Deep Learning Approach for Assessing Air Quality During COVID-19 Lockdown in Quito

During Quito's COVID-19 lockdown, concentrations of key pollutants such as CO, NO2, PM2.5, and SO2 dropped significantly by up to 64%, while ozone levels rose by over 26%. This stark shift in air quality was quantified using advanced Deep Learning (DL) models, specifically Long-Short Term Memory (LSTM) and Bidirectional Recurrent Neural Networks (BiRNN), which outperformed traditional Machine Learning methods like Gradient Boosting Machines (GBM). The study highlights the potential of DL-based Weather Normalized Models (WNMs) for assessing the impact of events on urban pollution. Critical predictors included temporal factors (time of day, Julian Day) and meteorological variables (relative humidity, wind speed, solar radiation), offering insights into the complex interplay between human activity and environmental conditions.

Chau, P. N., Zalakeviciute, R., Thomas, I., & Rybarczyk, Y. (2022). Deep Learning Approach for Assessing Air Quality During COVID-19 Lockdown in Quito. Frontiers in Big Data, 5. https://doi.org/10.3389/fdata.2022.842455