Using meteorological normalisation to detect interventions in air quality time series

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Air quality interventions are challenging to detect in time series data due to atmospheric complexity.
• Meteorological normalization is crucial for robust intervention analysis and trend assessment.
• Machine learning offers advanced methods for analyzing complex environmental data.

Purpose of the Study:

• To apply a machine learning-based meteorological normalization technique to air quality data.
• To assess the technique's effectiveness in identifying the impact of known interventions on pollutant concentrations.
• To demonstrate the technique's utility in complex urban and port environments.

Main Methods:

• A random forest machine learning algorithm was employed for meteorological normalization.
• The technique was applied to routinely collected air quality monitoring data (SO2, NOx, NO2).
• Data from two distinct locations (Dover port city, London urban) with known interventions were analyzed.

Main Results:

• The technique clearly identified the impact of marine fuel sulfur limits on sulfur dioxide (SO2) in Dover.
• Normalized time series for nitrogen oxides (NOx and NO2) in London revealed changes linked to primary emissions.
• Significant features in pollutant trends were illuminated by normalization, which were not apparent in raw data.

Conclusions:

• Meteorological normalization using random forest is a flexible and effective tool for air quality intervention analysis.
• The technique successfully highlights intervention impacts obscured by meteorological variability.
• It is suitable for diverse applications in air quality management and research.