ADVANCED AIR POLLUTION Edited by Farhad Nejadkoorki Advanced Air Pollution Edited by Farhad Nejadkoorki Published by InTech Janeza Trdine 9, 51000 Rijeka, Croatia Copyright 2011 InTech All chapters are Open Access articles distributed under the Creative CommonsNon Commercial Share Alike Attribution 3.0 license, which permits to copy,distribute, transmit, and adapt the work in any medium, so long as the originalwork is properly cited. After this work has been published by InTech, authorshave the right to republish it, in whole or part, in any publication of which theyare the author, and to make other personal use of the work. Any republication, referencing or personal use of the work must explicitly identify the original source. Statements and opinions expressed in the chapters are these of the individual contributors and not necessarily those of the editors or publisher. No responsibility is acceptedfor the accuracy of information contained in the published articles. The publisherassumes no responsibility for any damage or injury to persons or property arising outof the use of any materials, instructions, methods or ideas contained in the book.

Publishing Process Manager Ivana Lorkovic Technical Editor Teodora Smiljanic Cover Designer Jan Hyrat Image Copyright Stphane Bidouze, 2010. Used under license from Shutterstock.com First published July, 2011 Printed in Croatia A free online edition of this book is available at www.intechopen.com Additional hard copies can be obtained from orders@intechweb.org Advanced Air Pollution, Edited by Farhad Nejadkoorki p.cm.ISBN 978-953-307-511-2 free online editions of InTech Books and Journals can be found atwww.intechopen.com Contents PrefaceIX Part 1Air Quality Monitoring1 Chapter 1Observational Study of Black Carbon in the North Suburb of Nanjing, China3 Lili Tang, Shengjie Niu, Mingliang Yan, Xuwen Li, Xiangzhi Zhang, Yuan Zhu, Honglei Shen, Minjun Xu and Lei Tang Chapter 2Air Pollution Monitoring Using Fuzzy Logic in Industries21 Seyed Ebrahim Vahdat andForoogh Mofid Nakhaee Chapter 3Indoor Nitrogen Oxides31 Silvia Vilekov Chapter 4Particularities of Formation and Transport of Arid Aerosol in Central Asia51 Galina Zhamsueva, Alexander Zayakhanov, Vadim Tsydypov, Alexander Ayurzhanaev, Ayuna Dementeva,Dugerjav Oyunchimeg and Dolgorsuren Azzaya Chapter 5Assessment of Areal Average Air Quality Level over Irregular Areas: a Case Study of PM10 Exposure Estimation in Taipei (Taiwan)67 Hwa-Lung Yu, Shang-Chen Ku, Chiang-Hsing Yang,Tsun-Jen Cheng and Likwang Chen Chapter 6Evaluation of Hexamethylene Diisocyanateas an Indoor Air Pollutant and Biological Assessment of Hexamethylene Diamine in the Polyurethane Factories81 Seyedtaghi Mirmohammadi, M. Hakimi. Ibrahim and G. N. Saraji Chapter 7Assessment on the Ozone Air Pollution in a Medium Metropolitan Area: Seville (Spain)99 Jose A. Adame, Antonio Lozano, Juan Contreras and Benito A. de la MorenaVIContents Chapter 8Investigation on the Carbon Monoxide Pollution over Peninsular Malaysia Caused by Indonesia Forest Fires from AIRS Daily Measurement115 Jasim M. Rajab, K. C. Tan, H. S. Lim and M. Z. MatJafri Chapter 9Air Quality Plans for the Northern Region of Portugal: Improving Particulate Matter and Coping with Legislation137 C. Borrego, A.Carvalho, E. S, S. Sousa, D. Coelho,M. Lopes, A. Monteiro and A.I Miranda Chapter 10Changeability of Air Pollution in Katowice Region (Central Europe, Southern Poland)159 Mieczysaw Leniok Chapter 11Catalytic Converters and PAH Pollution in Urban Areas177 Federico Valerio, Anna Stella, Mauro Pala, Daniele Balducci, Maria Teresa Piccardo and Massimo Cipolla Part 2Air Quality Modelling191 Chapter 12Imprecise Uncertainty Modelling of Air Pollutant PM10193 Danni Guo, Renkuan Guo, Christien Thiart and Yanhong Cui Chapter 13Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region213 Alberto Mendoza, Santosh Chandru, Yongtao Hu,Ana Y. Vanoye and Armistead G. Russell Chapter 14Uncertainty in Integrated Modelling of Air Quality239 Piotr Holnicki Chapter 15Evaluation of Regional Emission Control Based in Photochemical Air Quality Modelling261 ngel Rodrguez,1Santiago Saavedra, Mara Dios, Carmen Torres, Jos A. Souto, Juan Casares, Beln Soto and Jos L. Bermdez Part 3Advanced GIS Applications in Air Quality Management277 Chapter 16Development of GIS-aided Emission Inventory of Air Pollutants for an Urban Environment279 Sailesh N. Behera, Mukesh Sharma, Onkar Dikshit and S.P. Shukla Chapter 17Air Pollution, Modeling and GIS based Decision Support Systems for Air Quality Risk Assessment295 Anjaneyulu Yerramilli, Venkata Bhaskar Rao Dodla and Sudha Yerramilli ContentsVII Chapter 18Contribution of Geostatistics to the Study of Risks Related to Air Pollution325 Jacques Deraisme, Michel Bobbia and Chantal de FouquetChapter 19Spatial Interpolation Methodologies in UrbanAir Pollution Modeling: Application for the Greater Area of Metropolitan Athens, Greece341 Despina Deligiorgi and Kostas Philippopoulos Part 4New Techniques in Air Quality Management363 Chapter 20The Electrical Conductivity as an Index ofAir Pollution in the Atmosphere365 Nagaraja Kamsali, B.S.N. Prasad and Jayati Datta Chapter 21Quick and Economic Spatial Assessment of Urban Air Quality391 Panayotis C. Yannopoulos Chapter 22New Approaches for Urban and Regional Air Pollution Modelling and Management429 Salvador Enrique Puliafito, David Allende, Rafael Fernndez,Fernando Castro and Pablo Cremades Chapter 23A New Air Quality Index for Cities455 Lgia T. Silva and Jos F. G. Mendes Chapter 24An Analytical Application forthe Determination of Metals in PM10473 Tony Byrd, Mary Stack and Ambrose Furey Chapter 25Artificial Neural Networks - a Useful Tool in Air Pollution and Meteorological Modelling495 Primo Mlakar and Marija Zlata Bonar Chapter 26Indoor Air Control by Microplasma509 Kazuo Shimizu Chapter 27Method for Validation of Lagrangian Particle Air Pollution Dispersion Model Based on Experimental FieldData Set from Complex Terrain535 Botjan Grai, Primo Mlakar and Marija Zlata Bonar Chapter 28Air Quality and Bioclimatic Conditions withinthe Greater Athens Area, Greece - Developmentand Applications of Artificial Neural Networks557 Panagiotis Nastos, Konstantinos Moustris,Ioanna Larissi and Athanasios Paliatsos Preface Theair,asanelementarycomponentofthebiosphere,isaconveyorofmany pollutantsbothgasesandparticlesintheatmosphere.Airpollutionisastateofthe atmospherewithpredominantpresenceofhazardoussubstancesthatareharmfulto humansandanimals.Theair-bornepollutantsdegradetheairqualityandconstant exposure to polluted air may lead to several health problems such as cardiopulmonary disease, bronchitis, asthma, wheezing and coughing etc. Anthropogenic sources of air pollutionincludeactivitiesoftransport(airports,highwaysandshipemissions), power generation, oil refineries, waste incineration, industries and agriculture. Natural emissionscomefromareasources(sea,ocean,countrysideland,forests)andpoint sources (volcanoes and territorial cracks).In writing this book, we have assumed that readers are well acquainted with the very basic concepts of air pollution. The book covers all the important areas in the science of airpollutionandisdesignedforatleasttwokindsofreaders:a)studentsof intermediateandadvancedcoursesinairpollution;b)professional,researchersand practitioners of many areas of the field.Thisbookcontainsfourmainparts.Therehavebeenimprovementsinmonitoring pollutantsbyprovidingpredictionofthetemporalandspatialdistributionofactual pollutionlevels.Thesearedescribedindifferentstudiesunderpartonebothfor indoor and outdoor. The second part covers several studies on air pollution modeling. Deterministic,statisticalandmathematicalmodelsareexplained.Thethirdpart discusses a review of the most common Geographical Information System applications inairqualitymanagement. Finallyinpartfourthelatestdevelopmentsinairquality management are given through numerous studies.June, 2011 Farhad Nejadkoorki, PhD Yazd University,Iran Part 1 Air Quality Monitoring 1 Observational Study of Black Carbon in the North Suburb of Nanjing, China 1Lili Tang et al.* Nanjing University of Information Science & Technology, Nanjing,China 1. Introduction Blackcarbonisakindofamorphouscarbonparticles,whichwasformedbyincomplete combustionofcarbonaceoussubstance(Andreaeetal.,1984;BondandBergstrom,2006; Hansen and Rosen, 1984; Hansen et al., 1988; Penner et al., 1993). The main natural sources ofblackcarbonarevolcaniceruptions,forestfires,etc.Itsshape,sizeandoriginsare extremely variable depending on the different type of fuels, combustion processes, and the ageandhistoryofairmasses.Itmainlyresidesinthesubmicronparticlesizerange.This aerosolisusuallyseparatedintotwomaincomponents:oneiscalledtheorganiccarbon (OC)andtheothertheblack carbon(BC)component.Emissions fromnaturalsources such asforestfiresarelocalizedandincidental(BaiandWang,2005),whereasanthropogenic emissionsarewidespreadandcontinuous(Liuetal.,2010;Streetsetal.,2001;Xuetal., 2006). The increasing demand for energy in the past century has led to a great increase in the rateofBCemissions.Blackcarbonparticleshavereceivedincreasedinterestrecentlysince Hansenetal.(Hansenetal.,2000)indicatethatitisperhapsmorecosteffectivetocontrol blackcarbonparticlesthanCO2atthepresenttime.Blackcarbonhasmanyeffects(IPCC 2001; Lary et al., 1999), such as harming to human health, changing climatic characteristics, reducing visibility, affecting the atmospheric chemical reaction process, which has become a focus of scientific research in recent years. Although the effect of increasing BC in the atmosphere may be considerable, knowledge of BCconcentrations,distributions,characteristics,andpotentialeffectsisstillseriously lacking.Moreover,thesurfaceofBCparticlescontainsnumerousadsorptionsitesthatare capableofenhancingcatalyticprocesses.Astheresultofitscatalyticproperties,BCmay intervene in some important chemical reactions involving atmospheric sulfur dioxide (SO2), nitrogenoxides(NOx),ozone(O3)andothergaseouscompounds(Gundeletal.,1989). Despite the evident significances of BC in air chemistry and physics, information concerning their spatial and temporal variability is still quite limited. *1Shengjie Niu, Mingliang Yan, Xuwen Li, Xiangzhi Zhang, Yuan Zhu, Honglei Shen, Minjun Xu and Lei Tang Jiangsu Environmental Monitoring Center, Nanjing, China Jiangsu Institute of Meteorological Sciences, Nanjing, China Advanced Air Pollution 4 Consideringtheglobaldistributionofblackcarbon,theblackcarbonconcentrationsinthe northern hemisphere were significantly higher than that in the southern hemisphere, as well as there is higher black carbon concentration value in eastern China (Streets et al., 2001). Of bothnaturalandanthropogenicorigin,emissionsofblackcarbonrecentlyhavebeen estimated at 24Tg per year (Penner et al., 1993). China has higher black carbon emissions in recent years. We must make observation of black carbon in order to grasp the first primary data. The research work about the systematic observation of black carbon begun from 1990s in China, which had been carried out in most areas of China after 21 century. NanjingislocatedinthecenterofYangtzeRiverDeltaregion,whichhasdeveloped petrochemical, steel and other heavy industries. The air pollution in Nanjing is very serious, buttheobservationsofblackcarbonhavenotbeenreported.Theobservationalresultsof blackcarboninfineparticulatematterfromNovember2008toApril2010inthenorth suburbofNanjingwereusedtostudytheevolutionlawofblackcarbonaerosol.The observational data of SO2, NO2, HNO2 and O3 in October 2009 was also used to analyze the correlationsbetweenblackcarbonaerosolsandSO2;todiscussthecomplex-phase photochemical reaction between black carbon aerosols and gas. Moreover, the findings may also be useful in the formulation of strategic air pollution control in China and other rapidly developing countries. To understand anthropogenic air pollutant sources, loading, and their fateintheregion,itisimportanttomeasureBCfordifferentseasonsandcomparetheir characteristics.In this paper, an intensive air pollution health effects study was conducted. The characteristics of spatial variations of BC in PM2.5 concentrations during one more year are summarized, and the relationships between gas and BC in PM2.5 are discussed.2. Experimental methods 2.1 Measurement site description ThefieldstudywasconductedinnorthsuburbareaofNanjingCitywhichislocatedin Jiangsu Province, 290 km east of Shanghai, and 10 km south-southeast of the Yangtze River. Therearelargeregionalsourcesthatcanimpactthissite,includingYangziPetrochemical IndustriesCompany,Nanjingsteelplant,NanjingchemistryplantandHuanengelectric generating station (5-10 km to the east and northeast). The No. 328 national highway is just 1 kmintheeastofmeasurementsite.Averagedailytrafficflowisover42,000onthis highway.ThedominantlocalsourcesofBCareexpectedtobevehicularemissions, incomplete combustion of fossil fuels and biomass burning. Measurementsweretakenatauniversity10kmnorthofthecenterofNanjingcityina residential area, near the top of automatic meteorological observation station (elevation 5m). Samples were collected 4.5 m above the ground. 2.2 Sample collection and analysis PM2.5aerosolsampleswerecollectedfromNovember2008toApril2010.Samplingstarted and ended at around 8 a.m. and 8 p.m. every week; each sample was collected for 12 h. The sampler (Andersen Instruments, Smyrna, GA, USA) was equipped with a slot inlet of 2.5 m cut size and operated at a flow rate of 1.11m3/min. All the collection substrates were 2025 cmquartzfiberfilters,prebakedat550overnighttoremoveanyabsorbedorganic materials.Thefilterswerestoredandtransportedtothefieldinannealedaluminumfoil pouches.Afterthesamplingwascompleted,the2025cmfilterswerefoldedinhalfsuch Observational Study of Black Carbon in the North Suburb of Nanjing, China 5 thatthesidewiththesampleonitwasfoldedontoitselfandstoredintheannealed aluminum pouches. All the filter samples were stored at -4 in a freezer until analysis. OCandECconcentrationsweredeterminedusingathermal/opticalaerosolcarbon analyzer(SunsetLaboratory,ForestGrove,OR,USA)(BirchandCary,1996).Forthe determinationofOCandECconcentrations,afilterpunch1.45cm2insizewasremoved fromthe2025cmfilterandloaded intothe aerosolcarbonanalyzer.Thethermalanalysis conditions (i.e., temperature program and the type of atmosphere) employed were similar to thoseadoptedbyallresearchgroupsthatparticipatedinthecarbonanalysisoftheACE-Asiasamples(Schaueretal.,2003).DuplicatemeasurementsofOCandECweremadeon every sample, and the average relative standard deviations were 4% and 9% for OC and EC, respectively.Laboratoryandfieldblankswerecollectedatarateof10%todeterminethe limitofdetection(LOD);collocatedsampleswerecollectedatarateof10%todetermine precision;replicateanalysiswasperformedon10%ofthesamplestodetermineanalytical precision. 2.3 The aethalometer AmodelAE-21Aethalometer(MageeScientificInc.,BerkeleyCA)wasusedtomeasure aerosol black carbon (BC) in real time (Lavanchy et al., 1999; Liousse et al., 1993; Petzold and Niessner,1995; Ruellan and Cachier,2000; Sharma et al.,2002). It measures the attenuation of lighttransmittedthroughparticlesthataccumulateonaquartzfiberfilter(Hansenetal., 1984). A vacuum pump draws air through the instrument so that the particles continuously accumulate on the filter while being illuminated by visible light (incandescent light source). Theeffectiveoperationalwavelengthoftheaethalometeris880nm(Bodhaine1995; Lavanchyetal.,1999).ThemeasurementofBCbythistechniqueiscontroversialsincethis opticaltechniqueisbasedonlightabsorptionofatmosphericparticlessampledonfilters andMietheorytoestimateBCmassconcentrationsintheatmosphere.Thismethodis similarinprincipletotheCoefficientofHazemonitorsdevelopedover40yearsago (Hemeonetal.,1953),butismoresensitiveandstable.OpticalattenuationfromBCis calculated using the decrease in light transmission through the filter and the sample volume. TheAethalometerisnotdesignedtomeasure eitheraerosolorganiccarbonoratmospheric lightextinctionfromparticles.Theprinciple ofthemethod is describedindetailelsewhere (Hansen et al., 1984; Hansen and McMurry1990). The Aethalometer sample flow rate was 4 L-min-1.An2.5msizeinletfractionwasused.Thediskaging,powerfailureandother reason cause some data loss in the observation process, neglecting whole day data less than 18 hours per day (Lou et al., 2005), the effective data is 285 days. Thisstudywasalsoconductedtodeterminetherelationshipbetweenlightabsorption measurements of an aethalometer and thermally derived black carbon concentrations.2.4 Observation of SO2, NO2, O3 and HNO2 AR500systemwasusedtodeterminetheconcentrationoffourkindsofgases,whichhas doubleopticalDOASAQMsystemandproducedbySwedishOPSISAB.High-pressure xenon lamp (B class) was used as launchers light. A smooth spectrum emitted by the xenon lamp can covere the wavelength. The wavelength of UV and visible light is about 200 ~ 400 nm and about 400 ~ 700 nm, respectively. The measured gases are O3, SO2, NO2 and HNO2. Observationdatewasfrom1Octoberto31Octoberof2009.Accordingto75%ofdata capture rate, excluding whole hour data less than 45 min per hour and whole day date less than 18 h per day. Advanced Air Pollution 6 Meteorological and air quality data are monitored continuously at this location by the Office of Atmospheric Environment Observatory. 3. Results and discussion 3.1 Methods comparison: elemental carbon and black carbon ComparisonbetweentheAethalometerBCdataandtheECcomponentoftheEC/OC sampler data should show good agreement, since EC is expected to be the principal (visible) lightabsorbingcomponentinambientair(Gundeletal.,1984;Hansenetal.,1984;Hansen andRosen,1990).FiveminutesAethalometerBCdatawereaveragedintotwelvehour values temporally matching the 61 EC/OC samples. The relationship between BC and EC is showninFigure1.Thetwomethodswerefoundtobehighlycorrelated(R2=0.9492).Itis said that particulate black carbon measured by the Aethalometer method has been shown to beareliablesurrogateforparticulateelementalcarbonmeasuredbyThermo/Optical Reflectance at this location and time of year. 1000 2000 3000 4000 5000 6000 7000 800010002000300040005000Sunset Lab Optical-EC/(ng.m-3)Aethalometer-BC/(ng.m- 3) Fig. 1. Comparison between BC measurements made by EC/OC Analyzer and Aethalometer 3.2 Monthly variation of BC concentration Fig.2showsthevariationofmonthlyaverageblackcarbonaerosolconcentrationsduring theobservationperiod.TheBlackcarbonaerosolconcentrationinautumnandwinteris higherthanthatinspringandsummer.Taketheyear2009asexample,variationorder (fromhightolow)ofblackcarbonaerosolconcentrationsisOctober>May>November> February> December> June> April> September> August> July. ThemonthlyaverageblackcarbonaerosolconcentrationinNanjingreachesmaximumin September 2009 in the whole observation process. It is earlier comparing with other areas of the country (Qin et al., 2007; Zhao et al., 2008) The rainfall in Nanjing is very low in October 2009,anditiscontrolledbythehighatmosphericpressure.Thesereasonsleadtoair pollutants concentrate and difficult to spread and eliminate. Moreover Nanjing was located inthetransitionzonebetweensubtropicalzoneandtemperatezoneanddoesntadopt large-scaleunifiedheatingmeasureslikenorthcitiesofChinainwinter.Theemission inventoryofblackcarbonaerosolwontincreasesharplybecauseofheating,whichis anotherreasonfortherelativehighermonthlyaverageconcentrationofblackcarbon y=0.74x +0.22 R2=0.9492 n=61 Observational Study of Black Carbon in the North Suburb of Nanjing, China 7 2008-12 2009-22009-42009-62009-8 2009-102009-12 2010-22010-420004000600080001000012000140001600018000BC Concentration(ng/m3)Month Fig. 2. Variation of BC average concentration in PM2.5 in the air of north suburb of Nanjing aerosolinOctober.Bycontrast,theblackcarbonmonthlyaverageconcentrationsin NovemberandDecember2009decreaseby23.7%and55.1%respectivelyoverthesame periods,whichindicatesthat"onthepromotionofcropstrawintegratedreusedecision" issuedbyJiangsuEnvironmentalProtectionAdministrationandaseriesofairpollution controlpoliciesinthepracticalapplicationhasgainedtangibleeffects.However,theblack carbonaerosolconcentrationisstillhigherinNanjing,theblackcarbonaerocolloidal concentrations in November and December 2009 are (6127 3689) ng/m3 and (5714 2849) ng/m3 respectively. It is still necessary to reduce emissions of industrial waste gas, vehicular exhaust and to prohibit straw incineration strictly. TheminimumofmonthlyaverageblackcarbonaerosolconcentrationappearedinMarch 2010inthewholeobservationprocess.TheprecipitationinMarch2010increasesby40% overthesameperiod,reaching118mm,16precipitationdays,blackcarbonaerosolwas significantlyremovedbyrain.Wuetal(Wuetal.,2009)foundinthePearlRiverDeltaof China that black carbon aerosol concentrations in the rainy season were significantly lower thanthatindryseasonsinPanyudistrictofGuangzhoucity.From2004to2007,black carbonaerosolconcentrationsinthedryseasonsis10.4%higherthanthatintherainy seasons(Wuetal.,2009).FromthebeginningofMarch,theamountofsolarradiation reachingthegroundincreaseswiththegradualincreaseofsunlighthoursandthe atmosphericconvectionbecomestrong.Theblackcarbonaerosolconcentrationsdecrease because of the diffusion of black carbon aerosol with atmospheric convection. 3.3 Seasonal variation of BC concentrationNanjingislocatedbetweensubtropicalandtemperatetransitionzone,whichhasfour distinctiveseasons.Thereforethecriteriafortheclassificationofastronomyisusedto differentiate the Four Seasons, March, April and May are classified as spring, June, July and Augustareclassifiedassummer,September,OctoberandNovemberareclassifiedasfall, December and January, February in the following year are classified as winter. Fig. 3 shows theseasonalvariationofblackcarbonaerosolconcentration,where1to7inx-axis represents different seasons from the fall in 2008 to the spring in 2010. From the Fig. 3, black carbon aerosol concentrations in fall and winter were significantly higher than that in spring andsummer,whichisconsistentwithvariationcharacterizationofmonthlyaverageblack carbonaerosolconcentrations.Themaximumofblackcarbonaerosolconcentration appeared in the fall of 2008 during the whole observation period. It is the peak time of straw Advanced Air Pollution 8 Fig. 3. Variation of seasonal average BC concentrations in PM2.5 in the north suburb of Nanjingincineration from the November 12 to November 31. The straw incineration attributes to the increaseoftheblackcarbonaerosolconcentration.Theconcentrationsofblackcarbon aerosolsisrelativelylowerwhilewithoutstrawincineration.Hencethevariationofdaily blackcarbonaerosolconcentrationisgreaterinthefallof2008andthedatahasahigher dispersion degree. The concentration is 18000ng /m3 at 95%, but less than 3000ng/m3 at 5% with no difference in other seasons. 3.4 Diurnal variation of BC concentration The diurnal variation of BC concentration is characterized by a special pronounced double-peakpatternsduringtheobservationperiodinFig4.Inthecaseoftemperatureinversion persisting, human activities starting from 5 a.m. lead to the increase of black carbon aerosols emissions,andcentralizationandnodiffusionofalargenumberofpollutantsnearthe ground.Blackcarbonaerosolconcentrationintheairbegantoincreaseatthesametime. Auromeetetal(AuromeetandSerge2009)foundthattransportationwasanimportant Fig. 4. Diurnal variation of BC concentration in PM2.5 in the air of north suburb of Nanjing Observational Study of Black Carbon in the North Suburb of Nanjing, China 9 source to the black carbon aerosol in Toulon region of France. In this study, the 328 national highway is just 1 km far away from observation site, so the vehicular exhaust make a greater contribution to the concentration of black carbon aerosol. After7a.m.,withtheincreaseofsolarradiation,temperatureinversionstructurewas destroyedandatmosphericconvectionwasenhanced,blackcarbonaerosolconcentration decreases gradually. The strongest solar radiation and most intense atmospheric convection appear from 13 p.m. to 15 p.m., so the aerosol particle is not easy to concentrate. Hence the minimum of black carbon aerosol concentrations occurs at this time. Blackcarbonaerosolconcentrationgraduallyincreasedafter16p.m.,reachedmaximum againat18p.m.Solarradiationwasgraduallydecreased,andconvectionwasweakened during this period. Meanwhile, a large number of black carbon aerosols were emitted to the atmosphereagainbyhumanactivities;blackcarbonaerosolconcentrationbeginsto increase.ThesechangesofBlackcarbonaerosolconcentrationhavesimilaritywithotherprevious reports(Qin2001;Zhaoetal.,2008).Butthemaximumoftheblackcarbonaerosol concentration at 18 p.m. will last until the dawn of the following day and does not decrease withthereductionofhumanactivityduringthisperiod,whichisdifferentwithother reports.ItisbecauseoftheparticularweatherconditionsinNanjing.Thebook"Jiangsu weather" published by Jiangsu Province Bureau of Meteorology 90 years in the 20th century (JiangsuClimate1991)describedthatNanjinghasahighfrequencyoftemperature inversion.Thedayoftemperatureinversionwasdefinedthatitappearedtemperature inversiononetimeinaday,whichhasstrongtemperatureinversionfrequency.The structureoftemperatureinversionlayercanbeformedattowardeveningintheNanjing region. As the time passed by, strength of temperature inversion was continually increasing. Thepresenceofstrongtemperatureinversionlayercausedthatblack carbonaerosolswere not easy to diffuse and maintained at a stable concentrations level. (a)(b) Fig. 5. Variation of diurnal BC concentration in PM2.5 on weekdays(a) and weekend(b) in the air of north suburb of Nanjing Thehourlyaverageconcentrationsofblackcarbonaerosolonweekdaysandweekendare discussedrespectively.Fig.5showsthevariationofhourlyaverageblackcarbonaerosols concentration on weekdays. The tendency of black carbon aerosol concentration in Fig. 5 is consistentwiththeobservationperiod,buttheconcentrationvaluehasmoredramatic Advanced Air Pollution 10change.WhilethetendencyofblackcarbonaerosolconcentrationgiveninFig.5hassome differences, the appearing time of maximum of black carbon aerosol concentration at 95% is putoffonehour;theappearingtimeofmaximumofblackcarbonaerosolconcentrationat 75% is still at 7 a.m. The different concentration among different numbers demonstrated that humanactivitiesarerandomonweekendanditisirregularaboutblackcarbonaerosols emission.However,themaximumofblackcarbonaerosolsconcentrationstillappearsat6 a.m.to8a.m.andafter5p.m.duetothevariationoftemperature.Theappearingtimeof afternoonmaximumofblackcarbonaerosolsconcentrationisdelayedthreehourson weekend, which is related that people enjoy some form of recreation and go home late. 3.5 The pollution character of BC concentration in the north of NanjingFrom Fig. 4, black carbon aerosol concentration has no significant difference with its average value,whichindicatestheconcentrationofblackcarbonaerosolsisnotanormal distribution.Itisnotscientifictouseaveragevaluetorepresentthebackgroundvalue. Consideringtherequestofatmosphericbackground,wedefinetheconcentrationhaving maximum frequency as the background concentration of black carbon aerosols (Tang et al., 1999). Taking 100ng/m3 as frequency statistics step, draw the frequency distribution map of hourlyaverageconcentrationofblackcarbonaerosol.FromFig.6,thefrequency distributionofhourlyaverageblackcarbonaerosoldoesnothavenormalcharacterization, usingthelog-normaldistributionfunctiontofitthefrequencydistributioncharacteristics, resultsshowthathourlyaverageconcentrationofblackcarbonaerosolsat2920ng/m3 appearedmaximumfrequency.TheobservationsiteislocatedinthecampusofNanjing UniversityofInformationScience&TechnologyinthenorthsuburbofNanjing,wecan conclude the background concentration of black carbon aerosol is approximately 2920ng/m3 inthenorthsuburbofNanjing.Thestatisticsdataarenotaffectedbyasmallnumberof pollutedairmassesanddonotappearextremelyhighconcentrationofblackcarbon, therefore the value is most representative to the black carbon aerosols in the north suburb of Nanjing. ThebackgroundconcentrationsofblackcarbonaerosolsindifferentareasofChinahave significantdifferences.ThedetailedresultsarelistedinTab.1.TheblackcarboninPM2.5 was studied in this paper, which is the main difference with other papers. Li Yang et al (Li et al.,2005)observedtheblackcarboninPM2.5inXi'an,andtheresultsshowedthat 5000 10000 15000 20000 25000020406080100120CountBC Concentration( ng/m3) Fig. 6. Count distribution of hourly mean of BC in PM2.5 in the air of north suburb of Nanjing Observational Study of Black Carbon in the North Suburb of Nanjing, China 11 backgroundconcentrationofblackcarboninPM2.5isashighas4500ng/m3,butthevalue only reflects the black carbon concentration levels in PM2.5 of the fall in Xi'an region. There is a higher value of the black carbon aerosol concentration in the fall. It can be inferred that the whole-year background concentration in Xian region is lower than the value in the fall. Lou Shujuan et al (Lou et al., 2005) found that 90% of black carbon exists in the aerosol particles with the diameter less than 10m, while 82.16% of black carbon exists in the aerosol particles withthediameterlessthan2.5minBeijingregion.Taking80%asconversionfactor betweentheblackcarboninTSPandtheblackcarboninPM2.5inNanjing,background concentrationofblackcarboninTSPisupto3650ng/m3inthisstudy.Thevalueismuch higher than that in other studies. The air pollution by black carbon aerosol is very serious in Nanjing. Site name in China LocationDatePositionBackground concentration (ng/m3) Literature Wa liguan, Qinhai Global backgroundSeptember 1994- November 1995 TSP63 Tangjie et al. (1999) Wenjiang, Sichuan Suburb of city September 1999-August 2000 TSP2890 Qin Shiguang (2001) Lasa, Tibetan Suburb of city June 1998-September 1998 TSP400 Qin Shiguang (2001) Linan,Zhejiang Region backgroundAugust 2000- February 2001 TSP2350 Qin Shiguang (2001) Shang dianzi, Beijing Region backgroundAugust 1999-September 2000 TSP222 Qin Shiguang (2001) Xian, Shanxi Suburb of city September 2003-November 2003 PM2.54500 Liyang et al. (2005) Xining, Qinghai Suburb of city September 2005-February 2006 TSP2300 Zhao Yucheng et al. (2008) Nanjing, Jiangsu Suburb of city November 2008-April 2010 PM2.52920This paper Table 1. The background concentration of BC in other areas of China 3.6 A case study Nanjingmunicipalgovernmenthadinstitutedtheblueskyplaninordertocontroltheair pollutionandamelioratetheairqualitysince2003.Theaimofblueskyplanfor2009isto have320daysachievingsuperiorrankoftheatmosphericstatus(API100).Considering from the whole year of 2009, there are only 14 pollution days with the help of further some weatherconditionsinthefirstsixmonths.ButfromthebeginningofOctober,duetothe increasingly deterioration of atmospheric status, there are 16 pollution days within October in Nanjing. There into, it appears 15 consecutive pollution days from October 16 to October 30 in Nanjing. The air condition of Pukou district in which observation site is located is better than whole Nanjingregion,butthereareninedayswithairpollutionindexexceeding100,slight pollution,andtherearemanydayswiththeairpollutionindexaround95inFig.7.There Advanced Air Pollution 122009-10-6 2009-10-12 2009-10-18 2009-10-24 2009-10-30406080100120140APIDate Nanjing Puk ou Fig. 7. The comparison of air pollution in two different sitearenolarge-scaleinfrastructuralconstructioninPukoudistrict,buttherearemanylarge industries, such as Yangzi Petrochemical, Nanjing Chemical, Nanjing steel and so on in the north suburb of Nanjing. Therefore black carbon aerosols make an important contribution to air pollution in the north suburb of Nanjing. Fig. 8 shows the evolution sequence of hourly average black carbon aerosol concentration in October,thedataofOctober10and11werelost.FromFig.8,theconcentrationofblack carbon aerosol has significant changes. The air pollution is most serious in late October, the variationscopeofdiurnalblackcarbonaerosolconcentrationscanreach20000ng/m3,the maximum(26701ng/m3)occurredat8p.m.onOctober26.Thesearedeterminedbythe boundarylayerstructureandtheemissionsofpollutionsourcesinNanjingregion.The northcoldairarrivedatNanjingonOctober31,blackcarbonaerosolconcentration simultaneouslydecreasedto964ng/m3at20p.m.InFig.9,correlationanalysiswasused betweentheairpollutionindexandblackcarbonaerosolconcentration.Theresultof correlation analysis shows that they have positive relationship (the correlation coefficient R =0.689;psuch Advanced Air Pollution 200 ( ) ( ){ }| ,mMB z y d y z M = < (9) 3.iff is Lipschitz continuous on the whole space m , then the function is called globally Lipschitz continuous. Remark 3.8: For continuity requirements, Lipschitz continuous function is stronger than that ofthecontinuousfunctioninNewtoncalculusbutitisweakerthanthedifferentiable functioninNewtondifferentiabilitysense.Inotherwords,Lipschitz-continuitydoesnot warrantthefirst-orderdifferentiabilityeverywherebutitdoesmeannowhere differentiability.Lipschitz-continuitydoesnotguaranteetheexistenceofthefirst-order derivative everywhere, however, if exists somewhere, the value of the derivative is bounded since ( ) () f x f yMx y< (10) by recalling the definition of the Newton derivative ( ) ()( ) lim 'y xf x f yf xx y= (11) Similartotheconceptofstochasticprocessinprobabilitytheory,anuncertainprocess { } , 0tt isafamilyofuncertaintyvariablesindexedbyt andtakingvaluesinthestate spaceS . Definition 3.9:(Liu 2010, 2011) Let{ } , 0tC t be an uncertain process.1. 00 C = and all the trajectories of realizations are Lipschitz-continous; 2.{ } , 0tC t has stationary and independent increments; 3.everyincrement t s sC C+ isanormaluncertaintyvariablebwithexpectedvalue0and variance 2t , i.e., the uncertainty distribution of t s sC C+ is ( )11 exp3t s sC Cxzzt+ = + (12) Then{ } , 0tC t is called a canonical process. Remark 3.10: Comparing to Brownian motion process{ } , 0tB t in probability theory, which iscontinuousalmosteverywhereandnowhereisdifferentiable,whileLiu'scanonical process{ } , 0tC t isLipschitz-continuousandif { } , 0tC t isdifferentiablesomewhere,the derivative is bounded. Therefore{ } , 0tC t is smoother than{ } , 0tB t . Definition 3.11: (Liu, 2010, 2011) Suppose{ } , 0tC t is a canonical process, andf andg are some given functions, then ( ) ( ) , ,t t t td f t dt g t dC = +(13) is called an uncertain differential equation. A solution to the uncertain differential equation is the uncertain process{ } , 0tt satisfying it for any0. t > Imprecise Uncertainty Modelling of Air Pollutant PM10 201 Remark3.12:Since tdC and td areonlymeaningfulundertheumbrellaofuncertain integral, i.e., the an uncertain differential equation is an alternative representation of( ) ( )00 0, ,t tt s s sf s ds g s dC = + + (14) Definition3.13:Thegeometriccanonicalprocess{ } , 0tG t satisfiestheuncertain differential equation t t t tdG G dt G dC = + (15) has a solution( ) expt tG t C = + (16) where can be called the drift coefficient and0 > can be called the diffusion coefficient of the geometric canonical process{ } , 0tG t due to the roles played respectively. 4. Spatial interpolation and extrapolation via inverse distance approach Statistically,spatialinterpolationandextrapolationmodelingisactuallyakindoflinear regressionmodelingexercises,say,krigingmethodology.Consideringtheshortageof California PM10 data records, we will utilize a weighted linear combination approach, which wasfirstproposedbyShepard(1968).Theweightsaretheinversedistancesbetweenthe missingvaluecelltotheactualobservedPM10valuecells.Theweightconstructionisa deterministic method, which is neutral and does not link to any specific measure theory.It iswidelyusedinspatialpredictionsandmapconstructionsingeostatistics,butisnot probabilityoriented,rather,molecularmechanicsstimulated.Auniqueaspectof geostatisticsistheuseofregionalizedvariableswhicharevariablesthatfallbetween randomvariablesandcompletelydeterministicvariables.TheweightofanobservedPM10 value is inversely proportional to its distance from the estimated value. Let: ijCThe thj cellontheSitei ,( i representtheactualsitenumber), 1, 2, , 213 i =. Noteindexj pointstothecellwherenoPM10valueisrecorded,i.e.,missing value cell. ix Longitude of siteiiy Latitude of siteiijdThe distance between sitei and site j (where missing valuethj cell is located) ( ) ( )2 2ij j i j id x x y y = + ijwWeight, ( )( )10100 Cell,has no PMobs.1/ Cell,has PMobs.ijiji jwd i j= Then the inverse distance formula is, Advanced Air Pollution 202 2131ijij ijijiwC zw== (17) site 2045site 2774 site 2199site 2263 site 2997site 2914 site 2248 Fig. 4. The 7 sites from the selected 7 counties with completed 19 year observations of PM10 Imprecise Uncertainty Modelling of Air Pollutant PM10 203 We wrote a VBA Macro to facilitate the interpolations and the extrapolations to "fill" up the 2048 missing value cells in terms of the 1639 cells with PM10 values. With the interpolations andtheextrapolations,everysitehas19PM10valuesnow.Astowhethertheinverse distance approach can facilitate highly accurate predictions for each cell without a observed PM10value,weperformedare-interpolationandre-extrapolationscheme(bydeletinga truePM10record,thenfillitbytheremainingrecordsonebyone)toevaluatethemean squarevalueforerrorevaluation,thecalculatedmeanofsumoferrorsquaresis59.885, which is statistically significant (asymptotically).We plotted sites 2045, 2744, 2199, 2263, 2297, 2914, and 2248 (appeared in Fig. 3) respectively inFig.4.BycomparingFig.3and4,itisobviousthatonlySite2744thehazardlevel changed(movinguptonexthigherhazardlevel),whilethehazardlevelofothersixsites areunchanged.Thismaygiveanjustificationoftheinversedistanceapproach.Keepin mind,theaimofthisarticleisinvestigatewhetherthePM10levelischangedover1989to 200719-yearperiod.Thechangeisnotnecessarilybeaccuratebutreasonablycalculated because of the impreciseness features of PM10 complete records. 5. Uncertain analysis of site temporal pattern Oncetheinterpolationsandthe extrapolationsintermsoftheinversedistanceapproachis completed, a "complete" data set is available, containing 4047 data records of 213 sites over 19 years. The next task is for a given site, how to model the uncertain temporal pattern. It is obviousthatthe"complete"datasetcontainsimprecisenessuncertaintyduetothe interpolations and the extrapolations. We are unsure that the impreciseness uncertainty is of randomuncertainty,sothatwestilluseuncertainmeasuretheorytopursuethetemporal uncertainty modelling.RecallthattheDefinition3.13inSection3facilitatesauncertaingeometriccanonical process,{ } , 0tG t .Noticethat 00 G = maynotfitthedatarealitysothatweproposea modified uncertain geometric canonical process, { }*, 0tG t with 00 G > : ( )*0 0expt t tG G G G t C = = +(18) Note that *0ln lnt tG G t C = + + (19) Let *lnt ty G = , 0 0lnG = , then we have 0 t ty t C = + + , 1, 2, , 18 t = (20) Recall the relevant definitions in Section 3, we have [ ] 0tC = , and[ ]2tV C t =(21) But note that fors t < , [ ] ( ) ( )( )( )22t s s t s ss s t ss t sC C C C C CC C C Cs C C C = + = + = + (22) Advanced Air Pollution 204 Notice that the incrementt sC C is independent of sC , i.e., t sC is independent of sC . Therefore, [ ] ( )( ) ( ) { }( )1 2 , 1 21 2 1 21 12 21 21 2,min ,min 1 exp , 1 exp3 3t s st s st s s C CC CC C z z d z zz z d z zz zz z dt s s = = = + + (23) sinceif 1 and 2 areindependentuncertainvariableswithuncertaintydistributions 1 and 2 respectively,thenthejointuncertaintydistributionof( )1 2, is ( ) ( ) ( ){ }1 2 1 2, 1 2 1 2, min , z z z z = . Hence we obtain the expression of , s t : [ ]( )1 12 22 1 2, 1 2min 1 exp , 1 exp3 3s t s tz zC C s z z dt s s = = + + + (24) Then the thi "variance-covariance" matrix for uncertain vector ( )1 2 19, , , 'i i iy y y ( )2,19 19ii j k =(25) wherei isthesiteindex,, 1, 2, , 19 j k =aretheentrypairini matrix.Hencewehavea regression model (Draper and Smith, 1981, Guo et al., 2010, Guo, 2010). For the thi site, the regression model is 0 , it i i i i ty t C = + +(26) Then in terms of the weighted least square criterion we can define an objective function as ( ) ( )'10( , )i i i i i i iJ Y X Y X = (27) where 12 0191 11 2 1 19ii iiiiyyY Xy = = =

(28) We further notice that ( ), , 1, 12, 3, , 19ij i j i ji i ij i j i i ijr y yC C Cj = = + = + = (29) Imprecise Uncertainty Modelling of Air Pollutant PM10 205 Then it is reasonable to estimate iby 192118i ijjr == (30) Furthermore, we notice that ( )19221 18i ij ijr == (31) Also, we can evaluate ( )1 12 1 2, 1 2min 1 exp , 1 exp3 3ij kz zj z z dk j j = + + + (32) in terms of numerical integration, Then an estimate for i matrix is obtained: ( )2,19 19ii i j k =(33) Finally, we use the approximated objective function ( ) ( )'10 ( , )i i i i i i i iJ Y X Y X = (34) toobtainapairofestimates( )0,i i .Repeatthisestimationprocessuntilallthe213 weighted least square estimate( )0,i i are obtained. Recall the definition of coefficient iso that the sign and the absolute value of iindicates the geometric change over the 19 years. Since the estimation procedure of iinvolves all the spatial-temporalinformation,itisreasonabletohavethemplottedinakrigingmapto reveal the overall changes over 19-year period. 6. Kriging maps and time-change maps based on completed PM10 data Kriging map presentation is vital for a geostatistian's visualization, and maps reveal hidden information or the whole picture. A sample statistic is typically condensing the wide-spread informationintoanumericalpoint.While,akringingmapisactuallyamapstatistic(ora statisticalmap)whichcontainsinfinitelymanyinformationaggregatedfromlimited "sample"information(i.e.,observations).Krigingitselfisnotspecificallyprobability oriented,itisanotherweightedlinearcombinationprediction,butrequiresmore mathematicalassumptions.Infuzzygeostatistics,thefuzzykrigingschemehasalsobeen developed (Bardossy et al., 1990). Ordinarykriging(abbreviatedasOK)isalinearpredictor,seeCressie(1991)andMase (2011). The formula is ( )( )01Nj jjZ s Z s == (35) Advanced Air Pollution 206 where js arespatiallocationswithobservation ( )jZ s availableandthecoefficients jsatisfy the OK linear equation system ( )( )( )( ) ( ) ( )011, 1, 2, ,1Nj j i ijNjjs s s s i N == = ==

(36) The OK system is generated under the assumptions of an additive spatial model ( ) ( ) ( ) Z s s s = +(37) where(s)isthebasic(expected)spatialtrendand( ) s isaGaussianerror( )( )20, N s , i.e.,Gaussian variable with mean and variance ( ) ( ) ( )20, s V s s = = (38) respectively.Accordingly,thevariogram2 oftherandomerrorfunction( ) isjust defined by ( ) ( ) ( ) ( )22 h s h h = + (39) whereh istheseparatevectorbetweentwospatialpoints h + ands undertheisotropy assumption. 198919901991 Imprecise Uncertainty Modelling of Air Pollutant PM10 207 199219931994 199519961997 Advanced Air Pollution 208 199819992000 200120022003 Imprecise Uncertainty Modelling of Air Pollutant PM10 209 200420052006 2007 Fig. 5. Kriging Prediction Maps for PM10 in California 1989-2007. The213observationsitesnowhave19-yearPM10values,a"complete"datasetisnow available,containing4047datarecordsof213sitesover19years,andthenthe19ordinary krigingpred4ictionmapsaregeneratedforcomparisons.InFig.5,all19yearsofPM10 concentrationinCaliforniaStateareshown.Itisveryinterestingtoexaminethechangein PM10 concentrations through the 19 years, based upon the modelled complete 213 site data. In particular, 1998 shows to have an extremely low PM10 concentration. Although air quality isvariedovertheyears,butingeneral,thePM10concentrationisdecreasing,showingan improvement of air quality trend. Advanced Air Pollution 210 Fig. 6. Changes in PM10 values and the rate of change of PM10 in California between 1989 and 2007. As one can clearly see from Fig. 6, that PM10 concentration has clearly decreased over the 19 years, and air quality has improved remarkably over the years. The blue and green colours show negative changes, and red shows positive changes or near positive changes. Counties suchasSanDiego,Inyo,SantaBarbara,Imperial,stillshowanincreaseinPM10 concentration in the air, and indicate bad air quality. While Kern, Modoc, Siskiyou counties showthemostimprovementinairquality. TheleftmapinFig.6isPM10record difference between2007and1989ateachlocation,intotal231values,andthenadifferencemapis constructed. It is obvious that the difference map only utilizes 1989 and 2007 two-year PM10 records, 1990, 1991, ..., 2006 seventeen years' information do not participate the change map construction. The right map in Fig. 6 show completed, 1, 2, , 213ii = , the rate of change over 1989 to 2007 19-year period. Notethatthecalculationsof, 1, 2, , 213ii = involveallnineteenyearsbytemporal regression, the dependent variabley are estimated form the actual PM10 observations cross over all the available locations. Therefore, the rate of change parameter i at each individual location containsallspatial-temporalinformation.It isreasonabletosaytherateofchange parameter i isanaggregatestatisticforrevealingthe19-yearchangesover213locations. i krigingmapisthusdifferentfrom2007-1989krigingmaps.Thepositivesignof iindicatestheincreasingtrendinPM10concentration,whilethenegativesignf i indicates the decreasing trend in PM10 concentration. The absolute value of i reveals the magnitude ofchangeofPM10concentration.Itisworthtoreport,among213locations,193locations have negative i , while the negative i locations are 20 (9% approximately). 7. Discussion Air quality and health is always a central issue to public concern on the quality of life. In this chapter,weexaminedPM10levelsover19years,from1989to2007,intheCaliforniaState. Imprecise Uncertainty Modelling of Air Pollutant PM10 211 Facingthedifficulttaskofalackof"complete"PM10observationaldata,weutilisedthe inverse distance weight methodology to "fill" in the locations with missing values. By doing so,theimprecisenessuncertaintyisintroduced,whichisnotnecessarilyexplainedby probabilitymeasurefoundation.Wenotedthecharacterofaregionalizedvariablein geostatisticsandthereforeengageLiu's(2010,2011)uncertaintytheorytoaddressthe imprecisenessuncertainty.Inthiscase,wedevelopedaseriesofuncertainmeasuretheory foundedspatial-temporalmethodology,includingtheinversedistancescheme,thekriging scheme, and the geometric canonical process based weighted regression analysis in order to extract the change information from the incomplete 1989-2007 PM10 records. The use of the rate of change parameter alpha is a new idea and it is an aggregate change index utilized all spatial-temporal data information available. It is far better than classical change treatments. However,duetothelimitationsofourability,weareunabletodemonstratethedetailed uncertain measure based spatial analysis model. In the future research, we plan to develop a more solid uncertain spatial prediction methodology. 8. Acknowledgements IwouldliketothanktheCaliforniaAirResourcesBoardforprovidingtheairqualitydata used in this paper. This study is supported financially by the National Research Foundation of South Africa (Ref. No. IFR2009090800013) and (Ref. No. IFR2011040400096). 9. References Bardossy,A.;Bogardi,I.&Kelly,E.(1990).Krigingwithimprecise(fuzzy)variograms,I: Theory. Mathematical Geology, Vol. 22, pp. 6379. CaliforniaEnvironmentalProtectionAgencyAirResourcesBoard.(2010).AmbientAir Quality Standards (AAQS) for Particulate Matter. (www.arb.ca.gov) Cressie, N. (1991). Statistics for Spatial Data. Wiley-Interscience, John-Wiley & Sons Inc. New York. Deng,J.L.(1984).Greydynamicmodelinganditsapplicationinlong-termpredictionoffood productions. Exploration of Nature, Vol. 3, No. 3, pp. 7-43. Draper, N. & Smith, H. (1981). Applied Regression Analysis. 2nd Edition. John Wiley & Sons, Inc. New York. Electric,W.(1956).StatisticalQualityControlHandbook.WesternElctricCorporation, Indianapolis. EnvironmentalProtectionAgency(EPA).2010.NationalAmbientAirQualityStandards (NAAQS). U.S. Environmental Protection Agency. (www.epa.gov) Guo,D.;Guo,R.&Thiart,C.(2007).PredictingAirPollutionUsingFuzzyMembership Grade Kriging. Journal of Computers, Environment and Urban Systems. Editors: Andy P Jones and Iain Lake. Elsevier, Vol. 31, No. 1, pp. 33-51. ISSN: 0198-9715 Guo, D.; Guo, R. & Thiart, C. (2007). Credibility Measure Based Membership Grade Kriging. InternationalJournalofUncertainty,FuzzinessandKnowledge-BasedSystems.Vol.15, No. Supp. 2, (April 2007), pp. 53-66. B.D. Liu (Editor). ISSN 0218-4885 Guo,D.(2010).ContributionstoSpatialUncertaintyModellinginGIS.LambertAcademic Publishing (online.lap-publishing.com). ISBN 978-3-8433-7388-3 Guo,R.;Cui,Y.H.&Guo,D.(2010).UncertaintyStatistics.(SubmittedtoJournalof Uncertainty Systems, under review) Advanced Air Pollution 212 Guo,R.;Cui,Y.H.&Guo,D.(2010).UncertaintyLinearRegressionModels.(Submittedto Journal of Uncertainty Systems, under review) Guo,R.;Guo,D.&Thiart,C.(2010).LiusUncertaintyNormalDistribution.Proceedingsof the First International Conference on Uncertainty Theory, August 11-19, 2010, Urumchi &Kashi,China,pp191-207,Editors:DanA.Ralescu,JinPeng,andRenkuanGuo. International Consortium for Uncertainty Theory. ISSN 2079-5238 Kolmogorov,A.N.(1950)FoundationsoftheTheoryofProbability.TranslatedbyNathan Morrison. Chelsea, New York. Liu, B.D. (2007). Uncertainty Theory: An Introduction to Its Axiomatic Foundations. 2nd Edition. Springer-Verlag Heidelberg, Berlin. Liu, B.D. (2010). Uncertainty Theory: A Branch of Mathematics of Modelling Human Uncertainty. Springer-Verlag, Berlin. Liu, B.D. (2011). Uncertainty Theory, 4th Edition, 17 February, 2011 drafted version. Liu, S.F. & Lin, Y. (2006). Grey Information. Springer-Verlag, London. Mase,S.(2011).GeoStatisticsandKrigingPredictors,In:InternationalEncyclopediaof StatisticalScience.Editor:MiodragLovric,1stEdition,2011,LVIII,pp.609-612, Springer. Montgomery, D.C. (2001). Introduction to Statistical Quality Control. 4th Edition. John Wiley & Sons, Now York. Moore, R.E. (1966). Interval Analysis. Prentice-Hall, Englewood Cliff, NJ. ISBN 0-13-476853-1 Pawlak, Z. (1982). Rough Sets. International Journal of Computer and Information Sciences, Vol. 11, pp. 341-356. Shepard,D.(1968).Atwo-dimensionalinterpolationfunctionforirregularly-spaceddata. Proceedings of the 1968 ACM National Conference, pp. 517524. Utkin,L.V.&Gurov,S.V.(1998).Newreliabilitymodelsonthebasisofthetheoryof impreciseprobabilities.Proceedingsofthe5thInternationalConferenceonSoft Computing and Information Intelligent Systems, Vol. 2, pp. 656-659. Utkin,L.V.& Gurov,S.V.(2000).NewReliabilityModels Based onImpreciseProbabilities. AdvancedSignalProcessingTechnology,SoftComputing.FuzzyLogicSystems Institute(FLSI)SoftComputingSeries-Vol.1,pp.110-139,CharlesHsu(editor). Publisher, World Scientific. November 2000. ISBN 9789812792105 Walley,P.(1991).StatisticalReasoningwithImpreciseProbabilities.London:Chapmanand Hall. ISBN 0412286602 Zadeh, L. A. (1965). Fuzzy sets. Information and Control, Vol. 8, pp. 338-353. Zadeh, L. A. (1978). Fuzzy sets as a basis for a theory of possibility. Fuzzy Sets and Systems, Vol. 1, pp. 3-28. 13 Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region Alberto Mendoza1, Santosh Chandru2, Yongtao Hu2,Ana Y. Vanoye1 and Armistead G. Russell2 1Tecnolgico de Monterrey, Campus Monterrey,2Georgia Institute of Technology,1Mexico2United States of America 1. Introduction Airpollutioncontinuestobeanincreasingprobleminthelargestmetropolitanareasand regionalindustrialandcommercialcorridorsintheworld.ThisisalsothecaseinMexico. Current air quality trends in Mexico indicate that major urban centers continue to exceed the MexicanAmbientAirQualityStandards(MAAQS)forozone(O3)andparticulatematter with less than 10 microns of aerodynamic diameter (PM10), while other cities are starting to show warning signs of future air quality problems (Zuk et al., 2007). PM2.5 monitors are just starting to be deployed around the country, thus no extensive historical data is available on this pollutant. Someoftheurbancentersofconcernshareacommonairshedwithtwincitiesacrossthe internationalborderwiththeUnitedStatesofAmerica(USA),bringingadditional complexitytothestudyofairpollutiondynamicsintheregion.Inthissense,trans-boundaryairpollutionacrossUSAandMexicohasbecomearisingproblemdueto increasedcommercialandindustrialactivitiesintheborderregion.Trans-boundaryair pollutionhasbeenstudiedatdifferentlevelsindifferentareasoftheborderregion (Mukerjee, 2001). Two main areas can be identified as the ones that have drawn most of the attention. The first one is the Lower California Area: Tijuana/San Diego, Mexicali/Calexico-ImperialValley(Figure1).Here,mostoftheattentionhasbeenonprimaryPM(e.g., Osornio-Vargasetal.,1991;Chowetal.,2000;Sheyaetal.,2000;Kellyetal.,2010),with some studies addressing secondary pollutants (e.g., Zielinska et al., 2001). The second area is theairshedformedbyCiudadJuarez-ElPaso-SunlandPark.Perhaps,thisareaistheone that has received most of the attention regarding trans-boundary air pollution and in a more comprehensive fashion (Currey et al., 2005). Two of the key steps to improving air quality in a region are identifying, quantitatively, the emissions from sources that affect the area, and assessing how those emissions evolve in the atmosphere to impact pollutant concentrations. Both are difficult, and both can be subject to uncertainties.Airqualitymodelingiskeytobothstepsbecauseitprovidesameanstodo Advanced Air Pollution 214 Fig. 1. Location of the twin cities of Mexicali-Calexico and Tijuana-San Diego in the Mexico-US border region. bothinaconsistent,supportablefashion(Russell&Dennis,2000).Armedwithsuch information,policymakerscanthenidentifyenvironmentallyandeconomicallyeffective strategies to improve air quality (McMurry et al., 2004). Asindicated,trans-boundaryairpollutionhasbeenstudiedatdifferentlevelsindifferent areasoftheUS-Mexicoborderregion.However,limitedmodelingstudiesexistwhere comprehensivechemistry-transportairqualitymodels(CTMs)havebeenappliedata regionalleveltounderstandtrans-boundaryairpollutionintheMexicali-ImperialValley borderregion.InthepresentstudyweusedaCTMtodescribepollutantformationand transportaroundtheMexicali-ImperialValleyborderarea,aswellastoestimatesource contributions to O3 and PM2.5. Even though the principal attention in this study was on the Mexicali-ImperialValleyarea,wealsoexpandedourattentionoutsidethisareatotrack downpollutanttransportfrommajorurbancentersandpointsourcesoutsideit,butclose enough to affect the air quality of the valley (e.g., Tijuana in Mexico, and San Diego and Los Angeles [LA] in the USA). 2. Past air pollution studies in the Mexicali-Imperial Valley Border Area Tijuana-San Diego has been a border economical belt for a long time. However, over the last 15 years, Mexicali has been one of the fastest-growing cities in Mexico in terms of industrial development,jobcreation,andenergydemand(Quintero-Nezetal.,2006).Thishas resulted in that Mexicali on the Mexican side of the border is non-compliant with respect to CO,O3andPM10MAAQS,asCalexicoisinnon-attainmentforPM10,PM2.5,andO3USA standards. Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 215 HarmfulcontaminantsintheMexicali-ImperialValleyborderregionoriginatefroma numberofsources(Sweedleretal.,2003;Quintero-Nezetal.,2006),includingmotor vehicles,farms,powerplants(naturalgasfiredandgeothermal),andfactories.Light manufacturingoperations,wastedisposalsites,mining,andaggregatehandlingarealso locatedneartheborder.Inparticular,poorlymaintainedvehiclescontributeheavilytothe levels of CO, NOx (NO+NO2), and hydrocarbons (HCs) in the air; driving on unpaved roads contributesheavilytoPMemissions.Burningoftrash,tires,andothermaterialsarealso sources of PM, SO2, and CO. Several studies have been conducted to understand the composition, spatial variability, and sourcesofairpollutionintheMexicali-ImperialValleyregion.CerroPrieto,thelargest geothermalplantinLatinAmerica(720MW)islocated~30kmtothesouthofdowntown Mexicali.Sinceitstartedoperationsinthe1970s,H2Semissionsandtransportfromthis facilitytoMexicaliandImperialValleyhasbeenaconcern(Gudiksenetal.,1980;Deane, 1984).However,atmosphericconversionofH2StoSO2wasestimatedasnotsignificant.A majoreffortto understandPM10pollution inMexicaliandImperialValleywasundertaken intheearly1990s(Chowetal.,2000;Chow&Watson,2001;Watson&Chow,2001).This study demonstrated that PM10 in the region is mainly composed of crustal material (50% to 62%ofthemass)andorganicmatter(over25%ofthemass).Receptormodelinggave evidencethatpollutiontransportfromLAtoMexicaliandCalexicocouldbeaconcern.In addition, preliminary pollutant flux estimates indicated that the total PM10 flux from Mexico totheUSAwasabout1.5timesthetotalfluxfromtheUSAtoMexico.PM10levelsin Mexicali are consistently higher than in Imperial Valley, however wind patterns tend tobe inahigherpercentagefromthenorth.Otherstudieshavealsogivenevidenceofthe potentialtransportofemissionsoriginatinginMexicaliandImperialValleytoareastothe northliketheGrandCanyonNationalPark(Eatoughetal.,2001).However,theseresults have relied on the use of receptor models rather than comprehensive CTMs. Morerecently,thefactthattheMexicaliValleyandImperialValleycontinuetoexperience highairpollutantslevelsmadeitrelevanttoconductan integratedstudyoftheairquality problemintheregion.Partialresultsofthisintegratedstudyhavebeenpublished elsewhere,particularlyonlevelsandchemicalcompositionoffinePM(Mendozaetal., 2010),chemicalspeciationandsourceapportionmentofVOCs(Mendozaetal.,2009), mobile source emissions characterization using a mobile laboratory (Zavala et al., 2009), and numericalexperimentstoaddressthemeteorologicalpatternsthatfosterairpollution episodes(Vanoye&Mendoza,2009).Herewepresentourfindingsontheapplicationofa regionalthree-dimensionalcomprehensiveCTMtotheMexicali-Calexicoborderregionto followthedynamicsofgas-phaseandparticulate-phaseairpollutants.Particularemphasis is placed on the relevance of understanding trans-boundary air pollutants transport and the implications on emission control strategies on both sides of the border. 3. Description of the modeling system and its application 3.1 Modeling platform Three-dimensionalCTMscontinuetobethemostscientificallysoundtooltoassesshow emissions from multiple sources impact air quality (Russell & Dennis, 2000). The modeling effortinthisstudyconsistedintheapplicationofanadvancedairqualityandemissions modeling system to the border region to assess how particular sources impact O3 and PM2.5 levels.Specifically,weappliedanextendedversionoftheModels-3suite,includingthe Advanced Air Pollution 216 CommunityMultiscaleAirQualitymodel-CMAQ-(Byun&Ching,1999)forairquality modeling, the PSU/NCAR 5th generation Mesoscale Meteorological model -MM5- (Grell et al., 1995) for meteorological modeling, and the Sparse Matrix Operator Kernel for Emissions model -SMOKE- (Houyoux & Vukovich, 1999) was used to process emissions. 3.2 Modeling domains Themodelingsystemwasappliedusingnestedgrids(Figure2).Atthecoarserscale, horizontally, 36 km 36 km grid cells were used. This mother domain is the same as the one defined by the Regional Planning Organizations (RPOs) that oversee other major modeling effortsinNorthAmerica(i.e.,VISTAS,WRAP,CENRAP,LADCO,MANE-VU).Atthe horizontalmid-resolutionlevel,weincludeda12km12kmgrid(Figure3).Thecoarse gridsystemallowsrelativelyrapidsimulationtosetappropriateboundaryconditionsfor the finer grid, serves to stabilize the meteorological and air quality model solutions, and to considerlong-rangetransportfromveryparticularsources.A4km4kmgridwas specifiedintheMexicaliareaforsimulationsthatsuggestedthatfinescalefeaturesexisted andcouldnotbeaccuratelyrepresentedusingthecoarsergrids(i.e.,capturethedynamics attheurbanscaleinMexicali-Calexico).Inthiswork,weonlypresenttheresultsobtained withthe12kmx12kmgrid,whichareonesthanprovidedetailsonthemid-range pollutants transport in the border region of interest. Details on the extent of each modeling domain are presented in Table 1. The horizontal resolution (including grid nesting) was kept consistent between MM5, SMOKE and CMAQ. 3.3 Episodes selection The adequate selection of modeling episodes constitutes a fundamental part of the modeling process.Ifrepresentativeepisodesarenotselectedadequately,themodelingresultsmight notcharacterizeeffectivelythemeteorologicalfeaturesthatfosterhighpollutionlevel episodes. Here we used Classification and Regression Tree (CART) Analysis (Breiman et al., 1998)astheformalstatisticaltooltoselectthemodelingepisodesofinterest.Inessence, CART is a recursive binary partition technique. It divides a set of observations in subgroups takingasreferencethevalueofaparticularvariabledefinedbytheuser(e.g.,maximum daily ozone concentration). Each partition in the decision tree is conducted to minimize the classification error of the decision variable. This technique has demonstrated its capacity to help in the selection of days with similar meteorological conditions that give rise to similar pollutionlevels,usingaformalprocedureandeliminatingtheeffectsofmeteorological variability (Kenski, 2004). CARTwasappliedtoobtaindecisiontreestoclassifydailymaximumO3,COandPM10 concentrations (separately). The database used was composed of observations (chemical and meteorological)takenbyairqualitystationsintheborderregionfortheyears2001and 2002.ThepurposewastogroupdayswithsimilarO3,CO, andPM10levelsandinfluenced bysimilarmeteorologicalcondition.TheresultsobtainedfromCARTapplicationwere comparedagainsttimeseriesplotstocorroboratethattheepisodesselectedinfact represented a continuum of days with relatively high pollutant concentrations levels. One of the parameters that can be manipulated while applying the CART technique is the number offinalbinsthatthedecisiontreewillhave.Typically,asthenumberofbinincreasesthe errorisreduced;however,ifthenumberofbinincreasestheprobabilityofhaving consecutive days in a bin decreases and thus it is harder to construct episodes for air quality Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 217 Fig. 2. Horizontal resolution of the nested modeling domain. Fig. 3. Location of the 12 km and 4 km horizontal modeling domains. Advanced Air Pollution 218 Grid IDOrigin (x, y) in km Horizontal domain (# columns, # rows) USMEX36(2736.0, 2088.0)(148,112) USMEX12(2232.0, 1160.0)(84,75) USMEX4(1908.0, 756.0)(63,54) Table 1. Modeling domains specifications. Origin coordinates are based on a Lambert Conformal Conic projection with centre lat. and long. as 40 and 97 degrees, respectively modelingpurposes.Aconvenientnumberofconsecutivedaysforamodelingepisodeis between 10 and 15, so with this in mind the number of bins was varied until decision trees withlowclassificationerrorsandhighnumberofconsecutivedaysinthebinswere obtained. BasedontheCARTAnalysisapplication,thefollowingmodelingepisodesweredefined: August 18-27, 2001 and July 17-25, 2001 for high O3 events, and January 6-16, 2002 for high COandPMeventsthataretypicalduringautumnandwintertimes.Additional detailson the episode selection process can be found elsewhere (Vanoye & Mendoza, 2009). 3.4 Emissions modeling SMOKE is a computational engine used to generate the gridded emissions inventory, and its mainpurposeistospeciateandallocatespatiallyandtemporallyareaandpointemissions andtocoupleemissionestimationtoolsformobileandbiogenicemissionstospatialand temporal allocation routines. BaseemissionsinventorydatafortheUSAside oftheborderwereobtainedfromthe2001 USNationalEmissionsInventory(NEI)preparedfortheCleanAirInterstateRule(CAIR). Emissions for the Mexican side came from combining the 1999 BRAVO Mexican inventories (Pitchfordetal.,2004)withthe1999SixBorderStatesMexicaninventory(MNEI)(ERGet al.,2004).BiogenicemissionsforbothsidesoftheborderwerepreparedusingBEIS3 (Vukovich&Pierce,2002),andUSAmobileemissionswerepreparedusingMOBILE6(US EPA, 2003). Mobile emissions for Mexico were directly obtained from BRAVO and MNEI. The emissions inventory generated considers O3 and PM precursors, as well as primary PM emissions and some toxics (particularly VOC species). The modeling episodes selected were notthesameonesasthebaseyearsusedtoderivetherawemissionsinventoriesusedfor the Mexican side of the border; thus, scaling was needed to update the emissions (e.i., MNEI base year is 1999 and modeling years for our applications were 2001 and 2002). This scaling wasbasedprimarilyonpopulationgrowth.VOCsspeciationwasconductedbasedonthe chemicalmechanismselectedfortheCTMapplication:SAPRC-99chemistry(Carter,2000). Spatialsurrogateratiosusedtoallocateemissionsonbothsidesoftheborderconsidered population, highways, total railroads, airport points, and marine ports. AsanexampleoftheresultsobtainedfromtheapplicationofSMOKE,Figure4illustrates COandbiogenicisopreneemissioninventoriesforthe12kmresolutiondomain.Itcanbe seen,forexample,thattheCOemissionsinventorycontainstheexpectedspatialstructure (mainroadsareclearlyshownandemissionsfollowgeneralpopulationpatterns).Overall, mobilesourcescontributeto~65%oftheNOxand~30%oftheVOCsemittedintheLA area;areasourcesrepresent~15%oftheNOx and25%oftheVOCsemittedinLA.Values forSanDiegoaresimilarastheonesforLA.Incontrast,34%oftheNOxand13%ofthe VOCsemittedinMexicalicomefrommobilesources;areasources(includingnon-road Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 219 mobilesources)represent37%oftheNOxand51%oftheVOCsemittedinMexicali.In Tijuana, mobile sources emit 61% of the NOx and 23 of the VOCs, whilst area sources emit 29% of the NOx and 60% of the VOCs. PM in the Mexicali-Calexico region comes from area sources that are dominated by wood-fuel combustion, agricultural burning, and paved and unpaved road dust. Fig. 4. Examples of (a) CO and (b) biogenic isoprene emissions allocated in the 12 km resolution modeling domain. 3.5 Meteorological modeling MM5 (Grell et al., 1995) version 3.7 was the meteorological model used here to develop the fieldsneededtodrivetheCTMsimulationsandtoprovidemeteorologicalinformation neededtoestimatemeteorological-variableemissions(e.g.,biogenicemissionsdependon solar radiation and temperature, while mobile emissions depend on temperature). MM5 is a non-hydrostaticmesoscalemeteorologicalmodelwithgridnestingandfour-dimensional dataassimilationcapabilities.Herewebrieflydescribethemodelsetupandtheinputdata usedtorunthemodel.Additionaldetails,includingmodelperformancestatisticsonthe MM5 application, can be found elsewhere (Vanoye & Mendoza, 2009). MM5wasrunwith34verticallayerswiththetopofthedomainsetat70mb;horizontal resolutionwasdescribedearlier.Followingasetofsensitivitytests,theMM5 parameterizationconfigurationthatgavethebeststatisticalperformanceofthemodelfor theJulyandAugustepisodesispresentedinTable2.Ofnote,thePleim-XiuLandSurface modelistherecommendedschemebytheUSEnvironmentalProtectionAgency(USEPA) andistheonethathasdemonstratedtogivethebestmeteorologicalfieldsforCMAQ (Olerud & Sims, 2003; Morris et al, 2004). Another advantage of the Pleim-Xiu scheme is that itallowsusingCMAQsdrydepositionschemewhichistechnicallysuperiortothe conventional Wesley scheme. MM5wasexecutedenablingitsfour-dimensionaldataassimilationcapabilitiesforthe36 kmand12kmdomains.One-waynestingwasselectedasthewayMM5transferred information from the outer grids to the inner grids. Finally, a relaxation scheme was chosen forthemanipulationoftheboundaryconditions,i.e.thefiveoutermostpointsareusedto damp the information flowing from the boundaries to the inner domain. InitialandboundaryconditionwerepreparedusingtheNationalCenterforAtmospheric Research(NCAR)/NationalCentersforEnvironmentalPrediction(NCEP)Etaanalyses Advanced Air Pollution 220 data.ThedataconsistofregionalmeteorologicalanalysesforNorthAmericabasedonthe output of the Eta model, which generates data every 12 hours from observations of over 600 stations in the region. To complement this information and increase the effectiveness of the dataassimilationstep,additionalobservationswithatemporalresolutionof6hourswere extractedfromNCARarchives,throughitsDataSupportSectionoftheScientific ComputingDivision.Thisincludedobservationsfromsurfaceandmarinestations,aswell as from aerial soundings. Basic landuse, vegetation cover and topography was also obtained fromNCAR.LanduseinformationwasbasedontheUSGS24categories,andtopographic resolutions of 10 min, 5 min, 2 min, and 30 sec were used. Parameter IDDescriptionSelected option IMPHYSExplicity Moisture SchemeMix Phase MPHYSTBL IntrinsicExponentforCalculating IMPHYS UseLook-uptableformoist physics ICUPACumulus SchemesGrell IBLTYPPlanetary Boundary Layer Pleim-Xiu FRADRadiation Cooling of AtmosphereRapid Radiative Transfer Model ISOILMultilayer Soil Temperature ModelPleim-Xiu Land Surface Model ISHALLOShallow Convection OptionNo Shallow Convection Table 2. MM5 parameterization options that gave the best model performance for the simulation of meteorological conditions in the Mexicali-Imperial Valley border area. 3.6 CMAQ application 3.6.1 Base case simulations CMAQisanEulerianphotochemicalmodelthatsimulatestheemissions,transport,and chemical transformations of gases and PM in the troposphere (Byun & Ching, 1999). Similar to other photochemical models, CMAQ solves the species conservation equation: ( ) ( )ii i i iCC C R Et= + + +u K (1) where,Ciistheconcentrationofspeciesi,uisthewindfield,Kistheeddydiffusivity tensor,Riisthenetrateofgenerationofspeciei,andEiistheemissionrateofspeciesi. MeteorologicalparameterssuchasuandKineq.1,aswellastemperatureandhumidity fields come from the MM5 application, while emission rates from SMOKE. CMAQ contains state-of-the-sciencedescriptionsofatmosphericprocessesandhasaone-atmosphere approachforfollowingthedynamicsofgas-phaseandparticulatematterpollutants.The latter is an important characteristic to assess simultaneously O3 and aerosols. CMAQ, as MM5, allows for grid nesting. The horizontal grid structure used was described earlier.Theverticalstructureofalldomainshas13layerswithitstopatabout15.9km above ground. Seven layers are below 1 km and the first layer thickness is set at 18 meters. Initialandboundaryconditionsusedforthemotherdomainwerethesameastheones suggestedbyotherRPOapplications(Russell,2008).Resultsfromthesimulationswere comparedtodatafromground-basedmonitorsforNO2,O3,CO,SO2,PM10,andPM2.5. ObservationaldatawasobtainedfromtheCaliforniaAirResourcesBoard(CARB)for Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 221 monitoringstationsintheStateofCalifornia,andfromMexicanbordermunicipalitiesof interest(i.e.,TijuanaandMexicali).ObservationaldatawerealsoobtainedfromUSEPAs Air Quality Data system. In each episode, the first two days were considered ramp-up days and were not further used for additional analysis. CMAQhasbeenusedextensivelytostudyairpollutantdynamicsinthecontinentalUSA (e.g.,Tagarisetal.,2007;Liaoetal.,2007)andinparticularregionsofthatcountry(e.g., Dennisetal.,2010;Ying&Krishnan,2010),aswellasinothercountriesaroundtheworld (e.g.,Cheetal.,2011;Imetal.,2011).OnlyoneadditionalapplicationusingCMAQasthe CTM has looked at trans-boundary air dynamics in the USA-Mexico border using fine scale grid resolutions. Choi et al. (2006) looked at high PM events over the sister cities of Douglas, Arizona(USA)andAguaPrieta,Sonora(Mexico).Inthatapplication,modelperformance wasacceptable,anditwasconcludedthatsecondaryprocessescontributedmarginallyto the modeled PM events. Primary local sources dominated high PM events. 3.6.2 Sensitivity analysis Sensitivityanalysisisanimportanttoolthatcanbeusedtounderstandtheimpactsof emissionsfromvarioussourcesonambientairconcentrationsofspecificpollutants.The abilitytoconductsensitivityanalysesinanefficientfashioniscriticaltoobtainrobust descriptionsoftheresponsefieldsofpollutantconcentrationstochangesinmodelinputs (particularlyemissions),whichthenareusedinsourceattributionanalysesandcontrol strategydesign(e.g.,Berginetal.,2007).Amongthedifferentchoicestoestimatethe sensitivity fields, the direct decoupled method for three-dimensional models (DDM-3D) has proven to be superior to other techniques (Yang et al., 1997; Hakami et al., 2003). DDM-3D is an implementation of the Decoupled, Direct Method (Dunker, 1984; Dunker et al., 2002) for sensitivityanalysis.TheversionofCMAQusedinourapplicationswasextendedwith DDM-3D (Cohan et al., 2005). The method directly calculates the response of model outputs (concentrations) to parameters and inputs, i.e., the semi-normalized sensitivities Sij: iijjcSe=(2) where ci is the concentration of species i and ej is the relative perturbation on parameter j pj-(e.g.,NOxemissions)fromitsnominalvaluepj(i.e.,ei=pj/pj).Thisisanefficientapproachfordirectlyassessingthesensitivityofmodelresultstovariousinputsand parameters, and replaces the need to use the traditional brute force approach of re-running a modelaftermodifyingaparameter.Moreimportantly,itdoesnotsufferfromnumerical noiseproblemsthatcanoverwhelmbruteforceapproaches.Inaddition,itisalinear method.Inpriorstudies,theatmosphericchemistryhasbeenfoundtorespondrelatively linearly for emissions changes on the order of 25% or more (Dunker et al., 2002; Hakami et al., 2004). Ofparticularinterestwastoexplorethesensitivityduetovariationsontheemissions inventory.TheimplementationofDDMtoCMAQallowsdefiningspatial-andsource-specificemissionscategoriesastheinputbeingperturbed,andinasinglemodelrunthe sensitivities of all species tracked by the model to changes in a set of emissions sources can becalculated.Toaccomplishthis,theregionofinterestwasdividedintothreedifferent areas:Mexicali-Calexico(abbreviatedasMXC),Tijuana-Tecate-SanDiego(abbreviatedas Advanced Air Pollution 222 TSD),andLosAngeles-Riverside-Orange-Ventura(abbreviatedasLAR).O3sensitivitiesto area-, mobile-, and point-source emissions of NOx and VOC were calculated for each of the regions defined for both summer episodes. Additionally, PM2.5 sensitivities to changes in the same source categories were calculated. 4. Results 4.1 Air quality model performance Domain-wide episode performance statistics were determined to ascertain the confidence of thesimulationresults.Table3presentstheaveragemodelperformanceforthe12km domainintermsofMeanBiasError(MBE),RootMeanSquaredError(RMSE),Mean NormalizedBias(MNB),andMeanNormalizedError(MNE).Establishedperformance guidelinesindicatethatthemodelshouldhaveaMNBof515%,andaMNE3035%for O3 (Tesche et al., 1990). Based on these guidelines, CMAQ performed well in predicting the observed O3 concentrations. No guidelines exist for the rest of the gas-phase species, though theresultsarecomparabletoresultsobtainedbyothersusingdifferentCTMs(e.g., Mendoza-Dominguez&Russell,2001).Overall,thegas-phasespeciesresultsindicatea tendency of the model to underestimate the pollutant concentrations. This is in line with the resultsobtainedfrommobilelaboratorymeasurementsthatindicateanunderestimationof theofficialemissionsinventoryforMexicali(Zavalaetal.,2009).PMprovedtobemore difficult to simulate correctly, which is a known setback of current CTMs (Russell, 2008). MBERMSEMNB (%)MNE (%) August-01 O31.64E-031.60E-020.2119.7 CO3.52E-016.56E-0118.663.2 NOx1.25E-022.52E-0234.974.2 SO21.40E-035.40E-0319.487.7 PM2.56.76E+009.14E+0036.939.2 PM103.00E+013.57E+0176.276.2 July-01 O39.19E-021.76E-0118.361.2 CO1.43E+001.97E+0022.056.8 NOx6.18E+007.58E+0031.060.0 SO25.31E+006.53E+0029.260.2 PM2.57.10E+008.72E+0033.363.0 PM108.12E+009.97E+0035.359.5 January-02 O32.86E-037.76E-037.014.1 CO7.33E-011.33E+0026.573.1 NOx3.99E-027.98E-0234.676.0 SO21.10E-035.40E-0329.674.4 PM2.52.54E+001.01E+014.637.2 PM102.53E+013.21E+0156.161.7 Table 3. Average performance metrics for the 12 km domain during the diferent episodes modeled. MBE and RMSE are in ppmv for gas-phase species and g/m3 for PM species. Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 223 Peak O3 concentrations in Calexico during the August episode were observed at Ethel Street (MBE 4.0 ppbv, MNB 3.0%), and East Calexico (MBE 5.0 ppbv, MNB 12.0%) sites. Calexico andMexicali,beingadjacenttoeachotherintheborderregion,experiencesimilarO3 concentrations. The inability to capture the minimums in Ethel Street and Calexico East sites canbeattributedtothefactthatboththeselocationsarelocatedclosetoroadways,hence experience strong O3 sinks in the night time due to its reaction with NO which the model is unable to capture using the 12 km grid structure (Figure 6). 00.020.040.060.080.10.120.141 16 31 46 61 76 91 106 121 136 151 166 181 196 211 226 241Time (hr)O3 Conc. (ppmv)00.010.020.030.040.050.060.070.080.090.11 16 31 46 61 76 91 106 121 136 151 166 181 196 211 226 241Time (hr)O3 Conc. (ppmv)ObservedSimulateda b Fig. 6. Observed vs. simulated O3 concentrations at representative sites in Calexico during August 2001 at (a) Ethel Street and (b) Calexico East site. Time scale represents hours of simulation. Also,theLARareawastestedformodelperformancewithrespecttospatialaswellas temporalvariability(Figure7).Thisregionwaschosenbecauseofthehighdensityof monitoringstationslocatedinit,whichcangiveamorerealisticcomparisonwiththe modeled average hourly concentration values in the region. During the August episode, the peakO3concentrationsoccurredonAugust26th,2001:189ppbvattheAzusasiteand190 ppbv at the Glendora Laurel station, respectively. These sites are located in the San Gabriel Valleyandcomeunderthesame12kmgridcell.Simulatedconcentrationscorrelatedwell withtheobservedconcentrationsatAzusa(MBE5.0ppbv,MNB12.2%)andGlendora Laurel(MBE0.0ppbv,MNB3.5%)onmostdays.FortheJanuaryepisode,January12and 13, 2001 were the high concentration days with peaks of ~60 ppbv in the Los Angeles area, and ~75 ppbv in the Mexicali area. 00.020.040.060.080.10.120.140.160.180.21 16 31 46 61 76 91 106 121 136 151 166 181 196 211 226 241Time (hr)O3 (ppmv)00.020.040.060.080.10.120.140.160.180.21 16 31 46 61 76 91 106 121 136 151 166 181 196 211 226 241Time (hr)O3 (ppmv)ObservedSimulateda b Fig. 7. Observed vs. simulated O3 concentrations at representative sites in LA, during August 2001 at (a) Azusa, (b) Glendora Laurel. Time scale represents hours of simulation. Advanced Air Pollution 224 4.2 Modeled pollutant concentration fields ResultingO3fieldsfortheJulyandAugustepisodesillustratetheinfluenceofregional transportacrossthedomain.DuringtheJuly2001episode,apeakof125ppbvO3was simulated in the LA area on July 23, 23:00 hrs UTC (Figure 8d). The plume from LA can be seentransportedtowardstheeast(Figure8a-c).Plumesofupto78ppbvO3emergefrom San Diego-Tijuana and travel eastwards and reach the Mexicali-Calexico region (Figure 8 a-c). Peak PM2.5 concentrations of over 50 g/m3 were simulated in the LA area on July 15th, whilePM2.5concentrationsdidnotexceed15g/m3inTijuana-SanDiegoandMexicali-Calexico during the July episode. OnAugust24th(20:00hrsUTC),strongO3plumesstartedtodevelop,andplumesfrom Mexicali-Calexico and Los Angeles almost converged (Figure 9a). At the same time plumes fromTijuana-SanDiegobuildupaswellandmoveeastwardstowardsMexicali-Calexico (Figure9b).SimilarpatternsstarttoemergeonAugust25th(19:00hrsUTC)(Figure9c); peaksreach162ppbvintheLAareaonAugust26th(00:00hrsUTC)(Figure9d).Apeak concentrationof162ppbvisreachedabout30kmnorthwestfromtheGlendoraLaurelsite where a simulated peak of 144 ppbv is reported. Similar to the July episode, O3 plumes from SanDiego-TijuanaborderareaaretransportedeastwardtowardsMexicali-Calexicoduring the August episode as well. Peak PM2.5 concentration of 100 g/m3 are seen close to the LA area on August 25th. In theMexicali-Calexico region, Mexicali showed a peak of 42g/m3 on August 25th (14:00 hrs UTC). a bcd Fig. 8. Regional dynamics of O3 plumes during the July 2001 episode (see text for details). Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 225 a bc d Fig. 9. Regional dynamics of O3 plumes during the August 2001 episode (see text for details). In addition to O3, the dynamics of other primary (e.g., CO, NO2, and SO2) and secondary PM species (e.g., sulfate in fine PM) are also of interest when analyzing the output data obtained fromCMAQ(Figure10).FortheAugustepisode,COconcentrationspeakaround7PM (PDT),withLAshowingthehighestconcentration,followedbySanDiego-Tijuana.CO levelsinMexicali-Calexicoarelowerandmorelocalized.Ingeneral,NO2distributionis verysimilartothatofCO,highlightingtheimportanceofmobilesourceemissions.SO2 emissionsarehighestintheTijuanaregion.Thus,withthewindblowinginthenortheast directionduringthemorninghours,muchoftheSO2istransportedinlandintotheSan Diegoregion.Consequently,thesulfateaerosolshaveahighregionaleffectencompassing the whole of San Diego region, and also showing its effect on Imperial Valley and Mexicali during late evening hours. As PM concentrations are a major concern during the winter season, we limit our discussion oftheJanuary2002episodetoPM2.5.Apeakof188g/m3wassimulatedonJanuary12, 2002(18:00hrsUTC)nearLA.ThemovementofregionalPM2.5plumesisrepresentedin Figure11.PlumesfromSanDiego-Tijuana,LAandLasVegasmovetowardstheMexicali-Calexicoregionwithimpactsof10to35g/m3.PM2.5originatedintheUSAand transportedtoMexicali-Calexico,alongwithlocalfreshemissions,iscarriedfurther southeastinsideMexico.Mexicali-CalexicoshowspeakPM2.5concentrationof50g/m3. Primary organic mass was the main contributor to fine PM in LA (98 g/m3). The maximum contribution from primary organic matter to the fine PM in Mexicali-Calexico was 10 g/m3. Peak soil dust concentrations of 40 g/m3 were found in Pheonix and Las Vegas areas. The soil dust contributions from LAR, TSD and MXC range between 5-25 g/m3 (Figure 11). Advanced Air Pollution 226 a bc d Fig. 10. Concentration fields for gas-phase and aerosol species (August 27, 2001): a) CO, b) NO2, c) SO2, d) sulfate PM2.5. a bc d Fig. 11. Dynamics of PM2.5 plumes during the January 2002 episode (see text for details). Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 227 4.3 Source contribution 4.3.1 Source contribution to specific NOx and VOC emission sources: August episode Tounderstandsourcecontributionsinthemodelingdomain,sensitivityfieldswere estimatedusingCMAQ/DDM.ResultsarepresentedonlyfortheAugust2001episode; values for the July 2001 episode were similar. First, the sensitivity of the regional O3 field to changes in NOx or VOC emissions from specific sources is presented. The figures presented areresponsesurfacesandareinterpretedastheamountofincrementinpollutant concentrationper10%increaseinemissions(oramountofreductionper10%decreasein emissions)fromcertainsource.Thesensitivitycoefficientsarelinear(firstorder)innature andthuscanbeusedinthemannerdescribed.Ingeneral,isreasonabletoimplyalinear responseoverarangeofemissionsperturbations(30%)evenforspeciesthatitiswell known their non-linear response in the atmosphere (e.g., O3; Hakami et al., 2003). TheimpactofNOxemissionsfromMXCwasthehighestonAugust26,withsensitivity responsereaching9ppbvofO3per10%changeintheemissions(Figure12).Theareaof influenceofNOxemissionsextendsnorthwardsintoImperialValley,andpartiallyinto Arizona. The results indicate that down-wind the atmosphere is not NOx-inhibited; that is, an increase in NOx does not give as a response a decrease in O3 down-wind as has been the case in other areas of the Mexico-US border (Mendoza-Dominguez & Russell, 2001). Change inVOCemissionsfromsourcesinMXCproduceasmallerchangelocalizedinO3 concentrations (maximum of 3 ppbv per 10% change), indicating the benefits of NOx control over VOC control in the region. a) b) Fig. 12. Maximum sensitivity of O3 to (a) NOx emissions and to (b) VOC emissions from the MXC region during the August 2001 episode. Theinfluenceofemissionsfromothergeographiclocationswasalsotested.Figure13a illustratestheresponseofO3tochangesinNOxemissionsfrommobilesourceslocatedin the TSD area. The highest impact is almost 17 ppbv, occurring near San Diego, with a strong influenceintheMXCregionaswell.Thisresultindicatesthatemissioncontrols implementedinSanDiego(orincrementinemissions)willimpacttheMXCarea accordingly.OfinterestisalsothesmallNOx-inhibitedregionlocateddown-windofSan Diego,towardtheTijuanaborder,thatimpliesthatadecreaseinmobileemissionswill resultinanincreaseinO3concentrations.Incontrast,andasexpected,impactfromVOCs Advanced Air Pollution 228 emittedbymobilesourceslocatedintheTSDareaislimitedtolessthan2ppbvper10% changeinemissions(Figure13b).Thespatialextentofinfluenceisalsomorelimitedthan the sensitivity to NOx emissions, influencing the northern region of Imperial Valley County. a) b) Fig. 13. Maximum sensitivity of O3 to (a) mobile NOx emissions and to (b) mobile VOC emissions from the TSD region during the August 2001 episode. Finally,examplesensitivityvaluesduetochangesinNOxemissionsfrommobileand area sourcesfromtheLARregionarepresented(Figure14).Forthecaseofmobilesources,the incrementofNOxemissionsresultsinadecreaseinozone(~7.5ppbvper10%increasein emissions)indowntownLA,withacorrespondingincrease(~30ppbv)inneighboring countiesofVentura,Orange,andRiverside.Fromtheextentofthesensitivityfield,itis possible that under the right meteorological conditions, the influence can reach the Imperial Valleyarea.Ontheotherhand,thesensitivitytoareasourceNOxissmallerinvalueandextentbecauseareaemissionsaresmallerthanthemobileemissionsinSouthern California. Fig. 14. Maximum sensitivity of O3 to (a) mobile NOx emissions and (b) area NOx emissions located in the LAR area during the August 2001 episode. Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 229 4.3.2 Source contribution to overall emission sources: August episode When considering the overall emissions from mobile sources, the LAR area made an overall contributionof44ppbvonthesurroundingregioni.e.,eastofthecityofLAtowards GlendoraLaurelandAzuzaonAugust26(00:00hrsUTC)(Figure15c).Presenceofhigh concentrationsofNOxresultsinnegativesensitivitiesupto46ppbvinurbanLA(Figure 15a).AsseeninthebasecasesimulationswhereO3plumesfromLAR,MXCandTSD formedatriangleoversouthernCalifornia,theO3sensitivityfieldsextendstowardsMXC withincrementsofupto10ppbv(Figure15a).Duetothenortheasterlydirectionofthe winds,plumesalsoreachtheGrandCanyonNationalParkarea,againwithincrementsof about 10 ppbv (Figure 15d). LAR area sources contribute up to 8 ppbv of O3 in the Riverside area. MobiletrafficpassingthroughMexicalisbordercrossingsisofconcern.However,the overall mobile contribution to O3 is found to be small in the simulation results. The impact fromMexicalivehiclesaloneisverysmall,withapeakimpactofonly1.3ppbvover CalexicoandMexicali(Figure16a).Possibleemissioninventoryunderestimatescanbea potential reason for low simulated impacts, and should be further explored. a) b)c) d) Fig. 15. O3 sensitivity to LA mobile source emissions (see text for details). Advanced Air Pollution 230 The maximum impact from Calexico mobile sources is 2 ppbv of O3 seen over the Calexico region itself, and the border between California and Arizona (Figure 16b). The primary areas of mobile emissions are the two border crossing areas (seen in blue as negative sensitivities). AreasourcesinMXCcontributeasimulatedmaximumof8ppbvO3duringthesummer episode(Figure17a).TheareaofinfluencecanbeseenencompassingCalifornia,andthe borderregionsofCalifornia-Arizona.O3impactsupto4ppbvintheGrandCanyonarea can be attributed to area sources in the Mexicali-Calexico region (Figure 17b). a) b) Fig. 16. Contribution to O3 concentrations from (a) Mexicali and (b) Calexico mobile sources. a) b) Fig. 17. Contribution to O3 concentrations from MXC area sources (see text for details). AreasourcesfromTSDhaveapeakimpactof40ppbvofO3overtheSanDiegoareaand this plume is carried eastwards into the USA close to the border region. The contribution of Tijuana emissions extends to the southeast into inner Baja California and impacting up to 20 ppbvofO3(Figure18a,b).Also,onAugust26th,thesensitivityfieldofO3fromtheTSD regionextendseastwardstowardsCalexico,thusaddingO3tothealreadypollutedairin Calexico and Mexicali (Figure 18c,d). Tijuana mobile source impacts reach up to 6 ppbv on both sides of the border depending on the wind direction (Figure 19b,c). Since the dominant wind pattern is towards the northeast, Modeling the Dynamics of Air Pollutants:Trans-Boundary Impacts in the Mexicali-Imperial Valley Border Region 231 a) b)c) d) Fig. 18. O3 sensitivity to TSD area sources during the August 2001 episode (see text for details). O3istransportedthroughtheCalifornia-BajaCaliforniabordertowardsCalexico(Figure 19a,c). Tijuana mobile sources impacts up to 3 ppbv of O3 in Mexicali-Calexico (Figure 19a). It is also of interest the areas of negative sensitivity observed in downtown Tijuana of more than 3.0 ppbv. Mobile sources from San Diego contribute up to 26 ppbv of O3 in the region itself, and also over the park areas such as Anza