Explore the vast range of titles available.

An automated air quality forecasting system (AI‐Air) was developed to optimize and improve air quality forecasting for different typical cities, combined with the China Meteorological Administration Unified Atmospheric Chemistry Environmental Model (CUACE), and used in a typical inland city of Zhengzhou and a coastal city of Haikou in China. The performance evaluation results show that for the PM2.5 forecasts, the correlation coefficient (R) is increased by 0.07–0.13, and the mean error (ME) and root mean square error (RMSE) is decreased by 3.2–3.5 and 3.8–4.7 μg/m³. Similarly, for the O3 forecasts, the R value is improved by 0.09–0.44, and ME and RMSE values are reduced by 7.1–22.8 and 9.0–25.9 μg/m³, respectively. Case analyses of operational forecasting also indicate that the AI‐Air system can significantly improve the forecasting performance of pollutant concentrations and effectively correct underestimation, or overestimation phenomena compared to the CUACE model. Additionally, explanatory analyses were performed to assess the key meteorological factors affecting air quality in cities with different topographic and climatic conditions. The AI‐Air system highlights the potential of AI techniques to improve forecast accuracy and efficiency, and with promising applications in the field of air quality forecasting. Plain Language Summary Currently, artificial intelligence (AI) technology provides an innovative technological way to solve air quality problems with its excellent capability. This work develops an advanced automated air quality forecasting system (AI‐Air), based on the China Meteorological Administration Unified Atmospheric Chemical Environmental Model (CUACE). By comparing the forecasting results with the existing numerical models, the AI‐Air system shows its excellent performance in both overall performance evaluation and case‐specific forecasting. The AI‐Air system not only surpasses the conventional methods in forecasting accuracy but also demonstrates its fine forecasting ability in detail. In addition, this study provides an in‐depth discussion of the key factors affecting air quality in different types of cities and conducts a feature importance analysis. This analysis deepens the understanding of the intrinsic mechanisms of air quality changes in different urban environments and provides a scientific basis for formulating more precise air quality management strategies. Overall, the development and application of the AI‐Air system not only improves the science and accuracy of air quality prediction, but also provides strong technical support for urban environmental management and policy formulation. Key Points An automated ML system (AI‐Air) is developed for urban air quality forecasting Operational analyses show effective correction of under‐/overestimation phenomena Explanatory analyses explore key influencing factors in inland and coastal cities

Air-quality monitoring and forecasting are crucial for atmosphere pollution control and management. We propose an innovative data-driven framework for air quality index (AQI) prediction by integrating GNSS radio occultation (GNSS-RO) observation and weather modeling. Empowered by the state-of-the-art machine learning approach, our method can effectively predict regional AQI with a comparable accuracy much more quickly than the traditional numerical modeling and simulation approach. In a real case study using a representative region of China, our data-driven approach achieves a 2000 times speedup; meanwhile, the prediction error measured by rRMSE is only 2.4%. We investigate further the effects of different models, hyperparameters, and meteorological factors on the performance of our AQI prediction framework, and reveal that wind field and atmospheric boundary-layer height are important influencing factors of AQI. This paper showcases a direct application of GNSS-RO observation in assisting in forecasting regional AQI. From a machine learning point of view, it provides a new way to leverage the unique merits of GNSS atmospheric remote sensing technology with the help of the more traditional weather forecasting modeling approach.

The randomness, non-stationarity and irregularity of air quality index series bring the difficulty of air quality index forecasting. To enhance forecast accuracy, a novel model combining grey accumulated generating technique and Holt–Winters method is developed for air quality index forecasting in this paper. The grey accumulated generating technique is utilized to handle non-stationarity of random and irregular data series and Holt–Winters method is employed to deal with the seasonal effects. To verify and validate the proposed model, two monthly air quality index series from January in 2014 to December in 2016 collected from Shijiazhuang and Handan in China are taken as the test cases. The experimental results show that the proposed model is remarkably superior to conventional Holt–Winters method for its higher forecast accuracy.

The contribution of meteorology and emissions to long-term PM2.5 trends is critical for air quality management but has not yet been fully analyzed. Here, we used the combination of a machine learning model, statistical method, and chemical transport model to quantify the meteorological impacts on PM2.5 pollution during 2000–2018. Specifically, we first developed a two-stage machine learning PM2.5 prediction model with a synthetic minority oversampling technique to improve the satellite-based PM2.5 estimates over highly polluted days, thus allowing us to better characterize the meteorological effects on haze events. Then we used two methods to examine the meteorological contribution to PM2.5: a generalized additive model (GAM) driven by the satellite-based full-coverage daily PM2.5 retrievals and the Weather Research and Forecasting/Community Multiscale Air Quality (WRF/CMAQ) modeling system. We found good agreements between GAM estimations and the CMAQ model estimations of the meteorological contribution to PM2.5 on a monthly scale (correlation coefficient between 0.53–0.72). Both methods revealed the dominant role of emission changes in the long-term trend of PM2.5 concentration in China during 2000–2018, with notable influence from the meteorological condition. The interannual variabilities in meteorology-associated PM2.5 were dominated by the fall and winter meteorological conditions, when regional stagnant and stable conditions were more likely to happen and when haze events frequently occurred. From 2000 to 2018, the meteorological contribution became more unfavorable to PM2.5 pollution across the North China Plain and central China but were more beneficial to pollution control across the southern part, e.g., the Yangtze River Delta. The meteorology-adjusted PM2.5 over eastern China (denoted East China in figures) peaked in 2006 and 2011, mainly driven by the emission peaks in primary PM2.5 and gas precursors in these years. Although emissions dominated the long-term PM2.5 trends, the meteorology-driven anomalies also contributed -3.9 % to 2.8 % of the annual mean PM2.5 concentrations in eastern China estimated from the GAM. The meteorological contributions were even higher regionally, e.g., -6.3 % to 4.9 % of the annual mean PM2.5 concentrations in the Beijing-Tianjin-Hebei region, -5.1 % to 4.3 % in the Fenwei Plain,-4.8 % to 4.3 % in the Yangtze River Delta, and -25.6 % to 12.3 % in the Pearl River Delta. Considering the remarkable meteorological effects on PM2.5 and the possible worsening trend of meteorological conditions in the northern part of China where air pollution is severe and population is clustered, stricter clean air actions are needed to avoid haze events in the future.

This review provides a community's perspective on air quality research focusing mainly on developments over the past decade. The article provides perspectives on current and future challenges as well as research needs for selected key topics. While this paper is not an exhaustive review of all research areas in the field of air quality, we have selected key topics that we feel are important from air quality research and policy perspectives. After providing a short historical overview, this review focuses on improvements in characterizing sources and emissions of air pollution, new air quality observations and instrumentation, advances in air quality prediction and forecasting, understanding interactions of air quality with meteorology and climate, exposure and health assessment, and air quality management and policy. In conducting the review, specific objectives were (i) to address current developments that push the boundaries of air quality research forward, (ii) to highlight the emerging prominent gaps of knowledge in air quality research, and (iii) to make recommendations to guide the direction for future research within the wider community. This review also identifies areas of particular importance for air quality policy. The original concept of this review was borne at the International Conference on Air Quality 2020 (held online due to the COVID 19 restrictions during 18–26 May 2020), but the article incorporates a wider landscape of research literature within the field of air quality science. On air pollution emissions the review highlights, in particular, the need to reduce uncertainties in emissions from diffuse sources, particulate matter chemical components, shipping emissions, and the importance of considering both indoor and outdoor sources. There is a growing need to have integrated air pollution and related observations from both ground-based and remote sensing instruments, including in particular those on satellites. The research should also capitalize on the growing area of low-cost sensors, while ensuring a quality of the measurements which are regulated by guidelines. Connecting various physical scales in air quality modelling is still a continual issue, with cities being affected by air pollution gradients at local scales and by long-range transport. At the same time, one should allow for the impacts from climate change on a longer timescale. Earth system modelling offers considerable potential by providing a consistent framework for treating scales and processes, especially where there are significant feedbacks, such as those related to aerosols, chemistry, and meteorology. Assessment of exposure to air pollution should consider the impacts of both indoor and outdoor emissions, as well as application of more sophisticated, dynamic modelling approaches to predict concentrations of air pollutants in both environments. With particulate matter being one of the most important pollutants for health, research is indicating the urgent need to understand, in particular, the role of particle number and chemical components in terms of health impact, which in turn requires improved emission inventories and models for predicting high-resolution distributions of these metrics over cities. The review also examines how air pollution management needs to adapt to the above-mentioned new challenges and briefly considers the implications from the COVID-19 pandemic for air quality. Finally, we provide recommendations for air quality research and support for policy.

Addressing the constraints inherent in traditional primary Air Quality Index (AQI) forecasting models and the shortcomings in the exploitation of meteorological data, this research introduces a novel air quality prediction methodology leveraging machine learning and the enhanced modeling of secondary data. The dataset employed encompasses forecast data on primary pollutant concentrations and primary meteorological conditions, alongside actual meteorological observations and pollutant concentration measurements, spanning from 23 July 2020 to 13 July 2021, sourced from long-term air quality projections at various monitoring stations within Jinan, China. Initially, through a rigorous correlation analysis, ten meteorological factors were selected, comprising both measured and forecasted data across five categories each. Subsequently, the significance of these ten factors was assessed and ranked based on their impact on different pollutant concentrations, utilizing a combination of univariate and multivariate significance analyses alongside a random forest approach. Seasonal characteristic analysis highlighted the distinct seasonal impacts of temperature, humidity, air pressure, and general atmospheric conditions on the concentrations of six key air pollutants. The performance evaluation of various machine learning-based classification prediction models revealed the Light Gradient Boosting Machine (LightGBM) classifier as the most effective, achieving an accuracy rate of 97.5% and an F1 score of 93.3%. Furthermore, experimental results for AQI prediction indicated the Long Short-Term Memory (LSTM) model as superior, demonstrating a goodness-of-fit of 91.37% for AQI predictions, 90.46% for O3 predictions, and a perfect fit for the primary pollutant test set. Collectively, these findings affirm the reliability and efficacy of the employed machine learning models in air quality forecasting.

The impacts of weather forecast uncertainties have not been quantified in current air quality forecasting systems. To address this, we developed an efficient 2‐D convolutional neural network‐surface ozone ensemble forecast (2DCNN‐SOEF) system using 2‐D convolutional neural network and weather ensemble forecasts, and we applied the system to 216‐hr ozone forecasts in Shenzhen, China. The 2DCNN‐SOEF demonstrated comparable performance to current operating forecast systems and met the air quality level forecast accuracies required by the Chinese authorities up to 144‐hr lead time. Uncertainties in weather forecasts contributed 38%–54% of the ozone forecast errors at 24‐hr lead time and beyond. The 2DCNN‐SOEF enabled an “ozone exceedance probability” metric, which better represented the risks of air pollution given the range of possible weather outcomes. Our ensemble forecast framework can be extended to operationally forecast other meteorology‐dependent environmental risks globally, making it a valuable tool for environmental management. Plain Language Summary Weather forecasts are intrinsically uncertain, but the impacts of that uncertainty on air quality forecasts are not explicitly quantified in current air quality forecast systems. We proposed here a surface ozone ensemble forecast system, analogous to modern weather ensemble forecast systems, to represent the probability distribution of forecasted surface ozone concentrations given 30–50 possible future weather outcomes. The computation costs of this surface ozone ensemble forecast system were greatly reduced using deep learning techniques that emphasized the spatial patterns of weather. We showed that the surface ozone ensemble forecast system's accuracy met the Chinese operational requirements. However, half of the ozone forecast error was due to weather forecast uncertainties, which cannot be completely eliminated even with perfect pollutant emission estimates and chemistry models. This weather‐induced innate uncertainty in air quality forecasts should be considered for effective air quality management. Key Points We built a deep‐learning surface ozone ensemble forecast system to quantify pollution risks given the range of possible weather outcomes Deep‐learning models accentuating the spatial patterns of weather effectively represented the ozone‐meteorology relationship Weather forecast uncertainties contributed 38%–54% of the ozone forecast errors at 24‐hr lead time in Shenzhen

This paper describes the pre-operational analysis and forecasting system developed during MACC (Monitoring Atmospheric Composition and Climate) and continued in the MACC-II (Monitoring Atmospheric Composition and Climate: Interim Implementation) European projects to provide air quality services for the European continent. This system is based on seven state-of-the art models developed and run in Europe (CHIMERE, EMEP, EURAD-IM, LOTOS-EUROS, MATCH, MOCAGE and SILAM). These models are used to calculate multi-model ensemble products. The paper gives an overall picture of its status at the end of MACC-II (summer 2014) and analyses the performance of the multi-model ensemble. The MACC-II system provides daily 96 h forecasts with hourly outputs of 10 chemical species/aerosols (O3, NO2, SO2, CO, PM10, PM2.5, NO, NH3, total NMVOCs (non-methane volatile organic compounds) and PAN+PAN precursors) over eight vertical levels from the surface to 5 km height. The hourly analysis at the surface is done a posteriori for the past day using a selection of representative air quality data from European monitoring stations. The performance of the system is assessed daily, weekly and every 3 months (seasonally) through statistical indicators calculated using the available representative air quality data from European monitoring stations. Results for a case study show the ability of the ensemble median to forecast regional ozone pollution events. The seasonal performances of the individual models and of the multi-model ensemble have been monitored since September 2009 for ozone, NO2 and PM10. The statistical indicators for ozone in summer 2014 show that the ensemble median gives on average the best performances compared to the seven models. There is very little degradation of the scores with the forecast day but there is a marked diurnal cycle, similarly to the individual models, that can be related partly to the prescribed diurnal variations of anthropogenic emissions in the models. During summer 2014, the diurnal ozone maximum is underestimated by the ensemble median by about 4 μg m−3 on average. Locally, during the studied ozone episodes, the maxima from the ensemble median are often lower than observations by 30–50 μg m−3. Overall, ozone scores are generally good with average values for the normalised indicators of 0.14 for the modified normalised mean bias and of 0.30 for the fractional gross error. Tests have also shown that the ensemble median is robust to reduction of ensemble size by one, that is, if predictions are unavailable from one model. Scores are also discussed for PM10 for winter 2013–1014. There is an underestimation of most models leading the ensemble median to a mean bias of −4.5 μg m−3. The ensemble median fractional gross error is larger for PM10 (~ 0.52) than for ozone and the correlation is lower (~ 0.35 for PM10 and ~ 0.54 for ozone). This is related to a larger spread of the seven model scores for PM10 than for ozone linked to different levels of complexity of aerosol representation in the individual models. In parallel, a scientific analysis of the results of the seven models and of the ensemble is also done over the Mediterranean area because of the specificity of its meteorology and emissions. The system is robust in terms of the production availability. Major efforts have been done in MACC-II towards the operationalisation of all its components. Foreseen developments and research for improving its performances are discussed in the conclusion.

Air pollution is a global problem that threatens environmental sustainability and severely affects public health. Monitoring air quality and predicting future pollution levels are critical for creating effective environmental policies and enabling individuals to take precautions against air pollution. This study presents a long-term assessment of daily Air Quality Index (AQI) prediction using machine learning models based on meteorological and pollutant data collected in eastern Türkiye from 2016 to 2024. The dataset includes four major air pollutants (PM₁₀, SO₂, NO₂, O₃) and five meteorological variables (temperature, precipitation, relative humidity, wind direction, wind speed). Three models—eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Support Vector Machine (SVM)—were evaluated using the coefficient of determination (R²), root mean square error (RMSE) and mean absolute error (MAE) as performance metrics. Among these, XGBoost achieved the highest prediction accuracy (R² = 0.999, RMSE = 0.234, MAE = 0.158). The results demonstrate that ensemble-based machine learning approaches, particularly XGBoost, can effectively model AQI fluctuations using environmental predictors. These results provide valuable insights for air quality forecasting systems and suggest practical implications for regional air pollution management and early warning systems, supporting public health protection and the development of environmental health policies.

Wildfire smoke is one of the most significant concerns of human and environmental health, associated with its substantial impacts on air quality, weather, and climate. However, biomass burning emissions and smoke remain among the largest sources of uncertainties in air quality forecasts. In this study, we evaluate the smoke emissions and plume forecasts from 12 state-of-the-art air quality forecasting systems during the Williams Flats fire in Washington State, US, August 2019, which was intensively observed during the Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) field campaign. Model forecasts with lead times within 1 d are intercompared under the same framework based on observations from multiple platforms to reveal their performance regarding fire emissions, aerosol optical depth (AOD), surface PM2.5, plume injection, and surface PM2.5 to AOD ratio. The comparison of smoke organic carbon (OC) emissions suggests a large range of daily totals among the models, with a factor of 20 to 50. Limited representations of the diurnal patterns and day-to-day variations of emissions highlight the need to incorporate new methodologies to predict the temporal evolution and reduce uncertainty of smoke emission estimates. The evaluation of smoke AOD (sAOD) forecasts suggests overall underpredictions in both the magnitude and smoke plume area for nearly all models, although the high-resolution models have a better representation of the fine-scale structures of smoke plumes. The models driven by fire radiative power (FRP)-based fire emissions or assimilating satellite AOD data generally outperform the others. Additionally, limitations of the persistence assumption used when predicting smoke emissions are revealed by substantial underpredictions of sAOD on 8 August 2019, mainly over the transported smoke plumes, owing to the underestimated emissions on 7 August. In contrast, the surface smoke PM2.5 (sPM2.5) forecasts show both positive and negative overall biases for these models, with most members presenting more considerable diurnal variations of sPM2.5. Overpredictions of sPM2.5 are found for the models driven by FRP-based emissions during nighttime, suggesting the necessity to improve vertical emission allocation within and above the planetary boundary layer (PBL). Smoke injection heights are further evaluated using the NASA Langley Research Center's Differential Absorption High Spectral Resolution Lidar (DIAL-HSRL) data collected during the flight observations. As the fire became stronger over 3–8 August, the plume height became deeper, with a day-to-day range of about 2–9 km a.g.l. However, narrower ranges are found for all models, with a tendency of overpredicting the plume heights for the shallower injection transects and underpredicting for the days showing deeper injections. The misrepresented plume injection heights lead to inaccurate vertical plume allocations along the transects corresponding to transported smoke that is 1 d old. Discrepancies in model performance for surface PM2.5 and AOD are further suggested by the evaluation of their ratio, which cannot be compensated for by solely adjusting the smoke emissions but are more attributable to model representations of plume injections, besides other possible factors including the evolution of PBL depths and aerosol optical property assumptions. By consolidating multiple forecast systems, these results provide strategic insight on pathways to improve smoke forecasts.