A framework for interpreting air pollution sensor data: Bridging gaps in monitoring, analysis and community action.

Abstract

Compared to high-grade reference monitors, low-cost sensors are small in size with minimal power consumption. This has partly led to their proliferation, especially within low- and middle-income countries-amidst accuracy and precision concerns. The dynamic nature of air pollution, coming from multiple sources and fluctuating over time and space, further complicates accuracy issues. This then makes air pollution of a given area unique based on context. To address this challenge, this paper lays out a stepwise approach that makes air quality measurement interpretation more accessible, thereby increasing community action in air quality awareness. Using an urban network of 11 low-cost PM_2.5 sensors (LCS) and a Beta Attenuation Monitor (BAM) located across diverse land use types in Kampala, Uganda. Hourly and daily average PM_2.5 concentrations from both LCS and a BAM were compared against each other over a 13-month period from January 2022 to January 2023. Monitoring sites were classified into emission zones based on surrounding emission sources, guided by the AIRQO land-use descriptors for sensor locations. Site classification of either low or high-emission reflected local emission intensity rather than residential status alone. Low-emission sites were primarily low-density, upscale residential areas characterized by greater green space, paved roads, and limited nearby traffic or commercial activity. High-emission locations had substantial nearby emission sources, including commercial centers, business hubs, proximity to highways or major roads, and high-density residential areas with mixed land use. Sensors followed a similar pattern with the BAM for both the hourly and 24-hr measurements at lower concentrations, but dispersed at the higher values. A stronger linear relationship from the 24 hr data across all sites exists. Slopes are lower in urban areas (~0.60-1.05) compared to peri-urban areas (~1.10-1.23). Findings suggest the need for a transparent framework that makes air quality data more accessible for community action.Implications: This work shows the promise of low-cost sensors in air quality measurements, especially within resource-constrained communities, where a stronger linear relationship from the 24 hr data across all sites exists when compared to the BAM. It also demonstrates the importance of considering the different land use in air quality measurement interpretation, which strengthens the need to make air quality readings more accessible and digestible to the general public. Clearly conveying air quality information can empower communities and decision-makers on pollution control innovation.