Abstract
Tropospheric ozone (O3) is a greenhouse gas and an influential ground-level air pollutant, thus determining the importance of various factors related to O3 formation is essential. However, O3 concentrations simulated by the available climate models exhibit significant variances, indicating the insufficiency of such models in modeling the O3 formation process accurately. This study used machine learning to identify and understand the impact of various factors on O3 formation and thereby predict O3 concentrations under different climate change and emission reduction scenarios. We employed six supervised learning models to estimate O3 concentrations by using 14 meteorological and chemical variables. We found that deep neural network (DNN) and long short-term memory (LSTM) models could accurately predict O3 concentrations. The variable importance analyses emphasized the significant positive contribution of solar radiation to O3; meanwhile both nitrogen oxides and volatile organic compounds negatively contributed to O3 concentrations. Predicted O3 concentrations based on our DNN and LSTM models in different climate change scenarios showed that increase in water vapor content could offset the effects of raised temperature, and controlling anthropogenic gases, especially carbon monoxide, might have the most bearing on controlling O3 in the future.
Original language | English (US) |
---|---|
Article number | 119148 |
Journal | Atmospheric Environment |
Volume | 282 |
DOIs | |
State | Published - Aug 1 2022 |
Funding
The authors express deep gratitude for the support from the Ministry of Science and Technology of Taiwan (MOST 110-2111-M-001-009, 109-2111-M-001-010). The machine learning computations in this study were carried out using Academia Sinica (AS) GPU computing service. The authors thank AS for the support. The authors also acknowledge anonymous reviewers for the comments and suggestions. The authors express deep gratitude for the support from the Ministry of Science and Technology of Taiwan (MOST 110-2111-M-001-009, 109-2111-M-001-010). The machine learning computations in this study were carried out using Academia Sinica (AS) GPU computing service. The authors thank AS for the support. The authors also acknowledge anonymous reviewers for the comments and suggestions.
Keywords
- Climate change
- Deep neural network
- Long short-term memory
- Ozone formation
- Supervised learning
ASJC Scopus subject areas
- General Environmental Science
- Atmospheric Science