Democratizing Access to Accurate Air Quality Measurements with Open-Source Tools

Resource-constrained regions often face the highest levels of air pollution yet lack the infrastructure to monitor and mitigate its impacts effectively. Reliable air quality (AQ) data is critical for understanding and addressing the adverse effects of air pollution. While low-cost monitors can reduce equipment and maintenance costs, ensuring the accuracy of the collected data presents challenges, often requiring more expertise and effort than conventional reference equipment.

This presentation introduces open-source tools designed to address these challenges, democratizing access to accurate AQ monitoring. Focusing on outdoor PM_2.5 sensors, our work highlights three key innovations that enhance the reliability and accessibility of low-cost air quality monitoring systems.

Calibration Strategies Without a Reference Instrument
The accuracy of low-cost sensors can be significantly enhanced by calibrating them against a local reference instrument. However, many regions lack access to such equipment. To address this gap, AirGradient has developed multiple calibration methods, including:

• In-house calibration using a Federal Equivalent Method during monitor assembly.
• Calibration via background extraction from sensor networks (see Point 2).
• Application of a generalized correction formula, which is the focus of this discussion:
  The U.S. Environmental Protection Agency has devised a comprehensive correction formula for Plantower PM_2.5 sensors. This algorithm accounts for relative humidity effects and the sensors’ non-linear response at high concentrations. When applied to AirGradient sensors deployed in nine global locations, the formula delivered significant improvements in accuracy. Average R² values increased from 0.899 (raw) to 0.923 (corrected), while the normalized root mean square error (nRMSE) was reduced from 86% to 34%. 

Sensor Networks for Enhanced Calibration and Maintenance
Beyond granular pollution mapping, AQ sensor networks offer advanced capabilities for calibration and maintenance. A network of 16 AirGradient monitors deployed in Pai, Thailand, was used to extract regional background PM_2.5 concentrations. With no local reference instruments available, calibration data from Mae Hong Son (a city 50 km away) was leveraged to adjust sensor readings, demonstrating the potential for regional calibration strategies.

Automatic Detection of Sensor Failures
A robust AQ dataset requires the detection and removal of anomalous data. Using a combination of AirGradient-owned co-location datasets and user-contributed public data, we identified various artefacts, including outliers, missing data, extended periods of high-concentration readings, and physically implausible values. An automated data cleaning algorithm was developed to identify and flag these anomalies effectively. While the system reliably enhances data quality, challenges remain in distinguishing short-term real emission events from artefacts. The integration of duplicate PM sensors into individual monitors could help address this issue, providing further reliability.

By improving the accuracy, reliability, and usability of low-cost AQ sensors, these open-source tools empower a wide range of users —such as schools, environmental organizations, and local governments— to generate actionable AQ data. This democratization of air quality monitoring fosters local engagement and equips resource-constrained regions with the tools needed to combat air pollution effectively.