From sensor noise to spatial insight: A pipeline for detecting and visualizing air pollution events in urban environments

Abstract

Air pollution, and in particular fine particulate matter (PM_2.5), has significant health impacts, especially in urban environments. In this study, a comprehensive low-cost sensor analysis process of hourly PM_2.5 data is presented for the detection, characterization and visualization of pollution events in Athens, Greece. The methodology integrates data quality control, air pollution event detection and spatiotemporal analysis of air pollution dynamics using inverse distance weighting (IDW) spatial interpolation, centroid tracking and bootstrap-based uncertainty analysis. During the data quality control stage, 4.88% of the PM_2.5 observations were flagged as erroneous or anomalous. Between 2018 and 2022, a total of 424 air pollution events were identified, with 84% occurring during winter months. Most of the air pollution events occurred during winter evenings and night hours, possibly due to residential heating and temperature inversions. The methodology successfully captured known extreme events, including the August 2021 wildfire smoke episode affecting Athens. The pipeline demonstrates that dense low-cost sensor networks, when combined with systematic quality control, event detection, and uncertainty-aware spatial analysis, can provide actionable high-resolution information for urban air-pollution monitoring, urban environmental management, and early-warning applications.