Aug 2026· SN Computer Science· Vol 7· 0 citations· 61 references
TL;DR
The findings highlight the potential of HFL with fog computing as a scalable, cost-effective, and privacy-preserving solution for real-time urban air quality monitoring in India, particularly in resource-constrained smart city environments.
There are significant gaps that remain in terms of model interpretability and the ability to generalize across climate variations, so this article provides a relatively comprehensive overview of the application of federated learning in air quality forecasting and monitoring.
Yuhao Wu· Mathematical Modeling and Al...· 0 citations
Addressing the growing problem of urban air pollution requires that sophisticated air pollution monitoring and forecasting systems be incorporated into civil engineering practice. New technologies that support smarter and more sustainable urban environments include the Internet of Things (IoT), big data analytics, and machine learning (ML). These technologies allow for real-time data acquisition, high-resolution monitoring, and precise predictive optimization. Continuous, spatially dense air quality monitoring is made possible by IoT-based sensor networks, while big data frameworks can effectively integrate and analyse diverse, multi-source datasets. Meanwhile, ML and deep learning can further improve forecasting accuracy, enabling urban planners and civil engineers to foresee pollution trends and implement proactive mitigation measures. Notwithstanding these developments, challenges persist for data integration, sensor calibration, model transparency, and the reliability of inexpensive monitoring systems. To overcome these constraints, improvements are needed in terms of data accuracy, robust calibration techniques, and the implementation of interpretable ML models. This review emphasizes the vital role of civil engineers in promoting resilient and sustainable urban development, advancing effective air quality management, and safeguarding public health through interdisciplinary collaboration between data scientists and policymakers.
Prakash Pathak, Pradip Jadhao· Ecology, Economy and Society...· 0 citations
An end-to-end Internet of Things framework designed for real-time water quality monitoring and predictive pollution modeling and a hybrid machine learning architecture—combining Long Short-Term Memory (LSTM) networks for time-series forecasting and Random Forest models for anomaly classification—is proposed.
Parvathy Krishna V, G. S, Sahala Mehrin et al.· International Journal of Tec...· 0 citations
Air quality information is commonly communicated through the air quality index (AQI), which residents often access through the nearest available monitoring site. Motivated by this real-world use pattern, we evaluated how well different monitoring networks represent local AQI across California’s San Joaquin Valley, one of the nation’s most polluted air basins. We used the Global Historical Air Pollution (GHAP) PM2.5 dataset as a spatially continuous reference surface to distinguish point-level concentration agreement from nearest-site AQI representativeness. We found that although regulatory monitors showed the closest overall agreement with GHAP in absolute PM2.5 concentration, their sparse distribution limited their ability to capture spatial variation in AQI. In contrast, the dense low-cost sensor network produced higher match rates with GHAP-derived AQI categories than the regulatory monitor network at daily, weekly, and monthly scales. This contrast was most evident under polluted conditions, when stronger spatial heterogeneity increased the need for finer-scale monitoring coverage. Under such conditions, even uncalibrated low-cost sensors performed more consistently with GHAP in AQI classification than regulatory monitors despite substantial noise and artifacts in the raw data. These findings suggest that dense low-cost sensor networks can improve the spatial relevance of public-facing AQI information and community-scale air quality awareness.
Kate DeMarsh, Zeyi Moo, Peizhi Hao et al.· Environmental Research Lette...· 0 citations
Air pollution is a global environmental and health issue that has adverse effects on climate change and human health and contributes to ecosystem degradation and biodiversity loss. Air quality monitoring is therefore crucial to understand the sources and impacts of air pollution and to assess mitigation interventions. However, the number of operative monitoring stations are usually limited in air quality monitoring networks due to the high cost of the instrumentation used to monitor selected air pollutants. This limits the number of stations and results in spatial gaps. However, air pollution is highly heterogeneous.
The recent development of sensor technology has offered a broad range of gas and particulate matter sensors of small size and relatively low energy consumption and cost. These features make low-cost sensors (LCS) an attractive alternative to increase the spatial coverage of air quality networks and to deploy affordable networks in countries with lower resources. Nonetheless, LCS are faced with limitations such as cross-sensitivity and drift, while sensor networks are characterized by their scale, mixed quality, data volatility, interdependencies and correlations among others. These factors, in conjunction with the inherent complexity of the entire LCS network, lead to the emergence of novel metrological challenges. Moreover, some of the approaches used in sensor networks like the use of sensor redundancy for fault detection and self- or co-calibration, require a more holistic metrological assessment than traditional metrology methods.
These metrological challenges must be addressed to ensure the data quality and comparability of LCS networks. For that purpose, several sensor network metrology techniques are currently under development in the European Partnership on Metrology project "Fundamentals of Sensor Network Metrology" (FunSNM). Some of these techniques include uncertainty propagation in sensor networks using Bayesian approaches and mathematical tools like Laplace tools, sensor fusion, correlation analysis, self- and co-calibration methods using optimised spatial statistics and machine learning algorithms among others. Current outputs of the projects will be included in this poster, showing the essential role of sensor network metrology in advancing the adaptability and reliability of sensor networks.
M. Iturrate-García, M. Zaidan, A. Vedurmudi et al.· 150th anniversary of the Met...· 0 citations
