FILTER (VALAIR)
About the tool
FILTER (the Framework for Improving Low-cost Technology Effectiveness and Reliability) is a quality control and correction system developed within CitiObs to enhance the reliability of citizen-generated air quality data. With the rise of low-cost sensors (LCS) from platforms like sensor.community and PurpleAir, communities now have access to hyperlocal air quality insights. However, the inconsistent quality of LCS data often limits its integration into policy or research.
FILTER addresses this challenge with a multi-step data pipeline that evaluates and corrects outdoor PM2.5 measurements from LCS networks. The system performs range checks, detects outliers and constant values, tests for spatial correlation, and uses spatial similarity patterns to flag and correct data without needing co-location data or sensor-type information. The result is a harmonized dataset with clear quality flags and uncertainty estimates, empowering users to select data based on confidence levels.
FILTER supports CitiObs’s mission to improve the usability of citizen-collected data across governance levels. It helps bridge the gap between grassroots monitoring and institutional standards, directly contributing to the goals of the European Ambient Air Quality Directive (2024). FILTER has already been applied to over 500 million hourly PM2.5 measurements from 2018–2023, improving data accuracy by more than 50% (in median RMSE).
Future developments include expanding support for mobile LCS devices, developing an R package, and integrating FILTER into the CitiObs Knowledge Platform. Collaborations with Frontrunner and Alliance Citizen Observatories will ensure the tool is tailored for local use cases across Europe.
Explore the tool:
GitHub (Code & Resources): https://github.com/CitiObs/ValAir
Research Article: https://www.sciencedirect.com/science/article/pii/S030147972501076X
