Prediction and cause investigation of ozone based on a double-stage attention mechanism recurrent neural network

Abstract

● Used a double-stage attention mechanism model to predict ozone.

● The model can autonomously select the appropriate time series for forecasting.

● The model outperforms other machine learning models and WRF-CMAQ.

● We used the model to analyze the driving factors of VOCs that cause ozone pollution.

Ozone is becoming a significant air pollutant in some regions, and VOCs are essential for ozone prediction as necessary ozone precursors. In this study, we proposed a recurrent neural network based on a double-stage attention mechanism model to predict ozone, selected an appropriate time series for prediction through the input attention and temporal attention mechanisms, and analyzed the cause of ozone generation according to the contribution of feature parameters. The experimental data show that our model had an RMSE of 7.71 μg/m³ and a mean absolute error of 5.97 μg/m³ for 1-h predictions. The DA-RNN model predicted ozone closer to observations than the other models. Based on the importance of the characteristics, we found that the ozone pollution in the Jinshan Industrial Zone mainly comes from the emissions of petrochemical enterprises, and the good generalization performance of the model is proved through testing multiple stations. Our experimental results demonstrate the validity and promising application of the DA-RNN model in predicting atmospheric pollutants and investigating their causes.

Graphical abstract

Keywords

Ozone prediction / Deep learning / Time series / Attention / Volatile organic compounds

References

[1] AtkinsonR ( 2000). Atmospheric chemistry of VOCs and NO _x . Atmospheric Environment, 34( 12– 14): 2063– 2101

[2] AveryR J. (2006). Reactivity-based VOC control for solvent products: more efficient ozone reduction strategies. Environmental Science & Technology, 40( 16): 4845– 4850

[3] BasM D, OrtizJ, BallesterosL, MartorellS. (2017). Evaluation of a multiple linear regression model and SARIMA model in forecasting ⁷Be air concentrations. Chemosphere, 177 : 326– 333

[4] BengioY, SimardP, FrasconiP. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5( 2): 157– 166

[5] ByunD, SchereK L. (2006). Review of the governing equations, computational algorithms, and other components of the Models-3 community multiscale air quality (CMAQ) modeling system. Applied Mechanics Reviews, 59( 2): 51– 77

[6] ChenY, PhonevilayV, TaoJ, ChenX, XiaR, ZhangQ, YangK, XiongJ, XieJ. (2021). The face image super-resolution algorithm based on combined representation learning. Multimedia Tools and Applications, 80( 20): 30839– 30861

[7] ChinM, GinouxP, KinneS, TorresO, HolbenB N, DuncanB N, MartinR V, LoganJ A, HigurashiA, NakajimaT. (2002). Tropospheric aerosol optical thickness from the GOCART model and comparisons with satellite and Sun photometer measurements. Journal of the Atmospheric Sciences, 59( 3): 461– 483

[8] Delle MonacheL, NipenT, DengX X, ZhouY M, StullR. (2006). Ozone ensemble forecasts: 2. A Kalman filter predictor bias correction. Journal of Geophysical Research, 111( D5): D05308

[9] DingY, ChenZ, LuW, WangX. (2021). A CatBoost approach with wavelet decomposition to improve satellite-derived high-resolution PM_2.5 estimates in Beijing-Tianjin-Hebei. Atmospheric Environment, 249 : 118212

[10] DonnellyA, MisstearB, BroderickB. (2015). Real time air quality forecasting using integrated parametric and non-parametric regression techniques. Atmospheric Environment, 103 : 53– 65

[11] EfstathiouG A, ZoumakisN M, MelasD, LolisC J, KassomenosP ( 2013). Sensitivity of WRF to boundary layer parameterizations in simulating a heavy rainfall event using different microphysical schemes : e ffect on large-scale processes. Atmospheric Research, 132– 133: 125– 143

[12] EkM B, MitchellK E, LinY, RogersE, GrunmannP, KorenV, GaynoG, TarpleyJ D. (2003). Implementation of Noah land surface model advances in the national centers for environmental prediction operational mesoscale Eta model. Journal of Geophysical Research, 108( D22): 2002JD003296

[13] ErdemE, ShiJ. (2011). ARMA based approaches for forecasting the tuple of wind speed and direction. Applied Energy, 88( 4): 1405– 1414

[14] GaraudD, MalletV. (2011). Automatic calibration of an ensemble for uncertainty estimation and probabilistic forecast: application to air quality. Journal of Geophysical Research, 116( D19): D19304

[15] GeB Z, WangZ F, XuX B, WuJ B, YuX L, LiJ. (2014). Wet deposition of acidifying substances in different regions of China and the rest of East Asia: modeling with updated NAQPMS. Environmental Pollution, 187 : 10– 21

[16] HongS Y, NohY, DudhiaJ. (2006). A new vertical diffusion package with an explicit treatment of entrainment processes. Monthly Weather Review, 134( 9): 2318– 2341

[17] HuX, BelleJ H, MengX, WildaniA, WallerL A, StricklandM J, LiuY. (2017). Estimating PM_2.5 concentrations in the conterminous United States using the random forest approach. Environmental Science & Technology, 51( 12): 6936– 6944

[18] HuiL, LiuX, TanQ, FengM, AnJ, QuY, ZhangY, JiangM. (2018). Characteristics, source apportionment and contribution of VOCs to ozone formation in Wuhan, Central China. Atmospheric Environment, 192 : 55– 71

[19] JiaP, CaoN, YangS. (2021). Real-time hourly ozone prediction system for Yangtze River Delta area using attention based on a sequence to sequence model. Atmospheric Environment, 244 : 117917

[20] KampaM, CastanasE. (2008). Human health effects of air pollution. Environmental Pollution, 151( 2): 362– 367

[21] KavasseriR G, SeetharamanK. (2009). Day-ahead wind speed forecasting using f-ARIMA models. Renewable Energy, 34( 5): 1388– 1393

[22] LiuC, ZhangH, ChengZ, ShenJ, ZhaoJ, WangY, WangS, ChengY. (2021). Emulation of an atmospheric gas-phase chemistry solver through deep learning: case study of Chinese mainland. Atmospheric Pollution Research, 12( 6): 101079

[23] MlawerE J, TaubmanS J, BrownP D, IaconoM J, CloughS A. (1997). Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. Journal of Geophysical Research, 102( D14): 16663– 16682

[24] Molina-GómezN I, Diaz-ArevaloJ L, Lopez-JimenezP A. (2021). Air quality and urban sustainable development: the application of machine learning tools. International Journal of Environmental Science and Technology, 18( 4): 1029– 1046

[25] OuJ, Zheng J, LiR, HuangX, ZhongZ, ZhongL, LinH ( 2015). Speciated OVOC and VOC emission inventories and their implications for reactivity-based ozone control strategy in the Pearl River Delta region, China. Sci ence of the Total Environ ment, 530– 531: 393– 402

[26] QinY, SongD, ChengH, ChengW, JiangG, CottrellG W. (2017). A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction. Melbourne, Australia : 2627– 2633

[27] ReichsteinM, Camps-VallsG, StevensB, JungM, DenzlerJ, CarvalhaisN. (2019). Deep learning and process understanding for data-driven Earth system science. Nature, 566( 7743): 195– 204

[28] ShaoM, ZhangY, ZengL, TangX, ZhangJ, ZhongL, WangB. (2009). Ground-level ozone in the Pearl River Delta and the roles of VOC and NO_x in its production. Journal of Environmental Management, 90( 1): 512– 518

[29] TaoQ Liu F LiY SidorovD ( 2019). Air pollution forecasting using a deep learning model based on 1D convnets and bidirectional GRU. IEEE Access: Practical Innovations, Open Solutions, 7: 76690– 76698

[30] VautardR, BuiltjesP H J, ThunisP, CuvelierC, BedogniM, BessagnetB, HonoreC, MoussiopoulosN, PirovanoG, SchaapM, SternR, TarrasonL, WindP. (2007). Evaluation and intercomparison of ozone and PM₁₀ simulations by several chemistry transport models over four European cities within the CityDelta project. Atmospheric Environment, 41( 1): 173– 188

[31] WangB, YuanQ, YangQ, ZhuL, LiT, ZhangL. (2021). Estimate hourly PM_2.5 concentrations from Himawari-8 TOA reflectance directly using geo-intelligent long short-term memory network. Environmental Pollution, 271 : 116327

[32] WangH, XiangZ, WangL, JingS, LouS, TaoS, LiuJ, YuM, LiL, LinL, ChenY, WiedensohlerA, ChenC. (2018). Emissions of volatile organic compounds (VOCs) from cooking and their speciation: A case study for Shanghai with implications for China. Science of the Total Environment, 621 : 1300– 1309

[33] WangS, FangC, MaH, WangY, QinJ. (2014). Spatial differences and multi-mechanism of carbon footprint based on GWR model in provincial China. Journal of Geographical Sciences, 24( 4): 612– 630

[34] WongP Y, SuH J, LeeH Y, ChenY C, HsiaoY P, HuangJ W, TeoT A, WuC D, SpenglerJ D. (2021). Using land-use machine learning models to estimate daily NO₂ concentration variations in Taiwan, China. Journal of Cleaner Production, 317 : 128411

[35] XiaR, Chen Y, RenB ( 2022). Improved anti-occlusion object tracking algorithm using Unscented Rauch-Tung-Striebel smoother and kernel correlation filter. Journal of King Saud University - Computer and Information Sciences

[36] XuX, ZhangC, LiangY. (2021). Review of satellite-driven statistical models PM_2.5 concentration estimation with comprehensive information. Atmospheric Environment, 256 : 118302

[37] ZhangH, LinfordJ C, SanduA, SanderR. (2011). Chemical mechanism solvers in air quality models. Atmosphere (Toronto), 2( 3): 510– 532

[38] ZhangJ M, FengW J, YuanT Y, WangJ, SangaiahA K. (2022a). SCSTCF: spatial-channel selection and temporal regularized correlation filters for visual tracking. Applied Soft Computing, 118 : 108485

[39] ZhangJ M, SunJ, WangJ, LiZ P, ChenX. (2022b). An object tracking framework with recapture based on correlation filters and Siamese networks. Computers & Electrical Engineering, 98 : 107730

[40] ZhangY, WenX Y, JangC J. (2010). Simulating chemistry-aerosol-cloud-radiation-climate feedbacks over the continental US using the online-coupled Weather Research Forecasting Model with chemistry (WRF/Chem). Atmospheric Environment, 44( 29): 3568– 3582

RIGHTS & PERMISSIONS

Higher Education Press