Smart Sensor Networks for Urban Air Quality Management

Thesis Abstract

Urban air quality remains a critical challenge in rapidly expanding cities, where heterogeneous emission sources and dynamic meteorology complicate timely decision-making and public health protection. This study addresses the gap between dense sensor deployment and actionable, location-specific air quality information by designing and evaluating an integrated smart sensor network that supports real-time monitoring, data fusion, and decision-support for urban authorities. The aim is to develop a scalable ICT-driven framework that enhances measurement accuracy, spatiotemporal resolution, and governance insights for mitigating exposure to PM2.5, PM10, NO2, and O3. Specific objectives are (1) to engineer a network of low-cost, calibrated air quality sensors coupled with edge computing capabilities; (2) to develop data fusion and calibration algorithms that correct for sensor drift and environmental interference, achieving an overall accuracy within ±5 µg/m³ for PM2.5 and ±10 ppb for NO2 under urban conditions; (3) to implement a data assimilation scheme combining dense sensor data with fixed-reference stations and meteorological inputs to generate high-resolution pollution maps; (4) to evaluate the framework in a metropolitan area with diverse land-use patterns, traffic intensity, and industrial activity; (5) to assess the impact of the system on stakeholder decision-making, policy implementation, and public health advisories; and (6) to propose a set of governance and deployment guidelines for scalable urban air quality management. A mixed-methods approach is employed. The quantitative component uses a quasi-experimental design in which the sensor network is deployed across 12 pilot districts, sampling 150 mobile and stationary nodes over 12 months. Data collection instruments include low-cost electrochemical NO2 sensors, optical particle counters for PM, temperature and humidity sensors, and a central cloud platform for data ingestion. Calibration experiments are conducted against reference-grade monitors following identical calibration protocols, and performance is evaluated using regression analysis, Bland-Altman plots, and time-series cross-correlation. The data assimilation framework integrates sensor outputs with numerical air quality models through an ensemble Kalman filter, producing gridded pollutant concentration fields at 100 m resolution. Spatial statistics (Kriging, variogram analysis) and machine learning approaches (gradient boosting, random forests) are used for feature selection and fault detection. The qualitative component involves stakeholder interviews with urban planners, public health officials, and community advocates (n=30) to examine perceived usefulness, trust, and governance implications. Thematic analysis is applied to interview transcripts to identify determinants of adoption and barriers to action. The study also conducts a cost-benefit analysis to quantify health and productivity gains associated with improved information access. Data integrity is ensured through rigorous validity and reliability checks, including test-retest reliability for sensor readings and inter-rater reliability for interview coding. Ethical considerations follow institutional guidelines, with informed consent obtained and data anonymized. Key expected findings include (i) enhanced accuracy and coverage of urban air quality measurements through calibrated sensor fusion, (ii) high-resolution pollution maps enabling neighborhood-level exposure assessment and targeted mitigation, (iii) evidence that near-real-time data improves timely public advisories and compliance with exposure-reduction recommendations, and (iv) identified governance pathways and cost-effective deployment strategies that balance accuracy, equity, and scalability. The study anticipates that the edge-computing architecture will reduce data latency by 60% and lower data transmission costs by 35% relative to centralized models, while the ensemble assimilation will reduce root-mean-square error by 20–35% for key pollutants. Contributions to knowledge include a robust, replicable ICT-driven framework for urban air quality management that integrates low-cost sensor networks with advanced data assimilation, a validated calibration strategy for heterogeneous sensors under urban dynamics, and empirical evidence on governance implications and cost-effectiveness of smart ambient monitoring systems. The main conclusion is that integrated smart sensor networks, when combined with data fusion and stakeholder engagement, substantially improve city-scale awareness and responsiveness to air quality threats. Recommendations emphasize standardized calibration protocols, open data policies, scalable deployment templates, and ongoing collaboration among technologists, policymakers, and communities to sustain improvements in urban air quality and public health outcomes.