Global Land Use Regression Model for Nitrogen Dioxide Air Pollution

Spatial estimation of surface ozone concentrations in Quito Ecuador with remote sensing data, air pollution measurements and meteorological variables

Surface ozone is problematic to air pollution. It influences respiratory health. The air quality monitoring stations measure pollutants as surface ozone, but they are sometimes insufficient or do not have an adequate distribution for understanding the spatial distribution of pollutants in an urban area. In recent years, some projects have found a connection between remote sensing, air quality and health data. In this study, we apply an empirical land use regression (LUR) model to retrieve surface ozone in Quito. The model considers remote sensing data, air pollution measurements and meteorological variables. The objective is to use all available Landsat 8 images from 2014 and the air quality monitoring station data during the same dates of image acquisition. Nineteen input variables were considered, selecting by a stepwise regression and modelling with a partial least square (PLS) regression to avoid multicollinearity. The final surface ozone model includes ten independent variables and presents a coefficient of determination (R²) of 0.768. The model proposed help to understand the spatial concentration of surface ozone in Quito with a better spatial resolution.

Residential green and blue space associated with better mental health: a pilot follow-up study in university students

Previous research has suggested that natural urban environment (green space and blue space) benefit mental health, but only a few longitudinal studies have explored the underlying mechanisms. In this pilot study we aimed to examine mechanisms/variables mediating associations between residential green/blue space and symptoms of anxiety/depression in 109 Bulgarian students from Plovdiv university. The students were followed from the beginning to the end of the school year (October 2017 to May 2018). Residential green space was defined as the mean of the normalised difference vegetation index (NDVI) in circular buffers of 100, 300, and 500 m around their residences. Blue space was assessed based on its presence in the same buffers. Levels of anxiety/depression were assessed using the 12-item General Health Questionnaire. The investigated mediator variables included residential noise (LAeq) and air pollution (NO2), environmental annoyance, perceived restorative quality of the neighbourhood, neighbourhood social cohesion, physical activity, and sleep disturbance. Cross-sectional data (obtained at baseline) showed that higher NDVI correlated with better mental health only indirectly through higher physical activity and restorative quality. Longitudinal (follow-up) data showed improved mental health but no significant effect of mediator variables. Similarly, blue space correlated with better mental health in all models, but physical activity and restorative quality were significant mediator variables only in the cross-sectional analysis. Our findings support that green space and blue space are psychologically restorative features in urban environment. Future research should replicate these findings in the general population and employ longitudinal modelling tailored to the specific mechanisms under study.

A picture tells a thousand…exposures: Opportunities and challenges of deep learning image analyses in exposure science and environmental epidemiology

BACKGROUND

Artificial intelligence (AI) is revolutionizing our world, with applications ranging from medicine to engineering.

OBJECTIVES

Here we discuss the promise, challenges, and probable data sources needed to apply AI in the fields of exposure science and environmental health. In particular, we focus on the use of deep convolutional neural networks to estimate environmental exposures using images and other complementary data sources such as cell phone mobility and social media information.

DISCUSSION

Characterizing the health impacts of multiple spatially-correlated exposures remains a challenge in environmental epidemiology. A shift toward integrated measures that simultaneously capture multiple aspects of the urban built environment could improve efficiency and provide important insights into how our collective environments influence population health. The widespread adoption of AI in exposure science is on the frontier. This will likely result in new ways of understanding environmental impacts on health and may allow for analyses to be efficiently scaled for broad coverage. Image-based convolutional neural networks may also offer a cost-effective means of estimating local environmental exposures in low and middle-income countries where monitoring and surveillance infrastructure is limited. However, suitable databases must first be assembled to train and evaluate these models and these novel approaches should be complemented with traditional exposure metrics.

CONCLUSIONS

The promise of deep learning in environmental health is great and will complement existing measurements for data-rich settings and could enhance the resolution and accuracy of estimates in data poor scenarios. Interdisciplinary partnerships will be needed to fully realize this potential.

Estimation of PM2.5 mortality burden in China with new exposure estimation and local concentration-response function

The estimation of PM_2.5-related mortality is becoming increasingly important. The accuracy of results is largely dependent on the selection of methods for PM_2.5 exposure assessment and Concentration-Response (C-R) function. In this study, PM_2.5 observed data from the China National Environmental Monitoring Center, satellite-derived estimation, widely collected geographic and socioeconomic information variables were applied to develop a national satellite-based Land Use Regression model and evaluate PM_2.5 exposure concentrations within 2013-2015 with the resolution of 1 km × 1 km. Population weighted concentration declined from 72.52 μg/m³ in 2013 to 57.18 μg/m³ in 2015. C-R function is another important section of health effect assessment, but most previous studies used the Integrated Exposure Regression (IER) function which may currently underestimate the excess relative risk of exceeding the exposure range in China. A new Shape Constrained Health Impact Function (SCHIF) method, which was developed from a national cohort of 189,793 Chinese men, was adopted to estimate the PM_2.5-related premature deaths in China. Results showed that 2.19 million (2013), 1.94 million (2014), 1.65 million (2015) premature deaths were attributed to PM_2.5 long-term exposure, different from previous understanding around 1.1-1.7 million. The top three provinces of the highest premature deaths were Henan, Shandong, Sichuan, while the least ones were Tibet, Hainan, Qinghai. The proportions of premature deaths caused by specific diseases were 53.2% for stroke, 20.5% for ischemic heart disease, 16.8% for chronic obstructive pulmonary disease and 9.5% for lung cancer. IER function was also used to calculate PM_2.5-related premature deaths with the same exposed level used in SCHIF method, and the comparison of results indicated that IER had made a much lower estimation with less annual amounts around 0.15-0.5 million premature deaths within 2013-2015.

Using MAIAC AOD to verify the PM2.5 spatial patterns of a land use regression model

Accurate spatial information of PM_2.5 is critical for air pollution control and epidemiological studies. Land use regression (LUR) models have been widely used for predicting spatial distribution of ground PM_2.5. However, the predicted PM_2.5 spatial patterns of a LUR model has not been adequately examined due to limited ground observations. The increasing aerosol optical depth (AOD) products might be an approximation of spatially continuous observation across large areas. This study established the relationship between seasonal 1 km × 1 km MAIAC AOD and observed ground PM_2.5 in Beijing, and then seasonal PM_2.5 maps were predicted based on AOD. Seasonal LUR models were also developed, and both the AOD and LUR models were validated by hold-out monitoring sites. Finally, the spatial patterns of LUR models were comprehensively verified by the above AOD PM_2.5 maps. The results showed that AOD alone could be used directly to predict the spatial distribution of ground PM_2.5 concentration at seasonal level (R² ≥ 0.53 in model fitting and testing), which was comparable with the capability of LUR models (R² ≥ 0.81 in model fitting and testing). PM_2.5 maps derived from the two methods showed similar spatial trend and coordinated variations near traffic roads. Large discrepancies could be observed at urban-rural transition areas where land use characters varied quickly. Variable and buffer size selection was critical for LUR model as they dominated the spatial patterns of predicted PM_2.5. Incorporating AOD into LUR model could improve model performance in spring season and provide more reliable results during testing.

Probabilistic Forecasting of Nitrogen Dioxide Concentrations at an Urban Road Intersection

The concentration of nitrogen dioxide in the air along a major route in a large city is affected by very many factors, which are also interdependent. As an alternative to complicated deterministic models based on these complex processes, in this study a probabilistic model for predicting NO2 concentrations is proposed, using a simple accounting cluster-based method for determining probability distributions for tabulated values of ambient factors. Using the example of hourly values of NO2 concentration and data on wind speed and traffic flow for the main intersection in Wrocław (Poland), a model is constructed to predict the frequency of occurrence of concentrations in the form of a probability distribution, for given values of the input variables. The model was successfully verified on data for the first six months of 2018. A mean continuous rank probability score (CRPS) of 9.15 μg/m3 was obtained. In spite of the greater impact of traffic volume on urban NO2 concentrations, as measured by Pearson’s correlation coefficient, for instance, the model indicates that wind speed is also a very important factor—wind being the principal mechanism causing the evacuation of pollutants. This underlines the importance of sustainable city planning with regard to ensuring suitable conditions for the passage of air.

Satellite-Based Land-Use Regression for Continental-Scale Long-Term Ambient PM2.5 Exposure Assessment in Australia

Australia has relatively diverse sources and low concentrations of ambient fine particulate matter (<2.5 μm, PM_2.5). Few comparable regions are available to evaluate the utility of continental-scale land-use regression (LUR) models including global geophysical estimates of PM_2.5, derived by relating satellite-observed aerosol optical depth to ground-level PM_2.5 ("SAT-PM_2.5"). We aimed to determine the validity of such satellite-based LUR models for PM_2.5 in Australia. We used global SAT-PM_2.5 estimates (∼10 km grid) and local land-use predictors to develop four LUR models for year-2015 (two satellite-based, two nonsatellite-based). We evaluated model performance at 51 independent monitoring sites not used for model development. An LUR model that included the SAT-PM_2.5 predictor variable (and six others) explained the most spatial variability in PM_2.5 (adjusted R² = 0.63, RMSE (μg/m³ [%]): 0.96 [14%]). Performance decreased modestly when evaluated (evaluation R² = 0.52, RMSE: 1.15 [16%]). The evaluation R² of the SAT-PM_2.5 estimate alone was 0.26 (RMSE: 3.97 [56%]). SAT-PM_2.5 estimates improved LUR model performance, while local land-use predictors increased the utility of global SAT-PM_2.5 estimates, including enhanced characterization of within-city gradients. Our findings support the validity of continental-scale satellite-based LUR modeling for PM_2.5 exposure assessment in Australia.

Multiple pathways link urban green- and bluespace to mental health in young adults

BACKGROUND

A growing body of scientific literature indicates that urban green- and bluespace support mental health; however, little research has attempted to address the complexities in likely interrelations among the pathways through which benefits plausibly are realized.

OBJECTIVES

The present study examines how different plausible pathways between green/bluespace and mental health can work together. Both objective and perceived measures of green- and bluespace are used in these models.

METHODS

We sampled 720 students from the city of Plovdiv, Bulgaria. Residential greenspace was measured in terms of the Normalized Difference Vegetation Index (NDVI), tree cover density, percentage of green areas, and Euclidean distance to the nearest green space. Bluespace was measured in terms of its presence in the neighborhood and the Euclidean distance to the nearest bluespace. Mental health was measured with the 12-item form of the General Health Questionnaire (GHQ-12). The following mediators were considered: perceived neighborhood green/bluespace, restorative quality of the neighborhood, social cohesion, physical activity, noise and air pollution, and environmental annoyance. Structural equation modelling techniques were used to analyze the data.

RESULTS

Higher NDVI within a 300 m buffer around the residence was associated with better mental health through higher perceived greenspace; through higher perceived greenspace, leading to increased restorative quality, and subsequently to increased physical activity (i.e., serial mediation); through lower noise exposure, which in turn was associated with lower annoyance; and through higher perceived greenspace, which was associated with lower annoyance. Presence of bluespace within a 300 m buffer did not have a straightforward association with mental health owing to competitive indirect paths: one supporting mental health through higher perceived bluespace, restorative quality, and physical activity; and another engendering mental ill-health through higher noise exposure and annoyance.

CONCLUSIONS

We found evidence that having more greenspace near the residence supported mental health through several indirect pathways with serial components. Conversely, bluespace was not clearly associated with mental health.

Pathways linking residential noise and air pollution to mental ill-health in young adults

BACKGROUND

Recent years have seen growing, but still tentative, evidence of the potential associations of environmental noise and air pollution with mental disorders. In the present study, we aimed to examine the associations between residential noise and air pollution exposures and general mental health in young adults with a focus on underlying processes METHODS: We sampled 720 students (18-35 years) from one university in the city of Plovdiv, Bulgaria. Residential noise (L_Aeq; day equivalent noise level) and air pollution (NO₂) were assessed at participant's residential address by land use regression models. General mental health was measured with a short form of the General Health Questionnaire (GHQ). The following putative mediators were considered: annoyance from environmental pollution, sleep disturbance, restorative quality of the neighborhood, neighborhood social cohesion, and commuting/leisure time physical activity. Structural equation modeling was used to analyze the theoretically-indicated interplay between exposures, mediators, and GHQ.

RESULTS

We observed an association between higher L_Aeq and GHQ, in which environmental annoyance and neighborhood restorative quality emerged as key mediators. First, L_Aeq was associated with higher annoyance, and through it with lower restorative quality, and then in turn with lower physical activity, and thus with higher GHQ. Simultaneously, higher annoyance was associated with higher sleep disturbance, and thereby with higher GHQ. NO₂ had no overall association with GHQ, but it was indirectly associated with it through higher annoyance, lower restorative quality, and lower physical activity working in serial.

CONCLUSION

We found evidence that increased residential noise was related to mental ill-health through several indirect pathways. Air pollution was associated with mental health only indirectly.

Estimates of the Global Burden of Ambient [Formula: see text], Ozone, and [Formula: see text] on Asthma Incidence and Emergency Room Visits

BACKGROUND

Asthma is the most prevalent chronic respiratory disease worldwide, affecting 358 million people in 2015. Ambient air pollution exacerbates asthma among populations around the world and may also contribute to new-onset asthma.

OBJECTIVES

We aimed to estimate the number of asthma emergency room visits and new onset asthma cases globally attributable to fine particulate matter ([Formula: see text]), ozone, and nitrogen dioxide ([Formula: see text]) concentrations.

METHODS

We used epidemiological health impact functions combined with data describing population, baseline asthma incidence and prevalence, and pollutant concentrations. We constructed a new dataset of national and regional emergency room visit rates among people with asthma using published survey data.

RESULTS

We estimated that 9–23 million and 5–10 million annual asthma emergency room visits globally in 2015 could be attributable to ozone and [Formula: see text], respectively, representing 8–20% and 4–9% of the annual number of global visits, respectively. The range reflects the application of central risk estimates from different epidemiological meta-analyses. Anthropogenic emissions were responsible for [Formula: see text] and 73% of ozone and [Formula: see text] impacts, respectively. Remaining impacts were attributable to naturally occurring ozone precursor emissions (e.g., from vegetation, lightning) and [Formula: see text] (e.g., dust, sea salt), though several of these sources are also influenced by humans. The largest impacts were estimated in China and India.

CONCLUSIONS

These findings estimate the magnitude of the global asthma burden that could be avoided by reducing ambient air pollution. We also identified key uncertainties and data limitations to be addressed to enable refined estimation. https://doi.org/10.1289/EHP3766.

Data Integration for the Assessment of Population Exposure to Ambient Air Pollution for Global Burden of Disease Assessment

Air pollution is a leading global disease risk factor. Tracking progress (e.g., for Sustainable Development Goals) requires accurate, spatially resolved, routinely updated exposure estimates. A Bayesian hierarchical model was developed to estimate annual average fine particle (PM_2.5) concentrations at 0.1° × 0.1° spatial resolution globally for 2010-2016. The model incorporated spatially varying relationships between 6003 ground measurements from 117 countries, satellite-based estimates, and other predictors. Model coefficients indicated larger contributions from satellite-based estimates in countries with low monitor density. Within and out-of-sample cross-validation indicated improved predictions of ground measurements compared to previous (Global Burden of Disease 2013) estimates (increased within-sample R² from 0.64 to 0.91, reduced out-of-sample, global population-weighted root mean squared error from 23 μg/m³ to 12 μg/m³). In 2016, 95% of the world's population lived in areas where ambient PM_2.5 levels exceeded the World Health Organization 10 μg/m³ (annual average) guideline; 58% resided in areas above the 35 μg/m³ Interim Target-1. Global population-weighted PM_2.5 concentrations were 18% higher in 2016 (51.1 μg/m³) than in 2010 (43.2 μg/m³), reflecting in particular increases in populous South Asian countries and from Saharan dust transported to West Africa. Concentrations in China were high (2016 population-weighted mean: 56.4 μg/m³) but stable during this period.

Comparison of land use regression models for NO2 based on routine and campaign monitoring data from an urban area of Japan

BACKGROUND

Typically, land use regression (LUR) models have been developed using campaign monitoring data rather than routine monitoring data. However, the latter have advantages such as low cost and long-term coverage. Based on the idea that LUR models representing regional differences in air pollution and regional road structures are optimal, the objective of this study was to evaluate the validity of LUR models for nitrogen dioxide (NO₂) based on routine and campaign monitoring data obtained from an urban area.

METHOD

We selected the city of Suita in Osaka (Japan). We built a model based on routine monitoring data obtained from all sites (routine-LUR-All), and a model based on campaign monitoring data (campaign-LUR) within the city. Models based on routine monitoring data obtained from background sites (routine-LUR-BS) and based on data obtained from roadside sites (routine-LUR-RS) were also built. The routine LUR models were based on monitoring networks across two prefectures (i.e., Osaka and Hyogo prefectures). We calculated the predictability of the each model. We then compared the predicted NO₂ concentrations from each model with measured annual average NO₂ concentrations from evaluation sites.

RESULTS

The routine-LUR-All and routine-LUR-BS models both predicted NO₂ concentrations well: adjusted R²=0.68 and 0.76, respectively, and root mean square error=3.4 and 2.1ppb, respectively. The predictions from the routine-LUR-All model were highly correlated with the measured NO₂ concentrations at evaluation sites. Although the predicted NO₂ concentrations from each model were correlated, the LUR models based on routine networks, and particularly those based on all monitoring sites, provided better visual representations of the local road conditions in the city.

CONCLUSION

The present study demonstrated that LUR models based on routine data could estimate local traffic-related air pollution in an urban area. The importance and usefulness of data from routine monitoring networks should be acknowledged.

Long-term nitrogen dioxide exposure assessment using back-extrapolation of satellite-based land-use regression models for Australia

Assessing historical exposure to air pollution in epidemiological studies is often problematic because of limited spatial and temporal measurement coverage. Several methods for modelling historical exposures have been described, including land-use regression (LUR). Satellite-based LUR is a recent technique that seeks to improve predictive ability and spatial coverage of traditional LUR models by using satellite observations of pollutants as inputs to LUR. Few studies have explored its validity for assessing historical exposures, reflecting the absence of historical observations from popular satellite platforms like Aura (launched mid-2004). We investigated whether contemporary satellite-based LUR models for Australia, developed longitudinally for 2006-2011, could capture nitrogen dioxide (NO₂) concentrations during 1990-2005 at 89 sites around the country. We assessed three methods to back-extrapolate year-2006 NO₂ predictions: (1) 'do nothing' (i.e., use the year-2006 estimates directly, for prior years); (2) change the independent variable 'year' in our LUR models to match the years of interest (i.e., assume a linear trend prior to year-2006, following national average patterns in 2006-2011), and; (3) adjust year-2006 predictions using selected historical measurements. We evaluated prediction error and bias, and the correlation and absolute agreement of measurements and predictions using R² and mean-square error R² (MSE-R²), respectively. We found that changing the year variable led to best performance; predictions captured between 41% (1991; MSE-R² = 31%) and 80% (2003; MSE-R² = 78%) of spatial variability in NO₂ in a given year, and 76% (MSE-R² = 72%) averaged over 1990-2005. We conclude that simple methods for back-extrapolating prior to year-2006 yield valid historical NO₂ estimates for Australia during 1990-2005. These results suggest that for the time scales considered here, satellite-based LUR has a potential role to play in long-term exposure assessment, even in the absence of historical predictor data.

Satellite-Based Estimates of Daily NO2 Exposure in China Using Hybrid Random Forest and Spatiotemporal Kriging Model

A novel model named random-forest-spatiotemporal-kriging (RF-STK) was developed to estimate the daily ambient NO₂ concentrations across China during 2013-2016 based on the satellite retrievals and geographic covariates. The RF-STK model showed good prediction performance, with cross-validation R² = 0.62 (RMSE = 13.3 μg/m³) for daily and R² = 0.73 (RMSE = 6.5 μg/m³) for spatial predictions. The nationwide population-weighted multiyear average of NO₂ was predicted to be 30.9 ± 11.7 μg/m³ (mean ± standard deviation), with a slowly but significantly decreasing trend at a rate of -0.88 ± 0.38 μg/m³/year. Among the main economic zones of China, the Pearl River Delta showed the fastest decreasing rate of -1.37 μg/m³/year, while the Beijing-Tianjin Metro did not show a temporal trend ( P = 0.32). The population-weighted NO₂ was predicted to be the highest in North China (40.3 ± 10.3 μg/m³) and lowest in Southwest China (24.9 ± 9.4 μg/m³). Approximately 25% of the population lived in nonattainment areas with annual-average NO₂ > 40 μg/m³. A piecewise linear function with an abrupt point around 100 people/km² characterized the relationship between the population density and the NO₂, indicating a threshold of aggravated NO₂ pollution due to urbanization. Leveraging the ground-level NO₂ observations, this study fills the gap of statistically modeling nationwide NO₂ in China, and provides essential data for epidemiological research and air quality management.

Towards Personal Exposures: How Technology Is Changing Air Pollution and Health Research

PURPOSE OF REVIEW

We present a review of emerging technologies and how these can transform personal air pollution exposure assessment and subsequent health research.

RECENT FINDINGS

Estimating personal air pollution exposures is currently split broadly into methods for modeling exposures for large populations versus measuring exposures for small populations. Air pollution sensors, smartphones, and air pollution models capitalizing on big/new data sources offer tremendous opportunity for unifying these approaches and improving long-term personal exposure prediction at scales needed for population-based research. A multi-disciplinary approach is needed to combine these technologies to not only estimate personal exposures for epidemiological research but also determine drivers of these exposures and new prevention opportunities. While available technologies can revolutionize air pollution exposure research, ethical, privacy, logistical, and data science challenges must be met before widespread implementations occur. Available technologies and related advances in data science can improve long-term personal air pollution exposure estimates at scales needed for population-based research. This will advance our ability to evaluate the impacts of air pollution on human health and develop effective prevention strategies.

Blending Multiple Nitrogen Dioxide Data Sources for Neighborhood Estimates of Long-Term Exposure for Health Research

Exposure to traffic related nitrogen dioxide (NO₂) air pollution is associated with adverse health outcomes. Average pollutant concentrations for fixed monitoring sites are often used to estimate exposures for health studies, however these can be imprecise due to difficulty and cost of spatial modeling at the resolution of neighborhoods (e.g., a scale of tens of meters) rather than at a coarse scale (around several kilometers). The objective of this study was to derive improved estimates of neighborhood NO₂ concentrations by blending measurements with modeled predictions in Sydney, Australia (a low pollution environment). We implemented the Bayesian maximum entropy approach to blend data with uncertainty defined using informative priors. We compiled NO₂ data from fixed-site monitors, chemical transport models, and satellite-based land use regression models to estimate neighborhood annual average NO₂. The spatial model produced a posterior probability density function of estimated annual average concentrations that spanned an order of magnitude from 3 to 35 ppb. Validation using independent data showed improvement, with root mean squared error improvement of 6% compared with the land use regression model and 16% over the chemical transport model. These estimates will be used in studies of health effects and should minimize misclassification bias.