GAT-EGRU: A Deep Learning Prediction Model for PM2.5 Coupled with Empirical Modal Decomposition Algorithm

Abstract

With the rapid development of the economy and industry and the improvement of pollution monitoring, how to accurately predict PM2.5 has become an issue of concern to the government and society. In the field of PM2.5 pollution forecasting, a series of results have emerged so far. However, in the existing research field of PM2.5 prediction, most studies tend to predict short-term temporal series. Existing studies tend to ignore the temporal and spatial characteristics of PM2.5 transport, which leads to its poor performance in long-term prediction. In this paper, by optimizing previous PM2.5 deep learning prediction models, we propose a model GAT-EGRU. First, we add a spatial modular Graph Attention Network (GAT) and couple an Empirical Modal Decomposition algorithm (EMD), considering the temporal and spatial properties of PM2.5. Then, we use Gated Recurrent Unit (GRU) to filter spatio-temporal features for iterative rolling PM2.5 prediction. The experimental results show that the GAT-EGRU model has more advantages in predicting PM2.5 concentrations, especially for long time steps. This proves that the GAT-EGRU model outperforms other models for PM2.5 forecasting. After that, we verify the effectiveness of each module by distillation experiments. The experimental results show that each model module has an essential role in the final PM2.5 prediction results. The new model improves the ability to predict PM2.5 after a long time accurately and can be used as a practical tool for predicting PM2.5 concentrations.

Keywords

Air pollution forecasting / deep learning / spatial-temporal prediction / empirical modal decomposition

References

[1] Allen D M. Mean square error of prediction as a criterion for selecting variables. Technometrics, 1971, 13(3): 469-475.

[2] Bahdanau D, Cho K, Bengio Y (2014). Neural machine translation by jointly learning to align and translate. arXiv Preprint arXiv:14090473.

[3] Bai Y, Li Y, Wang X, Xie J, Li C. Air pollutants concentrations forecasting using back propagation neural network based on wavelet decomposition with meteorological conditions. Atmospheric Pollution Research, 2016, 7(3): 557-566.

[4] Chai T, Draxler R R. Root mean square error (RMSE) or mean absolute error (MAE)? Arguments against avoiding RMSE in the literature. Geoscientific Model Development, 2014, 7(3): 1247-1250.

[5] Chung J, Gulcehre C, Cho K, Bengio Y (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv Preprint arXiv:1412.3555.

[6] Fey M, Lenssen JE (2019). Fast graph representation learning with pytorch geometric. arXiv Preprint arXiv:190302428.

[7] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Computation, 1997, 9(8): 1735-1780.

[8] Hu J, Wang Y, Ying Q, Zhang H. Spatial and temporal variability of PM2.5 and PM10 over the North China Plain and the Yangtze River Delta, China. Atmospheric Environment, 2014, 95: 598-609.

[9] Huang N E, Shen S S. Hilbert-Huang Transform and Its Applications(2ed), 2014, Singapore: World Scientific Publishing.

[10] Jeong J I, Park R J, Woo J, Han Y, Yi S. Source contributions to carbonaceous aerosol concentrations in Korea. Atmospheric Environment, 2011, 45(5): 1116-1125.

[11] Kim Y, Fu J S, Miller T L. Improving ozone modeling in complex terrain at a fine grid resolution: Part I — Examination of analysis nudging and all PBL schemes associated with LSMs in meteorological model. Atmospheric Environment, 2010, 44(4): 523-532.

[12] Kipf T N, Welling M (2016). Semi-supervised classification with graph convolutional networks. arXiv Preprint arXiv:1609.02907.

[13] Li T, Shen H, Yuan Q, Zhang X, Zhang L. Estimating ground-level PM2. 5 by fusing satellite and station observations: A geo-intelligent deep learning approach. Geophysical Research Letters, 2017, 44(23): 11. 911–985, 993

[14] Li T, Shen H, Zeng C, Yuan Q, Zhang L. Point-surface fusion of station measurements and satellite observations for mapping PM2. 5 distribution in China: Methods and assessment. Atmospheric Environment, 2017, 152: 477-489.

[15] Li X, Chen X, Yuan X, Zeng G, León T, Liang J, et al. Characteristics of particulate pollution (PM2.5 and PM10) and their spacescale-dependent relationships with meteorological elements in China. Sustainability, 2017, 9(12): 2330.

[16] Li X, Peng L, Yao X, Cui S, Hu Y, You C, et al. Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation. Environmental Pollution, 2017, 231: 997-1004.

[17] Liu S, Hua S, Wang K, Qiu P, Liu H, Wu B, et al. Spatial-temporal variation characteristics of air pollution in Henan of China: Localized emission inventory, WRF/Chem simulations and potential source contribution analysis. Science of the Total Environment, 2018, 624: 396-406.

[18] Luan T, Guo X, Guo L, Zhang T. Quantifying the relationship between PM 2.5 concentration, visibility and planetary boundary layer height for long-lasting haze and fog-haze mixed events in Beijing. Atmospheric Chemistry and Physics, 2018, 18(1): 203-225.

[19] Ni X Y, Huang H, Du W P. Relevance analysis and short-term prediction of PM2.5 concentrations in Beijing based on multi-source data. Atmospheric Environment, 2017, 150: 146-161.

[20] Pan Z, Liang Y, Zhang J, Yi X, Yu Y, Zheng Y (2018). Hyperst-net: Hypernetworks for spatio-temporal forecasting. arXiv Preprint arXiv:1809.10889.

[21] Paszke A, Gross S, Massa F, Lerer A, Bradbury J, Chanan G, et al. (2019). Pytorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems. DOI: https://doi.org/10.48550/arXiv.1912.01703.

[22] Qi Y, Li Q, Karimian H, Liu D. A hybrid model for spatiotemporal forecasting of PM2.5 based on graph convolutional neural network and long short-term memory. Science of the Total Environment, 2019, 664: 1-10.

[23] Rilling G, Flandrin P, Goncalves P. On empirical mode decomposition and its algorithms. IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing, 2003, 3: 8-11.

[24] Snoek J, Larochelle H, Adams R P (2012). Practical bayesian optimization of machine learning algorithms. Advances in Neural Information Processing Systems. DOI: https://doi.org/10.48550/arXiv.1206.2944.

[25] Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez A N, et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems. DOI: https://doi.org/10.48550/arXiv.1706.03762.

[26] Veličković P, Cucurull G, Casanova A, Romero A, Lio P, Bengio Y (2017). Graph attention networks. arXiv Preprint arXiv:1710.10903.

[27] Wang H, Tan S, Wang Y, Jiang C, Shi G, Zhang M, et al. A multisource observation study of the severe prolonged regional haze episode over eastern China in January 2013. Atmospheric Environment, 2014, 89: 807-815.

[28] Wang H, Zhuang B, Chen Y, Li N, Wei D (2018). Deep inferential spatial-temporal network for forecasting air pollution concentrations. arXiv Preprint arXiv:1809.03964.

[29] Wang L, Liu Z, Sun Y, Ji D, Wang Y. Long-range transport and regional sources of PM2.5 in Beijing based on long-term observations from 2005 to 2010. Atmospheric Research, 2015, 157: 37-48.

[30] Wang S, Li Y, Zhang J, Meng Q, Meng L, Gao F (2020). PM2.5-GNN: A domain knowledge enhanced graph neural network for PM2.5 forecasting. In Proceedings of the 28th International Conference on Advances in Geographic Information Systems 163–166.

[31] Wen C, Liu S, Yao X, Peng L, Li X, Hu Y, et al. A novel spatiotemporal convolutional long short-term neural network for air pollution prediction. Science of the Total Environment, 2019, 654: 1091-1099.

[32] Xu J, Chen L, Lv M, Zhan C, Chen S, Chang J (2021). High-air: A hierarchical graph neural network-based air quality forecasting method. arXiv Preprint arXiv:2101.04264.

[33] Yang X, Wu Q, Zhao R, Cheng H, He H, Ma Q, et al. New method for evaluating winter air quality: PM2.5 assessment using Community Multi-Scale Air Quality Modeling (CMAQ) in Xi’an. Atmospheric Environment, 2019, 211: 18-28.

[34] Yi X, Zhang J, Wang Z, Li T, Zheng Y (2018). Deep distributed fusion network for air quality prediction. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining 965–973.

[35] Zhang G, Rui X, Fan Y. Critical review of methods to estimate PM2. 5 concentrations within specified research region. ISPRS International Journal of Geo-Information, 2018, 7(9): 368.

[36] Zhang R H, Li Q, Zhang R. Meteorological conditions for the persistent severe fog and haze event over eastern China in January 2013. Science China Earth Sciences, 2014, 57(1): 26-35.

[37] Zhang L, Liu P, Zhao L, Wang G, Liu J. Air quality predictions with a semi-supervised bidirectional LSTM neural network. Atmospheric Pollution Research, 2020, 12(1): 328-339.

[38] Zheng Y, Liu F, Hsieh HP (2013). U-air: When urban air quality inference meets big data. In Proceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining: 1436–1444.

[39] Zhou Y, Chang F, Chang L, Kao I, Wang Y. Explore a deep learning multi-output neural network for regional multi-step-ahead air quality forecasts. Journal of Cleaner Production, 2019, 209: 134-145.