基于分布式感知深度神经网络的高分辨率PM2.5值估算

doi:10.11821/dlxb202101015

Abstract

Abstract:

The individual exposure level of fine particulate matter (PM_2.5) is a key issue in the study of health effects. However, the lack of historical data and the small coverage of ground monitoring stations have hindered the development of related research. Based on the aerosol optical depth (AOD) provided by NASA remote sensing data, multi-source data such as ground monitoring and meteorological data were integrated for modeling to estimate near-ground PM_2.5 concentration. The results have wide spatial coverage, strong time continuity and low method cost. Based on the Beijing-Tianjin-Hebei-Shandong region in 2018, this paper introduces 50 features such as meteorological elements, NDVI, time nodes and spatial markers to analyze the relationship between AOD and PM_2.5. In view of the information loss caused by the single traditional interpolation method, the spatiotemporal multi-view interpolation method is used to improve the accuracy and coverage of interpolation. Considering the complex relationship caused by hysteresis of features, and the correlation and partial correlation between features, this paper uses a distributed perception deep neural network model (DP-DNN) to separately capture higher-order features between multiple-source features. The results show that: (1) The relative error of the spatiotemporal multi-view interpolation method is 27.5%, and the average proportion of missing data decreases from 52.1% to 4.84%. (2) In terms of time prediction, mean absolute error, relative error, mean square error and root mean square error of DP-DNN are 7.7 μg/m ³, 46.8%, 766.2 μg ²/m ⁶ and 26.9 μg/m ³, respectively. In terms of space prediction, they are 16.6 μg/m ³, 41.8%, 691.5 μg ²/m ⁶ and 26.6 μg/m ³, respectively. In aspects of accuracy, robustness, generalization ability and time consuming, the results are superior to linear statistical models and common deep learning architecture.

Key words: aerosol optical depth (AOD), PM_2.5 prediction, multi-view interpolation, distributed perception, deep learning, prediction of spatiotemporal migration

LIU Jiwei, MIN Suqin, JIN Mengdi. High resolution PM_2.5 estimation based on the distributed perception deep neural network model[J].Acta Geographica Sinica, 2021, 76(1): 191-205.

Figures/Tables 11

Tab. 1

Fig. 1

Fig. 2

Fig. 3

Tab. 2

Tab. 3

Fig. 4

Fig. 5

Tab. 4

Fig. 6

Fig. 7

References 48

[1]	Geng G N, Zhang Q, Martin R V, et al. Estimating long-term PM_2.5 concentrations in China using satellite-based aerosol optical depth and a chemical transport model. Remote Sensing of Environment, 2015,166:262-270.
[2]	Lee D, Mukhopadhyay S, Rushworth A, et al. A rigorous statistical framework for spatio-temporal pollution prediction and estimation of its long-term impact on health. Biostatistics, 2017,18(2):370-385.
[3]	Mukhopadhyay S, Sahu S K. A Bayesian spatiotemporal model to estimate long-term exposure to outdoor air pollution at coarser administrative geographies in England and Wales. Journal of the Royal Statistical Society: Series A (Statistics in Society), 2018,181(2):465-486.
[4]	Fang X, Zou B, Liu X, et al. Satellite-based ground PM_2.5 estimation using timely structure adaptive modeling. Remote Sensing of Environment, 2016,186:152-163.
[5]	van Donkelaar A, Martin R V, Brauer M, et al. Global estimates of fine particulate matter using a combined geophysical-statistical method with information from satellites, models, and monitors. Environmental Science & Technology, 2016,50(7):3762-3772. doi: 10.1021/acs.est.5b05833 pmid: 26953851
[6]	He Q Q, Huang B. Satellite-based mapping of daily high-resolution ground PM_2.5 in China via space-time regression modeling. Remote Sensing of Environment, 2018,206:72-83.
[7]	Wang Z F, Chen L F, Tao J H, et al. Satellite-based estimation of regional particulate matter (PM) in Beijing using vertical-and-RH correcting method. Remote Sensing of Environment, 2010,114(1):50-63.
[8]	Fu D S, Xia X G, Wang J, et al. Synergy of AERONET and MODIS AOD products in the estimation of PM_2.5 concentrations in Beijing. Scientific Reports, 2018,8(1):10174-10181. doi: 10.1038/s41598-018-28535-2 pmid: 29977000
[9]	Wang J, Christopher S A. Intercomparison between satellite-derived aerosol optical thickness and PM_2.5 mass: Implications for air quality studies. Geophysical Research Letters, 2003,30(21):2095-2098.
[10]	Engel-Cox J A, Holloman C H, Coutant B W, et al. Qualitative and quantitative evaluation of MODIS satellite sensor data for regional and urban scale air quality. Atmospheric Environment, 2004,38(16):2495-2509. doi: 10.1016/j.atmosenv.2004.01.039
[11]	Jia Songlin, Su Lin, Tao Jinhua, et al. A study of multiple regression method for estimating concentration of fine particulate matter using satellite remote sensing. China Environmental Science, 2014,34(3):565-573.
	[ 贾松林, 苏林, 陶金花, 等. 卫星遥感监测近地表细颗粒物多元回归方法研究. 中国环境科学, 2014,34(3):565-573.]
[12]	Zang Z L, Wang W Q, Cheng X H, et al. Effects of boundary layer height on the model of ground-level PM_2.5 concentrations from AOD: Comparison of stable and convective boundary layer heights from different methods. Atmosphere, 2017,8:104-121.
[13]	Gupta P, Christopher S A. Particulate matter air quality assessment using integrated surface, satellite, and meteorological products: Multiple regression approach. Journal of Geophysical Research Atmospheres, 2009,114:D14205. DOI: 10.1029/2008JD011497.
[14]	Koelemeijer R B A, Homan C D, Matthijsen J. Comparison of spatial and temporal variations of aerosol optical thickness and particulate matter over Europe. Atmospheric Environment, 2006,40(27):5304-5315.
[15]	Chen Hui, Li Qing, Zhang Yuhuan, et al. Estimations of PM_2.5 concentrations based on the method of geographically weighted regression. Acta Scientiae Circumstantiae, 2016,36(6):2142-2151.
	[ 陈辉, 厉青, 张玉环, 等. 基于地理加权模型的我国冬季PM_2.5遥感估算方法研究. 环境科学学报, 2016,36(6):2142-2151.]
[16]	He Q Q, Huang B. Satellite-based high-resolution PM_2.5 estimation over the Beijing-Tianjin-Hebei region of China using an improved geographically and temporally weighted regression model. Environmental Pollution, 2018,236:1027-1037. doi: 10.1016/j.envpol.2018.01.053 pmid: 29455919
[17]	Chen Hui, Li Qing, Wang Zhongting, et al. Utilization of MERSI and MODIS data to monitor PM_2.5 concentration in Beijing-Tianjin-Hebei and its surrounding areas. Journal of Remote Sensing, 2018,22(5):822-832.
	[ 陈辉, 厉青, 王中挺, 等. MERSI和MODIS卫星监测京津冀及周边地区PM_2.5浓度. 遥感学报, 2018,22(5):822-832.]
[18]	Jiang M, Sun W W, Yang G, et al. Modelling seasonal GWR of daily PM_2.5 with proper auxiliary variables for the Yangtze River Delta. Remote Sensing, 2017,9(4):346. DOI: 10.3390/rs9040346.
[19]	Wu J S, Yao F, Li W F, et al. VIIRS-based remote sensing estimation of ground-level PM_2.5 concentrations in Beijing-Tianjin-Hebei: A spatiotemporal statistical model. Remote Sensing of Environment, 2016,184:316-328.
[20]	Wu Y R, Guo J P, Zhang X Y, et al. Synergy of satellite and ground based observations in estimation of particulate matter in eastern China. Science of the Total Environment, 2012,433:20-30.
[21]	Yao L, Lu N. Spatiotemporal distribution and short-term trends of particulate matter concentration over China, 2006-2010. Environmental Science and Pollution Research, 2014,21(16):9665-9675. pmid: 24824505
[22]	Shao Qi, Chen Yunhao, Li Jing. Inversion of PM_2.5 concentration in Beijing based on satellite remote sensing and meteorological reanalysis data. Geography and Geo-Information Science, 2018,34(3):32-38.
	[ 邵琦, 陈云浩, 李京. 基于卫星遥感和气象再分析资料的北京市PM_2.5浓度反演研究. 地理与地理信息科学, 2018,34(3):32-38.]
[23]	Zou B, Chen J W, Zhai L, et al. Satellite based mapping of ground PM_2.5 concentration using generalized additive modeling. Remote Sensing, 2016,9(1):1. DOI: 10.3390/rs9010001.
[24]	Kloog I, Koutrakis P, Coull B A, et al. Assessing temporally and spatially resolved PM_2.5 exposures for epidemiological studies using satellite aerosol optical depth measurements. Atmospheric Environment, 2011,45(35):6267-6275.
[25]	Sorek-Hamer M, Strawa A W, Chatfield R B, et al. Improved retrieval of PM_2.5 from satellite data products using non-linear methods. Environmental Pollution, 2013,182:417-423. doi: 10.1016/j.envpol.2013.08.002 pmid: 23995022
[26]	Li T W, Shen H F, Zeng C, et al. Point-surface fusion of station measurements and satellite observations for mapping PM_2.5 distribution in China: Methods and assessment. Atmospheric Environment, 2017,152:477-489.
[27]	Lyu H, Dai T, Zheng Y F, et al. Estimation of PM_2.5 concentrations over Beijing with MODIS AODs using an artificial neural network. Scientific Online Letters on the Atmosphere, 2018,14:14-18.
[28]	Zou B, Wang M, Wan N, et al. Spatial modeling of PM_2.5 concentrations with a multifactoral radial basis function neural network. Environmental Science and Pollution Research, 2015,22(14):10395-10404. doi: 10.1007/s11356-015-4380-3 pmid: 25813644
[29]	Yao L, Lu N, Jiang S. Artificial Neural Network (ANN) for multi-source PM_2.5 estimation using surface, MODIS, and meteorological data. Proceedings of the 2012 International Conference on Biomedical Engineering and Biotechnology, 2012: 1228-1231.
[30]	Mei Bo, Tian Maozai. Analysis of influential factors on PM_2.5 in Beijing based on spatio-temporal model. Journal of Applied Statistics and Management, 2018,37(4):571-586.
	[ 梅波, 田茂再. 基于时空模型北京市PM_2.5浓度影响因素研究. 数理统计与管理, 2018,37(4):571-586.]
[31]	Ellett F S Jr, Ericson D P. Correlation, partial correlation, and causation. Synthese, 1986,67(2):157-173.
[32]	Wang W L, Zhao S L, Jiao L M, et al. Estimation of PM_2.5 concentrations in China using a spatial back propagation neural network. Scientific Reports, 2019,9:13788. DOI: 10.1038/s41598-019-50177-1. doi: 10.1038/s41598-019-50177-1 pmid: 31551510
[33]	Wang X Q, Wang F, Jia L L, et al. Retrieval and validation of aerosol optical depth using GF-1 WFV cameras data. Advances in Space Research, 2020,65(3):997-1007.
[34]	Yu W J, Nan Z T, Wu T H, et al. An effective interpolation method for MODIS land surface temperature on the Qinghai-Tibet Plateau. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015,8(9):4539-4550.
[35]	Safarpour S, Abdullah K, Lim H S, et al. Spatial interpolation of aerosol optical depth pollution: Comparison of methods for the development of aerosol distribution. ISPRS: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2017,4(4):237-244.
[36]	Sathe Y, Kulkarni S, Gupta P, et al. Application of moderate resolution imaging spectroradiometer (MODIS) aerosol optical depth (AOD) and weather research forecasting (WRF) model meteorological data for assessment of fine particulate matter (PM_2.5) over India. Atmospheric Pollution Research, 2019,10(2):418-434.
[37]	Beckerman B S, Jerrett M, Serre M, et al. A hybrid approach to estimating national scale spatiotemporal variability of PM_2.5 in the contiguous United States. Environmental Science & Technology, 2013,47(13):7233-7241. doi: 10.1021/es400039u pmid: 23701364
[38]	Chow J C, Chen L W A, Watson J G, et al. PM_2.5 chemical composition and spatiotemporal variability during the California Regional PM₁₀/PM_2.5 Air Quality Study (CRPAQS). Journal of Geophysical Research: Atmospheres, 2006,111(10). DOI: 10.1029/2005jd006457.
[39]	Lin G, Fu J, Jiang D, et al. Spatio-temporal variation of PM_2.5 concentrations and their relationship with geographic and socioeconomic factors in China. International Journal of Environmental Research and Public Health, 2013,11(1):173-186. pmid: 24362546
[40]	Gelencsér A, May B, Simpson D, et al. Source apportionment of PM_2.5 organic aerosol over Europe: Primary/secondary, natural/anthropogenic, and fossil/biogenic origin. Journal of Geophysical Research: Atmospheres, 2007, 112:17D23S043. DOI: 10.1029/2006JD008094.
[41]	Liu Y, Paciorek C J, Koutrakis P. Estimating regional spatial and temporal variability of PM_2.5 Concentrations using satellite data, meteorology, and land use information. Environmental Health Perspectives, 2009,117(6):886-892. doi: 10.1289/ehp.0800123 pmid: 19590678
[42]	Kloog I, Nordio F, Coull B A, et al. Incorporating local land use regression and satellite aerosol optical depth in a hybrid model of spatiotemporal PM_2.5 exposures in the Mid-Atlantic states. Environmental Science & Technology, 2012,46(21):11913-11921. doi: 10.1021/es302673e pmid: 23013112
[43]	National Oceanic and Atmospheric Administration Weather Prediction Center. The Heat Index Equation[EB/OL]. (2014-05-28) [2019-08-13]. http://www.wpc.ncep.noaa.gov/html/heatindex_equation.shtml.
[44]	National Aeronautics and Space Administration. MODIS Standard Collection 6.1 Update Executive Summary[EB/OL]. (2017-08-09) [2019-08-20]. https://atmosphere-imager.gsfc.nasa.gov/documentation/collection-61.
[45]	Dark Sky Docs: Data Sources[EB/OL]. (2019-09-04) [2019-08-24]. https://darksky.net/dev/docs/sources.
[46]	Tobler W R. A computer movie simulating urban growth in the Detroit Region. Economic Geography, 1970,46:234-240.
[47]	Swami A, Jain R. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 2013,12(10):2825-2830.
[48]	Wang X X, Brown D J. Boosting orthogonal least squares regression. Lecture Notes in Computer Science, 2004,3177:678-683.

类别	特征	单位	特征描述
PM_2.5特征	PM_2.5	μg/m³	近地面细颗粒物空气动力学粒径小于等于2.5 μm的颗粒物质量浓度的全日平均值
AOD特征	AOD		MODIS暗像元与深蓝算法融合提取0.55 μm处气溶胶光学厚度
NDVI特征	NDVI		MODIS近红外波段反射值与红光波段反射值之差比上两者之和
气象特征	温度	℃	近地面空气气温的全日平均值
	体感温度^[43]	℃	人体所感受到的冷暖程度,转换成同等温度的全日平均值
	温差	℃	给定日期内近地面空气气温最大温差
	日间温差	℃	给定日期内白天时段近地面空气气温最大温差
	体感温差	℃	给定日期内体感气温最大温差
	日间体感温差	℃	给定日期内白天时段体感气温最大温差
	云量		介于0和之间1的被云遮挡的天空百分比的全日平均值
	露点	℃	空气中所含的气态水达到饱和而凝结成液态水所需要降至的温度的全日平均值
	相对湿度		近地面相对湿度的全日平均值
	日照长度	h	给定日期的日照长度
	能见度	km	平均能见度的全日平均值
	阵风速度	m/s	近地面的阵风速度的全日平均值
	风速	m/s	近地面的风速的全日平均值
	风速角	°	以角度为单位的风的来向,取全日平均风向
	气压	hPa	面积上从海平面到大气上界空气柱的重量的全日平均值
	积雪强度	cm	积雪强度的全日平均值
	降雨强度	cm/h	降雨强度的全日平均值
时滞特征			滞后一日的气溶胶光学厚度、气温、云量、露点、相对湿度、日照长度、能见度、风速、气压、积雪强度、降雨强度
空间标识	经纬度		监测点站点经纬度
时间节点	月份		给定日期所在月份

评价标准	多视图		时间视图		空间视图		局部视图		全局视图
评价标准	RE(%)	MAE(μg/m³)	RE(%)	MAE(μg/m³)	RE(%)	MAE(μg/m³)	RE(%)	MAE(μg/m³)	RE(%)	MAE(μg/m³)
PM_2.5	21.3	7.3	68.0	24.7	19.1	6.8	17.8	5.9	16.2	5.2
AOD	24.6	113.4	76.8	283.9	12.3	60.2	20.9	110.6	22.6	76.2
AODs	52.7	136.6	104.5	357.2	50.6	116.7	33.7	116.4	87.6	151.2
NDVI	26.4	0.05	49.8	0.12	12.9	0.02	20.0	0.03	12.3	0.02
气象	12.5	\	36.3	\	15.5	\	15.5	\	21.8	\
平均结果	27.5	\	67.8	\	22.1	\	21.5	\	32.1	\

插补方法	插补前	多视图	时间视图	空间视图	局部视图	全局视图
AOD缺失	71.7	0.0	0.4	1.5	2.8	0.0
AODs平均缺失	66.3	0.0	2.9	21.6	42.1	0.0
NDVI缺失	58.6	1.0	24.3	52.2	55.5	34.4
气象数据平均缺失	48.3	22.7	27.3	29.2	37.5	29.4
PM_2.5缺失	15.9	0.5	10.5	3.8	15.5	10.5
平均结果	52.1	4.84	13.1	21.7	30.7	14.9

		DP-DNN			LSTM	GWR	B-OLSR	EN
		地域迁移检验
MAE	16.6		17.2	18.5	22.1	29.1	19.2	21.3
MAE_std	1.1		1.2	1.6	0.9	41.0	0.4	0.6
RE(%)	41.80		44.7	44.6	76.1	141.1	56.6	67.0
RE_std(%)	4.90		4.72	5.0	13.7	141.7	4.3	5.7
MSE	691.5		744.2	901.2	1026.6	63086.4	788.6	1008.9
MSE_std	100.4		103.7	181.1	176.1	275336.9	49.0	81.0
RMSE	26.6		27.2	29.9	31.9	100.3	28.1	31.7
RMSE_std	1.9		1.9	2.9	2.6	230.3	0.9	1.3
		时间预测检验
MAE	17.7		18.9	22.4	25.7	33.7	53.8	21.9
MAE_std	3.7		5.0	7.4	4.7	2.7	12.0	4.3
RE(%)	46.8		50.9	61.9	87.4	108.3	211.2	70.7
RE_std(%)	7.2		11.7	20.8	18.8	26.9	71.3	10.9
MSE	766.2		837.4	1126.8	1645.7	2228.3	3966.7	1038
MSE_std	355.6		446.7	789.5	712.1	532.6	1300	489.7
RMSE	26.9		27.4	31.9	39.6	46.9	62.1	31.3
RMSE_std	6.1		5.0	10.3	8.9	5.1	10.5	7.6
平均耗时(s)	4.2		3.1	3.3	26.4	3389.0	1.3	0.5

High resolution PM_2.5 estimation based on the distributed perception deep neural network model

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 11

References 48

Related Articles 0

Metrics

Comments

High resolution PM2.5 estimation based on the distributed perception deep neural network model

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 11

References 48

Related Articles 0

Metrics

Comments

High resolution PM_2.5 estimation based on the distributed perception deep neural network model