Accurate measurement of atmospheric chemicals is essential for understanding their impact on human health, climate, and ecosystems. Satellites provide a unique advantage by capturing data across the entire atmosphere, but their measurements often lack vertical details. Here, we introduce DeepSAT4D, an innovative method that efficiently reconstructs 4D chemical concentrations from satellite data. It achieves this by regenerating the dynamic evolution of vertical structure, intricately linked to complex atmospheric processes such as plume rise and transport, using advanced deep learning techniques. Its application with the Ozone Monitoring Instrument - Nitrogen Dioxide, a commonly used satellite product, demonstrates good agreement with ground-based monitoring sites in China from 2017 to 2021. Additionally, DeepSAT4D successfully captures emission reductions during 2020-pandemic shutdown. These findings emphasize DeepSAT4D’s potential to enhance our understanding of the complete atmospheric chemical composition and to provide improved assessments of its impact on human health and Earth’s ecosystem in the future.

Ground-level PM_2.5 data derived from satellites with machine learning are crucial for health and climate assessments. However, uncertainties persist due to the absence of spatially covered observations. To address this, we propose a novel testbed using nontraditional numerical simulations to evaluate PM_2.5 estimation across the entire spatial domain. The testbed emulates the general machine-learning approach by training the model with grids corresponding to ground monitoring sites and subsequently testing its predictive accuracy for other locations. Our approach enables comprehensive evaluation of various machine-learning methods' performance in estimating PM_2.5 across the spatial domain for the first time. Unexpected results are shown in the application in China, with larger absolute PM_2.5 biases found in densely populated regions with abundant ground observations across all benchmark models due to the higher baseline concentration, though the relative error (approximately 20 %) is smaller compared to that in rural areas (over 50 %). The imbalance in training samples, mostly from urban areas with high emissions, is the main reason, leading to significant overestimation due to the lack of monitors in downwind areas where PM_2.5 is transported from urban areas with varying vertical profiles. Our proposed testbed also provides an efficient strategy for optimizing model structure or training samples to enhance satellite-retrieval model performance. Integration of spatiotemporal features, especially with conventional neural network (CNN)-based deep-learning approaches like the residual neural network (ResNet) model, has successfully mitigated PM_2.5 overestimation (by 5–30 µg m⁻³) and the corresponding exposure (by 3 million people ⋅ µg m⁻³) in the downwind area over 9 years (2013–2021) compared to the traditional approach. Furthermore, the incorporation of 600 strategically positioned ground monitoring sites identified through the testbed is essential for achieving a more balanced distribution of training samples, thereby ensuring precise PM_2.5 estimation and facilitating the assessment of the associated impacts in China. In addition to presenting the retrieved surface PM_2.5 concentrations in China from 2013 to 2021, this study provides a testbed dataset derived from physical modeling simulations which can serve to evaluate the performance of data-driven methodologies, such as machine learning, in estimating spatial PM_2.5 concentrations for the community (Li et al., 2024a; https://doi.org/10.5281/zenodo.11122294).

Accurate air pollutant prediction is essential for addressing environmental and public health concerns. Air quality models like WRF-CMAQ provide simulations, but often show significant errors compared to observed concentrations. To identify the sources of these model biases, we applied the XGBoost machine learning algorithm to assess the performance of WRF-CMAQ in predicting air pollutants across two regions in China. XGBoost models trained with observations achieved high accuracy (R > 0.95), indicating that the selected features effectively capture pollutant variations. When trained on WRF-CMAQ inputs, XGBoost still improved performance but revealed biases linked to both model inputs (10–60%) and mechanisms (1–30%). Analysis identified previous-hour pollutant levels as the largest bias contributor, followed by meteorological variables. The study highlights the need for improving both model inputs and mechanisms to enhance future air quality predictions and support pollution control strategies.

Over the last two decades, substantial studies have been conducted to assess the feasibility of a multi-pollutant strategy for managing air quality in the United States. Given the inherent complexity of multi-pollutant air quality challenges, including fine particulate matter (PM_2.5), ozone (O₃), and air toxics, this paper undertook a multi-pollutant analysis at both national and local levels. Our analysis incorporated O₃ and PM_2.5 concentrations, air toxics that increase the risk of cancer, environmental justice (EJ) data, emissions data, and monitoring data. Initially, we identified counties across the continental U.S. with heightened multi-pollutant exposures and EJ concerns. Subsequently, a case study within the Detroit metropolitan area was conducted, revealing a clear overlap between multi-pollutant and EJ issues, underscoring the disproportionate burden on disadvantaged communities. The analysis of detailed emissions data unveiled potential co-control benefits in this region. Lastly, employing a proximity analysis method, we assessed environmental issues surrounding points of interest such as monitoring sites and emissions sectors, in the Detroit metropolitan area. The results demonstrated that monitoring sites with the highest monitoring value, alongside top-ranked emissions sectors such as electric utilities, coke ovens, and iron and steel production, were likely to exhibit elevated air pollutant concentrations/risks and associated EJ concerns in their vicinity.

Ground-level PM_2.5 data derived from satellites with machine learning are crucial for health and climate assessments, however, uncertainties persist due to the absence of spatially covered observations. To address this, we propose a novel testbed using untraditional numerical simulations to evaluate PM_2.5 estimation across the entire spatial domain. The testbed emulates the general machine-learning approach, by training the model with grids corresponding to ground monitor sites and subsequently testing its predictive accuracy for other locations. Our approach enables comprehensive evaluation of various machine-learning methods’ performance in estimating PM_2.5 across the spatial domain for the first time. Unexpected results are shown in the application in China, with larger PM_2.5 biases found in densely populated regions with abundant ground observations across all benchmark models, challenging conventional expectations and are not explored in the recent literature. The imbalance in training samples, mostly from urban areas with high emissions, is the main reason, leading to significant overestimation due to the lack of monitors in downwind areas where PM_2.5 is transported from urban areas with varying vertical profiles. Our proposed testbed also provides an efficient strategy for optimizing model structure or training samples to enhance satellite-retrieval model performance. Integration of spatiotemporal features, especially with CNN-based deep-learning approaches like the ResNet model, successfully mitigates PM_2.5 overestimation (by 5–30 µg m^-3) and corresponding exposure (by 3 million people • µg m^-3) in the downwind area over the past nine years (2013–2021) compared to the traditional approach. Furthermore, the incorporation of 600 strategically positioned ground-measurement sites identified through the testbed is essential to achieve a more balanced distribution of training samples, thereby ensuring precise PM_2.5 estimation and facilitating the assessment of associated impacts in China. In addition to presenting the retrieved surface PM_2.5 concentrations in China from 2013 to 2021, this study provides a testbed dataset derived from physical modeling simulations which can serve to evaluate the performance of data-driven methodologies, such as machine learning, in estimating spatial PM_2.5 concentrations for the community.