Assessing the COVID-19 Impact on Air Quality: A Machine Learning Approach

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Global research confirms COVID-19 lockdowns reduced air pollution.
• Quantifying this reduction accurately remains a challenge.
• Machine learning offers a promising approach for precise impact assessment.

Purpose of the Study:

• To assess the impact of the COVID-19 outbreak and subsequent lockdowns on air quality in Quito, Ecuador.
• To quantify the reduction in specific air pollutants (NO2, SO2, CO, PM2.5) during lockdown periods.
• To evaluate the effectiveness of machine learning models in predicting and measuring these changes.

Main Methods:

• Development and application of Gradient Boosting Machine (GBM) learning models.
• Cross-validation using four years of pre-lockdown air quality data for model accuracy assessment.
• Quantification of pollutant concentration changes during full and partial lockdown phases.

Main Results:

• Machine learning models demonstrated high accuracy in estimating pre-lockdown pollution levels.
• Significant air pollution reductions were observed: Nitrogen Dioxide (NO2) by -53%, Sulfur Dioxide (SO2) by -45%, Carbon Monoxide (CO) by -30%, and Particulate Matter (PM2.5) by -15%.
• Pollution levels in traffic-heavy areas decreased most significantly and returned to pre-pandemic levels after partial relaxation of measures.

Conclusions:

• Machine learning, specifically GBM, provides a reliable method for quantifying air quality changes during events like pandemics.
• Lockdowns led to substantial, measurable improvements in air quality in Quito.
• The study highlights the direct link between human activity, particularly traffic, and urban air pollution levels.