兰州市CMAQ近地面臭氧模拟结果的订正方法——基于机器学习方法

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (2288 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要为能更加准确地模拟出兰州市近地面臭氧浓度，在CMAQ （社区多尺度空气质量建模系统）的基础上，利用机器学习方法中的XGBoost （极限梯度提升）模型及LSTM （长短期记忆）神经网络模型建立近地面臭氧模拟结果的订正模型，并以两种方法为基础，利用误差变权倒数组合方法构建LSTM-XGBoost组合模型，以期进一步提高订正效果.本文选取兰州市4个国控站点（兰炼宾馆，铁路设计院，榆中校区，生物制品所）2019年7、8月环境空气质量监测数据及兰州市气象站同期气象数据，对CMAQ模拟的同时段兰州市近地面臭氧浓度进行订正.结果表明，CMAQ能够模拟出兰州市近地面臭氧浓度的空间及时间分布特征，但整体上对浓度有所低估.利用上述方法构建的订正模型中，LSTM-XGBoost组合模型的订正效果最好，臭氧相关性由CMAQ模拟的0.61~0.76提升至0.89~0.95，臭氧8h平均相关性由0.65~0.79提升至0.81~0.88，臭氧RMSE由44.83~70.17mg/m³提升至15.21~26.53mg/m³，臭氧8h平均RMSE由40.07~67.57mg/m³提升至14.24~28.54mg/m³.该研究表明利用机器学习方法对CMAQ模拟结果订正可行，可以改善环境空气质量模式模拟结果.

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	周恒左
	陈恒蕤
	廖鹏
	孔祥如
	潘峰
	杨宏

关键词 ： CMAQ, 近地面臭氧, 机器学习, LSTM, XGBoost, 误差变权倒数组合

Abstract：To better simulate the near-surface ozone concentration in Lanzhou, the XGBoost (eXtreme Gradient Boosting) model and the LSTM (Long and Short-Term Memory) neural network model in the machine learning method were used on the basis of CMAQ (The Community Multiscale Air Quality Modeling System) to establish a revised model of near-surface ozone simulation results, and the combined LSTM-XGBoost model was constructed based on the two methods with the combination of the inverse of error variance weights to further improve the revision effect. In this study, four national monitoring sites in Lanzhou (Lan Lian Hotel, Railway Design Institute, Yuzhong Campus, and Biological Products Institute) were selected, and ambient air quality monitoring data and meteorological data in July and August 2019 were used to revise the near-surface ozone concentrations simulated by CMAQ. Results showed that the CMAQ model could simulate the spatial and temporal distributions of the near-surface ozone concentrations in Lanzhou, but the concentrations was underestimated. Among the revised models mentioned above, the XGBoost combined model revised best. Compared with the simulation results of CMAQ, the correlation of ozone concentration improved from 0.61~0.76 to 0.89~0.95, the correlation of 8h ozone concentration improved from 0.65~0.79 to 0.81~0.88, the ozone RMSE improved from 44.83~70.17μg/m³ to 15.21~26.53μg/m³, 8h ozone RMSE improved from 40.07~67.57μg/m³ to 14.24~28.54μg/m³. This study indicated that it is feasible to revise the model simulation results using machine learning methods to improve the air quality model.

Key words： CMAQ ozone machine learning LSTM XGBoost error-variant weighted inverse combination

收稿日期: 2022-05-13

PACS:

X511

基金资助:兰州大学中央高校基本科研业务费专项资金自由探索项目（lzujbky-2017-65）

通讯作者: 杨宏,工程师,yanghongyh@lzu.edu.cn E-mail: yanghongyh@lzu.edu.cn

作者简介: 周恒左(1998-),男,辽宁丹东人,兰州大学硕士研究生,主要从事大气污染、大气环境等方面研究.发表论文1篇.

引用本文:

周恒左, 陈恒蕤, 廖鹏, 孔祥如, 潘峰, 杨宏. 兰州市CMAQ近地面臭氧模拟结果的订正方法——基于机器学习方法[J]. 中国环境科学, 2022, 42(12): 5472-5483. ZHOU Heng-zuo, CHEN Heng-rui, LIAO Peng, KONG Xiang-rui, PAN Feng, YANG Hong. A revised approach to CMAQ near-surface ozone modelling results in Lanzhou-Based on machine learning methods. CHINA ENVIRONMENTAL SCIENCECE, 2022, 42(12): 5472-5483.

链接本文:

http://www.zghjkx.com.cn/CN/ 或 http://www.zghjkx.com.cn/CN/Y2022/V42/I12/5472

[1]	杨燕萍,王莉娜,杨丽丽,等.兰州市O3污染特征及气象因子相关性研究[J]. 环境与发展, 19,31(11):51-52. Yang Y P, Wang L N, Yang L L, et al. The relevance between the characteristics of O3 and meterological factors in LanZhou[J]. Environment and Development, 19,31(11):51-52.
[2]	李林耀,李雨娟,王晓琦,等.银川城区夏季臭氧污染成因分析[J]. 环境科学与技术, 021,44(S2):50-54. Lin L Y, Li Y J, Wang X Q, et al. Analysis of the causes of ozone pollution in Summer of YinChuan City[J]. Environmental Science and Technology, 021,44(S2):50-54.
[3]	Yu R L, Lin Y L, Zou J H, et al. Review on atmospheric ozone pollution in China:Formation, spatiotemporal distribution, precursors and affecting factors[J]. Atmosphere, 021,12(12):1675-1675.
[4]	赵伟,高博,卢清,等.2006~2019年珠三角地区臭氧污染趋势[J]. 环境科学, 021,42(1):97-105. Zhao W, Gao B, Lu Q, et al. Ozone pollution trend in the Pearl River Delta region during 2006~2019[J].Environmental Science, 021,42(1):97-105.
[5]	陈浪,赵川,关茗洋,等.我国大气臭氧污染现状及人群健康影响[J]. 环境与职业医学, 2017,34(11):1025-1030. Chen L, Zhao C, Guan M Y, et al. Ozone pollution in China and its adverse health effects[J]. Environmental and Occupational Medicine, 2017,34(11):1025-1030.
[6]	王倩,刘苗苗,杨建勋,等.2013~2019年臭氧污染导致的江苏稻麦产量损失评估[J]. 中国环境科学, 2021,41(11):5094-5103. Wang Q, Liu M M, Yang J X, et al. Yield losses of rice and winter wheat due to ground level ozone in Jiangsu Province from 2013 to 2019[J]. China Environmental Science, 2021,41(11):5094-5103.
[7]	赵俊日,肖昕,吴涛,等.空气质量数值预报优化方法研究[J]. 中国环境科学, 2018,38(6):2047-2054. Zhao J R, Xiao X, Wu T, et ai. A revised approach to air quality forecast based on Models-3/CMAQ[J]. China Environmental Science, 2018,38(6):2047-2054.
[8]	卢亚灵,李勃,范朝阳,等.空气质量预测模拟技术演变与发展研究[J]. 中国环境管理, 2021,13(4):84-92. Lu Y L, Li B, Fan C Y, et al. Evolution and development of air quality prediction and simulation technology[J]. China Environmental Management, 2021,13(4):84-92.
[9]	王占山,李晓倩,王宗爽,等.空气质量模型CMAQ的国内外研究现状[J]. 环境科学与技术, 2013,36(S1):386-391. Wang Z S, Li X Q, Wang Z S, et al. Application status of Models-3/CMAQ in environmental management[J]. Environmental Science and Technology, 2013,36(S1):386-391.
[10]	王茜,吴剑斌,林燕芬.CMAQ模式及其修正技术在上海市PM2.5预报中的应用检验[J]. 环境科学学报, 2015,35(6):1651-1656. Wang Q, Wu J B, Lin Y F. Implementation of a dynamic linear regression method on the CMAQ forecast of PM2.5 in Shanghai[J]. Journal of Environmental Science, 2015,35(6):1651-1656.
[11]	程兴宏,刁志刚,胡江凯,等.基于CMAQ模式和自适应偏最小二乘回归法的中国地区PM2.5浓度动力-统计预报方法研究[J]. 环境科学学报, 2016,36(8):2771-2782. Cheng X H, Diao Z G Hu J K, et al. Dynamical-statistical forecasting of PM2.5 concentration based on CMAQ model and adapting partial least square regression method in China[J]. Journal of Environmental Science, 2016,36(8):2771-2782.
[12]	芦华,谢旻,吴钲,等.基于机器学习的成渝地区空气质量数值预报PM2.5订正方法研究[J]. 环境科学学报, 2020,40(12):4419-4431. Lu H, Xie M, Wu Z, et al. Adjusting PM2.5 perdiction of the numerical air quality forecast model based on machine learning methods in Chengyu region[J]. Journal of Environmental Science, 2020,40(12):4419-4431.
[13]	陶燕,朱礼波,王洪新,等.兰州市大气污染特征及其成因探讨[J]. 环境科学与技术, 2014,37(8):87-92. Tao Y, Zhu L B, Wang H X, et al. Characteristics of air pollution and its causes in Lanzhou[J]. Environmental Science and Technology, 2014,37(8):87-92.
[14]	马秉吉,马玉霞,周建丁,等.兰州市近地面臭氧污染特征及影响因素分析[C]. 第35届中国气象学会年会S17气候环境变化与人体健康, 2018:105. Ma B J, Ma Y X, Zhou J D, et al. Analysis of near-ground-level ozone pollution characteristics and influencing factors in Lanzhou City[C]. The 35th Annual Meeting of the Chinese Meteorological Society S17Climate Environmental Change and Human Health, 2018:105.
[15]	刘玉红.中国臭氧污染时空分布特征及影响因素研究[D]. 大连:大连理工大学, 2021. Liu Y H. Research on the spatiotemporal distribution characteristics and influencing factors of ozone pollution in China[D]. Dalian:Dalian University of Technology, 2021.
[16]	卢秀娟,巨天珍,谢顺涛,等.兰州地区大气臭氧时空变化及影响因素研究[J]. 地球与环境, 2018,46(4):355-363. Lu X J, Ju T Z, Xie S T, et al. Temporal and spatial variation of atomspheric ozone in Lanzhou and its influencing factors[J]. Earth and Environment, 2018,46(4):355-363.
[17]	GB 3095-2012环境空气质量标准[S]. GB 3095-2012 Ambient air quality standards[S].
[18]	Agarwal S, Sharma S, Suresh R, et al. Air quality forecasting using artificial neural networks with real time dynamic error correction in highly polluted regions[J]. Science of the Total Environment, 2020,735:139454.
[19]	马明月.基于数值模式的兰州市主城区臭氧前体物减排情景研究[D]. 兰州:兰州大学, 2021. Ma M Y. Study on emission reduction scenario of ozone precursors in main urban area of Lanzhou bsaed on the numerical model[D]. Lanzhou:Lanzhou University, 2021.
[20]	马丹阳.兰州市石化生产行业对主城区近地面臭氧污染的数值模拟研究[D]. 兰州:兰州大学, 2020. Ma D Y. The simulation of petrochemical industry in Lanzhou City on the ozone pollution near the ground in the main urban area[D]. Lanzhou:Lanzhou University, 2020.
[21]	尹伊.基于CMAQ模式、MEIC清单对兰州市主城区臭氧污染特征的初步研究[D]. 兰州:兰州大学, 2019. Yin Y. Ozone pollution characteristics in the main urban area of Lanzhou City based on CMAQ model and MEIC[D]. 兰州:兰州大学, 2019.
[22]	Chen S H, Sun W Y.A One-dimensional time dependent cloud model[J]. Journal of the Meteorological Society of Japan. Ser. II, 2002,80(1):99-118.
[23]	康俊锋,黄烈星,张春艳,等.多机器学习模型下逐小时PM2.5预测及对比分析[J]. 中国环境科学, 2020,40(5):1895-1905. Kang J F, Huang L X, Zhang C Y, et al. Hourly PM2.5 prediction and its comparative analysis under multi-machine learning model[J]. China Environmental Science, 2020,40(5):1895-1905.
[24]	Liu L L, Liu H C, Zhu C F. Thunderstorm weather analysis based on XGBoost algorithm[J]. ISPRS-International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2020, XLII-3/W10:261-264.
[25]	Vuong P H, Dat T T, Mai T K, et al. Stock-price forecasting based on XGBoost and LSTM[J]. Computer Systems Science and Engineering, 2022,40(1):237-246.
[26]	Qian J C, Zhu M F, Zhao Y N, et al. Short-term wind speed prediction with a two-layer attention-based LSTM[J]. Computer Systems Science and Engineering, 2021,39(2):197-209.
[27]	亓浩雲,王晓琦,程水源.人为排放及气象因素对空气质量影响的定量分析——以疫情期间邢台市为例[J]. 中国环境科学, 2022,42(8):3512-3521. Qi H Y, Wang X Q, Cheng S Y. Quantitative analysis of the impact of anthropogenic emissions and meteorological factors on air quality:Cases during the epidemic in Xingtai City[J]. China Environmental Science, 2022,42(8):3512-3521.
[28]	岳艳春,黄廷祝.误差倒数变权组合预测方法[J]. 电子科技大学学报, 2007,(S1):349-351. Yue Y C, Huang T Z. A method for error reciprocal variable weight combined forecasting[J]. Journal of the University of Electronic Science and Technology of China, 2007,(S1):349-351.
[29]	James W B, Armistead G R. PM and light extinction model performance metrics, goals, and criteria for three-dimensional air quality models[J]. Atmospheric Environment, 2006,40(26):4946-4959.
[30]	姜少飞,邬天骥,彭翔,等.基于XGBoost特征提取的数据驱动故障诊断方法[J]. 中国机械工程, 2020,31(10):1232-1239. Jiang S F, Wu T J, Peng X, et al. Data driven fault diagnosis method bsaed on XGBoost feature extraction[J]. China Mechanical Engineering, 2020,31(10):1232-1239.