Air pollution profoundly affects health, especially in vulnerable newborns. This study delineates regions of air pollution sensitivity and examines their correlation with the incidence of toddler pneumonia on Java Island in 2023 through geographic clustering techniques, association analysis, and ordinal logistic regression. Java Island was chosen because of its low Air Quality Index (AQI) and elevated rates of toddler pneumonia cases. Data on air pollutants (PM2.5, NO2, O3, SO2, CH4), climatic variables, and sensitivity factors (industrial density, population density) were utilized to create clusters according to levels of air pollution vulnerability. The clustering outcomes indicated that the Spatial Generalized Fuzzy C-Means (SGFCM) method with three clusters yielded the most favorable evaluation findings. Clusters exhibiting elevated and moderate air pollution were predominantly located in the western and eastern parts of Java Island. This study illustrates the importance of the correlation between air pollution sensitivity clusters and the incidence of toddler pneumonia. Regions inside moderate or high clusters are 2.67 times more susceptible to elevated toddler pneumonia prevalence than those in low pollution clusters. This research can assist stakeholders in developing policies for managing air pollution, specifically regarding its effects on toddler pneumonia.
Agricultural sustainability faces the dual challenges of reducing carbon emissions while ensuring food production. Climate-smart agriculture (CSA), an innovative paradigm to address these challenges, has yet to be widely implemented in China due to variations in local agricultural sustainability levels. To this end, this study begins by assessing agricultural sustainability through the construction of a comprehensive indicator system for agricultural carbon emission efficiency (ACEE), integrating the objectives of grain production growth and multi-source emission reductions. We then analyze the spatiotemporal evolution trend of ACEE across 31 Chinese provinces (2010–2022) using the super-efficiency SBM and kernel density estimation. Furthermore, guided by a constructed CSA analytical framework, we integrate the Geodetector model and dynamic qualitative comparative analysis (QCA) to explore the explanatory factors and enhancement pathways of ACEE via three dimensions, including CSA technology, policy support, and social environment. The results show that: (1) Chinese average ACEE remains broadly stable, with spatial distribution exhibiting an asymmetric pattern of “high-value contraction” and “low-value stability”. Internal disparities in ACEE remain evident within the three major regions, with varying degrees of polarization. (2) The adoption levels of three CSA technologies, water-saving irrigation, straw-return, and no-tillage planting, show relatively strong explanatory power for spatial disparities in ACEE. Moreover, the explanatory power of factor interactions significantly exceeds their independent contributions. (3) Four differentiated configuration pathways enhance ACEE under the CSA framework: pathways driven by “policy and environment”, “technology and policy”, “technology, policy, and environment”, and “technology”. Our findings provide profound policy insights for the systematic governance of CSA and sustainable agricultural development.
Artificial intelligence (AI) has rapidly advanced as a key analytical tool for processing complex datasets across disciplines, including environmental and health research. Meteorological data is increasingly used to understand and mitigate health risks, often linked to climate change. This rapid review aims to synthesise studies involving AI for meteorological data applications in health research to better understand its use. PubMed, Web of Science, and Scopus were systematically searched from 2020 to 2025 following a standardised framework in line with PRISMA-RR guidelines. Eligible studies included any empirical design involving human-related health research where AI techniques were applied to meteorological data. Two reviewers independently screened studies, extracted data, and synthesised findings narratively. Twelve studies met eligibility criteria. Despite heterogeneity in study design and sample size, most examined the impact of extreme pollution levels and temperature variations on the prevalence and severity of respiratory, bacterial, and other diseases. AI methods primarily included Random Forest models, as well as time-series and clustering analyses. Model performance was commonly evaluated using sensitivity; however, methodological justification was often insufficiently recorded. Overall, findings suggest that incorporating meteorological variables enhances the prediction of health outcomes, although detailed population characteristics were frequently underreported. This restricts the generalisability and applicability of AI-driven health models incorporating meteorological data, with stronger methodological rigour and clearer reporting standards needed to support reliable future development.
Air pollution shows a major environmental health issue, leading to various negative health consequences. Nonetheless, limited research exists on how air pollution impacts mortality risks and the number of years of life lost (YLLs) from early death in Thailand. This study seeks to analyze the relationship between short-term exposure to air pollution – including particulate matter with aerodynamic diameters up to 2.5 (PM2.5) and 10 microns (PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), group-level ozone (O3), and carbon monoxide (CO) – and mortality and YLLs in Thailand using data from January 2017 to December 2021. A case-crossover approach using conditional Poisson regression was employed to assess the province-specific effects of air pollution on mortality. The same design and a Gaussian regression model investigated the province-specific relationships of air pollution with YLLs. The province-specific estimates were combined using a random-effects meta-analysis to determine the nationwide association. The results showed that an interquartile range (IQR) increase in PM10, PM2.5, and O3 at lag 0–3, as well as NO2, SO2, and CO at lag 0–2, was linked to a 4.12
With rapid industrialization and urbanization, multipollutant exposure to coexisting atmospheric pollutants has become a major global public health concern. Multipollutant exposure may pose modified harm to human health compared with single-pollutant exposure, depending on pollutant combinations, concentrations, and population susceptibility. In China, which currently faces a complex challenge of compound air pollution, the emissions reduction potential of pollution control measures exhibits a fluctuating downward trend, and the mechanisms underlying the compound synergistic effects of multi-pollutants still require further in-depth investigation. This study provides bibliometric analysis and narrative review of the current status of research on multipollutant exposure in China, outlines the sources and characteristics of air pollutants, and focuses on elucidating the health effects of multipollutant exposure to particulate matter (e.g., PM2.5, PM10) and gaseous pollutants (e.g., O3, NO2, SO2, CO) on the human respiratory and cardiovascular systems, as well as associated pathogenic mechanisms. It compares and analyzes the strengths and limitations of mainstream air quality assessment methods (Air Quality Index, Air Quality Health Index) both domestically and internationally, focusing on quantitative methods for assessing health risks resulting from multipollutant exposure to air pollutants. This review also introduces the principles, application scenarios, and key findings of generalized additive models, interaction effects models, and meta-analysis models. Overall, it aims to provide comprehensive theoretical support and a reference framework for preventing and controlling health risks associated with multipollutant exposure to air pollutants, thereby enabling more precise exposure management, and advancing related research.
Urban air quality dynamics in secondary and emerging cities are as yet underexplored despite the growing exposure risks. This research focuses on seasonal anomalies, persistence behaviour and spatial inequality in air quality of eight major urban centres of the North-east India using daily average Air Quality Index (AQI) over a period from 2022 to 2024. In this study, the analysis incorporates descriptive statistics, autocorrelation diagnostics, modified Mann- Kendall trend detection, Seasonal Anomaly Index (SAI), inter-urban inequality assessment, and hierarchical clustering in order to identify temporal and spatial pollution regimes. Results show a strong heterogeneity of AQI burden across NE India, where Agartala, Guwahati and Imphal show significantly higher levels of pollution when compared to hill cities such as Aizawl, Gangtok, and Shillong. Exposure analysis shows that more than 40–50
Residential biomass combustion in rural communities is recognised as a major contributor to air pollution, representing the second-largest source of gaseous emissions and the primary source of atmospheric particulate matter. In many African countries, including South Africa, residential emission inventories and air quality assessments rely largely on international emission factors (EFs), introducing considerable uncertainty due to regional differences in fuel characteristics, combustion technologies, and household energy-use practices. Generating locally derived EFs is therefore essential for improving the accuracy of emission inventories and supporting effective air quality management and climate policy development. This study quantified the EFs of CO₂, CO, NO, SO₂, and PM₁₀ from the combustion of the 12 predominantly preferred fuelwood species used for household energy in rural Limpopo Province, South Africa. Controlled combustion experiments were conducted using a traditional three-legged cookstove in a simulated village kitchen under laboratory conditions. The mean EFs determined for CO₂, CO, NO, SO₂, and PM₁₀ were 1389 ± 149, 131.38 ± 91.4, 1.22 ± 0.63, 5.16 ± 4.9, and 64.7 ± 76.3 g kg⁻¹, respectively. Modified combustion efficiency (MCE) exhibited a statistically significant negative association with the EFs of CO and SO₂, indicating that lower combustion efficiency was associated with higher emissions of these pollutants. In contrast, fuel moisture content was not significantly associated with the EFs of CO₂, NO, or PM₁₀, while no statistically significant relationship was observed between fuel nitrogen (N) content and NO EFs. Furthermore, no statistically significant differences were detected among the investigated fuelwood species in the individual EFs of CO₂, CO, NO, SO₂, and PM₁₀. This study provides the first laboratory-derived, species-specific EF dataset for the predominantly preferred fuelwood species used in rural Limpopo Province. The findings reduce uncertainties associated with the application of international default EFs and provide a robust basis for improving residential biomass emission inventories, air quality modelling, and the development of evidence-based air pollution mitigation and climate management strategies in South Africa and other regions with similar household energy-use patterns.
Major sources of atmospheric microplastic originate from human activities (e.g. construction, driving, laundry, and agriculture dust), along with sea-spray. Due to their small size (< 5 mm), they are easily transported long distances, being found in both urban and remote environments. This research examines the atmospheric deposition of microparticles on a seasonally-touristed island. Nantucket Island, USA, is a unique study site due to its proximity to ocean-atmospheric exchanges and its high influx of seasonal visitors. Atmospheric microparticles were collected at two contrasting sites and analyzed for amount and type from October 2021 to September 2022. Samples were collected each month and processed, counted, and measured. Microparticle flux ranged from 4.6 to 62.2 mp/m2/d and were composed mostly of fibers, > 75 μm. The Hatchery site had an average and s.d. of 28.2 ± 13.3 mp/m2/d and the Station site had an average and s.d. of 15.7 ± 8.2 mp/m2/d. Concentrations during the tourist months were 1.5 times greater than the non-tourist months.
Vehicular emission is an important source of particulate matter (PM2.5), black Carbon (BC) and brown Carbon (BrC) in urban areas, but the annual trends of vehicular emissions of PM2.5, BC and BrC pollutants have not been clearly identified in some urban areas. This study evaluates emission factors of vehicles-related PM2.5, BC and BrC via measurements of a road tunnel in Pyongyang, DPR Korea, in 2019 and 2024, respectively. The proportion of diesel vehicles in the tunnel is much higher than that in other urban road tunnels, with 10.2
To evaluate the health impacts of PM2.5 exposure on sensitive and vulnerable populations in South Korea and to identify flexion points—concentration levels at which the risk of environmental diseases increases significantly. A retrospective cohort study using nationwide health insurance claims data linked with environmental monitoring data. Statistical analysis included Cox proportional hazards regression and piecewise linear regression to estimate disease risks and identify flexion points. South Korea, using data from the National Health Insurance Service–National Sample Cohort (NHIS-NSC) and national air quality monitoring stations. A total of 1,031,517 individuals aged 0–79 years who were covered by the national health insurance system between 2017 and 2019 and had no prior diagnosis of the target diseases during the 2016 washout period. Primary outcomes were the incidence and exacerbation of four environmental diseases: chronic obstructive pulmonary disease (COPD), asthma, stroke, and heart failure. Exacerbation was defined as hospitalization or emergency department visits due to these diseases. Higher PM2.5 exposure was significantly associated with increased incidence and exacerbation of all four diseases. Flexion points were identified at 30 µg/m³ for incidence and between 24 and 26 µg/m³ for exacerbation. Subgroup analyses showed lower flexion points among elderly participants. The findings suggest that current PM2.5 standards in South Korea may not sufficiently protect vulnerable populations. Operational flexion points identified in this study may serve as evidence-based thresholds to guide targeted public health interventions and air quality regulations.
This work developed and tested Random Forest Regression (RFR) and Long Short-Term Memory (LSTM)models to predict PM2.5 concentrations and respiratory illness cases in Phrae Province, northern Thailand. In 2020–2023, daily atmospheric pollutants and meteorological parameters were analyzed alongside monthly hospital data on asthma, bronchitis, COPD, and lung cancer to examine the relationship between air pollution and respiratory health and to evaluate machine learning (ML) models. PM2.5 correlated positively with nitrogen dioxide (NO2) and ozone (O3) and negatively with relative humidity, especially during the rainy season. PM2.5 was strongly linked to bronchitis and asthma, especially during the dry season, when biomass-burning activities led to the highest PM2.5 concentrations. The best predictive model for PM2.5 concentrations and respiratory illness cases was RFR, which found no statistically significant short-term link between PM2.5 and lung cancer. This suggests that long-term cumulative exposure and other risk factors are more strongly linked to lung cancer than short-term variations in air pollution. LSTM detected episodic PM2.5 peaks. However, the limited monthly dataset (48 observations) and considerable temporal variability of patient records hindered LSTM’s respiratory illness prediction. So, consider these data exploratory rather than conclusive. Results suggest a hierarchy of decision support: RFR for normal PM2.5 forecasting, LSTM for peak event early warning when longer, continuous time-series data are available, and dry-season public health interventions for bronchitis surveillance and prevention. Local air quality management and respiratory health planning can be integrated with ML-based environmental forecasts using the framework to assist evidence-based pollution reduction and public health preparedness decisions.
To explore the transport pathways and potential source regions of fine particulate matter (PM2.5) in the severe polluted hotspots of Beijing-Tianjin-Hebei region (Handan and Xingtai), 72-hour backward trajectories were simulated for 2017, 2020, and 2023 using Meteoinfo software. Trajectory cluster analysis, weighted potential source contribution function (WPSCF) and weighted concentration-weighted trajectory (WCWT) analysis were applied. Analysis of the vertical distribution of air pollution transport pathways in Handan showed that near-surface transport was predominantly short-range. Transport pathway analysis indicated that during winter, the trajectory with the highest PM2.5 concentration for Handan (113.5 µg/m3) originated from Shijiazhuang, while that for Xingtai (113.6 µg/m3) originated from Cangzhou, both passing through the Hebei-Shandong border area. WPSCF and WCWT analysis revealed the spatial extent of major PM2.5 source regions was largest in 2017, decreased in 2020, but expanded again in 2023. The WPSCF/WCWT results of PM2.5 demonstrated that the largest potential areas were identified in winter and were mainly distributed at the junction of Hebei, Henan, and Shandong Provinces, and a larger area covering multiple provinces (Hebei, Henan, Shandong, and Shanxi Provinces) for Xingtai. The PM2.5 pollution in Handan and Xingtai exhibited high homogeneity and mutual influence. These findings underscore the necessity of implementing regional joint prevention and control measures to mitigate PM2.5 pollution in Handan and Xingtai.
Airborne dust is a significant source of toxic metals in urban settings, but there is a lack of integrated evaluations of the level of pollution, ecological risk, and human health risk in semi-arid urban settings. This study examined Cd, Pb, Cu, Cr, and Zn concentrations in airborne dust collected from four locations in Zakho City during the wet and dry seasons. The sampling represented seasonally accumulated deposited/fallout airborne dust rather than event-specific dust-storm samples. The measured metal concentrations generally followed the order Zn > Cu > Pb > Cr > Cd. Due to the limited unreplicated sampling design, the observed spatial and seasonal variations should be interpreted as preliminary descriptive trends rather than definitive statistical patterns. The pollution indices (CF, EF, Igeo, and PLI) showed that the anthropogenic enrichment of Zn (CF up to 4.7; EF > 3) and Cd (CF up to 2.9) was localized, whereas Pb, Cu, and Cr were in low contamination ranges. The pollution load index values at each site were less than 1, indicating a low total multi-metal pollution burden. Within the limitations of this screening-level assessment, Cd showed the highest contribution to the ecological risk indices among the analyzed metals, while the potential ecological risk index remained below 150 at all locations. However, the inhalation-only screening assessment showed HQ and HI values many-fold lower than threshold levels, indicating no apparent non-carcinogenic risk through the inhalation pathway under the applied assumptions.
The distribution of air pollution in urban environments is greatly influenced by its source and the complexity of the urban environment, which determine the pollution concentrations in individual locations. Traffic is one of the main pollution sources in cities and, especially particulate matter (PM) poses a serious risk for public health. This study applies magnetic biomonitoring using moss bags to establish an extensive urban canopy layer monitoring network (n = 52) to study the distributions and concentrations of magnetic PM. Moss bags were placed at street and roof levels within the grid plan area of Turku, Finland, for 62 days in late autumn. Samples were analysed for mass-specific magnetic susceptibility (χ), hysteresis parameters and elemental components. At street level, traffic was the primary source of magnetic PM and related elements. At higher altitudes, these concentrations were reduced and mainly replaced by crustal elements. Magnetic measurements were also applied to test for equivalence with the CAR-FMI dispersion model and to assess the representativeness of an air quality monitoring station in the city centre. Inconsistencies were observed between the model and magnetic PM concentrations at street level. At rooftop level, the model estimates were higher than indicated by the magnetic PM. The magnetic measurements complemented measurements derived from supplementary monitoring sensors and Turku monitoring station. The representativeness of the monitoring station should be re-evaluated, and recategorizing should be considered. Magnetic biomonitoring using moss bags is an unparalleled tool for monitoring airborne PM pollution and a reliable tool for evaluating monitoring stations and dispersion model performance.
This study examines two major dust storm events over Cameroon, in February 2021 and February 2023, using a multi-dataset framework combining ERA5 reanalysis, CAMS atmospheric products, GPCP and MODIS satellite observations to characterize their meteorological drivers and transport pathways. The analysis shows that both events were associated with suppressed precipitation, low relative humidity, strong northeasterly Harmattan winds, and high surface temperatures, creating favorable conditions for dust uplift and long-range transport. Spatial patterns reveal a coherent southward extension of dust plumes from northern Cameroon toward central and southern regions, with the 2023 event exhibiting a stronger and more persistent aerosol signal than the 2021 event. The results also show that aerosol loading evolves progressively across the domain, confirming a transported dust plume rather than isolated local emission. The main precursors of dust outbreaks were detectable about 24-48 hours before peak aerosol loading, suggesting a useful predictability window for threshold-based early warning, and emphasizing the urgent need for enhanced early warning systems, improved air quality monitoring, sustainable land management, and public awareness programs to mitigate adverse effects of dust storms.
The objective of this study is to examine the impact of renewable and non-renewable energy consumption, income, internet use and urbanization on CO₂ emissions in 38 OECD countries during over the period 1995–2022 using the heterogeneous panel quantile regression technique. The study addresses the existing gap in the literature by analyzing the relationship between income and renewable energy, as well as the Environmental Kuznets Curve (EKC) at specified quantile levels. The findings show that the EKC is more dominant in countries with low CO₂ emissions. Furthermore, the negative relationship between renewable energy consumption and CO2 emissions becomes considerably more significant in quantiles with higher CO2 emissions. The research clearly showed that while urbanisation is an important positive factor in increasing CO2 emissions, the use of the internet as an indicator of digitalization has a decreasing impact on CO2 emissions. Our study provides valuable insights for policymakers to identify effective strategies for renewable energy adoption, digitalization, and urbanization policies to ensure sustainable development while addressing environmental degradation. The results of this study indicate that in order to achieve green economic development, it is essential for governments to prioritize the utilization of renewable energy sources and the advancement of sustainable urbanization practices. Additional robustness checks based on MMQR and SIVQR estimations confirm the reliability of the empirical findings across alternative quantile estimation techniques.
Pollution is a growing problem with dire effects on public health; therefore, accurate prediction of Air Quality Index (AQI) is paramount for both urban development and public health strategy. Here, we present GreenAirOps, a production-ready MLOps system that ingests multithsource environmental data, automatizes preprocessing and feature extraction, and combines Random For- est and XGBoost with ensemble learning for prediction of low-latency, near real-time AQI. This production pipeline has full MLOps pipeline capabilities: Data versioning with DVC; experiment tracking and model registration with MLflow; automated retraining and deployment using GitHub Actions. The deployment infrastructure utilizes Docker containers run on an AWS environ- ment and is designed for production grade. The specific contributions of this work are: an optimized ensemble learning system for low-latency AQI prediction; an end-to-end MLOps system which guarantee reproducibility and operational capability; and a set of production grade functions for auto-retraining, model health monitoring and auto-rollback.
PM2.5, a fine particle, is sufficiently dangerous for classification as a Class-1 carcinogen, and it remarkably contributes to premature death. This study aimed to assess population exposure to PM2.5 in Seoul, South Korea, using data from air quality monitoring stations (AQMSs) and sensor-based monitoring instruments (SAMIs). Five AQMSs were verified using co-location tests, and population distribution data were segmented by sex and age in hourly grids. The population-weighted average concentration (PWAC) was calculated using occupancy rates and indoor-to-outdoor PM2.5 concentration ratios. Premature death was estimated using a concentration–risk function and compared to those of the WHO (World Health Organization) guidelines. In addition, machine-learning models, including random forest (RF) and a support vector machine (SVM), were used for PM2.5 estimation, achieving high accuracy (R² =0.83, RMSE = 10.1 µg/m³). The PWAC was found to be approximately 8
The evaluation of personal exposure to PM is a critical component of risk assessment: one of the main challenges is represented by the difficulty in simultaneously obtaining extensive and accurate personal exposure data for large populations. This study provides an empirical comparison between personal PM2.5 and PM10 exposure measurements and routinely available ambient data from fixed monitoring stations and air quality models in a hybrid working scenario. Time-activity information was also explored to identify activity patterns descriptively associated with larger discrepancies between personal and ambient estimates. To achieve this goal, a complex scenario consisting of a working day from home and a working day from office was considered. Personal exposure data were compared with daily average ambient data retrieved for both the residential and workplace addresses of each subject, using three different approaches: (i) Address-Specific; (ii) Residential-Address; (iii) Workplace-Address. According to the results, ambient data may provide useful proxies for average exposures whereas their ability to represent individual exposure was limited. Exceptions were observed for PM2.5 ambient data from monitoring for the Address-Specific and Residential-Address references. The main discrepancies were observed at high personal exposure levels, where ambient data showed a strong tendency to underestimate. Finally, larger discrepancies were descriptively associated with selected activity patterns, including indoor free time away from home, personal care, and secondhand smoke exposure.
