Normet package- NORmalising METeorological Impacts on Air Quality  

Speaker:  Congbo Song 

Chair(S): WUHU FENG

wednesday november 20 2024, 13:00

ABSTRACT

Assessing the effectiveness of air quality intervention policies is a crucial task in air quality management. However, this assessment is challenging because air quality is influenced by both emissions and meteorological factors. Machine learning-based weather normalization has been developed to decouple emission-related changes in air quality from observed data. In this study, we enhance existing weather normalization models and introduce a new model, ‘normet’, by incorporating Automated Machine Learning (AutoML) on both R and Python platforms. AutoML helps to select the optimal model and hyperparameters, improving the accuracy of meteorologically normalized concentrations. Additionally, we implement time series decomposition and an ML-based synthetic control method within ‘normet’ to better interpret air quality time series data and support air quality intervention studies. This open-source package on both R and Python platforms advances analytical capabilities in air quality management.