Over the past decade, low-cost sensor (LCS) networks for air quality monitoring have surged across Europe. LCSs have changed air quality monitoring by allowing citizen-led, hyperlocal data collection. Driven largely by citizen science initiatives, platforms like sensor.community and PurpleAir now provide vast datasets at fine spatial and temporal resolutions. However, despite their growing popularity, LCS data often suffer from inconsistency, poor quality, and lack of standardization. This makes it difficult to integrate them into official environmental assessments or use them confidently in policymaking and scientific research.

In response, we introduce FILTER — the Framework for Improving Low-cost Technology Effectiveness and Reliability — a multi-step data pipeline designed to evaluate, flag, and correct crowdsourced outdoor PM₂₅ (particulate matter with a diameter ≤2.5 μm) data from various LCS networks. Developed in the context of CitiObs (a Horizon Europe project that aims to enhance citizen-generated environmental observations), FILTER is grounded in a simple philosophy: make the best use of the data we have, without compromising on scientific integrity.

FILTER provides both quality flagging and value correction. It systematically screens LCS data through five stages: range check, constant value detection, outlier detection, spatial correlation testing, and spatial similarity assessment. It also corrects measurements where appropriate, using only spatial patterns — no direct co-location or information about the LCS type is required. The final output is a harmonized, corrected dataset with clear quality labels and associated uncertainty bounds, allowing users to prioritize accuracy based on their application.

FILTER directly supports the CitiObs mission to enhance Citizen Observatories and integrate them into multi-level governance. Through improving the trustworthiness of LCS data, FILTER makes it easier for citizens, researchers, and policymakers to rely on community-driven air quality data for real decisions, supporting the new European Ambient Air Quality Directive (2024)[1]. It bridges the gap between grassroots monitoring efforts and institutional data requirements, a key CitiObs goal.

For validation, we applied FILTER to more than 400 LCSs located near official monitoring stations from diverse environments across Europe. The result is a corrected, quality-labeled dataset that improves LCS readings, achieving reductions of over 50% in median Root Median Square Error (RMSE).

We applied FILTER to over 500 million hourly measurements from two of Europe’s largest community-based networks, i.e.,  sensor.community and PurpleAir,  covering 2018 to 2023. The output dataset and the complete MATLAB code for applying FILTER are openly accessible via Figshare (https://doi.org/10.6084/m9.figshare.27195720.v1). Whether you are working on urban air quality forecasting, health exposure mapping, or environmental justice, this is a resource built to empower the community.