Exposure-response associations between chronic exposure to fine particulate matter and risks of hospital admission for major cardiovascular diseases: population based cohort study.

OBJECTIVE: To estimate exposure-response associations between chronic exposure to fine particulate matter (PM2.5) and risks of the first hospital admission for major cardiovascular disease (CVD) subtypes. DESIGN: Population based cohort study. SETTING: Contiguous US. PARTICIPANTS: 59 761 494 Medicare fee-for-service beneficiaries aged ≥65 years during 2000-16. Calibrated PM2.5 predictions were linked to each participant's residential zip code as proxy exposure measurements. MAIN OUTCOME MEASURES: Risk of the first hospital admission during follow-up for ischemic heart disease, cerebrovascular disease, heart failure, cardiomyopathy, arrhythmia, valvular heart disease, thoracic and abdominal aortic aneurysms, or a composite of these CVD subtypes. A causal framework robust against confounding bias and bias arising from errors in exposure measurements was developed for exposure-response estimations. RESULTS: Three year average PM2.5 exposure was associated with increased relative risks of first hospital admissions for ischemic heart disease, cerebrovascular disease, heart failure, cardiomyopathy, arrhythmia, and thoracic and abdominal aortic aneurysms. For composite CVD, the exposure-response curve showed monotonically increased risk associated with PM2.5: compared with exposures ≤5 µg/m3 (the World Health Organization air quality guideline), the relative risk at exposures between 9 and 10 µg/m3, which encompassed the US national average of 9.7 µg/m3 during the study period, was 1.29 (95% confidence interval 1.28 to 1.30). On an absolute scale, the risk of hospital admission for composite CVD increased from 2.59% with exposures ≤5 µg/m3 to 3.35% at exposures between 9 and 10 µg/m3. The effects persisted for at least three years after exposure to PM2.5. Age, education, accessibility to healthcare, and neighborhood deprivation level appeared to modify susceptibility to PM2.5. CONCLUSIONS: The findings of this study suggest that no safe threshold exists for the chronic effect of PM2.5 on overall cardiovascular health. Substantial benefits could be attained through adherence to the WHO air quality guideline.

Seasonal Temperature Variability and Mortality in the Medicare Population.

BACKGROUND: Seasonal temperature variability remains understudied and may be modified by climate change. Most temperature-mortality studies examine short-term exposures using time-series data. These studies are limited by regional adaptation, short-term mortality displacement, and an inability to observe longer-term relationships in temperature and mortality. Seasonal temperature and cohort analyses allow the long-term effects of regional climatic change on mortality to be analyzed. OBJECTIVES: We aimed to carry out one of the first investigations of seasonal temperature variability and mortality across the contiguous United States. We also investigated factors that modify this association. Using adapted quasi-experimental methods, we hoped to account for unobserved confounding and to investigate regional adaptation and acclimatization at the ZIP code level. METHODS: We examined the mean and standard deviation (SD) of daily temperature in the warm (April-September) and cold (October-March) season in the Medicare cohort from 2000 to 2016. This cohort comprised 622,427,230 y of person-time in all adults over the age of 65 y from 2000 to 2016. We used daily mean temperature obtained from gridMET to develop yearly seasonal temperature variables for each ZIP code. We used an adapted difference-in-difference approach model with a three-tiered clustering approach and meta-analysis to observe the relationship between temperature variability and mortality within ZIP codes. Effect modification was assessed with stratified analyses by race and population density. RESULTS: For every 1°C increase in the SD of warm and cold season temperature, the mortality rate increased by 1.54% [95% confidence interval (CI): 0.73%, 2.15%] and 0.69% (95% CI: 0.22%, 1.15%) respectively. We did not see significant effects for seasonal mean temperatures. Participants who were classified by Medicare into an "other" race group had smaller effects than those classified as White for Cold and Cold SD and areas with lower population density had larger effects for Warm SD. DISCUSSION: Warm and cold season temperature variability were significantly associated with increased mortality rates in U.S. individuals over the age of 65 y, even after controlling for seasonal temperature averages. Warm and cold season mean temperatures showed null effects on mortality. Cold SD had a larger effect size for those who were in the racial subgroup other, whereas Warm SD was more harmful for those living in lower population density areas. This study adds to the growing calls for urgent climate mitigation and environmental health adaptation and resiliency. https://doi.org/10.1289/EHP11588.

The Impact of Exposure Measurement Error on the Estimated Concentration-Response Relationship between Long-Term Exposure to PM2.5 and Mortality.

BACKGROUND: Exposure measurement error is a central concern in air pollution epidemiology. Given that studies have been using ambient air pollution predictions as proxy exposure measures, the potential impact of exposure error on health effect estimates needs to be comprehensively assessed. OBJECTIVES: We aimed to generate wide-ranging scenarios to assess direction and magnitude of bias caused by exposure errors under plausible concentration-response relationships between annual exposure to fine particulate matter [PM ≤2.5µm in aerodynamic diameter (PM2.5)] and all-cause mortality. METHODS: In this simulation study, we use daily PM2.5 predictions at 1-km2 spatial resolution to estimate annual PM2.5 exposures and their uncertainties for ZIP Codes of residence across the contiguous United States between 2000 and 2016. We consider scenarios in which we vary the error type (classical or Berkson) and the true concentration-response relationship between PM2.5 exposure and mortality (linear, quadratic, or soft-threshold-i.e., a smooth approximation to the hard-threshold model). In each scenario, we generate numbers of deaths using error-free exposures and confounders of concurrent air pollutants and neighborhood-level covariates and perform epidemiological analyses using error-prone exposures under correct specification or misspecification of the concentration-response relationship between PM2.5 exposure and mortality, adjusting for the confounders. RESULTS: We simulate 1,000 replicates of each of 162 scenarios investigated. In general, both classical and Berkson errors can bias the concentration-response curve toward the null. The biases remain small even when using three times the predicted uncertainty to generate errors and are relatively larger at higher exposure levels. DISCUSSION: Our findings suggest that the causal determination for long-term PM2.5 exposure and mortality is unlikely to be undermined when using high-resolution ambient predictions given that the estimated effect is generally smaller than the truth. The small magnitude of bias suggests that epidemiological findings are relatively robust against the exposure error. In practice, the use of ambient predictions with a finer spatial resolution will result in smaller bias. https://doi.org/10.1289/EHP10389.

Air Pollution and Autism Spectrum Disorder in Israel: A Negative Control Analysis.

BACKGROUND: Residual confounding is a major concern for causal inference in observational studies on air pollution-autism spectrum disorder (ASD) associations. This study is aimed at assessing confounding in these associations using negative control exposures. METHODS: This nested case-control study included all children diagnosed with ASD (detected through 31 December 2016) born during 2007-2012 in Israel and residing in the study area (N = 3,843), and matched controls of the same age (N = 38,430). We assigned individual house-level exposure estimates for each child. We estimated associations using logistic regression models, mutually adjusted for all relevant exposure periods (prepregnancy, pregnancy, and postnatal). We assessed residual confounding using postoutcome negative control exposure at age 28-36 months. RESULTS: In mutually adjusted models, we observed positive associations with ASD for postnatal exposures to NOx (odds ratio per interquartile range, 95% confidence interval: 1.19, 1.02-1.38) and NO2 (1.20, 1.00-1.43), and gestational exposure to PM2.5-10 (1.08, 1.01-1.15). The result for the negative control period was 1.04, 0.99-1.10 for PM2.5, suggesting some residual confounding, but no associations for PM2.5-10 (0.98, 0.81-1.18), NOx (1.02, 0.84-1.25), or NO2 (0.98, 0.81-1.18), suggesting no residual confounding. CONCLUSIONS: Our results further support a hypothesized causal link with ASD that is specific to postnatal exposures to traffic-related pollution.

Associations Between Prenatal Exposure to Air Pollution and Congenital Hypothyroidism.

Adequate thyroid hormone availability is required for normal brain development. Studies have found associations between prenatal exposure to air pollutants and thyroid hormones in pregnant women and newborns. We aimed to examine associations of trimester-specific residential exposure to common air pollutants with congenital hypothyroidism (CHT). All term infants born in Israel during 2009-2015 were eligible for inclusion. We used data on CHT from the national neonatal screening lab of Israel, and exposure data from spatiotemporal air pollution models. We used multivariable logistic regression models to estimate associations of exposures with CHT, adjusting for ethnicity, socioeconomic status, geographical area, conception season, conception year, gestational age, birth weight, and child sex. To assess residual confounding, we used postnatal exposures to the same pollutants as negative controls. The study population included 696,461 neonates. We found a positive association between third-trimester nitrogen oxide exposure and CHT (per interquartile-range change, odds ratio = 1.23, 95% confidence interval: 1.08, 1.41) and a similar association for nitrogen dioxide. There was no evidence of residual confounding or bias by correlation among exposure periods for these associations.

An impact of air pollution on moderate to severe relapses among multiple sclerosis patients.

BACKGROUND: Multiple Sclerosis (MS) is a chronic inflammatory disease of the central nervous system with both a genetic and environmental component. OBJECTIVE: In the current study, we examined an association between incidence of MS moderate to severe relapses and exposure to air pollutants and meteorological exposures. METHODS: We enrolled MS patients in Southern Israel during 2000-2017. Exposure assessment relied on satellite-based model of exposure to particulate matter of size <2.5 and 10 microns (PM2.5, PM10) and temperature at a spatial resolution of 1 km (Kloog et al., 2015). The information on exposure to nitrogen dioxide (NO2), sulfur dioxide (SO2) and ozone (O3) levels was completed from the database of the monitoring stations. We analyzed the data using a semi-ecological approach. The monthly incidence of MS-related relapses requiring hospitalization as a function of environmental factors was analyzed by time-series technique, adjusting to sex, age and smoking. We also used a case-crossover approach to compare environmental exposure of a patient on the day of the relapse with the exposure on the relapse-free days. All estimates were adjusted to the heat index and were divided by IQR. RESULTS: There were 287 MS patients in the study, with an average age of 52.8 ± 16.7 years, 37% of them (107) being under 40. Mostly female (66.2%), and 13.6% of the patients smoking (47% non-smoking and 39.4% unknown). PM2.5 was independently associated with MS relapses within the non-smoking population [Relative Risk (RR)=1.28, 95%CI:1.01-1.62]. O3 was found adversely associated with MS relapses among patients younger than 40 [RR=1.58, 95%CI 1.03-4.43]. Based on the case-crossover approach, relapses were associated with elevated levels of PM10 and NO2 in all subjects [Odds Ratio (OR)=1.05, 95%CI:1.00-1.11; OR=1.85, 95%CI: 1.28-2.68, respectively]. An adverse association with PM2.5 was observed in non-smokers [OR 1.12, 95%CI 1.00-1.25]. CONCLUSIONS: The findings show that MS relapses are adversely associated with an ambient exposure to PM and NO2.

Air pollution and meteorological conditions during gestation and type 1 diabetes in offspring.

BACKGROUND: Growing evidence indicates that air pollution is capable of disrupting the immune system and therefore, might be associated with an onset of Type 1 diabetes (T1D). OBJECTIVES: We explored possible links of T1D with ambient exposures in the population of southern Israel, characterized by hot and dry climate and frequent dust storms. METHODS: We conducted a matched nested case-control study where exposure to environmental pollutants during gestation in T1D cases was compared to that of healthy children. Up to 10 controls were matched to every case by age, gender and ethnicity, in all 362 cases and 3512 controls. Measurements of pollutants' concentrations, nitrogen dioxide (NO2), sulphur dioxide (SO2), ozone (O3) and particulate matter of size less than 10 and 2.5 µm in diameter (PM10 and PM2.5), as well as the mean daily measurements of meteorological conditions were obtained from the local monitoring stations. The association between T1D and pollution, solar radiation (SR), temperature and relative humidity was adjusted for socioeconomic status, temperature, maternal age and pre-gestational maternal DM, using conditional logistic regression. The environmental exposures were presented as indicators of quartiles averaged over whole pregnancy and by trimesters. RESULTS: Exposure to ozone and solar radiation during gestation were both associated with the T1D in offspring, although at borderline significance. Compared to the lowest quartile, the odds ratio (OR) for exposure to 3rd and 4th quartile of O3 was equal 1.61 (95%CI: 0.95; 2.73) and 1.45 (95%CI: 0.83; 2.53), respectively. Likewise, the ORs for exposure to SR were equal 1.83 (95%CI: 0.92; 3.64), 2.54 (95%CI: 1.21; 5.29) and 2.06 (95%CI: 0.95; 4.45) for to 2nd, 3rd and 4th quartiles, respectively. Exposure to SO2 followed a dose-response pattern, but was not statistically significant. Other environmental factors were not independently related to T1D. Analysis of exposures one year prior to the disease onset indicated a positive association between T1D and SR. CONCLUSIONS: We showed that exposure to high ozone levels and solar radiation during gestation might be related to the T1D. More scientific evidence needs to accumulate to support the study findings.

Extreme temperature and out-of-hospital-cardiac-arrest. Nationwide study in a hot climate country.

BACKGROUND: Out-of-hospital-cardiac arrest (OHCA) is frequently linked to environmental exposures. Climate change and global warming phenomenon have been found related to cardiovascular morbidity, however there is no agreement on their impact on OHCA occurrence. In this nationwide analysis, we aimed to assess the incidence of the OHCA events attended by emergency medical services (EMS), in relation to meteorological conditions: temperature, humidity, heat index and solar radiation. METHODS: We analyzed all adult cases of OHCA in Israel attended by EMS during 2016-2017. In the case-crossover design, we compared ambient exposure within 72 h prior to the OHCA event with exposure prior to the four control times using conditional logistic regression in a lag-distributed non-linear model. RESULTS: There were 12,401 OHCA cases (68.3% were pronounced dead-on-scene). The patients were on average 75.5 ± 16.2 years old and 55.8% of them were males. Exposure to 90th and 10th percentile of temperature adjusted to humidity were positively associated with the OHCA with borderline significance (Odds Ratio (OR) =1.20, 95%CI 0.97; 1.49 and OR 1.16, 95%CI 0.95; 1.41, respectively). Relative humidity below the 10th percentile was a risk factor for OHCA, independent of temperature, with borderline significance (OR = 1.16, 95%CI 0.96; 1.38). Analysis stratified by seasons revealed an adverse effect of exposure to 90th percentile of temperature when estimated in summer (OR = 3.34, 95%CI 1.90; 3.5.86) and exposure to temperatures below 10th percentile in winter (OR = 1.75, 95%CI 1.23; 2.49). Low temperatures during a warm season and high temperatures during a cold season had a protective effect on OHCA. The heat index followed a similar pattern, where an adverse effect was demonstrated for extreme levels of exposure. CONCLUSIONS: Evolving climate conditions characterized by excessive heat and low humidity represent risk factors for OHCA. As these conditions are easily avoided, by air conditioning and behavioral restrictions, necessary prevention measures are warranted.

The effect of exposure to particulate matter during pregnancy on lower respiratory tract infection hospitalizations during first year of life.

BACKGROUND: Lower respiratory tract infections (LRTI) in early life, including pneumonia, bronchitis and bronchiolitis, can lead to decreased lung function, persistent lung damage and increased susceptibility to various respiratory diseases such as asthma. In-utero exposure to particulate matter (PM) during pregnancy may disrupt biological mechanisms that regulate fetal growth, maturation and development. We aimed to estimate the association between intrauterine exposure to PM of size < 2.5 µm in diameter (PM2.5) and incidence of LRTIs during the first year of life. METHODS: A retrospective population-based cohort study in a population of mothers and infants born in Soroka University Medical Center (SUMC) in the years 2004-2012. All infants < 1 year old that were hospitalized due to LRTIs were included. The main exposure assessment was based on a hybrid model incorporating daily satellite-based predictions at 1 km2 spatial resolution. Data from monitoring stations was used for imputation of main exposure and other pollutants. Levels of environmental exposures were assigned to subjects based on their residential addresses and averaged for each trimester. Analysis was conducted by a multivariable generalized estimating equation (GEE) Poisson regression. Data was analyzed separately for the two main ethnic groups in the region, Jewish and Arab-Bedouin. RESULTS: The study cohort included 57,331 deliveries that met the inclusion criteria. Overall, 1871 hospitalizations of infants < 1 year old due to pneumonia or bronchiolitis were documented. In a multivariable analysis, intrauterine exposure to high levels of PM2.5 (> 24 µg/m3) in the first and second trimesters was found to be adversely associated with LRTIs in the Arab-Bedouin population (1st trimester, RR = 1.31, CI 95% 1.08-1.60; 2nd trimester: RR = 1.34, CI 95% 1.09-1.66). CONCLUSION: Intrauterine exposure to high levels of PM2.5 is associated with a higher risk of hospitalizations due to lower respiratory tract infections in Arab-Bedouin infants.

Temperature and preeclampsia: Epidemiological evidence that perturbation in maternal heat homeostasis affects pregnancy outcome.

INTRODUCTION: This study aims to determine the association between temperature and preeclampsia and whether it is affected by seasonality and rural/urban lifestyle. METHODS: This cohort study included women who delivered at our medical center from 2004 to 2013 (31,101 women, 64,566 deliveries). Temperature values were obtained from a spatiotemporally resolved estimation model performing predictions at a 1×1km spatial resolution. In "Warm" pregnancies >50% of gestation occurred during the spring-summer period. In cold pregnancies >50% of gestation occurred during the fall and winter. Generalized estimating equation multivariable models were used to estimate the association between temperature and incidence of preeclampsia. RESULTS: 1) The incidence of preeclampsia in at least one pregnancy was 7% (2173/64,566); 2) during "warm" pregnancies, an elevation of one IQR of the average temperature in the 1st or the 3rd trimesters was associated with an increased risk to develop preeclampsia [patients with Jewish ethnicity: 1st trimester: relative risk (RR) of 2.38(95%CI 1.50; 3.80), 3rd trimester 1.94(95%CI 1.34;2.81); Bedouins: 1st trimester: RR = 2.91(95%CI 1.98;4.28), 3rd trimester: RR = 2.37(95%CI 1.75;3.20)]; 3) In "cold" pregnancies, an elevation of one IQR of average temperature was associated with a lower risk to develop preeclampsia among patients with Bedouin-Arab ethnicity RR = 0.68 (95% CI 0.49-0.94) for 1st trimester and RR = 0.62 (95% CI 0.44-0.87) for 3rd trimester. CONCLUSIONS: 1) Elevated averaged temperature during the 1st or 3rd trimesters in "warm" pregnancies confer an increased risk for the development of preeclampsia, especially in nomadic patients; 2) Of interest, during cold pregnancies, elevated averaged temperature was associated with a lower risk to develop preeclampsia for nomadic patients. 3) These findings suggest temperature might be associated with perturbations in maternal heat homeostasis resulting in reallocation of energy resources and their availability to the fetus that may increase the risk for preeclampsia. This observation is especially relevant in the context of global warming and its effects on maternal/fetal reproductive health.

Estimating Daily PM2.5 and PM10 over Italy Using an Ensemble Model.

Spatiotemporally resolved particulate matter (PM) estimates are essential for reconstructing long and short-term exposures in epidemiological research. Improved estimates of PM2.5 and PM10 concentrations were produced over Italy for 2013-2015 using satellite remote-sensing data and an ensemble modeling approach. The following modeling stages were used: (1) missing values of the satellite-based aerosol optical depth (AOD) product were imputed using a spatiotemporal land-use random-forest (RF) model incorporating AOD data from atmospheric ensemble models; (2) daily PM estimations were produced using four modeling approaches: linear mixed effects, RF, extreme gradient boosting, and a chemical transport model, the flexible air quality regional model. The filled-in MAIAC AOD together with additional spatial and temporal predictors were used as inputs in the three first models; (3) a geographically weighted generalized additive model (GAM) ensemble model was used to fuse the estimations from the four models by allowing the weights of each model to vary over space and time. The GAM ensemble model outperformed the four separate models, decreasing the cross-validated root mean squared error by 1-42%, depending on the model. The spatiotemporally resolved PM estimations produced by the suggested model can be applied in future epidemiological studies across Italy.

Anthropogenic or non-anthropogenic particulate matter: Which one is more dangerous and how to differentiate between the effects?

The authors have observed that the function linking health outcomes with exposure to particulate-matter (PM) follows a biphasic pattern. It peaks around levels of PM10≤100 µg/m3, then weakens and rises again at PM10 levels in the range of hundreds. This could be due to a different nature of PM, the first peak reflecting a stronger anthropogenic and the second - weaker non-anthropogenic particles' effect. The current analysis is focused at the biphasic pattern on the association between PM levels with BG and asthma exacerbations. Pollutants were assessed by local monitoring stations and a satellitebased model. Local weekends/holidays were used to define nonanthropogenic levels of pollutants featured by lower Nitrogen Dioxide, the proxy for anthropogeneity. The association of PM10 with health outcomes within 24-48h lag was explored using spline functions of generalized additive models. Analysis of 546,420 BG tests (43,569 subjects) showed an almost linear association of PM10 with asthma with BG during the days with anthropogenic activity and no trend on other days. Analysis of asthmatic exacerbations within 1576 children showed no heterogeneity in association with PM10 by anthropogeneity levels, possibly indicating a mechanical impact on alveolar as the main trigger for exacerbations rather than PM10 chemical composition.

Predicting Fine-Scale Daily NO2 for 2005-2016 Incorporating OMI Satellite Data Across Switzerland.

Nitrogen dioxide (NO2) remains an important traffic-related pollutant associated with both short- and long-term health effects. We aim to model daily average NO2 concentrations in Switzerland in a multistage framework with mixed-effect and random forest models to respectively downscale satellite measurements and incorporate local sources. Spatial and temporal predictor variables include data from the Ozone Monitoring Instrument, Copernicus Atmosphere Monitoring Service, land use, and meteorological variables. We derived robust models explaining ∼58% (R2 range, 0.56-0.64) of the variation in measured NO2 concentrations using mixed-effect models at a 1 × 1 km resolution. The random forest models explained ∼73% (R2 range, 0.70-0.75) of the overall variation in the residuals at a 100 × 100 m resolution. This is one of the first studies showing the potential of using earth observation data to develop robust models with fine-scale spatial (100 × 100 m) and temporal (daily) variation of NO2 across Switzerland from 2005 to 2016. The novelty of this study is in demonstrating that methods originally developed for particulate matter can also successfully be applied to NO2. The predicted NO2 concentrations will be made available to facilitate health research in Switzerland.

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, PM2.5) and coarse particles (PM between 2.5 and 10 µm, PM2.5-10) at 1-km2 grid for 2013-2015 using a machine learning approach, the Random Forest (RF). Separate RF models were defined to: predict PM2.5 and PM2.5-10 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 PM2.5, respectively. Model fitting was less optimal for PM2.5-10, 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.

Correcting Measurement Error in Satellite Aerosol Optical Depth with Machine Learning for Modeling PM2.5 in the Northeastern USA.

Satellite-derived estimates of aerosol optical depth (AOD) are key predictors in particulate air pollution models. The multi-step retrieval algorithms that estimate AOD also produce quality control variables but these have not been systematically used to address the measurement error in AOD. We compare three machine-learning methods: random forests, gradient boosting, and extreme gradient boosting (XGBoost) to characterize and correct measurement error in the Multi-Angle Implementation of Atmospheric Correction (MAIAC) 1 × 1 km AOD product for Aqua and Terra satellites across the Northeastern/Mid-Atlantic USA versus collocated measures from 79 ground-based AERONET stations over 14 years. Models included 52 quality control, land use, meteorology, and spatially-derived features. Variable importance measures suggest relative azimuth, AOD uncertainty, and the AOD difference in 30-210 km moving windows are among the most important features for predicting measurement error. XGBoost outperformed the other machine-learning approaches, decreasing the root mean squared error in withheld testing data by 43% and 44% for Aqua and Terra. After correction using XGBoost, the correlation of collocated AOD and daily PM2.5 monitors across the region increased by 10 and 9 percentage points for Aqua and Terra. We demonstrate how machine learning with quality control and spatial features substantially improves satellite-derived AOD products for air pollution modeling.