View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Elsevier - Publisher Connector AtmosphericPollutionResearch5(2014)759Ͳ768 AAtm spheric PPollution RResearch www.atmospolres.com The 2013 severe haze over the Southern Hebei, China: PM2.5 composition and source apportionment ZheWei1,Li–TaoWang1,Ming–ZhangChen2,YanZheng3 1DepartmentofEnvironmentalEngineering,SchoolofCityConstruction,HebeiUniversityofEngineering,Handan,Hebei056038,China 2SchoolofCivilEngineering,ShijiazhuangRailwayInstitute,No.17North2nd–RingEastRoad,Shijiazhuang,Hebei050043,China 3TheEnvironmentalMonitoringCenterofHandan,Handan,Hebei056002,China ABSTRACT PM2.5sampleswerecollectedandanalyzedforthefirsttimeinHandanCity,whichwaslistedinthetop4pollutedcities inChina,duringDecember2012toJanuary2013whentherecord–breakingseverehazepollutionhappened.Positive MatrixFactorizationmethod(PMF)wasappliedtounderstandmajorsourcestotheseverehazepollutionoverthiscity. –3 The daily average concentration of PM2.5 was 160.1ʅgm , which was 2.1 times of the National Ambient Air Quality –3 2– StandardsofChina(ClassII,AnnualAverageLevel)fordailyaveragePM2.5of75ʅgm .SO4 wasthemostabundantion + – (15.4%),followedbyNH4 andNO3 .Theyaccountedfor39.5%ofPM2.5.Eightfactorswereidentifiedbypositivematrix factorization(PMF)model.Themajorsourceswerecoalcombustionsource(25.9%),secondarysource(21.8%),industry source (16.2%), Ba, Mn and Zn source (12.7%), motor vehicle source (7.7%), road dust source (10.9%), K+, As and V –1 source(6.3%)andfueloilcombustionsource(2.5%).Themeanvalueofextinctioncoefficient(Bext)was682.1Mm and –1 CorrespondingAuthor: thelargestcontributortoBextwasammoniumsulfatewiththemeanvalueof221.0Mm ,accountedfor32.4%ofthe Li–Tao Wang B . ext :+86Ͳ310Ͳ8578755 :+86Ͳ310Ͳ8578755 Keywords:PM2.5,OC,Bext,PMF,sourceapportionment :[email protected] ArticleHistory: Received:07February2014 Revised:06June2014 Accepted:08June2014 doi:10.5094/APR.2014.085 1.Introduction 2.ExperimentalandAnalysisMethodology Extremely severe, widespread haze occurred in Central and 2.1.Sampling Eastern China during January 2013. The haze event attracted a wide attention locally and worldwide, because the haze in this Handan is located in the intersection area of Hebei, Shanxi, month was the most serious pollution episode since 1961. Henan,andShandong.Thesamplingsiteislocatedonthetopofa According to the statistics published by the Ministry of 12–mhighbuildingatSchoolofCityConstruction,HebeiUniversity Environmental Protection for January 2013, the top ten polluted of Engineering (Figure 1). The sampling period was from DecemͲ cities are Xingtai, Shijiazhuang, Baoding, Handan, Langfang, ber1, 2012 to January31, 2013. Daily PM2.5 samples were Hengshui,Jinan,Tangshan,BeijingandZhengzhou,sevenofwhich collected continuously from 8:00 am to 7:30 am of the next day arewithinHebei. usingaHighVolumePM2.5airsampler(ThermoScientificCo.)with 20.3×25.4cmquartzfilters(0.2ʅmofporesize).Flowratewas Veryfewstudieswerepursuedtostudythecharacteristicsof 1.13m3min–1andatotalof55sampleswerecollected. particulate matter in Handan, especially for fine particles. In the recentstudies,Wangetal.(2012),Weietal.(2013)andWanget 2.2.Chemicalcomponentsanalysis al.(2014)appliedtheCMAQmodelandusedtheso–called“Brute Force” method (BFM) (Dunker et al., 1996) to simulate the air Ionic analyses. A circular portion of the quartz filter (1cm2) was pollution over the southern Hebei cities and address the major extracted into 10mL ultrapure water (18.2Mɏcm) to determine sourcesofPM2.5.Wangetal.(2012)andWangetal.(2014)found the concentration of water–soluble inorganic ions. The extracted that local sources contributed 64.2% to PM2.5 in Handan in solution was filtered through a microporous membrane filter December 2007, 69.2% in January and 63.0% in February, 2013, (0.45ʅmofporesize),andstoredinarefrigeratorat–18°Cuntil respectively. However, to the best of our knowledge, there is no chemical analysis. An ion chromatograph (Dionex, DX–600, USA) researchfocusingonthechemicalcomponentsofPM2.5inHandan was used to measure the water–soluble cations. Another ion City.Inthisstudy,wesampledandanalyzedPM2.5duringthemost chromatoͲgraph (Dionex, ICS–2100, USA) was used to measure polluted period in Handan and applied PMF, a widely implied water–solubleanions. receptor model, to indentify source apportionment of PM2.5, to supportthepolicy–makinginairpollutioncontrolinthiscity. Carbon analysis. The total carbon, organic carbon and element carbonwereanalyzedinTsinghuaUniversitythroughmethodology ofTORusedThermal/OpticalCarbonAnalyzer(Model2001A)that ©Author(s)2014.ThisworkisdistributedundertheCreativeCommonsAttribution3.0License. Wei et al. – Atmospheric Pollution Research (APR) 760 was produced by Desert Research Institute (DRI). A piece of 0.01ʅgmL–1,respectively.Theprecisionwas<5%forTCand<10% processed sample filter (0.53cm2) was placed in an environment forOCandEC.TheMDLwas0.82ʅgcm–2forTOCand0.20ʅgcm–2 with pure He gas without O2, and was heated progressively at forTEC,respectively. 140°C, 280°C, 480°C, and 580°C, first to determine organic carboncontentsOC1,OC2,OC3,andOC4,respectively.Then,inan Elemental analyses. ICP–MS was calibrated by standard injection 2 environment with 2% O2 and 98% He, the sample was further (r >0.99) and the samples were analyzed in triplicate. Relative heatedprogressivelyat580°C,740°C,and840°Ctodeterminethe standarddeviationwaskeptbelow5%.Internalstandardrecovery elementalcarboncontentsEC1,EC2,andEC3. wascontrolledattherangefrom80%to120%. Elemental analyses. This study used ICP–MS to analyze the 2.4.PositiveMatrixFactorization(PMF) concentrationsoftraceelements(i.e.,Ti,V,Cr,Mn,Ni,Cu,Zn,As, Se,, Sr Cd, Ba and Pb). A piece of filter (1cm2) was digested for PMF is a multivariate factor analysis tool that decomposes a 25min at 190 °C with 8mL concentrated nitric acid (BV–III) and matrix of speciated sample data into two matrices, factor 0.5mL H2O2 (30%) in a Teflon microwave digestion tank (CEM, contributions and factor profiles, which then need to be MARS). Then, the bulk solution was moved into a 100mL flask. interpretedbyananalystastowhatsourcetypesarerepresented High–puritywaterwasusedtomakeupthevolume.Finally,10mL usingmeasuredsourceprofileinformation,winddirectionanalysis, solution was centrifuged for 10min to remove any suspended andemissioninventories(U.S.EPA,2008).Themethodisdescribed solidsbeforeinstrumentalanalyses. ingreaterdetailbyPaateroandTapper(1994)andPaatero(1997). PMFhasbeensuccessfullyappliedtoidentifythepollutantsources 2.3.QualityAssurance/QualityControl)(QA/QC (Leeetal.,1999;Chueintaetal.,2000;Qinetal.,2006;Brownet al.,2007;KimandHopke,2008;Yangetal.,2013). Sampling and weighing. There were several low–rise buildings aroundthesamplingsite.Therewasnotabigemissionsourcein The brief principle is described in the following: A speciated thisarea.Toensuretheflowof1.13m3min–1,flowofthesampling datasetcanbeviewedasadat matrixXofnbymdimensions,in equipmentwascheckedeverymonth.Thesiliconeoilwasreplaced whichnisthenumberofsamplesandmisthenumberofchemical every15daystoensurethesamplingrateofPM2.5.Thesequartz speciesmeasured.SoX=GF+E,G=n×pandF=p×m,wherepisthe filterswerebakedat450°Cfor4hours.Then,filterswereputina numberofpollutantsources,Gisthematrixof sourcecontribution, chamber at 25°C and 50% of relatively humidity for 24h until F is the matrix of species profile, and E is theresidual matrix for sampling.Thesesampledfilterswereputinthesamechamberfor eachsample/species.ThegoalofPMFistoidentifythenumberof 24h until the second weigh. Then, filters were preserved in a pollutantsourcesFandG[seeEquations(1)–(3)].ThePMFsolution refrigeratorthatsetat–20°C.Allprocedureswerestrictlyquality minimizes the object function Q [Equation (3)], based upon the controlledtoavoidanypossiblecontaminationofthesamples. uncertainties (u). Then PMF could calculate the G and F. The productofGandFcanexplainthesystematicvariationsinX. The sampling method was found to be effective since the measured concentrations agreed well with those measured using ௣ TEOM 1405D. In this period, three blank filters were preserved ܺ௜௝ ൌ෍݃௜௞ ݂௞௝ ൅݁௜௝ (1) under the same environment were analyzed to guarantee the ௞ୀଵ accuracyandreliabilityofsampling.Thecalculatedconcentrations fortheblankfilterswereusedtoadjusttheconcentrationsofthe ௣ measuredcompositions. (2) ݁௜௝ ൌܺ௜௝ െ෍݃௜௞ ݂௞௝ Ionicanalysesandcarbonanalysis.Glasscontainerswerewashed, ௞ୀଵ and soaked in ultrapure water for over 24 h. Then, they were ௡ ௠ cleaned with supersonic wavesfor 45min andwere washed two (times with ultrapure water (18.2Mɏcm). The method detection ܳൌ෍෍ൣ݁௜௝Ȁݑ௜௝൧ (3 + + + 2+ 2+ 2– – – – limit(MDL)ofNH4 ,Na ,K ,Mg ,Ca ,SO4 ,Cl ,NO2 andNO3 ௜ୀଵ ௝ୀଵ were 0.01, 0.001, 0.001, 0.004, 0.005, 0.01, 0.005, 0.01 and Figure1.LocationofthesamplingsiteinHandancity(Ÿ indicatesHebeiUniversityofEngineering,and*indicatesthesamplingsite). Wei et al. – Atmospheric Pollution Research (APR) 761 Inthisstudy,USEPAPMF3.0wasusedtoanalyzethesource average concentrations of PM2.5 in December and January were apportionment. The 10% of concentrations was used as the 135.1ʅgm–3and190.1ʅgm–3,whichare1.8and2.5timesofthe –3 uncertainty of each species in the data set. If the concentration dailyPM2.5limitvalue(75ʅgm )setbytheNationalAmbientAir was less than or equal to MDL provided, 1/2 MDL was used to QualityStandardinChina(CNAAQS)(MEP,2012),respectively. replace the concentration, the uncertainty was calculated using 5/6 MDL. Signal to noise (S/N) [see Equation