Outlier Detection in Urban Air Quality Sensor Networks

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Sensor Technology

Background:

• Low-cost urban air quality sensor networks are vital for assessing air pollutant spatio-temporal variability.
• These sensors often produce erroneous data, including outliers, which challenge traditional detection methods due to urban environmental dynamics.

Purpose of the Study:

• To develop and validate a novel outlier detection method for nitrogen dioxide (NO2) concentrations from low-cost urban sensors.
• To address the limitations of existing methods in handling the complex spatio-temporal variations of urban air quality data.

Main Methods:

• A spatio-temporal classification approach was developed, dividing a year of hourly NO2 data into 16 classes.
• Classes were defined by station type (urban background vs. traffic), day type (weekday vs. weekend), and time of day.
• Outliers were identified within each class using statistical analysis based on the truncated normal distribution of NO2 observations.

Main Results:

• The novel method identified 0.1-0.5% of outliers in the Eindhoven low-cost air quality sensor network data.
• Detected outliers may represent measurement errors or significant air pollution events, requiring expert evaluation for definitive classification.

Conclusions:

• The proposed spatio-temporal classification method effectively detects outliers in urban NO2 data from low-cost sensors.
• This method preserves the inherent spatio-temporal variability crucial for accurate air quality analysis in urban settings.