A novel machine learning method for evaluating the impact of emission sources on ozone formation

Ambient ozone air pollution is one of the most important environmental challenges in China today, and it is particularly significant to identify pollution sources and formulate control strategies. In present study, we proposed a novel method of positive matrix factorization-SHapley Additive explanation (PMF-SHAP) for evaluating the impact of emission sources on ozone formation, which can quantify the main emission sources of ozone pollution. In this method, we first used the PMF model to identify the source of volatile organic compounds (VOCs), and then quantified various emission sources using a combination of machine learning (ML) models and the SHAP algorithm. The R-2 of the optimal ML model in this method was as high as 0.96, indicating that the prediction performance was excellent. Furthermore, we explored the impact of different emission sources on ozone formation, and found that ozone formation in Shenzhen was more affected by VOCs, of which vehicle emission sources may have the greatest impact. Our results suggest that the appropriate combination of traditional models with ML models can well address environmental pollution problems. Moreover, the conclusions obtained based on the PMF-SHAP method were different from the traditional ozone formation potential (OFP) results, providing valuable clues for related mechanism studies.

Automated summary

In this study, a novel method called positive matrix factorization-SHapley Additive explanation (PMF-SHAP) was proposed to evaluate the impact of emission sources on ozone formation and quantify the main sources of ozone pollution. The results showed that ozone formation in Shenzhen was more affected by volatile organic compounds, with vehicle emission sources possibly having the greatest impact.