Quantifying multi-year ozone precursor contributions over the Pearl River Delta Region of China using a deep learning-based response surface model

Abstract

Understanding the characteristics of O₃ precursor contributions over multiple years is crucial for designing effective O₃ control strategies over the Pearl River Delta (PRD) region of China. In this study, a deep learning-based response surface model (DeepRSM) was developed and applied over the PRD (DeepRSM-PRD) to identify and quantify the main features of O₃ regimes and regional contributions in the core PRD over multiple years (2019–2021). The Out-of-Sample (OOS) validation results indicated that DeepRSM-PRD effectively predicted the nonlinear response of O₃ to emission controls, maintaining validity across non-training periods. Our study revealed that O₃ generation was sensitive to volatile organic compounds (VOC) in the core PRD in 2019, with nitrogen oxides (NO_x)-limited regimes emerging in most major cities in 2020 and 2021. Further investigation into source contributions showed that in our model domain, O₃ formation in central cities of the PRD was primarily driven by local contributions and was susceptible to influence from nearby cities. With small emission reductions, VOC contributions predominantly drive O₃ production in Guangzhou and Shenzhen. However, NO_x emissions were identified as the primary contributors in all central city receptors when anthropogenic emissions were removed, sharing 59.5%–69.3% in 2019, 64.4%–72.3% in 2020, and 62.75%–73.2% in 2021. Our results highlight the need for a high focus on NO_x emissions control in the core PRD. In addition, for Guangzhou and Shenzhen, VOC reduction also plays a crucial role in the initial stages of modest emission reductions.

Graphical abstract

Keywords

Air quality modeling / O ₃ control strategy / Source contribution / Deep learning / Response surface model / Pearl River Delta Region

Highlight

	● A deep learning-based response surface model for the PRD was first developed.
	● Multi-year O3 source contributions in the core PRD were assessed using the model.
	● From 2019 to 2021, the production of O3 became more sensitive to NO x emissions.
	● NO x emissions from local and nearby areas were the main driver of O3 formation.

References

[1] Borduas-Dedekind N, Naidoo M, Zhu B Q, Geddes J, Garland R M. (2023). Tropospheric ozone (O₃) pollution in Johannesburg, South Africa: exceedances, diurnal cycles, seasonality, O_x chemistry and O₃ production rates. Clean Air Journal, 33(1): 1–16

[2] Chatani S, Amann M, Goel A, Hao J, Klimont Z, Kumar A, Mishra A, Sharma S, Wang S X, Wang Y X. . (2014). Photochemical roles of rapid economic growth and potential abatement strategies on tropospheric ozone over south and east Asia in 2030. Atmospheric Chemistry and Physics, 14(17): 9259–9277

[3] Chatani S, Shimadera H, Itahashi S, Yamaji K. (2020). Comprehensive analyses of source sensitivities and apportionments of PM_2.5 and ozone over Japan via multiple numerical techniques. Atmospheric Chemistry and Physics, 20(17): 10311–10329

[4] Chen J H, Dong H, Zhang Z L, Quan B Q, Luo L. (2023). High-spatiotemporal-resolution estimation of ground-level ozone in China based on machine learning. Atmosphere, 15(1): 34

[5] Chen Y, Zhu Y, Lin C J, Arunachalam S, Wang S X, Xing J, Chen D H, Fan S J, Fang T T, Jiang A Q. (2022). Response surface model based emission source contribution and meteorological pattern analysis in ozone polluted days. Environmental Pollution, 307: 119459

[6] Chen Y T, Zhu L Y, Wang S H, Henze D K, Fu T M, Zhang L, Wang X L. (2025). Unraveling the complexities of ozone and PM_2.5 pollution in the Pearl River Delta region: impacts of precursors emissions and meteorological factors, and effective mitigation strategies. Atmospheric Pollution Research, 16(2): 102368

[7] Clappier A, Belis C A, Pernigotti D, Thunis P. (2017). Source apportionment and sensitivity analysis: two methodologies with two different purposes. Geoscientific Model Development, 10(11): 4245–4256

[8] Do K, Mahish M, Yeganeh A K, Gao Z Q, Blanchard C L, Ivey C E. (2023). Emerging investigator series: a machine learning approach to quantify the impact of meteorology on tropospheric ozone in the inland southern California. Environmental Science: Atmospheres, 3(8): 1159–1173

[9] Dong Z X, Xing J, Zhang F F, Wang S X, Ding D, Wang H L, Huang C, Zheng H T, Jiang Y Q, Hao J M. (2023). Synergetic PM_2.5 and O₃ control strategy for the Yangtze River Delta, China. Journal of Environmental Sciences, 123: 281–291

[10] Du X H, Tang W, Cheng M M, Zhang Z Z, Li Y, Li Y, Meng F. (2022). Modeling of spatial and temporal variations of ozone-NO_x-VOC sensitivity based on photochemical indicators in China. Journal of Environmental Sciences, 114: 454–464

[11] Dunker A M, Yarwood G, Ortmann J P, Wilson G M. (2002). Comparison of source apportionment and source sensitivity of ozone in a three-dimensional air quality model. Environmental Science & Technology, 36(13): 2953–2964

[12] Fang T T, Zhu Y, Jang J, Wang S X, Xing J, Chiang P C, Fan S J, You Z Q, Li J Y. (2020). Real-time source contribution analysis of ambient ozone using an enhanced meta-modeling approach over the Pearl River Delta region of China. Journal of Environmental Management, 268: 110650

[13] Fang T T, Zhu Y, Wang S X, Xing J, Zhao B, Fan S J, Li M H, Yang W W, Chen Y, Huang R L. (2021). Source impact and contribution analysis of ambient ozone using multi-modeling approaches over the Pearl River Delta region, China. Environmental Pollution, 289: 117860

[14] Feng R, Zheng H J, Zhang A R, Huang C, Gao H, Ma Y C. (2019). Unveiling tropospheric ozone by the traditional atmospheric model and machine learning, and their comparison: a case study in Hangzhou, China. Environmental Pollution, 252: 366–378

[15] Ge S J, Wang S J, Xu Q, Ho T. (2021). Source apportionment simulations of ground-level ozone in southeast Texas employing OSAT/APCA in CAMx. Atmospheric Environment, 253: 118370

[16] Hakami A, Odman M T, Russell A G. (2004). Nonlinearity in atmospheric response: a direct sensitivity analysis approach. Journal of Geophysical Research: Atmospheres, 109(D15): D15303

[17] Han X, Zhu L Y, Wang S L, Meng X Y, Zhang M G, Hu J. (2018). Modeling study of impacts on surface ozone of regional transport and emissions reductions over North China Plain in summer 2015. Atmospheric Chemistry and Physics, 18(16): 12207–12221

[18] Hu Y, Shi B W, Yuan X, Zheng C Z, Sha Q, Yu Y F, Huang Z J, Zheng J Y. (2023). VOC emission caps constrained by air quality targets based on response surface model: a case study in the Pearl River Delta region, China. Journal of Environmental Sciences, 123: 430–445

[19] Itahashi S, Hayami H, Uno I. (2015). Comprehensive study of emission source contributions for tropospheric ozone formation over east Asia. Journal of Geophysical Research: Atmospheres, 120(1): 331–358

[20] Jiang F, Guo H, Wang T J, Cheng H R, Wang X M, Simpson I J, Ding A J, Saunders S M, Lam S H M, Blake D R. (2010). An ozone episode in the Pearl River Delta: field observation and model simulation. Journal of Geophysical Research: Atmospheres, 115(D22): D22305

[21] Jiang Y Q, Ding D, Dong Z X, Liu S C, Chang X, Zheng H T, Xing J, Wang S X. (2023). Extreme emission reduction requirements for China to achieve world health organization global air quality guidelines. Environmental Science & Technology, 57(11): 4424–4433

[22] Jin J B, Zhu Y, Jang J, Wang S X, Xing J, Chiang P C, Fan S J, Long S C. (2021). Enhancement of the polynomial functions response surface model for real-time analyzing ozone sensitivity. Frontiers of Environmental Science & Engineering, 15(2): 31

[23] Kitagawa Y K L, Pedruzzi R, Galvão E S, De Araújo I B, de Almeida Alburquerque T T, Kumar P, Nascimento E G S, Moreira D M. (2021). Source apportionment modelling of PM_2.5 using CMAQ-ISAM over a tropical coastal-urban area. Atmospheric Pollution Research, 12(12): 101250

[24] Li J, Jang J C, Zhu Y, Lin C J, Wang S X, Xing J, Dong X Y, Li J Y, Zhao B, Zhang B Y. . (2023). Development of a recurrent spatiotemporal deep-learning method coupled with data fusion for correction of hourly ozone forecasts. Environmental Pollution, 335: 122291

[25] Li J B, Zhu Y, Kelly J T, Jang C J, Wang S X, Hanna A, Xing J, Lin C J, Long S C, Yu L. (2019). Health benefit assessment of PM_2.5 reduction in Pearl River Delta region of China using a model-monitor data fusion approach. Journal of Environmental Management, 233: 489–498

[26] Li J Y, Dai Y Z, Zhu Y, Tang X B, Wang S X, Xing J, Zhao B, Fan S J, Long S C, Fang T T. (2022a). Improvements of response surface modeling with self-adaptive machine learning method for PM_2.5 and O₃ predictions. Journal of Environmental Management, 303: 114210

[27] Li L, Li J Y, Qin M M, Xie X D, Hu J L, Zhang Y Q. (2024). Variations in summertime ozone in Nanjing between 2015 and 2020: roles of meteorology, radical chain length and ozone production efficiency. Frontiers of Environmental Science & Engineering, 18(11): 137

[28] Li Y, Lau A K H, Fung J C H, Zheng J Y, Zhong L J, Louie P K K. (2012). Ozone source apportionment (OSAT) to differentiate local regional and super-regional source contributions in the Pearl River Delta region, China. Journal of Geophysical Research: Atmospheres, 117(D15): D15305

[29] Li Z F, Zhu Y, Wang S X, Xing J, Zhao B, Long S C, Li M H, Yang W W, Huang R L, Chen Y. (2022b). Source contribution analysis of PM_2.5 using response surface model and particulate source apportionment technology over the PRD region, China. Science of the Total Environment, 818: 151757

[30] Mao J Y, Yan F H, Zheng L M, You Y C, Wang W W, Jia S G, Liao W H, Wang X M, Chen W H. (2022). Ozone control strategies for local formation- and regional transport-dominant scenarios in a manufacturing city in Southern China. Science of the Total Environment, 813: 151883

[31] MBSM(2021). Shenzhen Climate Bulletin. Shenzhen: Meteor-ological Bureau of Shenzhen Municipality

[32] Mckay M D, Beckman R J, Conover W J. (1979). Comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics, 21(2): 239–245

[33] Mu X, Wang S C, Jiang P, Wu Y L. (2023). Estimation of surface ozone concentration over Jiangsu Province using a high-performance deep learning model. Journal of Environmental Sciences, 132: 122–133

[34] Ou J M, Yuan Z B, Zheng J Y, Huang Z J, Shao M, Li Z K, Huang X B, Guo H, Louie P K K (2016). Ambient ozone control in a photochemically active region: short-term despiking or long-term attainment? Environmental Science & Technology, 50(11): 5720–5728

[35] Pei C L, Yang W Q, Zhang Y L, Song W, Xiao S X, Wang J, Zhang J P, Zhang T, Chen D H, Wang Y J. . (2022). Decrease in ambient volatile organic compounds during the COVID-19 lockdown period in the Pearl River Delta region, South China. Science of the Total Environment, 823: 153720

[36] Qu Y, An J L, Li J, Chen Y, Li Y, Liu X G, Hu M. (2014). Effects of NO_x and VOCs from five emission sources on summer surface O₃ over the Beijing-Tianjin-Hebei region. Advances in Atmospheric Sciences, 31(4): 787–800

[37] Sayeed A, Eslami E, Lops Y, Choi Y. (2022). CMAQ-CNN: a new-generation of post-processing techniques for chemical transport models using deep neural networks. Atmospheric Environment, 273: 118961

[38] Shen H Z, Sun Z, Chen Y L, Russell A G, Hu Y T, Odman M T, Qian Y, Archibald A T, Tao S. (2021). Novel method for ozone isopleth construction and diagnosis for the ozone control strategy of Chinese cities. Environmental Science & Technology, 55(23): 15625–15636

[39] Sillman S, Vautard R, Menut L, Kley D. (2003). O₃-NO_x-VOC sensitivity and NO_x-VOC indicators in Paris: results from models and atmospheric pollution over the paris area (ESQUIF) measurements. Journal of Geophysical Research: Atmospheres, 108(D17): 8563

[40] Tan J H, Duan J C, Ma Y L, He K B, Cheng Y, Deng S X, Huang Y L, Si-Tu S P. (2016). Long-term trends of chemical characteristics and sources of fine particle in Foshan City, Pearl River Delta: 2008–2014. Science of the Total Environment, 565: 519–528

[41] Tang M X, Huang X F, Yao P T, Wang R H, Li Z J, Liang C X, Peng X, Cao L M, Du K, Yu K Y. . (2023). How much urban air quality is affected by local emissions: a unique case study from a megacity in the Pearl River Delta, China. Atmospheric Environment, 299: 119666

[42] Wang M Y, Huang T, Wong D C, Ho K F, Dong G H, Yim S H L. (2021a). A new approach for health-oriented ozone control strategy: adjoint-based optimization of NO_x emission reductions using metaheuristic algorithms. Journal of Cleaner Production, 312: 127533

[43] Wang N, Lyu X, Deng X J, Huang X, Jiang F, Ding A J. (2019). Aggravating O₃ pollution due to NO_x emission control in eastern China. Science of the Total Environment, 677: 732–744

[44] Wang N, Xu J W, Pei C L, Tang R, Zhou D R, Chen Y N, Li M, Deng X J, Deng T, Huang X. . (2021b). Air quality during COVID-19 lockdown in the Yangtze River Delta and the Pearl River Delta: two different responsive mechanisms to emission reductions in China. Environmental Science & Technology, 55(9): 5721–5730

[45] Wang R Y, Wang L L, Sun J R, Zhang L, Li Y Y, Li K, Liu B Y, Zhang J X, Wang Y S. (2023). Maximizing ozone control by spatial sensitivity-oriented mitigation strategy in the Pearl River Delta region, China. Science of the Total Environment, 905: 166987

[46] Wang S X, Xing J, Jang C, Zhu Y, Fu J S, Hao J M. (2011a). Impact assessment of ammonia emissions on inorganic aerosols in East China using response surface modeling technique. Environmental Science & Technology, 45(21): 9293–9300

[47] Wang S X, Zhao B, Cai S Y, Klimont Z, Nielsen C P, Morikawa T, Woo J H, Kim Y, Fu X, Xu J Y. . (2014). Emission trends and mitigation options for air pollutants in East Asia. Atmospheric Chemistry and Physics, 14(13): 6571–6603

[48] Wang S Y, Zhang Y L, Ma J L, Zhu S Q, Shen J Y, Wang P, Zhang H L. (2021c). Responses of decline in air pollution and recovery associated with COVID-19 lockdown in the Pearl River Delta. Science of the Total Environment, 756: 143868

[49] Wang S Y, Zhu Y, Jang J C, Jiang M, Yue D L, Zhong L J, Yuan Y Z, Zhang M M, You Z Q. (2024). Modeling assessment of air pollution control measures and COVID-19 pandemic on air quality improvements over Greater Bay Area of China. Science of the Total Environment, 926: 171951

[50] Wang W N, Van Der A R, Ding J Y, Van Weele M, Cheng T H. (2021d). Spatial and temporal changes of the ozone sensitivity in China based on satellite and ground-based observations. Atmospheric Chemistry and Physics, 21(9): 7253–7269

[51] Wang X S, Zhang Y H, Hu Y T, Zhou W, Zeng L M, Hu M, Cohan D S, Russell A G. (2011b). Decoupled direct sensitivity analysis of regional ozone pollution over the Pearl River Delta during the PRIDE-PRD 2004 campaign. Atmospheric Environment, 45(28): 4941–4949

[52] Wei J, Li Z Q, Li K, Dickerson R R, Pinker R T, Wang J, Liu X, Sun L, Xue W H, Cribb M. (2022). Full-coverage mapping and spatiotemporal variations of ground-level ozone (O₃) pollution from 2013 to 2020 across China. Remote Sensing of Environment, 270: 112775

[53] Xiao X, Cohan D S, Byun D W, Ngan F. (2010). Highly nonlinear ozone formation in the Houston region and implications for emission controls. Journal of Geophysical Research: Atmospheres, 115(D23): D23309

[54] Xie Q Q, Tham Y J, Yu X Y, Wang Z, Ling Z H, Wang X M, Guo H, Wang T. (2023). Seasonal variations of O₃ formation mechanism and atmospheric photochemical reactivity during severe high O₃ pollution episodes in the Pearl River Delta Region. Atmospheric Environment, 309: 119918

[55] Xing J, Ding D, Wang S X, Zhao B, Jang C, Wu W J, Zhang F F, Zhu Y, Hao J M. (2018). Quantification of the enhanced effectiveness of NO_x control from simultaneous reductions of VOC and NH₃ for reducing air pollution in the Beijing-Tianjin-Hebei Region, China. Atmospheric Chemistry and Physics, 18(11): 7799–7814

[56] Xing J, Wang S X, Jang C, Zhu Y, Hao J M. (2011). Nonlinear response of ozone to precursor emission changes in China: a modeling study using response surface methodology. Atmospheric Chemistry and Physics, 11(10): 5027–5044

[57] Xing J, Wang S X, Zhao B, Wu W J, Ding D, Jang C, Zhu Y, Chang X, Wang J D, Zhang F F. . (2017). Quantifying nonlinear multiregional contributions to ozone and fine particles using an updated response surface modeling technique. Environmental Science & Technology, 51(20): 11788–11798

[58] Xing J, Zheng S X, Ding D, Kelly J T, Wang S X, Li S W, Qin T, Ma M Y, Dong Z X, Jang C. . (2020). Deep learning for prediction of the air quality response to emission changes. Environmental Science & Technology, 54(14): 8589–8600

[59] Xing J, Zheng S X, Li S W, Huang L, Wang X C, Kelly J T, Wang S X, Liu C, Jang C, Zhu Y. . (2022). Mimicking atmospheric photochemical modeling with a deep neural network. Atmospheric Research, 265: 105919

[60] XuBWangN Y ChenT QLi M (2015). Empirical evaluation of rectified activations in convolutional network. arXiv preprint arXiv: 1505.00853

[61] Xue J, Wang F T, Zhang K, Zhai H H, Jin D, Duan Y S, Yaluk E, Wang Y J, Huang L, Li Y W. . (2023). Elucidate long-term changes of ozone in Shanghai based on an integrated machine learning method. Frontiers of Environmental Science & Engineering, 17(11): 138

[62] Yang L F, Luo H H, Yuan Z B, Zheng J Y, Huang Z J, Li C, Lin X H, Louie P K K, Chen D H, Bian Y H. (2019a). Quantitative impacts of meteorology and precursor emission changes on the long-term trend of ambient ozone over the Pearl River Delta, China, and implications for ozone control strategy. Atmospheric Chemistry and Physics, 19(20): 12901–12916

[63] Yang W Y, Chen H S, Wang W D, Wu J B, Li J, Wang Z F, Zheng J Y, Chen D H. (2019b). Modeling study of ozone source apportionment over the Pearl River Delta in 2015. Environmental Pollution, 253: 393–402

[64] Zhang Z, Li Y C, Ding G Q, Fan B S, Da S Q, Xu Q H, Wang Y, Chi Z Q, Dong J, Liu C F. . (2022). Astronomical forcing of vegetation and climate change during the Late Pliocene–Early Pleistocene of the Nihewan Basin, North China. Quaternary International, 613: 1–13

[65] Zhao B, Wang S X, Xing J, Fu K, Fu J S, Jang C, Zhu Y, Dong X Y, Gao Y, Wu W J. . (2015). Assessing the nonlinear response of fine particles to precursor emissions: development and application of an extended response surface modeling technique v1.0. Geoscientific Model Development, 8(1): 115–128

[66] Zhao B, Zheng H T, Wang S X, Smith K R, Lu X, Aunan K, Gu Y, Wang Y, Ding D, Xing J. . (2018). Change in household fuels dominates the decrease in PM_2.5 exposure and premature mortality in China in 2005–2015. Proceedings of the National Academy of Sciences of the United States of America, 115(49): 12401–12406

[67] Zhao K H, Yuan Z B, Wu Y H, Huang J P, Yang F Y, Zhang X F, Huang D Z, Jiang R S. (2023). Identification of synergistic control for ozone and PM_2.5 pollution during a large-scale emission reduction in China. Atmospheric Research, 295: 107025

RIGHTS & PERMISSIONS

Higher Education Press 2025