Open research questions in Air Quality Monitoring and Forecasting

What the literature leaves open

• The need to handle non-linear relationships, noisy data, and sparse features in the data. The challenge of predicting daily maximum 8 h average O3 concentration and daily concentrations of NO and NO2.

  Attention-Based LSTM Forecasting of O3, NO2, and NO Concentrations in Eastern Santiago, Chile · 2026 · DOI

• Addressing the lack of temporal-based chronological splitting in existing approaches.

  Ensemble tree-based machine learning models for data-driven AQI prediction · 2026 · DOI

• The need to quantify station-to-grid uncertainty using nearest-station distance statistics and leave-one-out cross-validation. The challenge of separating fog from other weather conditions, such as precipitation.

  Classification and quantification of low-visibility events using deep learning over eastern China · 2026 · DOI

• Hardware such as LiDAR aging may lead to a decline in detection accuracy. The existing path planning algorithm may fail to adapt to complex traffic conditions. Dynamic obstacle avoidance is a challenge that needs to be addressed in subsequent work.

  A new method for detecting aerosols: combining atmospheric detection LiDAR technology with intelligent driving technology · 2026 · DOI

• Further research on dynamic obstacle avoidance will be pursued.

  A new method for detecting aerosols: combining atmospheric detection LiDAR technology with intelligent driving technology · 2026 · DOI

• The need for a combined approach to disentangle multiannual, multisite data. The limitation of traditional methods in identifying pollutant sources and addressing noise and sparse missing data.

  Disentangling Multiannual Air Quality Profiles Aided by Self-Organizing Map and Positive Matrix Factorization · 2025 · DOI

• Prior studies have explored the intersection of academic research and patented solutions, but they have yet to systematically link academic research with patented solutions. There is a need for a comprehensive understanding of the available technologies and identification of promising new directions for research and development.

  A Review of Academic and Patent Progress on Internet of Things (IoT) Technologies for Enhanced Environmental Solutions · 2025 · DOI

• Traditional air quality monitoring systems are often manual and labor-intensive. These systems cannot detect sudden environmental changes due to the lack of real-time sensing. There is a need for a system that can provide real-time monitoring and prediction of gas emissions in poultry farms.

  An IoT-Based System for Measuring Diurnal Gas Emissions of Laying Hens in Smart Poultry Farms · 2025 · DOI

• The need for low-cost and reliable CO2 monitoring systems for citizen scientists. The lack of suitable CO2 monitoring systems for agricultural land that can be used by non-expert users.

  Designing CO2 Monitoring System for Agricultural Land Utilizing Non-Dispersive Infrared (NDIR) Sensors for Citizen Scientists · 2025 · DOI

• The need for effective data pre-processing and feature selection. The challenge of selecting the optimal number of clusters for k-means clustering.

  A Two-Stage Machine Learning Framework for Air Quality Prediction in Hamilton, New Zealand · 2025 · DOI

• Lack of calibration of the AQM device. Inadequate awareness of air quality issues among residents and authorities. Insufficient funding for air quality monitoring and device production.

  Locating Low-Cost Air Quality Monitoring Devices in Low-Resource Regions Is Not Enough to Acquire Robust Air Quality Data Usable for Policy Decisions · 2025 · DOI

• The study identifies the limitation of using a small number of monitoring stations, which limits the robustness of the models and the spatial representativeness of the results. The restricted geographic coverage prevents a broader assessment of pollutant behavior throughout the metropolitan area.

  Modeling Air Pollution in Metropolitan Lima: A Statistical and Artificial Neural Network Approach · 2025 · DOI

• The paper discusses the need to address the contradictory metrics of range and sensitivity in sensor design.

  AI-Driven Sensing Technology: Review · 2024 · DOI

• The paper identifies the need for standardized calibration protocols to ensure the accuracy and reliability of low-cost sensor data. It highlights the challenges in calibrating low-cost sensors due to the variability in sensing materials, transducer designs, and environmental conditions.

  Challenges and Opportunities in Calibrating Low-Cost Environmental Sensors · 2024 · DOI

• There is a lack of understanding of ambient air quality in Hunan Province. The study identifies a gap in the prediction of atmospheric environmental quality in Hunan Province.

  Analysis and Prediction of Atmospheric Environmental Quality Based on the Autoregressive Integrated Moving Average Model (ARIMA Model) in Hunan Province, China · 2024 · DOI

• Existing algorithms may not provide sufficient robustness and convergence guarantees for federated learning on heterogeneous data.

  FedEHD: Entropic High-Order Descent for Robust Federated Multi-Source Environmental Monitoring · 2025 · DOI

• Explore hybrid approaches that integrate machine learning efficiency with CTM mechanistic rigor, - Explore hybrid models combining the interpretability of XGBoost with the temporal modeling capabilities of LSTM, - Extend these approaches to other pollutants and urban settings, - Integrate real-time data streams and multi-source predictors, such as satellite imagery and traffic flow…

  Attention-Based LSTM Forecasting of O3, NO2, and NO Concentrations in Eastern Santiago, Chile · 2026 · DOI

• Investigating other imputation methods for handling missing data - Exploring the application of the proposed framework to other datasets and regions - Developing more advanced machine learning models for AQI…

  Ensemble tree-based machine learning models for data-driven AQI prediction · 2026 · DOI

• Limited number and geographic coverage of ground-based pollution measurement stations - Getting trapped in local minima in gradient-based methods - High computational costs - Need for continuous objective…

  A Novel Evolutionary Deep Learning Approach for PM2.5 Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran · 2025 · DOI

• The study is limited to the city of Tehran, - The dataset used for training and testing is limited to the years 2014, 2015, and 2016, - The study does not consider other factors that may influence PM2.5 concentrations, - The proposed model may not be generalizable to other cities…

  A Novel Evolutionary Deep Learning Approach for PM2.5 Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran · 2025 · DOI

• Further research could investigate the use of the CO2 monitoring system in different environmental conditions - The text does not provide information on suggested future research…

  Designing CO2 Monitoring System for Agricultural Land Utilizing Non-Dispersive Infrared (NDIR) Sensors for Citizen Scientists · 2025 · DOI

• The study was conducted in only four classrooms, - The study was limited to a single school in San Jose, California, - The outdoor air dampers were set to a fixed position, - The supply and return airflow rates for each classroom were measured using…

  Impact of Teachers’ Decisions and Other Factors on Air Quality in Classrooms: A Case Study Using Low-Cost Air Quality Sensors · 2025 · DOI

• Further study is needed to investigate the impact of teacher decisions on IAQ in different types of classrooms, - Research is needed to develop more effective strategies for improving IAQ in schools, - Studies should be conducted to evaluate the effectiveness of low-cost air quality sensors in different…

  Impact of Teachers’ Decisions and Other Factors on Air Quality in Classrooms: A Case Study Using Low-Cost Air Quality Sensors · 2025 · DOI

• Lack of a means of calibrating the device, - No reference AQM instrument was available, - Inadequate funding, - Unavailability of trained manpower…

  Locating Low-Cost Air Quality Monitoring Devices in Low-Resource Regions Is Not Enough to Acquire Robust Air Quality Data Usable for Policy Decisions · 2025 · DOI

• Test the model in more regions to further evaluate its generalization ability, - Integrate more data sources, such as emission inventories and satellite observations, to enhance the predictive power of the model, - Explore the potential of the model in multi-pollutant prediction and long-term…

  Prediction of PM2.5 Concentration Based on Deep Learning, Multi-Objective Optimization, and Ensemble Forecast · 2024 · DOI

Most-cited papers in Air Quality Monitoring and Forecasting

• Tackling Climate Change with Machine Learning · OPUS 4 (Zuse Institute Berlin) · 2022 · 865 citations
• Improved 1 km resolution PM 2.5 estimates across China using enhanced space–time extremely randomized trees · Atmospheric chemistry and physics · 2020 · 685 citations
• Data-Driven Machine Learning in Environmental Pollution: Gains and Problems · Environmental Science & Technology · 2022 · 518 citations
• The ChinaHighPM10 dataset: generation, validation, and spatiotemporal variations from 2015 to 2019 across China · Environment International · 2020 · 510 citations
• A review of green artificial intelligence: Towards a more sustainable future · Neurocomputing · 2024 · 456 citations
• Multi-hour and multi-site air quality index forecasting in Beijing using CNN, LSTM, CNN-LSTM, and spatiotemporal clustering · Expert Systems with Applications · 2020 · 427 citations
• Deep Air Quality Forecasting Using Hybrid Deep Learning Framework · IEEE Transactions on Knowledge and Data Engineering · 2021 · 396 citations
• The AI gambit: leveraging artificial intelligence to combat climate change—opportunities, challenges, and recommendations · AI & Society · 2021 · 390 citations
• Machine Learning: New Ideas and Tools in Environmental Science and Engineering · Environmental Science & Technology · 2021 · 374 citations
• A Hybrid CNN-LSTM Model for Forecasting Particulate Matter (PM2.5) · IEEE Access · 2020 · 372 citations

Most recent work

• Interpretable machine learning framework for air quality prediction in Istanbul using Shapley additive explanations (SHAP) · Stochastic Environmental Research and Risk Assessment · 2026
• Harnessing deep learning for air pollution forecasting: trends, techniques, and future prospects · Artificial Intelligence Review · 2026
• Towards long-term air quality monitoring in Africa and the use of machine learning, low-cost monitors, reference instruments and satellite data to estimate the burden of disease – The case of Kampala, Uganda · Urban Climate · 2026
• Evaluating machine learning (RF, XGBoost) and statistical model (MLR) for PM10 and air quality prediction: A case from Kars, Türkiye · Atmospheric Pollution Research · 2026
• The reliability gap: Why high predictive accuracy doesn't guarantee stable feature importance · Marine Pollution Bulletin · 2026
• Room-Temperature Trace NO 2 Monitoring System Based on Two-Dimensional Heterostructures and Integrated with Deep Learning · ACS Sensors · 2026
• A comparative study on CMAQ-based predictors and reanalysis AOD for improving PM2.5 estimation in Thailand using machine learning · Atmospheric Environment · 2026
• Carbon emission prediction based on spatial–temporal pattern recognition and novel integrated vine copula · International Journal of Environmental Science and Technology · 2026
• Enhancing Accuracy of Indoor Air Quality Sensors via Automated Machine Learning Calibration · Atmospheric measurement techniques · 2026
• ACROPOLIS: Munich urban CO 2 sensor network · Atmospheric measurement techniques · 2026

Related topics in Environmental Science