Science AI Journal
Environmental Science · Research topic

Open research questions in Air Quality Monitoring and Forecasting

70 unresolved questions extracted from the limitations and future-work sections of 293 Air Quality Monitoring and Forecasting papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • However, the following limitations of this study should be noted: Although this study achieved encouraging pre- dictive performance, some limitations are noted: The analysis was based on data from one monitoring station at Visakhapatnam, so it is not necessarily representative of the city as a whole, which has a wide range of urban heterogeneity.

    A robust and uncertainty-aware machine learning framework for PM2.5 prediction in a coastal urban Indian environment: implications for sustainable air quality management · 2026 · DOI

  • This study provides many opportunities for further study and application in air quality forecast models. First, since deep-learning-based methods (e.g., LSTM and GRU) have shown generally lower error rates than traditional statistical models in terms of predicting air pollution levels, but are limited in the accuracy of their predictions due to the lack of data and assumptions used in their development, they have the potential to be used as complementary tools to traditional statistical models in early warning systems provided that sufficient contextual information is available and proper validation techniques are employed. Second, researchers should develop spatiotemporal models (e.g., CNN-LSTM models, attention-based neural networks) that can learn both temporal patterns and spatial relationships among different air pollution monitoring sites in order to improve the robustness of predictive models in urban areas where emissions vary greatly over space and time. Third, incorporating additional variables that describe the context in which an area's air pollution level is being measured (e.g., intensity, industrial production indicators, land-use characteristics, population density, extreme weather events) has the potential to increase the explanatory power of models and better represent the various factors that influence variability in PM10 levels. traffic Lastly, providing the results of PM10 forecasts to the public through digital decision-making tools (e.g., web-based dashboards, mobile apps), could allow for timely access to air quality information and potentially support preventative action by vulnerable populations (i.e., children, older adults, individuals with chronic respiratory or cardiovascular disease). If uncertainty associated with the models is communicated evaluated continuously, the integration of PM10 forecasts into digital decision-making tools has the potential to provide significant contributions to evidence-based public health and environmental management. and models clearly are Author contributions: All authors have contributed equally to work.

    PM10 Concentration Forecasting: A Comparative Evaluation of Deep Learning and Time Series Methods · 2026 · DOI

  • The present study, while providing valuable insights, has certain limitations that warrant discussion and serve as av- enues for future research. Firstly, the dataset is confined to one year (January 1, 2021, to December 31, 2021), which 136 X. Wang et al. Modeling the effect of pollutant gas on PM2.5 in China with computational intelligence limits the capture of long-term trends and potential sea- sonal variations beyond what is implicitly represented within a single year’s data. Secondly, the current mod- els do not explicitly incorporate external meteorological factors (e.g., temperature, relative humidity, wind speed, wind direction, precipitation, or dynamic wet and dry con- ditions). These factors are known to significantly influence the formation, dispersion, and concentration of PM2.5 and PGs. Their exclusion may introduce confounding effects on the observed correlations, and we acknowledge that our use of annual average data does not capture these crucial dynamic variations. Future work should integrate these meteorological variables to build more comprehensive and accurate predictive models. Thirdly, while the models were developed using data from 12 Chinese cities, direct quan- titative predictions beyond these cities would necessitate local data collection and model re-calibration to account for unique regional conditions. Another aspect to consider is the models’ performance on extreme cases of PM2.5 concentrations or pollutant events. Predicting these rare but critical high-pollution scenarios is often more challenging for data-driven models, especially if such extreme values are underrepresented in the training dataset. While our models aim to capture the general trends, their accuracy during severe pollution episodes might vary. Future work could focus on developing or applying models specifically tailored to predict extreme events, potentially by incorporating more data on such occurrences or employing robust statistical methods for outliers. Furthermore, environmental variables, including pollut- ant concentrations, often exhibit nonstationarity over ex- tended periods due to factors such as evolving emission sources, climatic changes, or policy interventions. While our study utilized one year of data, which limits the impact of long-term trends, the potential effect of nonstationarity on the relationships between PGs and PM2.5 over longer timescales or under different future conditions is a sig- nificant consideration. Future research should investigate methods to account for nonstationarity, such as adaptive modeling techniques or time series analysis specifically designed for nonstationary data, to ensure the long-term robustness and applicability of the models. Furthermore, this study was limited to four modeling paradigms. Future work would benefit from benchmark- ing these results against more advanced or hybrid ma- chine learning models, such as eXtreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM) networks, or hybrid neuro-genetic models, which have shown strong performance in other air quality forecasting studies. Last- ly, the study focuses on five primary PGs; incorporating other contributing factors, such as water-soluble ions or regional-specific emission sources, could further refine the models. Future research could also incorporate a more rig- orous and dedicated uncertainty and sensitivity analysis using advanced statistical or computational methods to precisely quantify the impact of input variable variations and model parameter uncertainties on PM2.5 predictions.

    Modeling the effect of pollutant gas on PM2.5 in China with computational intelligence · 2026 · DOI

  • investigate learning methodologies, incorporate supplementary machine including deep learning, spatiotemporal models, and hybrid frameworks, so as to enhance the understanding of complex pollutant interactions and urban heterogeneity. Adding secondary and trace pollutants to the list of pollutants and using highresolution data on weather, transportation, and land use, could make predictions even more accurate. Research could also examine scalable domain adaptation and transfer learning methodologies inadequate monitoring infrastructure, encompassing real-time model updating and ensemble methods that integrate several source-city models. Finally, in operational air quality management systems would allow for an understanding of how strong, long-lasting, and useful predictive forecasting may be for policy and public health actions.

    Explainable PSO-optimised machine learning models for multi-pollutant air quality forecasting in major African cities with transfer learning · 2026 · DOI

  • 103409 Fernandez-Bou AS, Ortiz-Partida JP, Dobbin KB, Flores-Landeros H, Bernacchi LA, Medellín-Azuara J (2021) Underrepresented, understudied, underserved: gaps and opportunities for advancing justice in disadvantaged communities.

    Artificial intelligence and advanced monitoring for air and water pollution control in the USA: opportunities, challenges, and policy directions · 2026 · DOI

  • The field of remote sensing is undergoing a transformative shift, moving toward the implementation of higher-resolution and higher-frequency atmospheric observations. An exemplary instrument in this evolution is TEMPO, which is designed to provide hourly geostationary observations of critical pollutants such as NO2, O3, and various aerosols. This capability allows for unprecedented tracking of diurnal emission patterns and the intricate transport dynamics of pollution at a remarkably fine resolution of 1 km2 (Zoogman et al., 2017).

    Remote-sensing technologies for air pollution monitoring in the USA: a comprehensive review · 2026 · DOI

  • Future research should focus on improving sensor robustness, expanding cloud- based monitoring, integrating automated self-calibration functions, and conducting an economic feasibility analysis to support the large-scale industrial adoption of domestic CEMS technology.

    Development and Preliminary Validation of a 69% Domestic Content (TKDN) Continuous Emission Monitoring System for Industrial Flue Gas Monitoring · 2026 · DOI

  • error address (1) To accumulation and limited interpretability in time series forecasting, this study proposes the DGC-FCM-Ridge model. A trend-driven dynamic granulation strategy is introduced to extract robust granule-level statistical features, which long-term multivariate 48 effectively suppress noise and enhance the stability of long - horizon prediction. (2) A fuzzy cognitive map is employed to model causal relationships among granules. The resulting cognitive features complement the statistical granule features, enabling the model to capture latent inter-granular dependencies that are difficult to represent explicitly. By integrating granule- level statistical features and FCM-based cognitive features within a regression framework, the proposed model achieves stable and consistent forecasting performance, while avoiding the prediction instability and overfitting risks commonly observed in deep learning–based long-term forecasting models. (3) Extensive experimental results demonstrate that the proposed approach outperforms multiple baseline methods, including ARIMA and LSTM, across various evaluation metrics in terms of both prediction accuracy and reliability. Furthermore, in the reduced feature space obtained through linear predictors exhibit superior granulation, simple complex compared with generalization performance nonlinear models, highlighting the effectiveness of combining fine-grained representations with interpretable modeling structures for long-term forecasting. Future work can be extended in three directions. First, global optimization strategies, such as hybrid particle swarm optimization algorithms, can be incorporated to improve the convergence behavior of FCM weight learning. Second, the proposed framework can be extended to multi-objective forecasting to better capture interactions among multiple variables. Third, by integrating online learning mechanisms with visual causal analysis, an intelligent forecasting system with real-time updating and dynamic monitoring capabilities can be developed, providing coherent and interpretable solutions for complex tasks in environmental monitoring and industrial process control.

    A Dynamic Grain and Semantic-FCM Model for Long-Term Multivariate PM2.5 Forecasting · 2026 · DOI

  • Multi-objective optimization in smart port systems addressing simultaneous security, mobility, efficiency, and environmental impact goals remains underexplored; the Marmara Region lacks specific quantitative validation of how intelligent traffic management systems and truck appointment systems simultaneously achieve these multiple objectives.

    Identifying pollution clusters in Türkiye’s Marmara Region with multi-layer self-organizing maps · 2026 · DOI

  • Collaborative frameworks and knowledge-sharing mechanisms between successful smart port implementations in Europe/North America and Asia-Pacific port operators have limited documented case studies; specific data on information transfer effectiveness and adaptation strategies for fleet assignment systems and real-time monitoring systems across regions is absent.

    Identifying pollution clusters in Türkiye’s Marmara Region with multi-layer self-organizing maps · 2026 · DOI

  • International standardization frameworks for smart transportation systems in ports that account for regional variations in port management conditions and operational challenges across different countries have not been systematically developed; specific standards for Turkish port operations in the Marmara Region require comparative benchmarking against European implementations.

    Identifying pollution clusters in Türkiye’s Marmara Region with multi-layer self-organizing maps · 2026 · DOI

  • Current smart port implementations predominantly focus on traffic efficiency and truck management with low-to-medium participation levels; the integration of autonomous and connected vehicle technologies with cargo tracking systems for intermodal smart ports in the Marmara Region and similar contexts remains underdeveloped and requires systematic case studies.

    Identifying pollution clusters in Türkiye’s Marmara Region with multi-layer self-organizing maps · 2026 · DOI

  • Smart transportation systems implementation in Asia-Pacific port regions, particularly in Pacific island nations where ports play critical socioeconomic roles, remains limited compared to European and North American ports; specific deployment of traffic monitoring systems, truck management systems, and real-time corridor control systems in these geographic contexts requires investigation.

    Identifying pollution clusters in Türkiye’s Marmara Region with multi-layer self-organizing maps · 2026 · DOI

  • The paper demonstrates that 1% improvement in overall accuracy benefits minority classes, with particular impact on soot misclassification affecting radiative forcing estimates in climate models. Direct quantification of how classification improvements in feldspar and soot discrimination propagate through cloud microphysics and ice nucleation parameterizations in regional and global climate simulations remains unexplored.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • While the dataset contains 18,827 labeled spectra across 20 aerosol types, generalization performance of the machine learning framework on SPMS measurements from different atmospheric environments, seasons, or geographic locations has not been evaluated. Cross-dataset validation is needed to assess whether the supervised and semi-supervised models maintain classification fidelity for soot and feldspars across diverse real-world deployment scenarios.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • The supervised stacked autoencoder achieved 91.1% overall accuracy, but performance degradation on chemically complex classes (feldspar cSA, feldspar cSOA, cellulose) with high reconstruction errors has not been systematically addressed. Architectural modifications to the autoencoder that explicitly handle compositionally heterogeneous spectra with elevated SSE values require investigation.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • Class imbalance remains a fundamental constraint for Soot (0.8% support) and biological particles (Bacteria, Snomax, Agar, Hazelnut), where limited training data prevents model development despite their high atmospheric significance. Targeted approaches must be developed to overcome the scarcity of labeled single-particle mass spectra for these underrepresented but scientifically critical aerosol types.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • Current semi-supervised learning frameworks (self-training SVM and mean teacher autoencoder) show limited effectiveness at leveraging abundant unlabeled SPMS data available in real-world atmospheric measurements. Targeted improvements to semi-supervised architectures are needed to maximize utilization of unlabeled spectral data for rare aerosol classes like Soot and biological particles.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • The study identifies feldspar discrimination as a key challenge due to chemical overlap between Na-feldspar, K-feldspar, feldspar cSA, and feldspar cSOA subtypes. Advanced feature engineering or architectural modifications to single-particle mass spectrometry models must be developed to capture subtle spectral differences that distinguish these compositionally similar mineral species.

    Leveraging machine learning to enhance aerosol classification using Single-Particle Mass Spectrometry · 2026 · DOI

  • Although the ensemble models achieved an impressive accuracy rate, studies have identified some major issues that need to be resolved. First, the studies were conducted using open-source datasets with very little time and space resolution; thus, these 166 Jurnal Kesehatan Lingkungan/10.20473/jkl.v18i2.2026.159-170 Vol. 18 No.2 April 2026 (159-170) datasets do not adequately represent the potential for variability in pollutant exposure for populations living in urban areas. Second, the primary focus of the study was on classification accuracy and did not provide an explicit representation of health outcomes (i.e., hospital admission and death rates). This study did not incorporate spatio-temporal dimensions, as the primary objective was to evaluate the performance of ensemble learning models in pollutant and health-risk classification based on historical exposure records. Future research should integrate spatial information (e.g., geographic pollutant distribution) and temporal variation (e.g., hourly, daily, or seasonal fluctuations) to enable spatial risk mapping, forecasting, and more comprehensive time-series environmental health surveillance. In future studies, researchers should incorporate spatio-temporal data and population vulnerability factors (age, socioeconomic status) to enable risk stratification. The potential to increase the interpretability of the model and to support the application of this type of model to real-world policy decisions will also be enhanced through the expansion of the model into a multi-source hybrid ensemble model. In addition to enhancing the potential of this type of model to be applied to real-world decision- making, collaborative efforts between environmental agencies and public health institutions will be needed to develop and implement the infrastructure necessary to support the implementation of these types of intelligent monitoring systems nationally (39,43–44).

    Ensemble Learning Approaches for Air Pollution Classification and Environmental Health Risk Assessment · 2026 · DOI

  • The model was trained on data from 2016–2022 and tested exclusively on 2023, but no analysis is provided of model performance under changing emission profiles (e.g., post-COVID emission recovery, policy-driven emission reductions, or seasonal emission shifts). Long-term temporal stability and adaptation to non-stationary emission patterns in air quality forecasting requires investigation.

    FuXi-Air: air quality forecasting based on emission-meteorology-pollutant multimodal machine learning · 2026 · DOI

  • The cross-attention mechanism for pollutant-meteorology-emission coupling in FuXi-Air is introduced as a key architectural innovation, but the paper does not provide visualization or analysis of learned attention weights to demonstrate which meteorological variables, emission types, or pollutant interactions are prioritized. Interpretability analysis of the cross-attention coupling is needed to validate the physical plausibility of learned relationships.

    FuXi-Air: air quality forecasting based on emission-meteorology-pollutant multimodal machine learning · 2026 · DOI

  • The frame interpolation model used within the joint forecasting framework to improve temporal resolution from coarser meteorological inputs to hourly predictions is mentioned but not detailed. The specific interpolation architecture, training methodology, and its contribution to overall forecast accuracy relative to other framework components remain unclear and require explicit ablation analysis.

    FuXi-Air: air quality forecasting based on emission-meteorology-pollutant multimodal machine learning · 2026 · DOI

  • The quantile forecasting strategy for uncertainty quantification in FuXi-Air is described as addressing extreme pollution conditions, but the paper does not specify which quantiles were selected, how the quantile loss function parameters were tuned, or how the uncertainty estimates perform during compound extreme events (e.g., simultaneous high PM2.5 and O3). Detailed validation of quantile prediction accuracy across different pollution percentiles and pollution episode types is needed.

    FuXi-Air: air quality forecasting based on emission-meteorology-pollutant multimodal machine learning · 2026 · DOI

  • The CAMS emission inventory data used in the model has a temporal resolution of one month, while meteorological data and pollutant observations operate at hourly resolution. This temporal mismatch in the multimodal dataset may obscure short-term emission fluctuations during pollution episodes; higher temporal resolution emission data (daily or hourly) should be integrated to better capture acute emission-pollution coupling dynamics.

    FuXi-Air: air quality forecasting based on emission-meteorology-pollutant multimodal machine learning · 2026 · DOI

Most-cited papers in Air Quality Monitoring and Forecasting

Most recent work

Find a gap in your own Air Quality Monitoring and Forecasting sub-topic

This page shows what the Air Quality Monitoring and Forecasting literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Environmental Science

Environmental Education and Sustainability
249 open questions
Sustainability and Climate Change Governance
174 open questions
Microplastics and Plastic Pollution
146 open questions
Species Distribution and Climate Change
122 open questions
Wildlife Ecology and Conservation
117 open questions
Air Quality and Health Impacts
117 open questions
Climate variability and models
117 open questions
Urban Green Space and Health
107 open questions
70 open questions have been extracted from the limitations and future-work passages of 293 Air Quality Monitoring and Forecasting papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.