Estimation of surface-level NO2 and O3 concentrations using TROPOMI data and machine learning over East Asia

Capabilities of satellite Geostationary Environment Monitoring Spectrometer (GEMS) NO2 data for hourly ambient NO2 exposure modeling

The Geostationary Environment Monitoring Spectrometer (GEMS) is the world's first geostationary instrument that monitors hourly gaseous air pollutant levels, including nitrogen dioxide (NO2). Using the first-of-its-kind capabilities of GEMS NO2 data, we examined how well GEMS NO2 levels can explain the spatiotemporal variabilities in hourly NO2 concentrations in the Republic of Korea for the year 2022. A correlation analysis between hourly GEMS NO2 levels and ground NO2 concentrations showed a higher spatial correlation [Pearson r = 0.56 (SD = 0.20)] than a temporal one [Pearson r = 0.42 (SD = 0.14)], on average. To take advantage of the enhanced spatial predictability of GEMS NO2 data, we employed a mixed effects model to allow hour-specific relationships between GEMS NO2 and NO2 concentrations on a given day in each region and subsequently estimated hourly NO2 concentrations in all urban and rural areas. The 10-fold cross validation demonstrated R2 = 0.72, mean absolute error (MAE) = 3.7 ppb, and root mean squared error (RMSE) = 5.5 ppb. The hourly variations of the relationships were attributed particularly to those of wind speed among meteorological parameters considered in this study. The spatial distributions of hourly estimated NO2 concentrations were highly correlated between hours [average r = 0.91 (SD = 0.06)]. Nonetheless, they represented the diurnal patterns of urban versus rural NO2 contrasts during the day [urban/rural NO2 ratios from 1.22 (5 p.m.) to 1.37 (12 p.m.)]. The newly retrieved GEMS NO2 data enable temporally as well as spatially resolved NO2 exposure assessment. In combination with the time-activity patterns of individual subjects, the GEMS NO2 data can generate 'sub-population' exposure estimates and therefore enhance health effect studies.

Understanding the variability of ground-level ozone and fine particulate matter over the Tibetan plateau with data-driven approach

The Tibetan Plateau, known as the "Third Pole", is susceptible to ground-level ozone (O3) and fine particulate matter (PM2.5) pollution due to its unique high-altitude environment. This study constructed random forest regression models using multi-source data from ground measurements and meteorological satellites to predict variations in ground-level O3 and PM2.5 concentrations and their influencing factors across seven major cities in the Tibetan Plateau over two-year periods. The models successfully reproduced O3 and PM2.5 levels with satisfactory R-squared values of 0.71 and 0.73, respectively. Results reveal combustion-related carbon monoxide (CO) and nitrogen dioxide (NO2) as the most substantial influences on O3 and PM2.5 concentrations. Solar radiation, geographical factors, and meteorological variables also played crucial roles in driving pollutant variations. Conversely, transport-related and human activity factors exhibited relatively lower significance. High O3 and PM2.5 pollution occurred during pre-monsoon and post-monsoon/winter seasons, driven by solar radiation and emissions, respectively. While CO consistently contributed across cities and seasons, key influencing factors varied locally. This study unveils the key driving forces governing air pollutant variations across the Tibetan Plateau, shedding light on complex atmospheric processes in this unique high-altitude region.

Artificial intelligence for geoscience: Progress, challenges, and perspectives

This paper explores the evolution of geoscientific inquiry, tracing the progression from traditional physics-based models to modern data-driven approaches facilitated by significant advancements in artificial intelligence (AI) and data collection techniques. Traditional models, which are grounded in physical and numerical frameworks, provide robust explanations by explicitly reconstructing underlying physical processes. However, their limitations in comprehensively capturing Earth's complexities and uncertainties pose challenges in optimization and real-world applicability. In contrast, contemporary data-driven models, particularly those utilizing machine learning (ML) and deep learning (DL), leverage extensive geoscience data to glean insights without requiring exhaustive theoretical knowledge. ML techniques have shown promise in addressing Earth science-related questions. Nevertheless, challenges such as data scarcity, computational demands, data privacy concerns, and the "black-box" nature of AI models hinder their seamless integration into geoscience. The integration of physics-based and data-driven methodologies into hybrid models presents an alternative paradigm. These models, which incorporate domain knowledge to guide AI methodologies, demonstrate enhanced efficiency and performance with reduced training data requirements. This review provides a comprehensive overview of geoscientific research paradigms, emphasizing untapped opportunities at the intersection of advanced AI techniques and geoscience. It examines major methodologies, showcases advances in large-scale models, and discusses the challenges and prospects that will shape the future landscape of AI in geoscience. The paper outlines a dynamic field ripe with possibilities, poised to unlock new understandings of Earth's complexities and further advance geoscience exploration.

Do Machine Learning Methods Improve Prediction of Ambient Air Pollutants with High Spatial Contrast? A Systematic Review

BACKGROUND & OBJECTIVE

The use of machine learning for air pollution modelling is rapidly increasing. We conducted a systematic review of studies comparing statistical and machine learning models predicting the spatiotemporal variation of ambient nitrogen dioxide (NO2), ultrafine particles (UFPs) and black carbon (BC) to determine whether and in which scenarios machine learning generates more accurate predictions.

METHODS

Web of Science and Scopus were searched up to June 13, 2024. All records were screened by two independent reviewers. Differences in the coefficient of determination (R2) and Root Mean Square Error (RMSE) between best statistical and machine learning methods were compared across categories of methodological elements.

RESULTS

A total of 38 studies with 46 model comparisons (30 for NO2, 8 for UFPs and 8 for BC) were included. Linear non-regularized methods and Random Forest were most frequently used. Machine learning outperformed statistical models in 34 comparisons. Mean differences (95% confidence intervals) in R2 and RMSE between best machine learning and statistical models were 0.12 (0.08, 0.17) and 20% (11%, 29%) respectively. Tree-based methods performed best in 12 of 17 multi-model comparisons. Nonlinear or regularization regression methods were used in only 12 comparisons and provided similar performance to machine learning methods.

CONCLUSION

This systematic review suggests that machine learning methods, especially tree-based methods, may be superior to linear non-regularized methods for predicting ambient concentrations of NO2, UFPs and BC. Additional comparison studies using nonlinear, regularized and a wider array of machine learning methods are needed to confirm their relative performance. Future air pollution studies would also benefit from more explicit and standardized reporting of methodologies and results.

High-spatial resolution ground-level ozone in Yunnan, China: A spatiotemporal estimation based on comparative analyses of machine learning models

Monitoring ground-level ozone concentrations is a critical aspect of atmospheric environmental studies. Given the existing limitations of satellite data products, especially the lack of ground-level ozone characterization, and the discontinuity of ground observations, there is a pressing need for high-precision models to simulate ground-level ozone to assess surface ozone pollution. In this study, we have compared several widely utilized ensemble learning and deep learning methods for ground-level ozone simulation. Furthermore, we have thoroughly contrasted the temporal and spatial generalization performances of the ensemble learning and deep learning models. The 3-Dimensional Convolutional Neural Network (3-D CNN) model has emerged as the optimal choice for evaluating the daily maximum 8-h average ozone in Yunnan Province. The model has good performance: a spatial resolution of 0.05° × 0.05° and strong predictive power, as indicated by a Coefficient of Determination (R2) of 0.83 and a Root Mean Square Error (RMSE) of 12.54 μg/m³ in sample-based 5-fold cross-validation (CV). In the final stage of our study, we applied the 3-D CNN model to generate a comprehensive daily maximum 8-h average ozone dataset for Yunnan Province for the year 2021. This application has furnished us with a crucial high-resolution and highly accurate dataset for further in-depth studies on the issue of ozone pollution in Yunnan Province.

Exploring the spatial effects and influencing mechanism of ozone concentration in the Yangtze River Delta urban agglomerations of China

Ground-level ozone (O3) pollution has emerged as a significant concern impacting air quality in urban agglomerations, primarily driven by meteorological conditions and social-economic factors. However, previous studies have neglected to comprehensively reveal the spatial distribution and driving mechanism of O3 pollution. Based on the O3 monitoring data of 41 cities in the Yangtze River Delta (YRD) from 2014 to 2021, a comprehensive analysis framework of spatial analysis-spatial econometric regression was constructed to reveal the driving mechanism of O3 pollution. The results revealed the following: (1) O3 concentrations in the YRD exhibited a general increasing and then decreasing trend, indicating an improvement in pollution levels. The areas with higher O3 concentration are mainly the cities concentrated in central and southern Jiangsu, Shanghai, and northern Zhejiang. (2) The change of O3 concentration and distribution is the result of various factors. The effect of urbanization on O3 concentrations followed an inverted U-shaped curve, which implies that achieving higher quality urbanization is essential for effectively controlling urban O3 pollution. Traffic conditions and energy consumption have significant direct positive influences on O3 concentrations and spatial spillover effects. The indirect pollution contribution, considering economic weight, accounted for about 35%. Thus, addressing overall regional energy consumption and implementing traffic source regulations are crucial paths for O3 pollution control in the YRD. (3) Meteorological conditions play a certain role in regulating the O3 concentration. Higher wind speed will promote the diffusion of O3 and increase the O3 concentration in the surrounding city. These findings provide valuable insights for designing effective policies to improve air quality and mitigate ozone pollution in urban agglomeration area.

A novel approach to deriving the fine-scale daily NO2 dataset during 2005-2020 in China: Improving spatial resolution and temporal coverage to advance exposure assessment

Surface NO2 pollution can result in serious health consequences such as cardiovascular disease, asthma, and premature mortality. Due to the extensive spatial variation in surface NO2, the spatial resolution of a NO2 dataset has a significant impact on the exposure and health impact assessment. There is currently no long-term, high-resolution, and publicly available NO2 dataset for China. To fill this gap, this study generated a NO2 dataset named RBE-DS-NO2 for China during 2005-2020 at 1 km and daily resolution. We employed the robust back-extrapolation via a data augmentation approach (RBE-DA) to ensure the predictive accuracy in back-extrapolation before 2013, and utilized an improved spatial downscaling technique (DS) to refine the spatial resolution from 10 km to 1 km. Back-extrapolation validation based on 2005-2012 observations from sites in Taiwan province yielded an R2 of 0.72 and RMSE of 10.7 μg/m3, while cross-validation across China during 2013-2020 showed an R2 of 0.73 and RMSE of 9.6 μg/m3. RBE-DS-NO2 better captured spatiotemporal variation of surface NO2 in China compared to the existing publicly available datasets. Exposure assessment using RBE-DS-NO2 show that the population living in non-attainment areas (NO2 ≥ 30 μg/m3) grew from 376 million in 2005 to 612 million in 2012, then declined to 404 million by 2020. Unlike this national trend, exposure levels in several major cities (e.g., Shanghai and Chengdu) continued to increase during 2012-2020, driven by population growth and urban migration. Furthermore, this study revealed that low-resolution dataset (i.e., the 10 km intermediate dataset before the downscaling) overestimated NO2 levels, due to the limited specificity of the low-resolution model in simulating the relationship between NO2 and the predictor variables. Such limited specificity likely biased previous long-term NO2 exposure and health impact studies employing low-resolution datasets. The RBE-DS-NO2 dataset enables robust long-term assessments of NO2 exposure and health impacts in China.

Explainable ensemble machine learning revealing the effect of meteorology and sources on ozone formation in megacity Hangzhou, China

Megacity Hangzhou, located in eastern China, has experienced severe O3 pollution in recent years, thereby clarifying the key drivers of the formation is essential to suppress O3 deterioration. In this study, the ensemble machine learning model (EML) coupled with Shapley additive explanations (SHAP), and positive matrix factorization were used to explore the impact of various factors (including meteorology, chemical components, sources) on O3 formation during the whole period, pollution days, and typical persistent pollution events from April to October in 2021-2022. The EML model achieved better performance than the single model, with R2 values of 0.91. SHAP analysis revealed that meteorological conditions had the greatest effects on O3 variability with the contribution of 57 %-60 % for different pollution levels, and the main drivers were relative humidity and radiation. The effects of chemical factors on O3 formation presented a positive response to volatile organic compounds (VOCs) and fine particulate matter (PM2.5), and a negative response to nitrogen oxides (NOx). Oxygenated compounds (OVOCs), alkenes, and aromatic of VOCs subgroups had higher contribution; additionally, the effects of PM2.5 and NOx were also important and increased with the O3 deterioration. The impact of seven emission sources on O3 formation in Hangzhou indicated that vehicle exhaust (35 %), biomass combustion (16 %), and biogenic emissions (12 %) were the dominant drivers. However, for the O3 pollution days, the effects of biomass combustion and biogenic emissions increased. Especially in persistent pollution events with highest O3 concentrations, the magnitude of biogenic emission effect elevated significantly by 156 % compared to the whole situations. Our finding revealed that the combination of the EML model and SHAP analysis could provide a reliable method for rapid diagnosis of the cause of O3 pollution at different event scales, supporting the formulation of control measures.

Estimating morning and evening commute period O3 concentration in Taiwan using a fine spatial-temporal resolution ensemble mixed spatial model with Geo-AI technology

Elevated levels of ground-level ozone (O3) can have harmful effects on health. While previous studies have focused mainly on daily averages and daytime patterns, it's crucial to consider the effects of air pollution during daily commutes, as this can significantly contribute to overall exposure. This study is also the first to employ an ensemble mixed spatial model (EMSM) that integrates multiple machine learning algorithms and predictor variables selected using Shapley Additive exExplanations (SHAP) values to predict spatial-temporal fluctuations in O3 concentrations across the entire island of Taiwan. We utilized geospatial-artificial intelligence (Geo-AI), incorporating kriging, land use regression (LUR), machine learning (random forest (RF), categorical boosting (CatBoost), gradient boosting (GBM), extreme gradient boosting (XGBoost), and light gradient boosting (LightGBM)), and ensemble learning techniques to develop ensemble mixed spatial models (EMSMs) for morning and evening commute periods. The EMSMs were used to estimate long-term spatiotemporal variations of O3 levels, accounting for in-situ measurements, meteorological factors, geospatial predictors, and social and seasonal influences over a 26-year period. Compared to conventional LUR-based approaches, the EMSMs improved performance by 58% for both commute periods, with high explanatory power and an adjusted R2 of 0.91. Internal and external validation procedures and verification of O3 concentrations at the upper percentile ranges (in 1%, 5%, 10%, 15%, 20%, and 25%) and other conditions (including rain, no rain, weekday, weekend, festival, and no festival) have demonstrated that the models are stable and free from overfitting issues. Estimation maps were generated to examine changes in O3 levels before and during the implementation of COVID-19 restrictions. These findings provide accurate variations of O3 levels in commute period with high spatiotemporal resolution of daily and 50m * 50m grid, which can support control pollution efforts and aid in epidemiological studies.

Using deep transfer learning and satellite imagery to estimate urban air quality in data-poor regions

Urban air pollution is a critical public health challenge in low-and-middle-income countries (LMICs). At the same time, LMICs tend to be data-poor, lacking adequate infrastructure to monitor air quality (AQ). As LMICs undergo rapid urbanization, the socio-economic burden of poor AQ will be immense. Here we present a globally scalable two-step deep learning (DL) based approach for AQ estimation in LMIC cities that mitigates the need for extensive AQ infrastructure on the ground. We train a DL model that can map satellite imagery to AQ in high-income countries (HICs) with sufficient ground data, and then adapt the model to learn meaningful AQ estimates in LMIC cities using transfer learning. The trained model can explain up to 54% of the variation in the AQ distribution of the target LMIC city without the need for target labels. The approach is demonstrated for Accra in Ghana, Africa, with AQ patterns learned and adapted from two HIC cities, specifically Los Angeles and New York.

Contrasting changes in ozone during 2019-2021 between eastern and the other regions of China attributed to anthropogenic emissions and meteorological conditions

Ozone pollution is one of the most severe air quality issues in China that poses a serious threat to human health and ecosystems. During 2019-2021, the maximum daily 8-h average ozone concentrations in eastern China (110-122.5°E, 26-42°N) and the rest of China (ROC) show different decreasing patterns, with ozone concentrations in eastern China decreasing by 14.9 μg/m3, which is much larger than 4.8 μg/m3 in ROC. Here, based on two independent methods, the atmospheric chemical transport model (GEOS-Chem) simulations and the machine learning (ML) model (LightGBM) predictions, the reasons for the differences in ozone changes between eastern China and ROC during the warm season (April to September) are investigated. According to the GEOS-Chem (LightGBM) results, changes in the meteorological conditions contributed to an ozone decrease by 7.3 (6.8) μg/m3 in eastern China due to decreased chemical production and an ozone decrease by 6.8 (7.0) μg/m3 in ROC attributed to the weakened horizontal and vertical advection. With the influence of meteorological factors excluded, the observations show that changes in anthropogenic emissions resulted in an ozone decrease by 7.6 (8.1) μg/m3 in eastern China and an ozone increase by 2.0 (2.2) μg/m3 in ROC, which is primarily induced by the changes in NOx emissions. The surface measurements and satellite retrievals also indicate that the reduction in NOx emissions in ROC is less efficient than that in the more developed eastern China, leading to contrasting changes in ozone concentrations between eastern China and ROC during 2019-2021. Our results highlight the critical need to reduce ozone precursor emissions in the rest regions of China apart from eastern China.

Long-term spatiotemporal variations in surface NO2 for Beijing reconstructed from surface data and satellite retrievals

Remote sensing data from the Ozone Monitoring Instrument (OMI) and the TROPOspheric Monitoring Instrument (TROPOMI) play important roles in estimating surface nitrogen dioxide (NO2), but few studies have compared their differences for application in surface NO2 reconstruction. This study aims to explore the effectiveness of incorporating the tropospheric NO2 vertical column density (VCD) from OMI and TROPOMI (hereafter referred to as OMI and TROPOMI, respectively, for conciseness) for deriving surface NO2 and to apply the resulting data to revisit the spatiotemporal variations in surface NO2 for Beijing over the 2005-2020 period during which there were significant reductions in nitrogen oxide emissions. In the OMI versus TROPOMI performance comparison, the cross-validation R2 values were 0.73 and 0.72, respectively, at 1 km resolution and 0.69 for both at 100 m resolution. The comparisons between satellite data sources indicate that even though TROPOMI has a finer resolution it does not improve upon OMI for deriving surface NO2 at 1 km resolution, especially for analyzing long-term trends. In light of the comparison results, we used a hybrid approach based on machine learning to derive the spatiotemporal distribution of surface NO2 during 2005-2020 based on OMI. We had novel, independent passive sampling data collected weekly from July to September of 2008 for hindcasting validation and found a spatiotemporal R2 of 0.46 (RMSE = 7.0 ppb). Regarding the long-term trend of surface NO2, the level in 2008 was obviously lower than that in 2007 and 2009, as expected, which was attributed to pollution restrictions during the Olympic Games. The NO2 level started to steadily decline from 2015 and fell below 2008's level after 2017. Based on OMI, a long-term and fine-resolution surface NO2 dataset was developed for Beijing to support future environmental management questions and epidemiological research.

NO2 emissions from oil refineries in the Mississippi Delta

Of the >17,943 thousand barrels per calendar day (bbl/d) of oil refining capacity located in the US, the Petroleum Administration for Defense District 3 (PADD-3) region has the largest number of refineries and accounts for >53 % (or 9607 tbbl/d) of all US oil refining capacity. Processing facilities in this area are mainly located on the Gulf of Mexico coast in Texas and Louisiana. This study selected a sub-region for analysis within the Mississippi River delta in the state of Louisiana between the cities of New Orleans and Baton Rouge. This region is characterized by intensive industrial activity connected with oil refining and related activities. The TROPOspheric Monitoring Instrument (TROPOMI) detected highly localized NO2 vertical column densities (VCDs) over the two largest US refineries in Baton Rouge (503,000 bbl/d) and Garyville (578,000 bbl/d). TROPOMI NO2 VCD over these stations were 100 μmol/m2 and 80 μmol/m2, respectively. A high correlation coefficient (r = 0.65, p < 0.05) was also found between TROPOMI NO2 and population density. Data from the National Emissions Inventory (NEI) showed high NOx emissions from refineries and other industries including coal-fired power generation, chemical, and aluminum processing plants. The results of the NO2 analysis are of practical interest for a comparative assessment of air pollution, as well as for the exchange of best practices in the field of low-waste fuel combustion technologies.

Estimate of near-surface NO2 concentrations in Fenwei Plain, China, based on TROPOMI data and random forest model

Nitrogen dioxide (NO2) concentration is a crucial indicator of ground-level air quality, and elevated concentrations can adversely affect human health and the atmospheric environment. In this study, we utilized Tropospheric Monitoring Instrument (TROPOMI) tropospheric NO2 vertical column density data (VCD) and multi-source geographic data to establish a random forest regression (RF) model that accurately estimates NO2 concentrations near the ground in the Fenwei Plain. The model addresses the inherent limitations of traditional ground-based monitoring and provides data support for analyzing regional pollution spatial and temporal characteristics. (1) The RF model based on TROPOMI and geographic data demonstrates high estimation accuracy, with monthly average RF model fit and validation coefficient of determination (R2) reaching 0.949 and 0.875, respectively. (2) A complex nonlinear relationship exists between near-surface NO2 concentration and multi-source geographic data. The RF model's estimations reveal clear seasonal and regional variations in near-surface NO2 concentration. Concentrations are generally highest in winter, followed by spring and autumn, and lowest in summer. The high NO2 concentrations are primarily mainly distributed in the plains and river valleys with low elevation and dense population density. The model estimation results also indicate that the estimated effect is better when the NO2 concentration fluctuates less and anthropogenic emission reduction measures significantly impact the NO2 concentration near the ground. (3) The population exposure risk results indicate that most cities in the Fenwei Plain face varying exposure risks. These findings offer valuable insights for regional NO2 pollution management.

Estimation of ground-level O3 concentration in the Yangtze River Delta region based on a high-performance spatiotemporal model MixNet

In recent years, the escalating ozone (O3) concentration has significantly damaged human health. The machine learning models are widely used to estimate ground-level O3 concentrations, but the spatial and temporal features in the data are less considered. To address the issue, this study proposed a novel framework named MixNet to estimate daily O3 concentration from 2020 to 2021 over the Yangtze River Delta. The MixNet utilized image convolution to extract the potential spatial information related to O3 fully. The temporal features were extracted by a Long Short-Term Memory (LSTM). A U-Net, a new jump connection method with an attention mechanism and residual blocks, facilitated a more comprehensive extraction of spatial features in the data. The extracted temporal and spatial features were fused to estimate ground-level O3. Meanwhile, a novel training method was proposed to enhance the accuracy of MixNet. The daily mean O3 maps have high validation results in comparison with ground-level O3 measurement, with R2 (RMSE) of 0.903 (14.511 μg/m3) for sample-based validation, 0.831 (19.036 μg/m3) for site-based validation, and 0.712 (25.108 μg/m3) for time-based validation. The season-average maps indicate that O3 concentration is summer > autumn > spring > winter. The highest value was 137.41 μg/m3 in the summer of 2021 over the Yangtze River Delta urban agglomeration, and the lowest value was 52.73 μg/m3 in winter 2020. The MixNet showed better performance compared with other models, and thus the "point-plane image thinking" will contribute to future studies in developing better methods to estimate atmospheric pollutants.

Assessment of variations in air quality in cities of Ecuador in relation to the lockdown due to the COVID-19 pandemic

This study analyzes the effect of lockdown due to COVID-19 on the spatiotemporal variability of ozone (O₃), sulfur dioxide (SO₂), and nitrogen dioxide (NO₂) concentrations in different provinces of continental Ecuador using satellite information from Sentinel - 5P. The statistical analysis includes data from 2018 to March 2021 and was performed based on three periods defined a priori: before, during, and after lockdown due to COVID-19, focusing on the provinces with the highest concentrations of the studied gases (hotspots). The results showed a significant decrease in NO₂ concentrations during the COVID-19 lockdown period in all the study areas: the Metropolitan District of Quito (DMQ) and the provinces of Guayas and Santo Domingo de los Tsáchilas. In the period after lockdown, NO₂ concentrations increased by over 20% when compared to the pre-lockdown period, which may be attributable to a shift towards private transportation due to health concerns. On the other hand, SO₂ concentrations during the lockdown period showed irregular, non-significant variations; however, increases were observed in the provinces of Chimborazo, Guayas, Santa Elena, and Morona Santiago, which could be partly attributed to the eruptive activity of the Sangay volcano during 2019-2020. Conversely, O₃ concentrations increased by 2-3% in the study areas; this anomalous behavior could be attributed to decreased levels of NO_x, which react with ozone, reducing its concentration. Finally, satellite data validation using the corresponding data from monitoring stations in the DMQ showed correlation values of 0.9 for O₃ data and 0.7 for NO₂ data, while no significant correlation was found for SO₂.

Deep learning mapping of surface MDA8 ozone: The impact of predictor variables on ozone levels over the contiguous United States

The limited number of ozone monitoring stations imposes uncertainty in various applications, calling for accurate approaches to capturing ozone values in all regions, particularly those with no in-situ measurements. This study uses deep learning (DL) to accurately estimate daily maximum 8-hr average (MDA8) ozone and examines the spatial contribution of several factors on ozone levels over the contiguous U.S. (CONUS) in 2019. A comparison between in-situ observations and DL-estimated MDA8 ozone values shows a correlation coefficient (R) of 0.95, an index of agreement (IOA) of 0.97, and a mean absolute bias (MAB) of 2.79 ppb, highlighting the promising performance of the deep convolutional neural network (Deep-CNN) at estimating surface MDA8 ozone. Spatial cross-validation also confirms the high spatial accuracy of the model, which obtains an R of 0.91, and IOA of 0.96 and an MAB of 3.46 ppb when it is trained and tested on separate stations. To interpret the black-box nature of our DL model, we use Shapley additive explanations (SHAP) to generate a spatial feature contribution map (SFCM), the results of which confirm an advanced ability of Deep-CNN to capture the interactions between most predictor variables and ozone. For instance, the model shows that solar radiation (SRad) SFCM, with higher values, enhances the formation of ozone, particularly in the south and southwestern CONUS. As SRad triggers ozone precursors to produce ozone via photochemical reactions, it increases ozone concentrations. The model also shows that humidity, with its low values, increases ozone concentrations in the western mountainous regions. The negative correlation between humidity and ozone levels can be attributed to factors such as higher ozone decomposition resulting from increased levels of humidity and OH radicals. This study is the first to introduce the SFCM to investigate the spatial role of predictor variables on changes in estimated MDA8 ozone levels.

Retrieval of hourly PM2.5 using top-of-atmosphere reflectance from geostationary ocean color imagers I and II

To produce real-time ground-level information on particulate matter with a diameter equal to or less than 2.5 μm (PM_2.5), many studies have explored the applicability of satellite data, particularly aerosol optical depth (AOD). However, many of the techniques used are computationally demanding; to overcome these challenges, machine learning(ML)-based research has been on the rise. Here, we used ML techniques to directly estimate ground-level PM_2.5 concentrations over South Korea using top-of-atmosphere (TOA) reflectance from the Geostationary Ocean Color Imager I (GOCI-I) and its next generation GOCI-II with improved spatial, spectral, and temporal resolutions. Three ML techniques were used to estimate ground-level PM_2.5 concentrations: random forest, light gradient boosting machine (LGBM), and artificial neural network. Three schemes were examined based on the input feature composition of the GOCI spectral bands: scheme 1 using all GOCI-I bands, scheme 2 using only GOCI-II bands that overlap with GOCI-I bands, and scheme 3 using all GOCI-II bands. The results showed that LGBM performed better than the other ML models. GOCI-II-based schemes 2 and 3 (determination of coefficient (R²) = 0.85 and 0.85 and root-mean-square-error (RMSE) = 7.69 and 7.82 μg/m³, respectively) performed slightly better than GOCI-I-based scheme 1 (R² = 0.83 and RMSE = 8.49 μg/m³). In particular, TOA reflectance at a new channel (380 nm) of GOCI-II was identified as the most contributing variable, given its high sensitivity to aerosols. The long-term estimation of PM_2.5 concentrations using the proposed models was examined for ground stations located in two major cities. GOCI-II-based models produced a more detailed spatial distribution of PM_2.5 concentrations owing to their higher spatial resolution (i.e., 250 m). The use of TOA reflectance data, instead of AOD and other aerosol products commonly used in previous studies, reduced the missing rate of the estimated ground-level PM_2.5 concentrations by up to 50%. Our results indicate that the proposed approach using TOA reflectance data from geostationary satellite sensors has great potential for estimating ground-level PM_2.5 concentrations for operational purposes.

Estimation of near-surface ozone concentration and analysis of main weather situation in China based on machine learning model and Himawari-8 TOAR data

Ozone (O₃) is an important greenhouse gas in the atmosphere. Stratospheric ozone protects human beings, but high near-surface ozone concentrations threaten environment and human health. Owing to the uneven distribution of ground-monitoring stations and the low time resolution of polar orbiting satellites, it is difficult to accurately evaluate the refinement and synergistic pollution of near-surface ozone in China. Besides, atmospheric circulation patterns also affect ozone concentrations greatly. In this study, a new generation of geostationary satellite is used to estimate the hourly near-surface ozone concentration with a spatial resolution of 0.05°. First, the Pearson correlation coefficient and maximum information coefficient were used to study the correlation between the top of atmospheric radiation (TOAR) of Himawari-8 satellite and O₃ concentration; seven TOAR channels were selected. Second, based on an interpretable deep learning model, the hourly ozone concentration in China from September 2015 to August 2021 was obtained using the TOAR-O₃ model. Finally, the self-organizing map method was used to determine six major summer weather circulation patterns in China. The results showed that (1) the near-surface O₃ concentration can be accurately estimated; the R² (RMSE: μg/m³) values of the daily, monthly, and annual tenfold cross validation results were 0.91 (12.74), 0.97 (5.64), and 0.98 (1.75), respectively. The feature importance of the model showed that the temperature, TOAR, and boundary layer height contributed 38 %, 22 %, and 13 %, respectively. (2) The O₃ concentration showed obvious spatiotemporal difference and gradually increased from 10:00 to 15:00 (Beijing time) every day. In most areas of China, O₃ concentration had increased significantly. (3) The O₃ concentration in northern China was the highest under the circulation pattern of the Meiyu front over the Yangtze River Delta, while in southern China, it was the highest under the circulation pattern of the northeast cold vortex controlling most of China.

Science communication on the public health risks of air pollution: a computational scoping review from 1958 to 2022

BACKGROUND

Air pollutants are a health risk for the entire population. Particulate matter (PM) including the smallest fraction, ultra-fine particles (UFP), therefore continue to be the focus of scientific research in this area. To protect the population from the harmful effects of exposure to PM, communication and information of research results are of special relevance as individuals with heightened awareness of the harms of poor air quality are more likely to take action to improve their exposure.

METHODS

We conducted a scoping review of the scientific literature on science communication of public health information about risks associated with air pollutants to generate an initial over-view of existing research in this field. We searched the PubMed and Scopus databases and analyzed the data using a structured topic modeling (STM) approach.

RESULTS

The existing scientific literature dates back to 1958 but increases significantly from the 1990s onwards. Publications are mainly found in the discipline of environmental research and are primarily concerned with health effects. It is often stated that adequate communication of the results to the public would be important, but specific approaches are rare. Overall, the topic of risk communication seems to be underrepresented for both air pollutants and UFP.

CONCLUSIONS

To protect public health, it is important to conduct more intensive science and risk communication related to scientific findings on the risks of air pollutants. For adequate communication and information, further research is needed to provide specific approaches that also involve the affected population and take different target groups into account. In addition, the effectiveness of communication efforts should also be analyzed.

Mapping contiguous XCO2 by machine learning and analyzing the spatio-temporal variation in China from 2003 to 2019

As China is the world's largest CO₂ emitter, it is important to understand the spatio-temporal variation of atmospheric CO₂ to reduce carbon emissions. Satellite remote sensing for carbon monitoring has been widely used and studied because of its long-term and large-scale characteristics. However, the satellite data results are very sparse with significant gaps due to narrow swath and other factors on CO₂ retrieval. The simple interpolation methods ignore the influential factors of CO₂ and loss the spatial resolution, which leads to the inability to quantify the spatio-temporal variation well. This study developed a machine learning method that considers carbon emissions, vegetation, and meteorology. Using the column-averaged dry-air mole fraction of CO₂ (XCO₂) data of SCIAMACHY, GOSAT, and OCO-2, we derived monthly-scale contiguous XCO₂ data across China from 2003 to 2019 with 0.25° resolution. The results showed a good agreement with the satellite measurements, with the bias and standard deviation of 0.11 and 1.38 ppmv for the validation dataset, respectively. Moreover, the results were consistent with the model simulation and in-situ sites, indicating the ability to reflect long-term spatio-temporal variation with a finer texture. We analyzed the spatial distribution, seasonal variation, and long-term trends of XCO₂ in China, revealing that the machine learning method has comparable performance to model simulations. The results showed that XCO₂ is dominated by anthropogenic emissions spatially and has a clear seasonal cycle, with a larger amplitude the further north. The long-term trend shows the XCO₂ increased by an average rate of 2.17 ppmv per year from 2003 to 2019 in China, which is consistent with the global. The method and data can further study the carbon cycle and climate change.

Neighborhood-scale ambient NO2 concentrations using TROPOMI NO2 data: Applications for spatially comprehensive exposure assessment

This study estimated long-term average ambient NO₂ concentrations using TROPOspheric Monitoring Instrument (TROPOMI) tropospheric NO₂ data and land use information at the spatial resolution of 500 m in California for the years 2018-2019. Our satellite-land use regression model demonstrated reasonably high predictive power with cross-validation (CV) R² = 0.76, mean absolute error (MAE) = 1.95 ppb, and root mean squared error (RMSE) = 2.51 ppb in a comparison between measured and estimated NO₂ concentrations. Exploiting the high-resolution NO₂ estimates, we further investigated the representativeness of ground NO₂ monitors for population exposures and examined the spatial variation of NO₂ in relation to parcel-level property data for exposure attributions. The ground NO₂ monitors were the most representative of population exposures in Los Angeles and San Diego counties, supported by population-weighted average NO₂ concentrations (satellite-derived estimations) similar to arithmetic average NO₂ concentrations (ground measurements). On the contrary, the exposure assessment using the ground monitors was the least representative and protective in Humboldt, San Luis Obispo, and Yolo counties with population-weighted average NO₂ greater than arithmetic average NO₂ by 82.2 % (1.85 ppb), 67.1 % (1.89 ppb), and 58.2 % (2.48 ppb), respectively. In a case study of LA County, we identified comparatively high NO₂ concentrations for the property types of food processing facilities and high-density residential complexes (such as high-rise apartments and apartments). This finding provides evidence that these emerging sources may be crucial to mitigate cumulative NO₂ exposures and subsequent health risks from a regulatory perspective.

A new perspective to satellite-based retrieval of ground-level air pollution: Simultaneous estimation of multiple pollutants based on physics-informed multi-task learning

Remote sensing of air pollution is essential for air quality management and health risk assessment. Many machine-learning-based retrieval models have been established for estimating various kinds of air pollutants. These methods mainly aimed to estimate a single pollutant (single-pollutant approach). However, different air pollutants interact with each other and are highly correlated. Building a unified model and conducting a joint retrieval of multiple pollutant can make a better use of these connections and improve the model efficiency. This study proposed a physics-informed multi-task deep neural network (phyMTDNN) for the joint retrieval of six main air pollutants, i.e., PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃. The relationships among these pollutants were used to design the physics-informed network structure and loss function. Top-of-atmosphere reflectance which can generate retrieval results at ultrahigh resolution was used as model input. Experiments for mainland China in 2019 showed that the proposed model successfully achieved simultaneous estimation of six air pollutants, with the cross-validated R² for the six pollutants varying from 0.72 to 0.90. The contrast experiments proved that physics-informed network structure contributed to the most of the model performance improvement. Compared to the single-pollutant approach, phyMTDNN ameliorated the model accuracy on traces gases retrieval. Furthermore, the modeling efficiency was largely improved in that a lot of repetitive work was avoided and modeling method was optimized. The proposed new multiple-pollutant retrieval frame can be applied to various fields for multi-variate retrieval or estimation.

Using machine learning approach to reproduce the measured feature and understand the model-to-measurement discrepancy of atmospheric formaldehyde

The solar absorption spectrometry in the infrared spectral region, using high-resolution Fourier transform infrared (FTIR) spectrometer, has been established as a powerful tool in atmospheric science. These observations cannot be performed continuously, for example, clouds prevent observations. On the other hand, chemical transport models give continuously data. Their results depend on the knowledge of emission inventories, the chemistry involved, and the meteorological fields, yielding to potential biases between measurements and simulations. In our study we concentrated on Formaldehyde (HCHO) and used machine learning approach to fill the gap between the observations, performed on an irregular time scale and having their measurement lacks, and model data, giving continuous data, but having potential variable biases. The proposed machine learning approach is based on the Light Gradient Boosting Machine (LightGBM) algorithm and created by using GEOS-Chem simulations, meteorological fields, emission inventory, and is referred to as the GEOS-Chem-LightGBM model. The results of established GEOS-Chem-LightGBM model have generated consistent HCHO predictions with the ground-based FTIR and satellite (OMI and TROPOMI) observations. In order to understand the GEOS-Chem model to measurement discrepancy, we have investigated the contribution of each input variable to GEOS-Chem-LightGBM model HCHO predictions through the SHapely Additive exPlanations (SHAP) approach. We found that the GEOS-Chem model underestimates the sensitivities of HCHO total column to most photochemical variables, contributing to lower amplitudes of diurnal cycle and seasonal cycle by the GEOS-Chem model. By correcting the model-to-measurement discrepancy, the sensitivities of HCHO total column to all variables by the GEOS-Chem-LightGBM became to be in good agreement with the FTIR observations. As a result, GEOS-Chem-LightGBM model has significantly improved the performance of HCHO predictions compared to the GEOS-Chem alone. The proposed GEOS-Chem-LightGBM model can be extendible to other atmospheric constituents obtained by various measurement techniques and platforms, and is expected to have wide applications.

A review of datasets and methods for deriving spatiotemporal distributions of atmospheric CO2

As the most abundant greenhouse gas, atmospheric carbon dioxide (CO₂) is considered one of the main attributors to climate change. Atmospheric CO₂ concentrations can be measured by ground-based monitoring networks, mobile monitoring campaigns, and carbon-observing satellites. However, the worldwide ground-based monitoring networks are composed of sparsely distributed sites and are inadequate to represent the spatiotemporal distributions of CO₂. Satellite-based remote sensing features repeated, long-term, and large-scale measurements, so it plays a crucial role in monitoring the global distributions of atmospheric CO₂. However, due to the presence of heavy clouds (or aerosols) and the limitation of satellite orbiting tracks, there exist large amounts of missing data in satellite retrievals. Various methods, including chemical transport models (CTMs), geostatistical methods, and regression-based models, have been employed to derive full-coverage spatiotemporal distributions of CO₂ based on the limited CO₂ measurements. This review summarizes the strengths and limitations of these methods. However, CTMs simulation results can have high uncertainty due to imperfect knowledge of the real world, and the interpolation accuracy of all geostatistical methods is limited by the large amount of data gaps in current satellite retrieved CO₂ products. To overcome these limitations, regression-based methods (especially machine learning models) have the ability to predict CO₂ with superior predictive performance, so this review also summarizes the framework of the machine learning approach. Leveraging the ongoing advancements of satellite instrumentation, the satellite-based CO₂ products have been improving dramatically in recent decades, and this review will describe and critically assess the advantages and disadvantages of the currently used systems in detail. For future improvements, we recommend the fusion of data from multiple satellite retrievals and CTMs by using machine learning algorithms in order to obtain even longer-term, larger-scale, finer-resolution, and higher-accuracy CO₂ datasets.

Contributions of meteorology to ozone variations: Application of deep learning and the Kolmogorov-Zurbenko filter

From hourly ozone observations obtained from three regions⸻Houston, Dallas, and West Texas⸻we investigated the contributions of meteorology to changes in surface daily maximum 8-h average (MDA8) ozone from 2000 to 2019. We applied a deep convolutional neural network and Shapely additive explanation (SHAP) to examine the complex underlying nonlinearity between variations of surface ozone and meteorological factors. Results of the models showed that between 2000 and 2019, specific humidity (38% and 27%) and temperature (28% and 37%) contributed the most to ozone formation over the Houston and Dallas metropolitan areas, respectively. On the other hand, the results show that solar radiation (50%) strongly impacted ozone variation over West Texas during this time. Using a combination of the Kolmogorov-Zurbenko (KZ) filter and multiple linear regression, we also evaluated the influence of meteorology on ozone and quantified the contributions of meteorological parameters to trends in surface ozone formation. Our findings showed that in Houston and Dallas, meteorology influenced ozone variations to a large extent. The impacts of meteorology on West Texas, however, showed meteorological factors had fewer influences on ozone variabilities from 2000 to 2019. This study showed that SHAP analysis and the KZ approach can investigate the contributions of the meteorological factors on ozone concentrations and help policymakers enact effective ozone mitigation policies.

Geostationary satellite-derived ground-level particulate matter concentrations using real-time machine learning in Northeast Asia

Rapid economic growth, industrialization, and urbanization have caused frequent air pollution events in East Asia over the last few decades. Recently, aerosol data from geostationary satellite sensors have been used to monitor ground-level particulate matter (PM) concentrations hourly. However, many studies have focused on using historical datasets to develop PM estimation models, often decreasing their predictability for unseen data in new days. To mitigate this problem, this study proposes a novel real-time learning (RTL) approach to estimate PM with aerodynamic diameters of <10 μm (PM₁₀) and <2.5 μm (PM_2.5) using hourly aerosol data from the Geostationary Ocean Color Imager (GOCI) and numerical model outputs for daytime conditions over Northeast Asia. Three schemes with different weighting strategies were evaluated using 10-fold cross-validation (CV). The RTL models, which considered both concentration and time as weighting factors (i.e., Scheme 3) yielded consistent improvement for 10-fold CV performance on both hourly and monthly scales. The real-time calibration results for PM₁₀ and PM_2.5 were R² = 0.97 and 0.96, and relative root mean square error (rRMSE) = 12.1% and 12.0%, respectively, and the 10-fold CV results for PM₁₀ and PM_2.5 were R² = 0.73 and 0.69 and rRMSE = 41.8% and 39.6%, respectively. These results were superior to results from the offline models in previous studies, which were based on historical data on an hourly scale. Moreover, we estimated PM concentrations in the ocean without using land-based variables, and clearly demonstrated the PM transport over time. Because the proposed models are based on the RTL approach, the density of in-situ monitoring sites could be a major uncertainty factor. This study identified that a high error occurred in low-density areas, whereas a low error occurred in high-density areas. The proposed approach can be operated to monitor ground-level PM concentrations in real-time with uncertainty analysis to ensure optimal results.

Evaluation of NOx emissions before, during, and after the COVID-19 lockdowns in China: A comparison of meteorological normalization methods

Meteorological normalization refers to the removal of meteorological effects on air pollutant concentrations for evaluating emission changes. There currently exist various meteorological normalization methods, yielding inconsistent results. This study aims to identify the state-of-the-art method of meteorological normalization for characterizing the spatiotemporal variation of NOx emissions caused by the COVID-19 pandemic in China. We obtained the hourly data of NO2 concentrations and meteorological conditions for 337 cities in China from January 1, 2019, to December 31, 2020. Three random-forest based meteorological normalization methods were compared, including (1) the method that only resamples meteorological variables, (2) the method that resamples meteorological and temporal variables, and (3) the method that does not need resampling, denoted as Resample-M, Resample-M&T, and Resample-None, respectively. The comparison results show that Resample-M&T considerably underestimated the emission reduction of NOx during the lockdowns, Resample-None generates widely fluctuating estimates that blur the emission recovery trend during work resumption, and Resample-M clearly delineates the emission changes over the entire period. Based on the Resample-M results, the maximum emission reduction occurred during January to February 2020, for most cities, with an average decrease of 19.1 ± 9.4% compared to 2019. During April of 2020 when work resumption initiated to the end of 2020, the emissions rapidly bounced back for most cities, with an average increase of 12.6 ± 15.8% relative to those during the strict lockdowns. Consequently, we recommend using Resample-M for meteorological normalization, and the normalized NO2 concentration dynamics for each city provide important implications for future emission reduction.