An end-to-end data analysis framework for real-time detection and source identification of pollution events via e-nose networks

Ambient air pollution contributes to an estimated 4.2 million premature deaths worldwide each year, underscoring the imperative for advanced air quality management tools. In heavily industrialized regions, diffuse fugitive leaks and concentrated stack emissions threaten public health and ecosystems, making real-time monitoring and accurate source apportionment indispensable. To address these challenges, we introduce an end-to-end data analysis framework that employs a distributed network of low-cost electronic noses (e-noses) to deliver spatially and temporally resolved emission monitoring. The real-time, high-resolution outputs of e-noses are deconvoluted into discrete emission events using a chemometric pipeline including principal component analysis (PCA), hierarchical cluster analysis (HCA), and multivariate curve resolution-alternating least squares (MCR-ALS). Each event is then characterized within a 5W attribution schema, which identifies what anomaly was detected, when, and where it occurred, why it arose, and who is responsible for mitigation, thereby creating a searchable database of pollution incidents. By combining real-time, low-cost sensing with robust multivariate analysis and stakeholder-centered reporting, this framework enables transparent emission accountability. In turn, it supports rapid mitigation responses and strengthens regulatory compliance in complex industrial regions.