1.珠海市公共气象服务中心,广东 珠海 519000
2.珠澳气象创新与应用研究中心,广东 珠海 519000
孙磊(1992年生),男;研究方向:空气污染预报与来源解析;E-mail:sunleimodel@gmail.com
蓝玉峰(1990年生),女;研究方向:空气污染预报与来源解析;E-mail:lyfnuist@163.com
纸质出版日期:2024-05-25,
网络出版日期:2024-03-20,
收稿日期:2023-12-17,
录用日期:2024-01-09
扫 描 看 全 文
孙磊,蓝玉峰,梁秀姬等.基于图卷积记忆网络对珠海臭氧时空预测[J].中山大学学报(自然科学版)(中英文),2024,63(03):48-59.
SUN Lei,LAN Yufeng,LIANG Xiuji,et al.The spatio-temporal prediction of ozone in Zhuhai based on graph convolutional memory network[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2024,63(03):48-59.
孙磊,蓝玉峰,梁秀姬等.基于图卷积记忆网络对珠海臭氧时空预测[J].中山大学学报(自然科学版)(中英文),2024,63(03):48-59. DOI: 10.13471/j.cnki.acta.snus.ZR20230043.
SUN Lei,LAN Yufeng,LIANG Xiuji,et al.The spatio-temporal prediction of ozone in Zhuhai based on graph convolutional memory network[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2024,63(03):48-59. DOI: 10.13471/j.cnki.acta.snus.ZR20230043.
臭氧(O
3
)已成为影响珠三角(乃至广东)空气质量达标的首要因素。数据驱动的统计模型(较数值模式)虽展现出改进的预报能力,但多数未能解析站点数据(非欧结构)之间的空间依赖性。本文基于珠海市6个环保国控站及其周边气象站监测数据,通过构建时空协同的图卷积记忆网络(GCN-LSTM)开展多站点未来3天逐小时O
3
质量浓度预报。结果表明:GCN_LSTM在不同预报时效均准确还原了O
3
的年、季节和昼夜变化特征,但对日变化的预报技巧随预报时效增加下降明显。通过与业务数值模式(GRACEs)和长短期记忆网络(LSTM)对比发现:GCN-LSTM表现最优,其72 h预报时效内RMSE和
R
均值分别为27.13 μg/m
3
和0.64,LSTM表现次之(RMSE=28.44 μg/m
3
;
R
=0.61),而GRACEs与统计模型存在明显差距(RMSE=40.93 μg/m
3
;
R
=0.33)。此外,相较于LSTM,GCN-LSTM全局考虑所有站点及其之间的相互联系,不仅将计算速度提高了71%,而且在不同站点的表现也更为优秀和稳定,同时捕捉秋季O
3
污染事件的能力也有所提高。最后,敏感性实验揭示出考虑相关性较高的变量作为预报因子可以提高模型能力。
Ozone(O₃) has become the primary factor affecting air quality over the Pearl River Delta and even the entire Guangdong Province. Although data-driven statistical models have shown improved forecast capabilities compared to numerical models, most of them operate grid-by-grid and cannot resolve the spatial dependence between site data of non-Euclidean structures. Based on in-situ measurements from national environmental stations and surrounding weath
er stations in Zhuhai, this study performs hourly O₃ concentration forecasts for up to three days over multiple sites by constructing a graph convolution memory network (GCN-LSTM). The results show that GCN_LSTM forecasts at different lead times could accurately reproduce the annual, seasonal, and diurnal variations of O
3
, but the capability of capturing daily variations decreases significantly with the increase in lead time. Further comparisons with the operational numerical model(GRACEs) and Long Short-Term Memory (LSTM) reveal that GCN-LSTM performs the best, with mean RMSE=27.13 μg/m
3
and
R
=0.64, LSTM is the second (RMSE=28.44 μg/m
3
;
R
=0.61), and GRACEs presents distinct results (RMSE = 40.93 μg/m
3
;
R
=0.33) in 72h forecasting. Compared with LSTM, GCN-LSTM considers all sites and their interconnections, it not only increases the calculation speed by 71% but also performs better and more stably over different sites. Moreover, it is also optimal for capturing O₃ pollution events in cold seasons. Additional sensitivity experiments reveal that considering more correlated variables improves forecasting capabilities.
臭氧时空预报机器学习图卷积记忆网络
ozone(O₃)spatial-temporal forecastmachine learninggraph convolution memory network
陈德辉,薛纪善,沈学顺,等,2012.我国自主研制的全球/区域一体化数值天气预报系统GRAPES的应用与展望[J].中国工程科学,14(9):46-54.
邓雪娇,邓涛,麦博儒,等,2016.华南区域大气成分业务数值预报GRACEs模式系统[J].热带气象学报,32(6):900-907.
赖安琪,陈晓阳,刘一鸣,等,2018.珠江三角洲高质量浓度PM2.5和O3复合污染特征[J].中山大学学报(自然科学版),57(4):30-36.
李婷苑,吴乃庚,邓雪娇,等,2021.华南区域大气成分数值模式GRACEs预报性能评估[J].热带气象学报,37(2):207-217.
李勋,秦墨梅,李琳,等,2020.基于空气质量模拟的江苏省大气污染物排放清单比较研究[J].南京信息工程大学学报(自然科学版),12(6):705-713.
祁柏林,郭昆鹏,杨彬,等,2021.基于GCN-LSTM 的空气质量预测[J].计算机系统应用,30(3):208-213.
孙弦,孙磊,聂会文,等,2023.两种空气质量数值模式的应用评估与集合改进研究[J].热带气象学报,39(3):361-373.
王自发,吴其重,GBAGUIDI A,等,2009.北京空气质量多模式集成预报系统的建立及初步应用[J].南京信息工程大学学报(自然科学版),1(1):19-26.
吴乃庚, 邓玉娇, 向昆仑, 等, 2020. 2019年广东省生态气象监测公报[Z].广州:
广东省气象局.
夏冬,谭浩波,陈玲,等,2012.用逐步回归法预报地面臭氧浓度[J].广东气象,34(1):47-50.
杨雨佳,肖庆来,陈健,等,2023.融合空间和统计特征的CNN-GRU臭氧浓度预测模型研究[J].南京大学学报(自然科学版),59(2):322-332.
张灿,蒋昌潭,罗财红,等,2017.气象因子对臭氧的影响及其在空气质量预报中的应用[J].中国环境监测,33(4):221-228.
张远航,郑君瑜,陈长虹,等,2020.中国大气臭氧污染防治蓝皮书(2020年)[C].北京:中国环境科学学会臭氧污染控制专业委员会.
赵燕,李大伟,翟宇虹,等,2022. 2014—2021年珠海市环境空气质量变化趋势及污染特征研究[J].环境科学与管理,47(12):144-149.
周广强,谢英,吴剑斌,等,2016.基于WRF-Chem模式的华东区域PM2.5预报及偏差原因[J].中国环境科学,8(9):2251-2259.
BASSETT R,YOUNG P J,BLAIR G S,et al,2020. A large ensemble approach to quantifying internal model variability within the WRF numerical model[J].J Geophys Res Atmos,125(7):e2019JD031286.
BRUNA J,ZAREMBA W,SZLAM A,et al,2013. Spectral networks and locally connected networks on graphs[EB/OL].arXiv:1312.6203.
CABANEROS S M,CALAUTIT J K,HUGHES B R,2019.A review of artificial neural network models for ambient air pollution prediction[J].Environ Model Softw,119(C):285-304.
EDER B,KANG D,MATHUR R,et al,2006. An operational evaluation of the Eta-CMAQ air quality forecast model[J].Atmos Environ,40(26):4894-4905.
GAO M,YIN L,NING J,2018. Artificial neural network model for ozone concentration estimation and Monte Carlo analysis[J].Atmos Environ,184:129-139.
HENZI A,ZIEGEL J F,GNEITING T,2021. Isotonic distributional regression[J].J R Stat Soc Ser B Stat Methodol,83(5):963-993.
HOCHREITER S,SCHMIDHUBER J,1997. Long short-term memory[J].Neural Comput,9(8):1735-1780.
KITAYAMA K,MORINO Y,YAMAJI K,et al,2019. Uncertainties in O3 concentrations simulated by CMAQ over Japan using four chemical mechanisms[J].Atmos Environ,198:448-462.
KRZYZANOWSKI M,COHEN A,2008. Update of WHO air quality guidelines[J].Air Qual Atmos Health,1(1):7-13.
LI M,LIU H,GENG G,et al,2017. Anthropogenic emission inventories in China:A review[J].Natl Sci Rev,4(6):834-866.
LU H,LYU X,CHENG H,et al,2019. Overview on the spatial-temporal characteristics of the ozone formation regime in China[J].Environ Sci Processe Impacts,21(6):916-929.
PAK U,KIM C,RYU U,et al,2018. A hybrid model based on convolutional neural networks and long short-term memory for ozone concentration prediction [J]. Air Qual Atmos Health,11(8):883-895.
QI Y,LI Q,KARIMIAN H,et al,2019. A hybrid model for spatiotemporal forecasting of PM2.5 based on graph convolutional neural network and long short-term memory[J].Sci Total Environ,664:1-10.
SUN L,LAN Y,JIANG R,2023. Using CNN framework to improve multi-GCM ensemble predictions of monthly precipitation at local areas:An application over China and comparison with other methods[J].J Hydrol,623:129866.
XIAO X,JIN Z,WANG S,et al,2022. A dual-path dynamic directed graph convolutional network for air quality prediction[J].Sci Total Environ,827:154298.
YU B,YIN H,ZHU Z,2017.Spatio-temporal graph convolutional networks:A deep learning framework for traffic forecasting[EB/OL].arXiv:1709.04875.
ZHANG J,DING W,2017. Prediction of air pollutants concentration based on an extreme learning machine:The case of Hong Kong[J].Int J Environ Res Public Health,14(2):114.
ZHANG Y,BOCQUET M,MALLET V,et al,2012. Real-time air quality forecasting,part I:History,techniques,and current status[J]. Atmos Environ,60:632-655.
0
浏览量
11
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构