Evaluation of Crowd-Sourced PM2.5 Measurements from Low-Cost Sensors for Air Quality Mapping in Stuttgart City

Abstract

AbstractExposure to particulate matter (PM) pollution poses a major risk to the environment and human health. Monitoring PM pollution is thus crucial to understand particle distribution and mitigation. There has been rapid development of low-cost PM sensors and advancement in the field of Internet of Things (IoT) that has led to the deployment of the sensors by technology-aware people in cities. In this study, we evaluate the stability and accuracy of PM measurements from low-cost sensors crowd-sourced from a citizen science project in Stuttgart. Long-term measurements from the sensors show a strong correlation with measurements from reference stations with most of the selected sensors achieving Pearson correlation coefficients of r > 0.7. We investigate the stability of the sensors for reproducibility of measurements using five sensors installed at different height levels and horizontal distances. They exhibit minor variations with low correlation of variation (CV) values of between 10 and 14%. A CV of ≤10% is recommended for low-cost sensors. In a dense network, the sensors enable extraction pollution patterns and trends. We analyse PM measurements from 2 years using space-time pattern analysis and generate two clusters of sensors that have similar trends. The clustering shows the relationship between traffic and pollution with most sensors near major roads being in the same cluster.