Deep learning-based spatio-temporal prediction and uncertainty assessment of urban PM2.5 distribution

doi:10.11947/j.AGCS.2024.20230071

Abstract

Abstract:

The goal of predicting PM2.5 concentration is to achieve a comprehensive perception of the PM2.5 distribution in the study area based on limited observations. Ideal prediction models are required to ensure both high accuracy and high reliability of the results. However, most of the existing studies prioritize the efforts to improve accuracy, which ignores the uncertainty of results caused by data and model. This greatly limits the reliability and potential availability of high-precision prediction results, making it difficult to assist practical applications such as air pollution control effectively. To overcome this problem, this paper proposes a PM2.5 concentration spatiotemporal distribution prediction model with coupled uncertainty assessment. The prediction module, mainly based on graph convolutional and recurrent networks, achieves high-precision prediction of PM2.5 concentration. Meanwhile, the uncertainty quantification module based on adversarial learning strategies and variational autoencoder is constructed to synchronously reveal the uncertainty level of the prediction results. Extensive evaluations of real-world dataset show that the proposed model can effectively balance the accuracy and reliability of PM2.5 concentration prediction results, providing scientific decision-making support for environmental management.

Key words: PM2.5, deep learning, uncertainty, geographical prediction

CLC Number:

P208

Huimin LIU, Chenwei ZHANG, Kaiqi CHEN, Min DENG, Chong PENG. Deep learning-based spatio-temporal prediction and uncertainty assessment of urban PM2.5 distribution[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(4): 750-760.

Figures/Tables 11

Fig. 1

Fig. 2

Tab. 1

Fig. 3

Tab. 2

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Tab. 3

Fig. 8

References 25

[1]	王冬冬, 李洪利. 空气污染中的经济因素分析与环境保护可持续发展策略研究[J]. 环境科学与管理, 2022, 47(12):179-183.
	WANG Dongdong, LI Hongli. Analysis of economic factors in air pollution and research on sustainable development strategy of environmental protection[J]. Environmental Science and Management, 2022, 47(12):179-183.
[2]	孙志豪, 崔燕平. PM2.5对人体健康影响研究概述[J]. 环境科技, 2013, 26(4):75-78.
	SUN Zhihao, CUI Yanping. An overview of PM2.5 impacts on human health[J]. Environmental Science and Technology, 2013, 26(4):75-78.
[3]	生态环境部.“十四五”生态环境监测规划[EB/OL]. [2021-12-28]. https://www.mee.gov.cn/ywgz/sthjjcgl/wlgl/202201/P020220121523953377742.pdf.
	Ministry of Ecology and Environment. “14th Five-Year” ecological environment monitoring plan[EB/OL]. [2021-12-28]. https://www.mee.gov.cn/ywgz/sthjjcgl/wlgl/202201/P020220121523953377742.pdf.
[4]	白瑞. 环境空气质量监测的实现与优化[J]. 环境与发展, 2017, 29(7):183-184.
	BAI Rui. Implementation and optimization of ambient air quality monitoring[J]. Environment and Development, 2017, 29(7):183-184.
[5]	刘鹏华, 姚尧, 梁昊, 等. 耦合卡尔曼滤波和多层次聚类的中国PM2.5时空分布分析[J]. 地球信息科学学报, 2017, 19(4):475-485.
	LIU Penghua, YAO Yao, LIANG Hao, et al. Analyzing spatiotemporal distribution of PM2.5 in China by integrating Kalman filter and multilevel clustering[J]. Journal of Geo-Information Science, 2017, 19(4):475-485.
[6]	CHUANG M T, CHOU C C K, LIN N H, et al. A simulation study on PM2.5 sources and meteorological characteristics at the northern tip of taiwan in the early stage of the Asian haze period[J]. Aerosol and Air Quality Research, 2017, 17(12):3166-3178.
[7]	HUANG Yeqi, DENG Tao, LI Zhenning, et al. Numerical simulations for the sources apportionment and control strategies of PM2.5 over Pearl River Delta, China, part I: Inventory and PM2.5 sources apportionment[J]. Science of the Total Environment, 2018, 634:1631-1644.
[8]	WANG Qing, LUO Kun, FAN Jianren, et al. Spatial distribution and multiscale transport characteristics of PM2.5 in China[J]. Aerosol and Air Quality Research, 2019, 19(9):1993-2007.
[9]	薛文博, 王金南, 杨金田, 等. 国内外空气质量模型研究进展[J]. 环境与可持续发展, 2013, 38(3):14-20.
	XUE Wenbo, WANG Jinnan, YANG Jintian, et al. Domestic and foreign research progress of air quality model[J]. Environment and Sustainable Development, 2013, 38(3):14-20.
[10]	HU Xuefei, WALLER L A, AL-HAMDAN M Z, et al. Estimating ground-level PM2.5 concentrations in the southeastern U.S. using geographically weighted regression[J]. Environmental Research, 2013, 121:1-10.
[11]	LU Tianjun, BECHLE M J, WAN Yanyu, et al. Using crowd-sourced low-cost sensors in a land use regression of PM2.5 in 6 US cities[J]. Air Quality, Atmosphere & Health, 2022, 15(4):667-678.
[12]	ISKANDARYAN D, RAMOS F, TRILLES S. Graph neural network for air quality prediction: a case study in Madrid[J]. IEEE Access, 2023, 11:2729-2742.
[13]	HAN Jindong, LIU Hao, XIONG Haoyi, et al. Semi-supervised air quality forecasting via self-supervised hierarchical graph neural network[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(5):5230-5243.
[14]	尹文君, 张大伟, 严京海, 等. 基于深度学习的大数据空气污染预报[J]. 中国环境管理, 2015, 7(6):46-52.
	YIN Wenjun, ZHANG Dawei, YAN Jinghai, et al. Deep learning based air pollutant forecasting with big data[J]. Chinese Journal of Environmental Management, 2015, 7(6):46-52.
[15]	BILODEAU C, JIN W, JAAKKOLA T, et al. Generative models for molecular discovery: recent advances and challenges[J]. Wiley Interdisciplinary Reviews: Computational Molecular Science, 2022, 12(5):e1608.
[16]	ZHANG Lei, ZHU Axing, LIU Junzhi, et al. An adaptive uncertainty-guided sampling method for geospatial prediction and its application in digital soil mapping[J]. International Journal of Geographical Information Science, 2023, 37(2):476-498.
[17]	ZHENG Yu, LIU Furui, HSIEH H P. U-air: when urban air quality inference meets big data[C]//Proceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Chicago: ACM Press, 2013: 1436-1444.
[18]	毛文婧, 王卫林, 焦利民, 等. 基于深度学习的中国连续空间覆盖PM2.5浓度预报[J]. 测绘学报, 2022, 51(3):361-372. DOI: 10.11947/j.AGCS.2022.20200385.
	MAO Wenjing, WANG Weilin, JIAO Limin, et al. Continuous spatial coverage PM2.5 concentration forecast in China based on deep learning[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(3):361-372. DOI: 10.11947/j.AGCS.2022.20200385.
[19]	邓敏, 谌恺祺, 石岩, 等. 多尺度空间同位模式挖掘的点过程分解方法[J]. 测绘学报, 2022, 51(2):258-268. DOI: 10.11947/j.AGCS.2022.20200548.
	DENG Min, CHEN Kaiqi, SHI Yan, et al. Point process decomposition method for multi-scale spatial co-location pattern mining[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(2):258-268. DOI: 10.11947/j.AGCS.2022.20200548.
[20]	CEMGIL T, GHAISAS S, DVIJOTHAM K, et al. The autoencoding variational autoencoder[J]. Advances in Neural Information Processing Systems, 2020, 33:15077-15087.
[21]	TESCHE T W, BERGSTROM R W. Use of digital terrain data in meteorological and air quality modelling[J]. Photogrammetric Engineering Remote Sensing, 1978, 44(12):1549-1559.
[22]	CHEN Kaiqi, DENG Min, SHI Yan. A temporal directed graph convolution network for traffic forecasting using taxi trajectory data[J]. ISPRS International Journal of Geo-Information, 2021, 10(9):624.
[23]	何占军, 邓敏, 蔡建南, 等. 顾及背景知识的多事件序列关联规则挖掘方法[J]. 武汉大学学报(信息科学版), 2018, 43(5):766-772.
	HE Zhanjun, DENG Min, CAI Jiannan, et al. A context-based association rules mining method for multiple event sequences[J]. Geomatics and Information Science of Wuhan University, 2018, 43(5):766-772.
[24]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45:5-32.
[25]	WANG Jinfeng, STEIN A, GAO Binbo, et al. A review of spatial sampling[J]. Spatial Statistics, 2012, 2:1-14.

数据源	数据类型	统计信息
PM2.5	站点数量	11
	时间间隔/h	1
	时间范围	2016-09-01—2016-09-30
POI	POI数量	574 361
路网	路网总长度/km	11 703
	高速路段长度/km	701
	交叉口数量	40 933
DSM	栅格面积/km²	1 942.7
气象数据	时间范围	2016-09-01—2016-09-30
浮动车轨迹	轨迹数量/条	246 982
	时间间隔/min	5
	时间范围	2016-09-01—2016-09-30

指标	RF	GCN	GRU	本文方法
RMSE	7.864	7.399	6.843	6.340
MAE	5.289	5.076	4.959	4.580
R²	0.762	0.774	0.799	0.854

增加站点	RMSE	MAE	R²	预测不确定性
—	13.289	9.379	0.311	0.822
1360A	13.173	9.131	0.322	0.811
1360A、1357A	11.590	8.561	0.476	0.732
1360A、1357A、1359A	8.359	5.853	0.727	0.688