Development and evaluation of a low-cost real-time air quality monitoring network for urban health assessment

Thesis Abstract

Rapid urbanisation has heightened exposure to ambient air pollutants, yet traditional monitoring networks are sparse and costly, limiting timely assessment of population health in cities. This study addresses the gap by developing and evaluating a low-cost real-time air quality monitoring network designed to enhance urban health surveillance and intervention capabilities. The aim is to design an integrated sensor network, deploy it in a representative urban district, and evaluate its performance against reference stations in terms of data quality, spatial representativeness, and health relevance. Specific objectives are to (i) design a modular, scalable low-cost sensor architecture capable of measuring PM2.5, PM10, NO2, O3 and CO with synchronized data logging; (ii) implement edge-computing protocols and cloud-based data fusion to generate real-time exposure maps; (iii) validate sensor outputs against regulatory reference monitors using regression and Bland-Altman analyses; (iv) assess the correlation between real-time exposure metrics and short-term health indicators such as emergency department visits and self-reported respiratory symptoms; and (v) evaluate the system’s usability for city authorities and identify policy-relevant recommendations for urban health management. The study adopts a mixed-methods approach grounded in the Exposure–Response framework and supported by the Ecological Systems Theory. A multi-site deployment will be conducted in the central district of a metropolitan area, comprising 20 low-cost sensor nodes co-located with a reference station and 40 mobile nodes mounted on public transport vehicles to capture spatial variability. Data collection spans 12 months, capturing seasonal fluctuations. Quantitative instruments include commercially available electrochemical and optical sensors calibrated pre-deployment and subjected to routine maintenance checks. Data streams (PM2.5, PM10, NO2, O3, CO) are streamed to a central database at 1-minute intervals, using time synchronization protocols and calibration coefficients derived from laboratory and field co-location exercises. Qualitative insights are obtained from interviews with 12 city health officials and focus groups with 6 community health workers to assess usability, perceived reliability, and decision-making utility. Analytical techniques encompass (i) regression analyses and concordance checks (Passing-Bovis and Bland-Altman) to evaluate sensor accuracy against reference monitors; (ii) time-series analyses, including ARIMA and transfer function-noise models, to quantify exposure trends and temporal associations with health outcomes; (iii) spatial interpolation (kriging) to generate high-resolution exposure maps; (iv) generalized linear models to examine associations between real-time pollutant metrics and health indicators, controlling for meteorological covariates; and (v) thematic analysis of stakeholder interviews to elucidate implementation barriers and enablers. A data governance framework will be established to ensure data quality, privacy, and ethical use. Sensitivity analyses will test the robustness of findings to sensor drift, missing data, and urban canyon effects. Expected findings include high concordance between low-cost sensor data and reference measurements for PM components under calibrated conditions, with systematic biases quantified and corrected via calibration functions. Real-time exposure maps are anticipated to reveal pronounced intra-urban variability correlating with traffic density, industrial activity, and meteorological conditions. Statistical models are expected to show positive associations between short-term pollutant exposures and acute health indicators, with lag structures identifying the most critical exposure windows. The study will produce a validated, scalable framework for deploying low-cost networks in similar urban contexts, including a reproducible data pipeline, calibration protocols, and decision-support dashboards. The study contributes to knowledge by demonstrating the feasibility and reliability of a cost-effective ambient monitoring approach capable of informing urban health management, bridging the gap between sensor technology and public health practice. It advances methodological discourse in exposure assessment by integrating sensor networks, real-time analytics, and health surveillance within an operational governance framework. Practical implications include guidance for city authorities on resource allocation, emergency response, and policy design aimed at reducing population-level health risks from air pollution. The main conclusions anticipate that a carefully calibrated low-cost network can provide actionable exposure insights, while recommendations emphasize standardized calibration, continuous quality assurance, stakeholder engagement, and scalable deployment strategies for broader urban applications.