully elucidated, making the exposure assessment all the more inaccurate. Clinical outcomes’ assessment, on the other hand, mostly relies on the access to medical records from hospital admissions and outpatients’ visits in clinics. This method differentiates by health care seeking behavior and is therefore, problematic in evaluation of an onset, duration, and severity of an outcome. In the current paper, we review a number of novel solutions aimed to mitigate the aforementioned biases. First, a hybrid satellite-based modeling approach provides daily continuous spatiotemporal estimations with improved spatial resolution of 1 × 1 km2 and 200 × 200 m2 grid, and thus allows a more accurate exposure assessment. Utilizing low-cost air pollution sensors allowing a direct measurement of indoor air pollution levels can further validate these models. Furthermore, the real temporal-spatial activity can be assessed by GPS tracking devices within the individuals’ smartphones. A widespread use of smart devices can help with obtaining objective measurements of some of the clinical outcomes such as vital signs and glucose levels. Finally, human biomonitoring can be efficiently done at a population level, providing accurate estimates of in-vivo absorbed pollutants and allowing for the evaluation of body responses, by biomarkers examination. We suggest that the adoption of these novel methods will change the research paradigm heavily relying on ecological methodology and support development of the new clinical practices preventing adverse environmental effects on human health.

Methods, availability, and applications of PM2.5 exposure estimates derived from ground measurements, satellite, and atmospheric models

Fine particulate matter (PM2.5) is a well-established risk factor for public health. To support both health risk assessment and epidemiological studies, data are needed on spatial and temporal patterns of PM2.5 exposures. This review article surveys publicly available exposure datasets for surface PM2.5 mass concentrations over the contiguous U.S., summarizes their applications and limitations, and provides suggestions on future research needs. The complex landscape of satellite instruments, model capabilities, monitor networks, and data synthesis methods offers opportunities for research development, but would benefit from guidance for new users. Guidance is provided to access publicly available PM2.5 datasets, to explain and compare different approaches for dataset generation, and to identify sources of uncertainties associated with various types of datasets. Three main sources used to create PM2.5 exposure data are ground-based measurements (especially regulatory monitoring), satellite retrievals (especially aerosol optical depth, AOD), and atmospheric chemistry models. We find inconsistencies among several publicly available PM2.5 estimates, highlighting uncertainties in the exposure datasets that are often overlooked in health effects analyses. Major differences among PM2.5 estimates emerge from the choice of data (ground-based, satellite, and/or model), the spatiotemporal resolutions, and the algorithms used to fuse data sources.Implications: Fine particulate matter (PM2.5) has large impacts on human morbidity and mortality. Even though the methods for generating the PM2.5 exposure estimates have been significantly improved in recent years, there is a lack of review articles that document PM2.5 exposure datasets that are publicly available and easily accessible by the health and air quality communities. In this article, we discuss the main methods that generate PM2.5 data, compare several publicly available datasets, and show the applications of various data fusion approaches. Guidance to access and critique these datasets are provided for stakeholders in public health sectors.

Machine learning models accurately predict ozone exposure during wildfire events

Epidemiologists use prediction models to downscale (i.e., interpolate) air pollution exposure where monitoring data is insufficient. This study compares machine learning prediction models for ground-level ozone during wildfires, evaluating the predictive accuracy of ten algorithms on the daily 8-hour maximum average ozone during a 2008 wildfire event in northern California. Models were evaluated using a leave-one-location-out cross-validation (LOLO CV) procedure to account for the spatial and temporal dependence of the data and produce more realistic estimates of prediction error. LOLO CV avoids both the well-known overly optimistic bias of k-fold cross-validation on dependent data and the conservative bias of evaluating prediction error over a coarser spatial resolution via leave-k-locations-out CV. Gradient boosting was the most accurate of the ten machine learning algorithms with the lowest LOLO CV estimated root mean square error (0.228) and the highest LOLO CV Rˆ² (0.677). Random forest was the second best performing algorithm with an LOLO CV Rˆ² of 0.661. The LOLO CV estimates of predictive accuracy were less optimistic than 10-fold CV estimates for all ten models. The difference in estimated accuracy between the 10-fold CV and LOLO CV was greater for more flexible models like gradient boosting and random forest. The order of estimated model accuracy depended on the choice of evaluation metric, indicating that 10-fold CV and LOLO CV may select different models or sets of covariates as optimal, which calls into question the reliability of 10-fold CV for model (or variable) selection. These prediction models are designed for interpolating ozone exposure, and are not suited to inferring the effect of wildfires on ozone or extrapolating to predict ozone in other spatial or temporal domains. This is demonstrated by the inability of the best performing models to accurately predict ozone during 2007 southern California wildfires.

Long-term effect of fine particulate matter on hospitalization with dementia

BACKGROUND

New evidence suggests that particulate matter less than 2.5 μm in diameter (PM_2.5) is associated with late-onset dementia (LOD). However, epidemiological studies for the entire population are lacking.

METHODS

We analyzed approximately 94 million follow-up records from fee-for-service Medicare records for 13 million Medicare beneficiaries residing in the southeastern United States (U.S.) from 2000 to 2013. We used spatially and temporally continuous PM_2.5 exposure data. To account for time-varying PM_2.5 levels, we applied an Andersen-Gill counting process proportional hazard model; we stratified our analyses by subtype of dementia and level of urbanization of residence.

RESULTS

During a median follow-up of 6 years, 1,409,599 hospitalizations with dementia occurred. The adjusted hazard ratio (HR) of hospitalization with dementia was 1.049 (95% confidence interval [CI], 1.048 to 1.051) per 1 μg/m³ increase in annual PM_2.5. The hazard ratio for vascular dementia was higher (HR, 1.086; 95% CI, 1.082 to 1.090). In large, the magnitude of the effect grew as the level of urbanization increased (HR, 1.036; 95% CI, 1.031 to 1.041 in rural areas versus HR, 1.052; 95% CI, 1.050 to 1.054 in metropolitan areas).

CONCLUSIONS

Long-term exposure to higher PM_2.5 was associated with increased hospitalizations with dementia.

An ensemble-based model of PM2.5 concentration across the contiguous United States with high spatiotemporal resolution

Various approaches have been proposed to model PM2.5 in the recent decade, with satellite-derived aerosol optical depth, land-use variables, chemical transport model predictions, and several meteorological variables as major predictor variables. Our study used an ensemble model that integrated multiple machine learning algorithms and predictor variables to estimate daily PM2.5 at a resolution of 1 km × 1 km across the contiguous United States. We used a generalized additive model that accounted for geographic difference to combine PM2.5 estimates from neural network, random forest, and gradient boosting. The three machine learning algorithms were based on multiple predictor variables, including satellite data, meteorological variables, land-use variables, elevation, chemical transport model predictions, several reanalysis datasets, and others. The model training results from 2000 to 2015 indicated good model performance with a 10-fold cross-validated R2 of 0.86 for daily PM2.5 predictions. For annual PM2.5 estimates, the cross-validated R2 was 0.89. Our model demonstrated good performance up to 60 μg/m3. Using trained PM2.5 model and predictor variables, we predicted daily PM2.5 from 2000 to 2015 at every 1 km × 1 km grid cell in the contiguous United States. We also used localized land-use variables within 1 km × 1 km grids to downscale PM2.5 predictions to 100 m × 100 m grid cells. To characterize uncertainty, we used meteorological variables, land-use variables, and elevation to model the monthly standard deviation of the difference between daily monitored and predicted PM2.5 for every 1 km × 1 km grid cell. This PM2.5 prediction dataset, including the downscaled and uncertainty predictions, allows epidemiologists to accurately estimate the adverse health effect of PM2.5. Compared with model performance of individual base learners, an ensemble model would achieve a better overall estimation. It is worth exploring other ensemble model formats to synthesize estimations from different models or from different groups to improve overall performance.

A Full-Coverage Daily Average PM2.5 Retrieval Method with Two-Stage IVW Fused MODIS C6 AOD and Two-Stage GAM Model

Current PM2.5 retrieval maps have many missing values, which seriously hinders their performance in real applications. This paper presents a framework to map full-coverage daily average PM2.5 concentrations from MODIS C6 aerosol optical depth (AOD) products and fill missing pixels in both the AOD and PM2.5 maps. First, a two-stage inversed variance weights (IVW) algorithm was adopted to fuse the MODIS C6 Terra and Aqua AOD products, which fills missing data in MODIS standard AOD data and obtains a high coverage daily average. After that, using the fused MODIS daily average AOD and ground-level PM2.5 in all grid cells, a two-stage generalized additive model (GAM) was implemented to obtain the full-coverage PM2.5 concentrations. Experiments on the Yangtze River Delta (YRD) in 2013–2016 were carefully designed to validate the performance of our proposed framework. The results show that the two-stage IVW could not only improve the spatial coverage of MODIS AOD against the original standard product by 230%, but could also keep its data accuracy. When compared with the ground-level measurements, the two-stage GAM can obtain accurate PM2.5 concentration estimates (R2 = 0.78, RMSE = 19.177 μg/m3, and RPE = 28.9%). Moreover, our method performs better than the inverse distance weighted method and kriging methods in mapping full-coverage daily PM2.5 concentrations. Therefore, the proposed framework provides a good methodology for retrieving full-coverage daily average PM2.5 concentrations from MODIS standard AOD products.

Estimation of daily PM10 and PM2.5 concentrations in Italy, 2013-2015, using a spatiotemporal land-use random-forest model

Particulate matter (PM) air pollution is one of the major causes of death worldwide, with demonstrated adverse effects from both short-term and long-term exposure. Most of the epidemiological studies have been conducted in cities because of the lack of reliable spatiotemporal estimates of particles exposure in nonurban settings. The objective of this study is to estimate daily PM₁₀ (PM < 10 μm), fine (PM < 2.5 μm, PM_2.5) and coarse particles (PM between 2.5 and 10 μm, PM_2.5-10) at 1-km² grid for 2013-2015 using a machine learning approach, the Random Forest (RF). Separate RF models were defined to: predict PM_2.5 and PM_2.5-10 concentrations in monitors where only PM₁₀ data were available (stage 1); impute missing satellite Aerosol Optical Depth (AOD) data using estimates from atmospheric ensemble models (stage 2); establish a relationship between measured PM and satellite, land use and meteorological parameters (stage 3); predict stage 3 model over each 1-km² grid cell of Italy (stage 4); and improve stage 3 predictions by using small-scale predictors computed at the monitor locations or within a small buffer (stage 5). Our models were able to capture most of PM variability, with mean cross-validation (CV) R² of 0.75 and 0.80 (stage 3) and 0.84 and 0.86 (stage 5) for PM₁₀ and PM_2.5, respectively. Model fitting was less optimal for PM_2.5-10, in summer months and in southern Italy. Finally, predictions were equally good in capturing annual and daily PM variability, therefore they can be used as reliable exposure estimates for investigating long-term and short-term health effects.

Interactive effects of changing stratospheric ozone and climate on tropospheric composition and air quality, and the consequences for human and ecosystem health

The composition of the air we breathe is determined by emissions, weather, and photochemical transformations induced by solar UV radiation. Photochemical reactions of many emitted chemical compounds can generate important (secondary) pollutants including ground-level ozone (O₃) and some particulate matter, known to be detrimental to human health and ecosystems. Poor air quality is the major environmental cause of premature deaths globally, and even a small decrease in air quality can translate into a large increase in the number of deaths. In many regions of the globe, changes in emissions of pollutants have caused significant changes in air quality. Short-term variability in the weather as well as long-term climatic trends can affect ground-level pollution through several mechanisms. These include large-scale changes in the transport of O₃ from the stratosphere to the troposphere, winds, clouds, and patterns of precipitation. Long-term trends in UV radiation, particularly related to the depletion and recovery of stratospheric ozone, are also expected to result in changes in air quality as well as the self-cleaning capacity of the global atmosphere. The increased use of substitutes for ozone-depleting substances, in response to the Montreal Protocol, does not currently pose a significant risk to the environment. This includes both the direct emissions of substitutes during use and their atmospheric degradation products (e.g. trifluoroacetic acid, TFA).

The sensitivity of satellite-based PM2.5 estimates to its inputs: Implications to model development in data-poor regions

Exposure to fine particulate matter (PM_2.5) has been associated with a wide range of negative health outcomes. The overwhelming majority of the epidemiological studies that helped establish such associations was conducted in regions with sufficient ground observations and other supporting data, i.e., the data-rich regions. However, air pollution health effects research in the data-poor regions, where pollution levels are often the highest, is still very limited due to the lack of high-quality exposure estimates. To improve our understanding of the desired input datasets for the application of satellite-based PM_2.5 exposure models in data-poor areas, we applied a Bayesian ensemble model in the southeast U.S. that was selected as a representative data-rich region. We designed four groups of sensitivity tests to simulate various data-poor scenarios. The factors considered that would influence the model performance included the temporal sampling frequency of the monitors, the number of ground monitors, the accuracy of the chemical transport model simulation of PM_2.5 concentrations, and different combinations of the additional predictors. While our full model achieved a 10-fold cross-validated (CV) R² of 0.82, we found that when reducing the sampling frequency from the current 1-in-3 day to 1-in-9 day, the CV R² decreased to 0.58, and the predictions could not capture the daily variations of PM_2.5. Half of the current stations (i.e., 30 monitors) could still support a robust model with a CV R² of 0.79. With 20 monitors, the CV R² decreased from 0.71 to 0.55 when 100% additional random errors were added to the original CMAQ simulations. However, with a sufficient number of ground monitors (e.g., 30 monitors), our Bayesian ensemble model had the ability to tolerate CMAQ errors with only a slight decrease in CV R² (from 0.79 to 0.75). With fewer than 15 monitors, our full model collapsed and failed to fit any covariates, while the models with only time-varying variables could still converge even with only five monitors left. A model without the land use parameters lacked fine spatial details in the prediction maps, but could still capture the daily variability of PM_2.5 (CV R² ≥ 0.67) and might support a study of the acute health effects of PM_2.5 exposure.

Satellite-based view of the aerosol spatial and temporal variability in the Córdoba region (Argentina) using over ten years of high-resolution data

Space-based observations offer a unique opportunity to investigate the atmosphere and its changes over decadal time scales, particularly in regions lacking in situ and/or ground based observations. In this study, we investigate temporal and spatial variability of atmospheric particulate matter (aerosol) over the urban area of Córdoba (central Argentina) using over ten years (2003-2015) of high-resolution (1 km) satellite-based retrievals of aerosol optical depth (AOD). This fine resolution is achieved exploiting the capabilities of a recently developed inversion algorithm (Multiangle implementation of atmospheric correction, MAIAC) applied to the MODIS sensor datasets of the NASA-Terra and -Aqua platforms. Results of this investigation show a clear seasonality of AOD over the investigated area. This is found to be shaped by an intricate superposition of aerosol sources, acting over different spatial scales and affecting the region with different yearly cycles. During late winter and spring (August-October), local as well as near- and long-range transported biomass burning (BB) aerosols enhance the Córdoba aerosol load, and AOD levels reach their maximum values (> 0.35 at 0.47μm). The fine AOD spatial resolution allowed to disclose that, in this period, AOD maxima are found in the rural/agricultural area around the city, reaching up to the city boundaries pinpointing that fires of local and near-range origin play a major role in the AOD enhancement. A reverse spatial AOD gradient is found from December to March, the urban area showing AODs 40 to 80% higher than in the city surroundings. In fact, during summer, the columnar aerosol load over the Córdoba region is dominated by local (urban and industrial) sources, likely coupled to secondary processes driven by enhanced radiation and mixing effects within a deeper planetary boundary layer (PBL). With the support of modelled AOD data from the Modern-Era Retrospective Analysis for Research and Application (MERRA), we further investigated into the chemical nature of AOD. The results suggest that mineral dust is also an important aerosol component in Córdoba, with maximum impact from November to February. The use of a long-term dataset finally allowed a preliminary assessment of AOD trends over the Córdoba region. For those months in which local sources and secondary processes were found to dominate the AOD (December to March), we found a positive AOD trend in the Córdoba outskirts, mainly in the areas with maximum urbanization/population growth over the investigated decade. Conversely, a negative AOD trend (up to -0.1 per decade) is observed all over the rural area of Córdoba during the BB season, this being attributed to a decrease of fires both at the local and the continental scale.

Modeling Wildfire Smoke Pollution by Integrating Land Use Regression and Remote Sensing Data: Regional Multi-Temporal Estimates for Public Health and Exposure Models

To understand the health effects of wildfire smoke, it is important to accurately assess smoke exposure over space and time. Particulate matter (PM) is a predominant pollutant in wildfire smoke. In this study, we develop land-use regression (LUR) models to investigate the impact that a cluster of wildfires in the northwest USA had on the level of PM in southern Alberta (Canada), in the summer of 2015. Univariate aerosol optical depth (AOD) and multivariate AOD-LUR models were used to estimate the level of PM2.5 in urban and rural areas. For epidemiological studies, it is also important to distinguish between wildfire-related PM2.5 and PM2.5 originating from other sources. We therefore subdivided the study period into three sub-periods: (1) Pre-fire, (2) during-fire, and (3) post-fire. We then developed separate models for each sub-period. With this approach, we were able to identify different predictors significantly associated with smoke-related PM2.5 verses PM2.5 of different origin. Leave-one-out cross-validation (LOOCV) was used to evaluate the models’ performance. Our results indicate that model predictors and model performance are highly related to the level of PM2.5, and the pollution source. The predictive ability of both uni- and multi-variate models were higher in the during-fire period than in the pre- and post-fire periods.

Satellite-Based Daily PM2.5 Estimates During Fire Seasons in Colorado

The western United States has experienced increasing wildfire activities, which have negative effects on human health. Epidemiological studies on fine particulate matter (PM2.5) from wildfires are limited by the lack of accurate high-resolution PM2.5 exposure data over fire days. Satellite-based aerosol optical depth (AOD) data can provide additional information in ground PM2.5 concentrations and has been widely used in previous studies. However, the low background concentration, complex terrain, and large wildfire sources add to the challenge of estimating PM2.5 concentrations in the western United States. In this study, we applied a Bayesian ensemble model that combined information from the 1 km resolution AOD products derived from the Multi-angle Implementation of Atmospheric Correction (MAIAC) algorithm, Community Multiscale Air Quality (CMAQ) model simulations, and ground measurements to predict daily PM2.5 concentrations over fire seasons (April to September) in Colorado for 2011-2014. Our model had a 10-fold cross-validated R 2 of 0.66 and root-mean-squared error of 2.00 μg/m3, outperformed the multistage model, especially on the fire days. Elevated PM2.5 concentrations over large fire events were successfully captured. The modeling technique demonstrated in this study could support future short-term and long-term epidemiological studies of wildfire PM2.5.

Predicting daily PM2.5 concentrations in Texas using high-resolution satellite aerosol optical depth

BACKGROUND AND OBJECTIVE

The regulatory monitoring data of particulate matter with an aerodynamic diameter <2.5μm (PM_2.5) in Texas have limited spatial and temporal coverage. The purpose of this study is to estimate the ground-level PM_2.5 concentrations on a daily basis using satellite-retrieved Aerosol Optical Depth (AOD) in the state of Texas.

METHODS

We obtained the AOD values at 1-km resolution generated through the Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm based on the images retrieved from the Moderate Resolution Imaging Spectroradiometer (MODIS) satellites. We then developed mixed-effects models based on AODs, land use features, geographic characteristics, and weather conditions, and the day-specific as well as site-specific random effects to estimate the PM_2.5 concentrations (μg/m³) in the state of Texas during the period 2008-2013. The mixed-effects models' performance was evaluated using the coefficient of determination (R²) and square root of the mean squared prediction error (RMSPE) from ten-fold cross-validation, which randomly selected 90% of the observations for training purpose and 10% of the observations for assessing the models' true prediction ability.

RESULTS

Mixed-effects regression models showed good prediction performance (R² values from 10-fold cross validation: 0.63-0.69). The model performance varied by regions and study years, and the East region of Texas, and year of 2009 presented relatively higher prediction precision (R²: 0.62 for the East region; R²: 0.69 for the year of 2009). The PM_2.5 concentrations generated through our developed models at 1-km grid cells in the state of Texas showed a decreasing trend from 2008 to 2013 and a higher reduction of predicted PM_2.5 in more polluted areas.

CONCLUSIONS

Our findings suggest that mixed-effects regression models developed based on MAIAC AOD are a feasible approach to predict ground-level PM_2.5 in Texas. Predicted PM_2.5 concentrations at the 1-km resolution on a daily basis can be used for epidemiological studies to investigate short- and long-term health impact of PM_2.5 in Texas.

Doubly Robust Additive Hazards Models to Estimate Effects of a Continuous Exposure on Survival

BACKGROUND

The effect of an exposure on survival can be biased when the regression model is misspecified. Hazard difference is easier to use in risk assessment than hazard ratio and has a clearer interpretation in the assessment of effect modifications.

METHODS

We proposed two doubly robust additive hazards models to estimate the causal hazard difference of a continuous exposure on survival. The first model is an inverse probability-weighted additive hazards regression. The second model is an extension of the doubly robust estimator for binary exposures by categorizing the continuous exposure. We compared these with the marginal structural model and outcome regression with correct and incorrect model specifications using simulations. We applied doubly robust additive hazard models to the estimation of hazard difference of long-term exposure to PM2.5 (particulate matter with an aerodynamic diameter less than or equal to 2.5 microns) on survival using a large cohort of 13 million older adults residing in seven states of the Southeastern United States.

RESULTS

We showed that the proposed approaches are doubly robust. We found that each 1 μg m increase in annual PM2.5 exposure was associated with a causal hazard difference in mortality of 8.0 × 10 (95% confidence interval 7.4 × 10, 8.7 × 10), which was modified by age, medical history, socioeconomic status, and urbanicity. The overall hazard difference translates to approximately 5.5 (5.1, 6.0) thousand deaths per year in the study population.

CONCLUSIONS

The proposed approaches improve the robustness of the additive hazards model and produce a novel additive causal estimate of PM2.5 on survival and several additive effect modifications, including social inequality.

Estimating the Causal Effect of Low Levels of Fine Particulate Matter on Hospitalization

BACKGROUND

In 2012, the EPA enacted more stringent National Ambient Air Quality Standards (NAAQS) for fine particulate matter (PM2.5). Few studies have characterized the health effects of air pollution levels lower than the most recent NAAQS for long-term exposure to PM2.5 (now 12 μg/m).

METHODS

We constructed a cohort of 32,119 Medicare beneficiaries residing in 5138 US ZIP codes who were interviewed as part of the Medicare Current Beneficiary Survey (MCBS) between 2002 and 2010 and had 1 year of follow-up. We considered four outcomes: all-cause hospitalizations, hospitalizations for circulatory diseases and respiratory diseases, and death.

RESULTS

We found that increasing exposure to PM2.5 from levels lower than 12 μg/m to levels higher than 12 μg/m is associated with increases in all-cause admission rates of 7% (95% CI = 3%, 10%) and in circulatory admission hazard rates of 6% (95% CI = 2%, 9%). When we restricted analysis to enrollees with exposure always lower than 12 μg/m, we found that increasing exposure from levels lower than 8 μg/m to levels higher than 8 μg/m increased all-cause admission hazard rates by 15% (95% CI = 8%, 23%), circulatory by 18% (95% CI = 10%, 27%), and respiratory by 21% (95% CI = 9%, 34%).

CONCLUSIONS

In a nationally representative sample of Medicare enrollees, changes in exposure to PM2.5, even at levels consistently below standards, are associated with increases in hospital admissions for all causes and cardiovascular and respiratory diseases. The robustness of our results to inclusion of many additional individual level potential confounders adds validity to studies of air pollution that rely entirely on administrative data.

Satellite-based high-resolution PM2.5 estimation over the Beijing-Tianjin-Hebei region of China using an improved geographically and temporally weighted regression model

Ground fine particulate matter (PM2.5) concentrations at high spatial resolution are substantially required for determining the population exposure to PM2.5 over densely populated urban areas. However, most studies for China have generated PM2.5 estimations at a coarse resolution (≥10 km) due to the limitation of satellite aerosol optical depth (AOD) product in spatial resolution. In this study, the 3 km AOD data fused using the Moderate Resolution Imaging Spectroradiometer (MODIS) Collection 6 AOD products were employed to estimate the ground PM2.5 concentrations over the Beijing-Tianjin-Hebei (BTH) region of China from January 2013 to December 2015. An improved geographically and temporally weighted regression (iGTWR) model incorporating seasonal characteristics within the data was developed, which achieved comparable performance to the standard GTWR model for the days with paired PM_2.5- AOD samples (Cross-validation (CV) R² = 0.82) and showed better predictive power for the days without PM_2.5- AOD pairs (the R² increased from 0.24 to 0.46 in CV). Both iGTWR and GTWR (CV R² = 0.84) significantly outperformed the daily geographically weighted regression model (CV R² = 0.66). Also, the fused 3 km AODs improved data availability and presented more spatial gradients, thereby enhancing model performance compared with the MODIS original 3/10 km AOD product. As a result, ground PM2.5 concentrations at higher resolution were well represented, allowing, e.g., short-term pollution events and long-term PM2.5 trend to be identified, which, in turn, indicated that concerns about air pollution in the BTH region are justified despite its decreasing trend from 2013 to 2015.

Design of an Air Pollution Monitoring Campaign in Beijing for Application to Cohort Health Studies

No cohort studies in China on the health effects of long-term air pollution exposure have employed exposure estimates at the fine spatial scales desirable for cohort studies with individual-level health outcome data. Here we assess an array of modern air pollution exposure estimation approaches for assigning within-city exposure estimates in Beijing for individual pollutants and pollutant sources to individual members of a cohort. Issues considered in selecting specific monitoring data or new monitoring campaigns include: needed spatial resolution, exposure measurement error and its impact on health effect estimates, spatial alignment and compatibility with the cohort, and feasibility and expense. Sources of existing data largely include administrative monitoring data, predictions from air dispersion or chemical transport models and remote sensing (specifically satellite) data. New air monitoring campaigns include additional fixed site monitoring, snapshot monitoring, passive badge or micro-sensor saturation monitoring and mobile monitoring, as well as combinations of these. Each of these has relative advantages and disadvantages. It is concluded that a campaign in Beijing that at least includes a mobile monitoring component, when coupled with currently available spatio-temporal modeling methods, should be strongly considered. Such a campaign is economical and capable of providing the desired fine-scale spatial resolution for pollutants and sources.

Assessing the Exposome with External Measures: Commentary on the State of the Science and Research Recommendations

The exposome comprises all environmental exposures that a person experiences from conception throughout the life course. Here we review the state of the science for assessing external exposures within the exposome. This article reviews (a) categories of exposures that can be assessed externally, (b) the current state of the science in external exposure assessment, (c) current tools available for external exposure assessment, and (d) priority research needs. We describe major scientific and technological advances that inform external assessment of the exposome, including geographic information systems; remote sensing; global positioning system and geolocation technologies; portable and personal sensing, including smartphone-based sensors and assessments; and self-reported questionnaire assessments, which increasingly rely on Internet-based platforms. We also discuss priority research needs related to methodological and technological improvement, data analysis and interpretation, data sharing, and other practical considerations, including improved assessment of exposure variability as well as exposure in multiple, critical life stages.

An example of aerosol pattern variability over bright surface using high resolution MODIS MAIAC: The eastern and western areas of the Dead Sea and environs

The extreme rate of evaporation of the Dead Sea (DS) has serious implicatios for the surrounding area, including atmospheric conditions. This study analyzes the aerosol properties over the western and eastern parts of the DS during the year 2013, using MAIAC (Multi-Angle Implementation of Atmospheric Correction) for MODIS, which retrieves aerosol optical depth (AOD) data at a resolution of 1km. The main goal of the study is to evaluate MAIAC over the study area and determine, for the first time, the prevailing aerosol spatial patterns. First, the MAIAC-derived AOD data was compared with data from three nearby AERONET sites (Nes Ziona - an urban site, and Sede Boker and Masada - two arid sites), and with the conventional Dark Target (DT) and Deep Blue (DB) retrievals for the same days and locations, on a monthly basis throughout 2013. For the urban site, the correlation coefficient (r) for DT/DB products showed better performance than MAIAC (r=0.80, 0.75, and 0.64 respectively) year-round. However, in the arid zones, MAIAC showed better correspondence to AERONET sites than the conventional retrievals (r=0.58-0.60 and 0.48-0.50 respectively). We investigated the difference in AOD levels, and its variability, between the Dead Sea coasts on a seasonal basis and calculated monthly/seasonal AOD averages for presenting AOD patterns over arid zones. Thus, we demonstrated that aerosol concentrations show a strong preference for the western coast, particularly during the summer season. This preference, is most likely a result of local anthropogenic emissions combined with the typical seasonal synoptic conditions, the Mediterranean Sea breeze, and the region complex topography. Our results also indicate that a large industrial zone showed higher AOD levels compared to an adjacent reference-site, i.e., 13% during the winter season.

Estimating PM2.5 Concentrations in the Conterminous United States Using the Random Forest Approach

To estimate PM_2.5 concentrations, many parametric regression models have been developed, while nonparametric machine learning algorithms are used less often and national-scale models are rare. In this paper, we develop a random forest model incorporating aerosol optical depth (AOD) data, meteorological fields, and land use variables to estimate daily 24 h averaged ground-level PM_2.5 concentrations over the conterminous United States in 2011. Random forests are an ensemble learning method that provides predictions with high accuracy and interpretability. Our results achieve an overall cross-validation (CV) R² value of 0.80. Mean prediction error (MPE) and root mean squared prediction error (RMSPE) for daily predictions are 1.78 and 2.83 μg/m³, respectively, indicating a good agreement between CV predictions and observations. The prediction accuracy of our model is similar to those reported in previous studies using neural networks or regression models on both national and regional scales. In addition, the incorporation of convolutional layers for land use terms and nearby PM_2.5 measurements increase CV R² by ∼0.02 and ∼0.06, respectively, indicating their significant contributions to prediction accuracy. A pair of different variable importance measures both indicate that the convolutional layer for nearby PM_2.5 measurements and AOD values are among the most-important predictor variables for the training process.

Airborne Fine Particles and Risk of Hospital Admissions for Understudied Populations: Effects by Urbanicity and Short-Term Cumulative Exposures in 708 U.S. Counties

BACKGROUND

Evidence of health risks associated with ambient airborne fine particles in nonurban populations is extremely limited.

OBJECTIVE

We estimated the risk of hospitalization associated with short-term exposures to particulate matter with an aerodynamic diameter < 2.5 μm (PM_2.5) in urban and nonurban counties with population ≥ 50,000.

METHODS

We utilized a database of daily cardiovascular- and respiratory-related hospitalization rates constructed from Medicare National Claims History files (2002-2006), including 28 million Medicare beneficiaries in 708 counties. Daily PM_2.5 exposures were estimated using the Community Multiscale Air Quality (CMAQ) downscaler. We used time-series analysis of hospitalization rates and PM_2.5 to evaluate associations between PM_2.5 levels and hospitalization risk in single-pollutant models.

RESULTS

We observed an association between cardiovascular hospitalizations and same-day PM_2.5 with higher risk in urban counties: 0.35% [95% posterior interval (PI): -0.71%, 1.41%] and 0.98% (95% PI: 0.73%, 1.23%) increases in hospitalization risk per 10-μg/m³ increment in PM_2.5 were observed in the least-urban and most-urban counties, respectively. The largest association for respiratory hospitalizations, a 2.57% (95% PI: 0.87%, 4.30%) increase per 10-μg/m³ increase in PM_2.5, was observed in the least-urban counties; in the most-urban counties, a 1.13% (0.73%, 1.54%) increase was observed. Effect estimates for cardiovascular hospitalizations were highest for smaller lag times, whereas effect estimates for respiratory hospitalizations increased as more days of exposure were included.

CONCLUSION

In nonurban counties with population ≥ 50,000, exposure to PM_2.5 is associated with increased risk for respiratory hospitalizations; in urban counties, exposure is associated with increased risk of cardiovascular hospitalizations. Effect estimates based on a single day of exposure may underestimate true effects for respiratory hospitalizations.

Comparing the Health Effects of Ambient Particulate Matter Estimated Using Ground-Based versus Remote Sensing Exposure Estimates

BACKGROUND

Remote sensing (RS) is increasingly used for exposure assessment in epidemiological and burden of disease studies, including those investigating whether chronic exposure to ambient fine particulate matter (PM2.5) is associated with mortality.

OBJECTIVES

We compared relative risk estimates of mortality from diseases of the circulatory system for PM2.5 modeled from RS with that for PM2.5 modeled using ground-level information.

METHODS

We geocoded the baseline residence of 668,629 American Cancer Society Cancer Prevention Study II (CPS-II) cohort participants followed from 1982 to 2004 and assigned PM2.5 levels to all participants using seven different exposure models. Most of the exposure models were averaged for the years 2002-2004, and one RS estimate was for a longer, contemporaneous period. We used Cox proportional hazards regression to estimate relative risks (RRs) for the association of PM2.5 with circulatory mortality and ischemic heart disease.

RESULTS

Estimates of mortality risk differed among exposure models. The smallest relative risk was observed for the RS estimates that excluded ground-based monitors for circulatory deaths [RR = 1.02, 95% confidence interval (CI): 1.00, 1.04 per 10 μg/m3 increment in PM2.5]. The largest relative risk was observed for the land-use regression model that included traffic information (RR = 1.14, 95% CI: 1.11, 1.17 per 10 μg/m3 increment in PM2.5).

CONCLUSIONS

We found significant associations between PM2.5 and mortality in every model; however, relative risks estimated from exposure models using ground-based information were generally larger than those estimated using RS alone.

Long-term Exposure to PM2.5 and Mortality Among Older Adults in the Southeastern US

BACKGROUND

Little is known about what factors modify the effect of long-term exposure to PM2.5 on mortality, in part because in most previous studies certain groups such as rural residents and individuals with lower socioeconomic status (SES) are under-represented.

METHODS

We studied 13.1 million Medicare beneficiaries (age ≥65) residing in seven southeastern US states during 2000-2013 with 95 million person-years of follow-up. We predicted annual average of PM2.5 in each zip code tabulation area (ZCTA) using a hybrid spatiotemporal model. We fit Cox proportional hazards models to estimate the association between long-term PM2.5 and mortality. We tested effect modification by individual-level covariates (race, sex, eligibility for both Medicare and Medicaid, and medical history), neighborhood-level covariates (urbanicity, percentage below poverty level, lower education, median income, and median home value), mean summer temperature, and mass fraction of 11 PM2.5 components.

RESULTS

The hazard ratio (HR) for death was 1.021 (95% confidence interval: 1.019, 1.022) per 1 μg m increase in annual PM2.5. The HR decreased with age. It was higher among males, non-whites, dual-eligible individuals, and beneficiaries with previous hospital admissions. It was higher in neighborhoods with lower SES or higher urbanicity. The HR increased with mean summer temperature. The risk associated with PM2.5 increased with relative concentration of elemental carbon, vanadium, copper, calcium, and iron and decreased with nitrate, organic carbon, and sulfate.

CONCLUSIONS

Associations between long-term PM2.5 exposure and death were modified by individual-level, neighborhood-level variables, temperature, and chemical compositions.

Estimation of daily PM10 concentrations in Italy (2006-2012) using finely resolved satellite data, land use variables and meteorology

Health effects of air pollution, especially particulate matter (PM), have been widely investigated. However, most of the studies rely on few monitors located in urban areas for short-term assessments, or land use/dispersion modelling for long-term evaluations, again mostly in cities. Recently, the availability of finely resolved satellite data provides an opportunity to estimate daily concentrations of air pollutants over wide spatio-temporal domains. Italy lacks a robust and validated high resolution spatio-temporally resolved model of particulate matter. The complex topography and the air mixture from both natural and anthropogenic sources are great challenges difficult to be addressed. We combined finely resolved data on Aerosol Optical Depth (AOD) from the Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm, ground-level PM₁₀ measurements, land-use variables and meteorological parameters into a four-stage mixed model framework to derive estimates of daily PM₁₀ concentrations at 1-km2 grid over Italy, for the years 2006-2012. We checked performance of our models by applying 10-fold cross-validation (CV) for each year. Our models displayed good fitting, with mean CV-R2=0.65 and little bias (average slope of predicted VS observed PM₁₀=0.99). Out-of-sample predictions were more accurate in Northern Italy (Po valley) and large conurbations (e.g. Rome), for background monitoring stations, and in the winter season. Resulting concentration maps showed highest average PM₁₀ levels in specific areas (Po river valley, main industrial and metropolitan areas) with decreasing trends over time. Our daily predictions of PM₁₀ concentrations across the whole Italy will allow, for the first time, estimation of long-term and short-term effects of air pollution nationwide, even in areas lacking monitoring data.

Evaluating the effectiveness of air quality regulations: A review of accountability studies and frameworks

Assessments of past environmental policies-termed accountability studies-contribute important information to the decision-making process used to review the efficacy of past policies, and subsequently aid in the development of effective new policies. These studies have used a variety of methods that have achieved varying levels of success at linking improvements in air quality and/or health to regulations. The Health Effects Institute defines the air pollution accountability framework as a chain of events that includes the regulation of interest, air quality, exposure/dose, and health outcomes, and suggests that accountability research should address impacts for each of these linkages. Early accountability studies investigated short-term, local regulatory actions (for example, coal use banned city-wide on a specific date or traffic pattern changes made for Olympic Games). Recent studies assessed regulations implemented over longer time and larger spatial scales. Studies on broader scales require accountability research methods that account for effects of confounding factors that increase over time and space. Improved estimates of appropriate baseline levels (sometimes termed "counterfactual"-the expected state in a scenario without an intervention) that account for confounders and uncertainties at each link in the accountability chain will help estimate causality with greater certainty. In the direct accountability framework, researchers link outcomes with regulations using statistical methods that bypass the link-by-link approach of classical accountability. Direct accountability results and methods complement the classical approach. New studies should take advantage of advanced planning for accountability studies, new data sources (such as satellite measurements), and new statistical methods. Evaluation of new methods and data sources is necessary to improve investigations of long-term regulations, and associated uncertainty should be accounted for at each link to provide a confidence estimate of air quality regulation effectiveness. The final step in any accountability is the comparison of results with the proposed benefits of an air quality policy.

IMPLICATIONS

The field of air pollution accountability continues to grow in importance to a number of stakeholders. Two frameworks, the classical accountability chain and direct accountability, have been used to estimate impacts of regulatory actions, and both require careful attention to confounders and uncertainties. Researchers should continue to develop and evaluate both methods as they investigate current and future air pollution regulations.

Study on the association between ambient temperature and mortality using spatially resolved exposure data

There are many studies that have posited an association between extreme temperature and increased mortality. However, most studies use temperature at a single station per city as the reference point to analyze deaths. This leads to exposure misclassification and usually the exclusion of exurban, small town, and rural populations. In addition, few studies control for confounding by PM_2.5, which is expected to induce upward bias. The high-resolution temperature and PM_2.5 data at a resolution of 1km² were derived from satellite images and other land use sources. To capture the nonlinear association of temperature with mortality we fit a piecewise linear spline function for temperature, with a change in slope at -1°C and 28°C, the temperature threshold at which mortality in Georgia, North Carolina, and South Carolina increases due to cold and heat, respectively. We conducted stratified analyses by age group, sex, race, education, and urban vs nonurban, as well as sensitivity analyses of different temperature threshold and covariate sets. We found a 0.19% (95% CI=-0.98, 1.34%) increase in mortality for each 1°C decrease in temperature below -1°C and a 2.05% (95% CI=0.87, 3.24%) increase in mortality for each 1°C increase in temperature above 28°C, a 79.8% larger effect size for heat compared to the station-based metric. The effect estimates relying on the monitoring stations were 0.09% (95% CI=-0.79, 0.95%) and 1.14% (95% CI=0.08, 1.57%) for the equivalent temperature changes. The estimates were not confounded by PM_2.5. Children under 15 years of age had the largest percentage increase per 1°C increase in temperature (8.19%, 95% CI=-0.38 to 17.49%) followed by Blacks (4.35%, 95% CI=2.22 to 6.53%). Higher education was a protective factor for the effect of extreme temperature on mortality. There was a suggestion that people in less urban areas were more susceptible to extreme temperature. The relationship between temperature and mortality was stronger when using exposure data with more spatial variability than using exposure data based on existing monitors alone.

Assessing PM2.5 Exposures with High Spatiotemporal Resolution across the Continental United States

A number of models have been developed to estimate PM2.5 exposure, including satellite-based aerosol optical depth (AOD) models, land-use regression, or chemical transport model simulation, all with both strengths and weaknesses. Variables like normalized difference vegetation index (NDVI), surface reflectance, absorbing aerosol index, and meteoroidal fields are also informative about PM2.5 concentrations. Our objective is to establish a hybrid model which incorporates multiple approaches and input variables to improve model performance. To account for complex atmospheric mechanisms, we used a neural network for its capacity to model nonlinearity and interactions. We used convolutional layers, which aggregate neighboring information, into a neural network to account for spatial and temporal autocorrelation. We trained the neural network for the continental United States from 2000 to 2012 and tested it with left out monitors. Ten-fold cross-validation revealed a good model performance with a total R(2) of 0.84 on the left out monitors. Regional R(2) could be even higher for the Eastern and Central United States. Model performance was still good at low PM2.5 concentrations. Then, we used the trained neural network to make daily predictions of PM2.5 at 1 km × 1 km grid cells. This model allows epidemiologists to access PM2.5 exposure in both the short-term and the long-term.

Satellite remote sensing in epidemiological studies

PURPOSE OF REVIEW

Particulate matter air pollution is a ubiquitous exposure linked with multiple adverse health outcomes for children and across the life course. The recent development of satellite-based remote-sensing models for air pollution enables the quantification of these risks and addresses many limitations of previous air pollution research strategies. We review the recent literature on the applications of satellite remote sensing in air quality research, with a focus on their use in epidemiological studies.

RECENT FINDINGS

Aerosol optical depth (AOD) is a focus of this review and a significant number of studies show that ground-level particulate matter can be estimated from columnar AOD. Satellite measurements have been found to be an important source of data for particulate matter model-based exposure estimates, and recently have been used in health studies to increase the spatial breadth and temporal resolution of these estimates.

SUMMARY

It is suggested that satellite-based models improve our understanding of the spatial characteristics of air quality. Although the adoption of satellite-based measures of air quality in health studies is in its infancy, it is rapidly growing. Nevertheless, further investigation is still needed in order to have a better understanding of the AOD contribution to these prediction models in order to use them with higher accuracy in epidemiological studies.

Acute effect of fine particulate matter on mortality in three Southeastern states from 2007-2011

Epidemiologic studies on acute effects of air pollution have generally been limited to larger cities, leaving questions about rural populations behind. Recently, we had developed a spatiotemporal model to predict daily PM2.5 level at a 1 km(2) using satellite aerosol optical depth (AOD) data. Based on the results from the model, we applied a case-crossover study to evaluate the acute effect of PM2.5 on mortality in North Carolina, South Carolina, and Georgia between 2007 and 2011. Mortality data were acquired from the Departments of Public Health in the States and modeled PM2.5 exposures were assigned to the zip code of residence of each decedent. We performed various stratified analyses by age, sex, race, education, cause of death, residence, and environmental protection agency (EPA) standards. We also compared results of analyses using our modeled PM2.5 levels and those imputed daily from the nearest monitoring station. 848,270 non-accidental death records were analyzed and we found each 10 μg/m(3) increase in PM2.5 (mean lag 0 and lag 1) was associated with a 1.56% (1.19 and 1.94) increase in daily deaths. Cardiovascular disease (2.32%, 1.57-3.07) showed the highest effect estimate. Blacks (2.19%, 1.43-2.96) and persons with education ≤ 8 year (3.13%, 2.08-4.19) were the most vulnerable populations. The effect of PM2.5 on mortality still exists in zip code areas that meet the PM2.5 EPA annual standard (2.06%, 1.97-2.15). The effect of PM2.5 below both EPA daily and annual standards was 2.08% (95% confidence interval=1.99-2.17). Our results showed more power and suggested that the PM2.5 effects on rural populations have been underestimated due to selection bias and information bias. We have demonstrated that our AOD-based exposure models can be successfully applied to epidemiologic studies. This will add new study populations in rural areas, and will confer more generalizability to conclusions from such studies.