ESCAP：交通擁擠對空氣污染物影響的建模工作文件（英文版）（31頁）.pdf

《ESCAP：交通擁擠對空氣污染物影響的建模工作文件（英文版）（31頁）.pdf》由會員分享，可在線閱讀，更多相關《ESCAP：交通擁擠對空氣污染物影響的建模工作文件（英文版）（31頁）.pdf（31頁珍藏版）》請在三個皮匠報告上搜索。

1、MODELING TRAFFIC CONGESTION EFFECTS ON AIR POLLUTANTSWORKING PAPERunsplashDisclaimer:The designations employed and the presentation of the material in this policy brief do not imply the expression of any opinion whatsoever on the part of the Secretariat of the United Nations concerning the legal sta

2、tus of any country,territory,city or area,or of its authorities,or concerning the delimitation of its frontiers or boundaries.Where the designation“country or area”appears,it covers countries,territories,cities or areas.Bibliographical and other references have,wherever possible,been verified.The Un

3、ited Nations bears no responsibility for the availability or functioning of URLs.The opinions,figures and estimates set forth in this publication should not necessarily be considered as reflecting the views or carrying the endorsement of the United Nations.The mention of firm names and commercial pr

4、oducts does not imply the endorsement of the United Nations.The shaded areas of the map indicate ESCAP members and associate members.*The designations employed and the presentation of material on this map do not imply the expression of any opinion whatsoever on the part of the Secretariat of the Uni

5、ted Nations concerning the legal status of any country,territory,city or area or of its authorities,or concerning the delimitation of its frontiers or boundaries.Acknowledgement:This policy brief on“Modelling Traffic Congestion Effects on Air Pollutants”has been issued in January 2023.Preparation of

6、 this paper is coordinated by the Economic and Social Commission for Asia and the Pacific,through its Environment and Development Division,under the supervision of Sangmin Nam,Director,Environment and Development Division,Curt Garrigan,Chief,Environment and Development Policy Section and authorship

7、by Matthew Perkins,Environmental Affairs Officer,Divya Nair and Vinayak Dixit.This document also benefitted from contributions by Abigail Smith and Mervin Chin,whose inputs were integral to the production of this report.Financial support by the Republic of Korea is gratefully acknowledged.For furthe

8、r information on this policy brief,please address your inquiries to:Environment and Development Division United Nations Economic and Social Commission for Asia and the Pacific(ESCAP)Email:ESCAP-EDDun.org United Nations publication Copyright United Nations 2023 Abstract 2Methods 3 Study Area 3 Data 3

9、 Road Network 3 Speed Data 4 Air Pollutants 4 Modeling 6 Work Flow 6 Congestion Index 7 Correlation Analysis 8 Linear Regression 8 Copula Function 9Results 11 Correlation Analysis Results 11 Linear Regression Results 13 Morning Peak Hours 13 Evening Peak Hours 14 Copula Results 15References 16 Appen

10、dix 171Table of Contents2Air pollution is universally recognised as one of the most pressing environmental challenges in the Asia-Pacific region.In recent years,this impact has risen at an alarming rate and has resulted in an increased premature death and threatened the livelihoods and sustainable d

11、evelopment in region.This trend is expected to continue,particularly in fast-growing cities with urban population exponentially.Though biomass burning is a significant air pollution source in Chiang Mai within the northern region,traffic-related air pollution has also become one of the major sources

12、 of air pollutants such as NO2,CO,and SO.This paper comprehensively examines the relationship between traffic congestion effect and ambient air pollution through conducting quantitative modelling,correlational analysis,linear regression analysis,and Copula analysis.Results from the hourly air pollut

13、ant concentration data from two stations in Chiang Mai from 2017 to 2021 reveal that Congestion Index(CI)is positively correlated with air pollutants during morning and evening peak hours.This positive dependence indicated that traffic congestion does lead to higher air pollutant concentrations duri

14、ng rush hours.ABSTRACTunsplashChiang Mai is the second largest city in Thailand.The Thailand Meteorological Department1 divides the climate into three seasons:hot(summer)season(March to May),rainy season(May to October),and cool(winter)season (November to February).There are two rush hour periods(7:

15、00 8:00a.m.and 5:00 6:00p.m.).It is found that the traffic contributed approximately 90 per cent of total CO emissions during rush hours2.3Figure 1:The road network of study area(Chiang Mai)The road network of Chiang Mai including the latitude and longitude coordinates of each link,speed limit,lengt

16、h,name,highway types is obtained from OpenStreetMap(OSM)data sets3.There are seven main highway types in OSM,namely motorway,trunk,primary,secondary,tertiary,unclassified,and residential.The process of ac-quiring road networks data is detailed in the Appendix download_network.py.METHODSStudy AreaDAT

17、ARoad Network4The speed data of Chiang Mai is collected from Intelligent Traffic Information Center(iITC).The probe-data provides vehicle unique ID,latitude and longitude coordinates,time stamp and speed from 2017 to 2021.By using the vehicle latitude and longitude coordinates,we can determine which

18、 road the vehicle is on and the speed limit of the road.The example of speed data:We collected hourly air pollutant concentration data from two stations in Chiang Mai from 2017 to 2021(Figure 2).The atmospheric pollutants are PM2.5,PM10,CO,NO2,SO2,O3.Table 1 The example of air pollutant data(contd o

19、n next page)Speed DataAir Pollutants5Figure 2:The locations of air quality monitoring stations in Chiang Mai6Firstly,as shown in Figure 3,the correlation analysis is conducted to find the relationship between CI and air pollutants for 24 hours a day of each month.The results indicate that during the

20、 morning and evening peak hours,most air pollutants are positively correlated with CI,i.e.,the more congested the road,the higher the air pollution concentration.During the off-peak period,most air pollutants and CI become weakly positively or negatively correlated.Therefore,in the second step,linea

21、r regression is applied only to the peak period between air pollutants and CI.We find that the during peak hours,traffic congestion does lead to higher air pollutant concentrations,but the R2 values are small.Finally,the Copula functions are used to find the dependency between CI and air pollutants

22、during peak hours for each group each month.Figure 3:The flow chart of the CI and air pollutants analysisMODELINGWork Flow7Congestion Index(CI)4 is adopted to represent the beginning of the delay by comparing the current traffic speed and the free-flow speed.The free-flow speed(Vf)is defined as the

23、maximum speed during the off-peak period.As shown in Equation(1),the congestion index of a link i at time t(CIit)is computed as the ratio of its delay and the free flow travel time/km.Because the station-level CI is calculated by averaging the CI of the links around the station,the distance from the

24、 station to the link significantly affects the value of the station-level CI.We set the distance from the station to the link as 400m,800m,1000m and 2000m,and the correlation analysis results are used to determine the appropriate distance.Congestion Index8Figure 4:Different distances between station

25、s and linksThe correlations between CI and air pollutants are investigated by using Pearson correlation coefficient(Equation(2).Correlation analysis is carried out with hourly air pollutant concentrations and CI for each month.where x and y are the means of x and y variables.The linear regression be

26、tween CI and air pollutants are used in this study.The linear regression is conducted with rush hour air pollutant concentrations and CI for each month,and R2 is used to measure the fitness.Congestion IndexLinear Regression9Copulas5 are functions that associates multivariate distribution functions o

27、f random variables with their one-dimensional marginal distribution functions.The dependence structures are measured by correlation coefficients:Pearsons product-moment correlation coefficient,Kendalls tau,and Spearmans rho.Consider a d-dimensional cumulative distribution function(CDF),F,with margin

28、als F1,.,Fd.Then there exists a copula,C,such thatfor allThe marginal distributions of air pollutants and CI are fitted by the empirical cumulative distribution function(CDF),i.e.,the data(u1,u2)are transformed as the uniformly distributed random variables with support on 0,1.The proposed study uses

29、 Gaussian,Clayton,Frank,Gumbel and Joe copulas functions.Then,the best fitted copula is selected based on the log-likelihood and Akaike information criterion(AIC).Bivariate Gaussian copulaThe bivariate Gaussian copula is one of the most commonly used elliptical copulas.The function with correlation

30、function is given by:where is the standard normal distribution and is the Pearsons liner correlation coefficient.The Frank copulaThe Frank copula function6 is defined as:The generator function is expressed as .The correlation between the random variables can be positive and negative.Copula Functions

31、10The Clayton copulaThe Clayton copula function7 is defined as:The generator function is expressed as .The Clayton copula only accounts for the positive dependence,especially for the lower tail of the random variables.The Gumbel copulaThe Gumbel copula function8 is expressed as:The generator functio

32、n is expressed as .Compared with the Clayton copula,The Gumbel copula only account for the positive dependence,but especially for the upper tail of the random variables.The Joe copulaThe Joe copula function is proposed by Joe9,10 with the generator function and the function is expressed as:Similar w

33、ith Gumbel copula,the Joe copula only accommodates positive dependence especially for the upper tail of the random variables.11RESULTSThe correlation coefficient is used to test the correlation between individual air pollutants(PM2.5,PM10,CO,NO2,SO2,O3,CO)and CI on an hourly basis per month.By compa

34、ring the different distances from the station to the links(400m,800m,1000m,2000m),we found that 1000m is a suitable threshold.As shown in Figure 5,except for Figure 5a,all plots show roughly two peaks of correlation coefficient between PM2.5 and CI,(b)at 10am and 20pm,(c)at 8am and 20pm,(d)at 6am an

35、d 22pm.Since the morning and evening peak time periods are 7am-8am and 17pm-18pm,the correlation between CI and PM2.5 is better represented in Figure 5c.5a5b5c5dFigures 5a,5b,5c,5d:Correlation results Notes:The correlation results with different distances between station and links(a)400m,(b)800m,(c)

36、1000m,(d)2000mCorrelation Analysis Results12In Figure 5d,there is an unreasonable fluctuation at 18 pm-20pm(0.11 at 18pm,-0.4 at 19pm,0.09 at 20pm),probably due to the long setting distance,the CI of other links affects the value of 35t station setting.The distance of 1000m between station and links

37、 is selected in the following analysis.Figure 6 The correlation analysis results for different months at different stationsAs shown in the Figure 6,the result show that CI and air pollutants are positively correlated during peak hours and are not correlated or even negatively correlated during off-p

38、eak hours.More correlation analysis results can be found in the Appendix.The results confirm our intuition that the more congested the traffic is,the more air pollutants are emitted by vehicles.Therefore,we did a linear regression analysis of CI and air pollutants in the morning and evening peak per

39、iods.13Linear Regression ResultsThe morning peak hour is set to 7am to 8am.As shown in Figure 7,the air pollutants such as PM10,O3,PM2.5 are showing the positive correlation with CI.The range of R-squared is 0.02 to 0.06,indicating that the linear relationship between air pollutants and CI is insign

40、ificant,which may be mainly due to the lack of data from other regional stations and the influence of weather.Figure 7:The morning peak hour linear regression resultsMorning Peak Hours14The evening peak hour is set to 17pm to 18pm.As shown in Figure 8,the air pollutants such as PM2.5,PM10 and NO2 ar

41、e showing the positive correlation with CI.Figure 8:The evening peak hour linear regression resultsEvening Peak Hours15Copula ResultsCopulas are functions that associates multivariate distribution functions of random variables with their one-dimensional marginal distribution functions.The marginal d

42、istributions of air pollutants and CI are fitted by the empirical cumulative distribution function(CDF),i.e.,the data(u1,u2)are transformed as the uniformly distributed random variables with support on 0,1.As shown in Figure 9,the first two plots are scatterplots of the original data being transform

43、ed into a uniform distribution and the fitted scatterplot,respectively,and the third is the isolines of probability density of the empirical copula functions and the estimated ones.Taking CI and PM2.5 as an example,the results indicate that there is no significant relationship between CI and air pol

44、lutant.Figure 9:Copula results161 TMD,“The Climate of Thailand.”2015.Online.Available:https:/www.tmd.go.th/en/arch ive/thailand_climate.pdf2 S.Sathitkunarat,P.Wongwises,R.Pan-Aram,and M.Zhang,“Carbon monoxide emission and concentration models for Chiang Mai urban area,”Advances in Atmospheric Scienc

45、es,vol.23,no.6,pp.901908,2006.3 O.Contributors,“OpenStreetMap.2019,”URL http:/planet.openstreetmap.org.Online,2019.4 D.J.Nair,F.Gilles,S.Chand,N.Saxena,and V.Dixit,“Characterizing multicity urban traffic conditions using crowdsourced data,”PLoS ONE,vol.14,no.3,p.e0212845,Mar.2019,doi:10.1371/journal

46、.pone.0212845.5 M.Sklar,“Fonctions de repartition an dimensions et leurs marges,”Publ.inst.statist.univ.Paris,vol.8,pp.229231,1959.6 M.J.Frank,“On the simultaneous associativity ofF(x,y)andx+y F(x,y),”Aequationes mathematicae,vol.19,no.1,pp.194226,1979.7 D.G.Clayton,“A model for association in bivar

47、iate life tables and its application in epidemiological studies of familial tendency in chronic disease incidence,”Biometrika,vol.65,no.1,pp.141151,1978.8 E.J.Gumbel,“Bivariate exponential distributions,”Journal of the American Statistical Association,vol.55,no.292,pp.698707,1960.9 H.Joe,“Parametric

48、 families of multivariate distributions with given margins,”Journal of multivariate analysis,vol.46,no.2,pp.262282,1993.10 H.Joe,Multivariate models and multivariate dependence concepts.CRC Press,1997.References17AppendixCorrelation analysis results1819Linear analysis results202122Copula function re

49、sults23Air quality monitoring stationsScripts24download_network.pyOutputDirectory outputExample Usage./download_network.py-north_lat 14.17-west_lon 99.96-south_lat 13.46-east_lon 101.033-save_path network_files/mapper.py25OutputDirectory outputFile outputOptional argument outputDirectory output26Exa

50、mple Usage./mapper.py-nodes_shapefile./network_files/nodes.shp-edges_shapefile./network_files/edges_w_stations.shp-data_directory./data/Generating edge shapefiles with station information./mapper.py-nodes_shapefile./network_files/nodes.shp-edges_shapefile./network_files/edges.shp-station_list./station_location.csv-edge_with_station_save_path./network_files/edges_w_stations.shpmapped_CI.pyOutput27CI_air.pyOutputcorrelation_analysis.pyOutput28regression_analysis.pyOutputcopula.r29Output