Modelling calibration uncertainty in networks of environmental sensors

Smith, Michael Thomas; Ross, Magnus; Ssematimba, Joel; Álvarez, Mauricio A; Bainomugisha, Engineer; Wilkinson, Richard

Abstract

Networks of low-cost environmental sensors are becoming ubiquitous, but often suffer from poor accuracies and drift. Regular colocation with reference sensors allows recalibration but is complicated and expensive. Alternatively, the calibration can be transferred using low-cost, mobile sensors. However, inferring the calibration (with uncertainty) becomes difficult. We propose a variational approach to model the calibration across the network. We demonstrate the approach on synthetic and real air pollution data and find it can perform better than the state-of-the-art (multi-hop calibration). In Summary: The method achieves uncertainty-quantified calibration, which has been one of the barriers to low-cost sensor deployment.

Citation

Smith, M. T., Ross, M., Ssematimba, J., Álvarez, M. A., Bainomugisha, E., & Wilkinson, R. (2023). Modelling calibration uncertainty in networks of environmental sensors. Journal of the Royal Statistical Society: Series C, 72(5), 1187-1209. https://doi.org/10.1093/jrsssc/qlad075

• Journal Article Type: Article
• Acceptance Date: Jul 17, 2023
• Online Publication Date: Aug 24, 2023
• Publication Date: Aug 24, 2023
• Deposit Date: Oct 13, 2023
• Publicly Available Date: Aug 25, 2024
• Journal: Journal of the Royal Statistical Society Series C: Applied Statistics
• Print ISSN: 0035-9254
• Electronic ISSN: 1467-9876
• Publisher: Wiley
• Peer Reviewed: Peer Reviewed
• Volume: 72
• Issue: 5
• Article Number: qlad075
• Pages: 1187-1209
• DOI: https://doi.org/10.1093/jrsssc/qlad075
• Keywords: air pollution, Bayesian modelling, calibration, Gaussian processes, low-cost sensors, variational inference
• Public URL: https://nottingham-repository.worktribe.com/output/25057392
• Publisher URL: https://academic.oup.com/jrsssc/advance-article-abstract/doi/10.1093/jrsssc/qlad075/7250331?redirectedFrom=fulltext