A scalable framework for harmonizing, standardization, and correcting crowd-sourced low-cost sensor PM_2.5 data across...

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Science

Background:

• Citizen-operated low-cost air quality sensors (LCSs) offer expanded monitoring but face challenges in data quality and standardization.
• Integrating crowd-sourced data into official assessments requires robust methods for quality control and validation.

Purpose of the Study:

• To introduce FILTER, a scalable framework for unifying, correcting, and enhancing the reliability of crowd-sourced PM2.5 data from various LCS networks.
• To establish data quality assessment steps and benchmarks using official air quality data.

Main Methods:

• Developed a five-step data quality assessment: range check, constant value detection, outlier detection, spatial correlation, and spatial similarity.
• Modeled PM2.5 spatial correlation and similarity against geographic distance using official data as benchmarks.
• Determined seasonal optimal distance thresholds for in-situ data correction.

Main Results:

• Identified seasonal variations in PM2.5 spatial correlation and similarity across Europe.
• Established seasonal optimal distance thresholds for PM2.5 data correction, ranging from ~11.5 km (DJF) to ~20 km (JJA).
• Validated the FILTER framework on European-scale data from sensor.community and PurpleAir, processing over 521 million hourly timestamps.

Conclusions:

• The FILTER framework effectively unifies, corrects, and enhances the reliability of crowd-sourced PM2.5 data.
• Results demonstrate the potential for integrating validated crowd-sourced LCS data into regulatory applications and scientific research.
• Seasonal variability in spatial representativeness influences optimal data correction strategies.