Spatial clustering in air quality data accuracy: assessment of low-cost sensors in Lisbon using a novel correlation-distance index

Abstract

Effective air quality monitoring is essential for ensuring urban environmental comfort and protecting public health, especially within smart-city frameworks. In Lisbon, the performance of low-cost NO₂ and PM 10 sensors was assessed using a novel Correlation-Distance Index (CDI) coupled with Principal Component Analysis (PCA) weighting, alongside a demographic-weighted spatial association technique, benchmarked against reference monitoring stations. The analysis began by quantifying the inverse relationship between sensor–sensor correlation and Euclidean distance. This metric was then refined through the integration of PCA-derived weights and local population density, resulting in a composite CDI-PCA score assigned to each sensor. Spatial clustering of these scores identified zones of underperforming sensors, particularly in densely populated downtown and western neighborhoods. When these clusters were compared with census-block population data, there was a statistically significant association between CDI patterns and population density (χ²=406.5, p

Bibliographic Information