期刊
ATMOSPHERIC RESEARCH
卷 278, 期 -, 页码 -出版社
ELSEVIER SCIENCE INC
DOI: 10.1016/j.atmosres.2022.106339
关键词
NO2; Deep learning; Ensemble model; Spatiotemporal distribution; Random forest
资金
- National Natural Science Founda- tion of China [22076129, 42007197]
- Sichuan Key R D Project [2020YFS0055]
- Chengdu Major Technology Application and Demonstration Project [2020-YF09-00031-SN]
This study explores the use of deep learning models to predict spatiotemporal distributions of surface NO2 in China. The ensemble FC-AE-RES-CNN model outperforms the random forest model, suggesting that deep learning models have the potential to replace traditional machine learning models in environmental mapping applications.
While deep learning is an emerging trend in modeling spatiotemporal dynamics of air pollution, designing a high-performance architecture with various neural network components is difficult. This study aims to identify a suitable deep learning architecture for deriving spatiotemporal distributions of surface NO2 across China. Based on satellite retrievals and geographic variables, we compared the performance of a series of deep learning models and a benchmark model (i.e., random forest) in predicting daily NO2 during 2016-2017 on a 0.1 degrees grid. The deep learning models tested were hybrids of full connection layers (FC), autoencoders (AE), residual networks (RES), and convolution layers (CNN). Their ensemble forms were also comparatively evaluated. The integrated results of the sample-, cell-, and date-based holdout tests show that the ensemble FC-AE-RES-CNN model (R2 = 0.71; RMSE = 9.9 mu g/m3) outperformed the random forest (R2 = 0.67; RMSE = 10.8 mu g/m3), but none of the singleform networks did (R2 ranged from 0.56 to 0.61, and RMSE ranged from 11.6 to 12.2 mu g/m3). The superior performance of the ensemble FC-AE-RES-CNN model could be attributed to its lower bias and lower variance, which benefited from the sophisticated model structure and the ensemble strategy. Consequently, the ensemble FC-AE-RES-CNN model predictions demonstrated a spatiotemporal distribution of surface NO2 with higher fidelity. In light of the comprehensive comparisons, we expect that ensemble deep learning models will be substituting for traditional machine learning models in environmental spatiotemporal mapping applications.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据