A Comprehensive Review of Deep Learning Approaches for Air Quality Pollution Analysis and Prediction

Abstract

Due to the rapid urbanization and industrialization, and the escalated vehicular emissions, air pollution has resulted in a critical global environmental and public health issue. The complexity of the spatiotemporal dynamics of pollutants is usually not well represented by conventional statistical and deterministic models. This paper presents a systematic review of recent deep learning methods for air quality prediction, including LSTM, GRU, CNN, autoencoders, and hybrid CNN-LSTM models. In contrast to the previous surveys that are confined to the performance evaluation of the models, the review takes a critical look at data integration strategies, the spatiotemporal modeling processes, the uncertainty estimation, the real time monitoring applications, and the feasible issues of the deployment. The research synthesizes the recent literature (20182025) findings and identifies the research gaps concerning the interpretability, generalization, and infrastructure requirements. In the review, the authors also discuss the application of deep learning to the process of smart city planning, health risk identification, and environmental policy-making. It shows that the hybrid deep learning models are much more effective than the traditional ones, and present-day challenges of the explainability and scalability of the solutions are necessary to implement deep learning models in the real-world setting.

Introduction

Rapid industrialization, urbanization, population growth, increased vehicle use, and fossil fuel consumption have significantly degraded air quality worldwide. Key pollutants include PM2.5, PM10, NO2, SO2, CO, O3, and VOCs, which lead to severe health issues such as respiratory diseases, cardiovascular problems, neurological disorders, and premature deaths. Air pollution also affects agriculture, ecosystems, and overall quality of life. Effective monitoring and forecasting of air quality are crucial for environmental management, public health, and policy-making.

Traditional Methods of Air Quality Analysis
Conventional approaches rely on statistical, regression-based, or deterministic atmospheric models. While useful, these methods struggle with nonlinear, dynamic, and large-scale air pollution data due to assumptions, complex parameter tuning, and limited flexibility.

Deep Learning for Air Quality Prediction
The rise of high-resolution environmental data, IoT sensors, and satellite imagery has enabled data-driven approaches. Deep learning (ANNs, CNNs, RNNs, LSTM, GRU) excels at modeling nonlinear and spatiotemporal patterns in air quality data.

• RNNs/LSTM/GRU: Capture long-term temporal dependencies in pollution data.

• CNNs: Model spatial patterns across monitoring points.

• Hybrid CNN-RNN systems: Combine spatial and temporal modeling for improved predictive accuracy.
  Advantages include robustness to noise, scalability, and the ability to integrate multi-source data (ground stations, traffic, satellite, weather). Accurate forecasts support early warning systems and preventive measures for public health.

Literature Insights and Applications

• Hybrid models (CNN-LSTM, CNN-BiLSTM, Conv1D-BiLSTM) improve AQI and pollutant forecasting accuracy.

• Incorporating spatial correlations, uncertainty quantification, and multi-source data enhances prediction stability and reliability.

• Applications include agriculture planning, smart city pollution management, and real-time monitoring during emergencies.

Research Gaps
Despite progress, challenges remain:

• Model interpretability and explainability.

• Cross-regional generalization.

• Handling heterogeneous or incomplete data.

• Practical deployment in real-time environmental governance and smart city frameworks.

Contributions of Current Review

1. Structured comparison of deep learning architectures for spatiotemporal air quality modeling.

2. Synthesis of hybrid and uncertainty-aware models.

3. Analysis of implementation challenges, including computational demands and interpretability.

4. Integration with IoT, satellites, and smart city frameworks.

5. Identification of future research directions emphasizing explainable AI and scalable deployment.

Air Pollution Overview

• Sources: Natural (wildfires, dust storms, volcanic activity) and anthropogenic (vehicles, factories, power plants, agriculture).

• Impacts: Human health (respiratory/cardiovascular), agriculture, ecosystems, biodiversity, acid rain, and environmental degradation.

• Monitoring: Essential for identifying hotspots, guiding policies, issuing health alerts, and evaluating mitigation strategies.

Monitoring Systems

• Fixed stations: High-precision, limited coverage.

• Portable sensors: Flexible, real-time exposure measurement.

• Satellite remote sensing: Large-scale coverage, tracking long-term and transboundary pollution.

• Technological advances: IoT and affordable sensors enable real-time, dense monitoring networks; remote sensing supplements inaccessible areas.

Data for Deep Learning Models

• Ground pollutant measurements (PM2.5, PM10, NO2, O3).

• Meteorological data (temperature, humidity, wind, precipitation).

• Satellite imagery for spatial distribution.

• Traffic and industrial emissions.

• Multi-source data fusion enhances temporal and spatial predictions for effective air quality management.

Conclusion

To sum up, air quality pollution is a form of global challenge with grave consequences on the health of the population, environmental sustainability and socio-economic development that require sophisticated analytical and predictive solutions to this problem. Conventional air quality monitoring and forecasting methods would not be sufficient to detect the complex, nonlinear and dynamic interactions between pollutants, meteorological factors, source of emissions and geographic location. Deep learning models have proven to be a strong and efficient solution since they allow extracting features automatically, including heterogeneous data sources, and precisely predicting the level of the pollutant in space and time. Convolutional neural networks, long short term memory networks, autoencoders, and hybrid networks have proven effective in air quality prediction, real-time prediction, anomalies, and source location as well as health risk assessment. These features help in active environmental management, through ease of early warning mechanism, mitigation plans, and formulation of policy based on evidence. However, the application of the deep learning-based air quality systems is still limited by the challenges such as scarce data availability and quality, high computational and infrastructure requirements, model interpretability, and the inability to extrapolate predictions to a different area and time. To overcome these constraints, data needs to be enhanced, and scalable computing infrastructure will need to be invested in, as well as explainable artificial intelligence methods to increase transparency and trust. Even with these problems, the incorporation of deep learning into air quality analysis is an important step in the direction of smart and data-driven environmental regulation.

References

[1] K. Chan, P. Matthews, and K. Munir, “Time Series Forecasting for Air Quality with Structured and Unstructured Data Using Artificial Neural Networks,” Atmosphere (Basel)., vol. 16, no. 3, pp. 1–17, 2025, doi: 10.3390/atmos16030320. [2] S. Alkayal, H. Almisbahi, S. Baowidan, and E. Alkayal, “Air Pollution Trends and Predictive Modeling for Three Cities with Different Characteristics Using Sentinel-5 Satellite Data and Deep Learning,” Atmosphere (Basel)., vol. 16, no. 2, 2025, doi: 10.3390/atmos16020211. [3] I. Ganguli, M. Nakum, B. Das, and N. Kshetrimayum, “Comprehensive Analysis of Air Quality Trends in India Using Machine Learning and Deep Learning Models,” ICDCN 2025 - Proc. 26th Int. Conf. Distrib. Comput. Netw., no. January 2025, pp. 313–318, 2025, doi: 10.1145/3700838.3703681. [4] T. Istiana et al., “Causality Analysis of Air Quality and Meteorological Parameters for PM2.5 Characteristics Determination: Evidence from Jakarta,” Aerosol Air Qual. Res., vol. 23, no. 9, p. 230014, Sep. 2023, doi: 10.4209/aaqr.230014. [5] R. Zrieq, S. Kamel, F. Al-Hamazani, S. Boubaker, R. Attili, and M. J. Araúzo-Bravo, “Spatial-Temporal Forecasting of Air Pollution in Saudi Arabian Cities Based on a Deep Learning Framework Enabled by AI,” Toxics, vol. 13, no. 8, pp. 1–31, 2025, doi: 10.3390/toxics13080682. [6] H. Madokoro and S. Nix, “Multimodal Particulate Matter Prediction: Enabling Scalable and High-Precision Air Quality Monitoring Using Mobile Devices and Deep Learning Models,” Sensors, vol. 25, no. 13, pp. 1–23, 2025, doi: 10.3390/s25134053. [7] S. Jayaraman, N. T, A. S, and S. G, “Enhancing urban air quality prediction using time-based-spatial forecasting framework,” Sci. Rep., vol. 15, no. 1, pp. 1–15, 2025, doi: 10.1038/s41598-024-83248-z. [8] A. K. Rad, M. J. Nematollahi, A. Pak, and M. Mahmoudi, “Predictive modeling of air quality in the Tehran megacity via deep learning techniques,” Sci. Rep., vol. 15, no. 1, pp. 1–20, 2025, doi: 10.1038/s41598-024-84550-6. [9] I. Zandi, A. Jafari, and A. Lotfata, “Enhancing PM2.5 Air Pollution Prediction Performance by Optimizing the Echo State Network (ESN) Deep Learning Model Using New Metaheuristic Algorithms,” Urban Sci., vol. 9, no. 5, 2025, doi: 10.3390/urbansci9050138. [10] A. Bekkar, B. Hssina, S. Douzi, and K. Douzi, “Air-pollution prediction in smart city, deep learning approach,” J. Big Data, vol. 8, no. 1, pp. 1–21, 2021, doi: 10.1186/s40537-021-00548-1. [11] A. Dairi, F. Harrou, S. Khadraoui, and Y. Sun, “Integrated Multiple Directed Attention-Based Deep Learning for Improved Air Pollution Forecasting,” IEEE Trans. Instrum. Meas., vol. 70, 2021, doi: 10.1109/TIM.2021.3091511. [12] C. Guo, G. Liu, and C. H. Chen, “Air Pollution Concentration Forecast Method Based on the Deep Ensemble Neural Network,” Wirel. Commun. Mob. Comput., vol. 2020, 2020, doi: 10.1155/2020/8854649. [13] Y. Zhou, F. J. Chang, L. C. Chang, I. F. Kao, and Y. S. Wang, “Explore a deep learning multi-output neural network for regional multi-step-ahead air quality forecasts,” J. Clean. Prod., vol. 209, pp. 134–145, 2019, doi: 10.1016/j.jclepro.2018.10.243. [14] B. S. Freeman, G. Taylor, B. Gharabaghi, and J. Thé, “Forecasting air quality time series using deep learning,” J. Air Waste Manag. Assoc., vol. 68, no. 8, pp. 866–886, 2018, doi: 10.1080/10962247.2018.1459956. [15] S. J et al., “Smart Agriculture: Integrating Air Quality Monitoring with Deep Learning for Process Optimization,” Scalable Comput. Pract. Exp., vol. 26, no. 3, pp. 1005–1016, 2025, doi: 10.12694/scpe.v26i3.4190. [16] J. Mohammad and M. A. Kashem, “Deep Learning Approach for Evaluating Air Pollution Using the RFM Model,” Nat. Environ. Pollut. Technol., vol. 24, no. 2, 2025, doi: 10.46488/NEPT.2025.v24i02.D1718. [17] I. Ayus, N. Natarajan, and D. Gupta, “Comparison of machine learning and deep learning techniques for the prediction of air pollution: a case study from China,” Asian J. Atmos. Environ., vol. 17, no. 1, 2023, doi: 10.1007/s44273-023-00005-w. [18] B. Ghose and L. Anthopoulos, “A Deep Learning based Air Quality Prediction Technique Using Influencing Pollutants of Neighboring Locations in Smart City,” vol. 28, no. 8, pp. 799–826, 2022, doi: 10.3897/jucs.78884. [19] I. Mokhtari, W. Bechkit, H. Rivano, and M. R. Yaici, “Uncertainty-Aware Deep Learning Architectures for Highly Dynamic Air Quality Prediction,” IEEE Access, vol. 9, pp. 14765–14778, 2021, doi: 10.1109/ACCESS.2021.3052429. [20] F. Naz et al., “Comparative Analysis of Deep Learning and Statistical Models for Air Pollutants Prediction in Urban Areas,” IEEE Access, vol. 11, no. June, pp. 64016–64025, 2023, doi: 10.1109/ACCESS.2023.3289153. [21] S. Hameed, A. Islam, K. Ahmad, S. B. Belhaouari, J. Qadir, and A. Al-Fuqaha, “Deep learning based multimodal urban air quality prediction and traffic analytics,” Sci. Rep., vol. 13, no. 1, pp. 1–19, 2023, doi: 10.1038/s41598-023-49296-7. [22] G. T. M. N. Anwar, G. T. M. Navi Anwar, H. P. Gunasekar, J. Krishnan, and R. Sekar, “Air quality monitoring and assessment using IoT,” IOP Conf. Ser. Mater. Sci. Eng., vol. 955, no. 1, Nov. 2020, doi: 10.1088/1757-899X/955/1/012006. [23] S. Aguilar-Gomez, B. Dwyer, J. Graff Zivin, and M. Neidell, “This is Air: The ‘Non-Health’ Effects of Air Pollution,” SSRN Electron. J., 2022, doi: 10.2139/ssrn.4062400. [24] A. Kurasz, G. Y. H. Lip, S. Dobrzycki, and ?. Ku?ma, “A Breath of Trouble: Unraveling the Impact of Air Pollution on Atrial Fibrillation,” J. Clin. Med., vol. 13, no. 23, Dec. 2024, doi: 10.3390/jcm13237400. [25] X. Wang, G. Wu, Z. Xiang, and J. Zhang, “Air Pollution and Media Slant: Evidence from Chinese Corporate News,” Emerg. Mark. Financ. Trade, vol. 58, no. 10, pp. 2880–2894, 2022, doi: 10.1080/1540496X.2021.2013196. [26] K. Vohra et al., “Rapid rise in premature mortality due to anthropogenic air pollution in fast-growing tropical cities from 2005 to 2018,” Sci. Adv., vol. 8, no. 14, 2022, doi: 10.1126/sciadv.abm4435. [27] M. J. Wolf et al., “New Insights for Tracking Global and Local Trends in Exposure to Air Pollutants,” 2022, doi: 10.1021/acs.est.1c08080. [28] A. Dun, Y. Yang, and F. Lei, “A Novel Hybrid Model Based on Spatiotemporal Correlation for Air Quality Prediction,” Mob. Inf. Syst., vol. 2022, 2022, doi: 10.1155/2022/9759988. [29] L. Montalvo, D. Fosca, D. Paredes, M. Abarca, and C. Saito, “An Air Quality Monitoring and Forecasting System for Lima City With Low-Cost Sensors and Artificial Intelligence Models,” vol. 4, no. July, pp. 1–15, 2022, doi: 10.3389/frsc.2022.849762. [30] P. O. Czechowski et al., “Financing Costs and Health Effects of Air Pollution in the Tri-City Agglomeration,” Front. Public Heal., vol. 10, no. March, pp. 1–10, 2022, doi: 10.3389/fpubh.2022.831312. [31] U. Im et al., “Reviewing the links and feedbacks between climate change and air pollution in Europe,” no. September, pp. 1–19, 2022, doi: 10.3389/fenvs.2022.954045. [32] Q. Zhang, Y. Han, V. O. K. Li, and J. C. K. Lam, “Deep-AIR: A Hybrid CNN-LSTM Framework for Fine-Grained Air Pollution Estimation and Forecast in Metropolitan Cities,” IEEE Access, vol. 10, pp. 55818–55841, 2022, doi: 10.1109/ACCESS.2022.3174853. [33] A. Choudhary et al., “Time Series Simulation and Forecasting of Air Quality Using In-situ and SatelliteBased Observations over an Urban Region,” Nat. Environ. Pollut. Technol., vol. 21, no. 3, pp. 1037–1048, 2022, doi: 10.46488/NEPT.2022.v21i03.018. [34] M. G. Vivanco et al., “Assessment of the Effects of the Spanish National Air Pollution Control Programme on Air Quality,” 2021. [35] A. Dutta and W. Jinsart, “Air Pollution in Indian Cities and Comparison of MLR , ANN and CART Models for Predicting PM 10 Concentrations in Guwahati , India,” vol. 15, no. 1, 2021. [36] V. Thi, T. Minh, T. T. Tin, and T. T. Hien, “PM 2 . 5 Forecast System by Using Machine Learning and WRF Model , A Case Study?: Ho Chi Minh City ,” vol. 21, no. 12. [37] G. Huang et al., “Large scale air pollution prediction with deep convolutional networks,” Sci. China Inf. Sci., vol. 64, no. 9, 2021, doi: 10.1007/s11432-020-2951-1. [38] M. Arsov, E. Zdravevski, R. Corizzo, N. Koteli, and K. Mitreski, “Short-term air pollution forecasting based on environmental factors and deep learning models,” vol. 21, pp. 15–22, 2020, doi: 10.15439/2020F211. [39] L. Vitali et al., “A standardized methodology for the validation of air quality forecast applications (F-MQO): lessons learnt from its application across Europe,” Geosci. Model Dev., vol. 16, no. 20, pp. 6029–6047, 2023, doi: 10.5194/gmd-16-6029-2023. [40] Y. Yarragunta, S. Srivastava, D. Mitra, and H. Chandra, “Influence of forest fire episodes on the distribution of gaseous air pollutants over,” GIScience Remote Sens., vol. 57, no. 2, pp. 190–206, 2020, doi: 10.1080/15481603.2020.1712100. [41] Y. A. Y. Okkao?lu and E. G. M. E. Yücel, “Estimation and forecasting of ­ PM 10 air pollution in Ankara via time series and harmonic regressions,” Int. J. Environ. Sci. Technol., vol. 17, no. 8, pp. 3677–3690, 2020, doi: 10.1007/s13762-020-02705-0. [42] S. Lolli, Y. C. Chen, S. H. Wang, and G. Vivone, “Impact of meteorological conditions and air pollution on COVID-19 pandemic transmission in Italy,” Sci. Rep., vol. 10, no. 1, pp. 1–15, 2020, doi: 10.1038/s41598-020-73197-8. [43] Di. Madaan, R. Dua, P. Mukherjee, and B. Lall, “Vayuanukulani: Adaptive memory networks for air pollution forecasting,” Glob. 2019 - 7th IEEE Glob. Conf. Signal Inf. Process. Proc., 2019, doi: 10.1109/GlobalSIP45357.2019.8969343. [44] H. Wang, B. Zhuang, Y. Chen, N. Li, and D. Wei, “Deep inferential spatial-temporal network for forecasting air pollution concentrations,” arXiv, 2018. [45] S. He, S. Tang, Y. Xiao, and R. A. Cheke, “Stochastic Modelling of Air Pollution Impacts on Respiratory Infection Risk,” Bull. Math. Biol., vol. 80, no. 12, pp. 3127–3153, 2018, doi: 10.1007/s11538-018-0512-5. [46] R. San José, J. L. Pérez, L. Pérez, and R. M. Gonzalez Barras, “Effects of climate change on the health of citizens modelling urban weather and air pollution,” Energy, vol. 165, pp. 53–62, 2018, doi: 10.1016/j.energy.2018.09.088. [47] J. Bousquet et al., “POLLAR: Impact of air POLLution on Asthma and Rhinitis; A European Institute of Innovation and Technology Health (EIT Health) project,” Clin. Transl. Allergy, vol. 8, no. 1, pp. 1–13, 2018, doi: 10.1186/s13601-018-0221-z.

Copyright

Copyright © 2026 Sanjana Tiwari, Vishal Sharma. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.