AtmosphericPollutionResearch1(2010)102Ͳ111
Atmospheric Pollution Research www.atmospolres.com MethodologicalguideforimplementationoftheAERMODsystem withincompletelocaldata
LeonorMariaTurtosCarbonell,MadeleineSanchezGacita,JosedeJesusRiveroOliva, LarisaCurbeloGarea,NorbertoDiazRivero,EliezaMenesesRuiz
CUBAENERG7A,Calle20#4111%18Ay47,Playa,11300,CiudaddelaHabana,Cuba
ABSTRACT The stateͲofͲtheͲart air quality modelling system AERMOD includes an updated treatment of turbulence and Keywords: dispersionintheplanetaryboundarylayerforflowovertheflatandcomplexterrain,scalingconcepts,andsurface AERMOD andelevatedsources.However,thedatarequirementsarehigherthanforthescreeningandtheISCST3models,in Airpollution particularmeteorological,topographyandlandusedataforthemodellingdomain. Dispersionmodelling Mixingheight This work presents a methodology for implementation of the AERMOD modelling system when local data is Gaussianmodels incompleteandisnotintheformatrequiredbythemodel,whichisatypicalsituationinmanycountries,particularly indevelopingones.Inaddition,themaincomputationaltoolsdevelopedtoachievethisobjectivewerepresented: ArticleHistory: LandUse.xls, which takes the place of AERSURFACE; AERMET+, a version of AERMET that runs without upper air Received:06January2010 soundingmeteorologicaldata;andSD_Aermet,whichconvertsthesurfacemeteorologicaldatatoaformataccepted Revised:22March2010 byAERMET. Accepted:24March2010 Threerepresentativecasestudieswerealsopresented,andthemodellingresultsofacasestudywerecomparedto CorrespondingAuthor: measurementdata.TheworkconcludeswithadiscussionofthepossibilityofusingtheAERMODmodelinCubaand LeonorMariaTurtosCarbonell in other countries, even when some input data is absent and climatological conditions differ from the medium Tel:+537Ͳ206Ͳ2065 latitudesforwhichthemodelwasdevelopedfor. Fax:+537Ͳ204Ͳ1188 EͲmail:[email protected]
©Author(s)2010.ThisworkisdistributedundertheCreativeCommonsAttribution3.0License. doi:10.5094/APR.2010.013 1.Introduction dispersion in the planetary boundary layer for flow over the flat andcomplexterrain. Air quality assessment by integrating measurement techͲ niques and modelling tools is a crucial element in pollution AERMOD adopts the ISCST3’s input/output architecture, mitigation.However,inmanycountriessystematicmeasurements ensuringthatthesourcesandatmosphericprocessesmodelledby for monitoring and evaluation of air quality are not available, the ISCST3 can still be handled. Therefore, all the work done to mainlyduetolackofresources. implementISCST3(Turtosetal.,2007a)isastartingpointforthe implementationofAERMOD. Additionally, modelling works are not an effective manageͲ ment tools in many countries due to lack of regulations. In AERMOD, like its predecessor ISCST3, is open–source developingcountrieslikeCuba,theregulatoryframeworkisbased software.Thesourcecode,usermanuals,modelformulation,and on screening models, which are many years behind the stateͲofͲ test cases are available for public access as anonymous user at theͲart dispersion models and they generally yield inaccurate http://www.epa.gov/scram001/dispersion_prefrec.htm and diͲ predictions. spersion_alt.htmrespectively.Beingopen–sourcehasfacilitatedits steadyimprovementandensuredthatitcanbeusedindeveloping The implementation of high–resolution models at the local countries at no additional cost, after making the necessary scale, such as AERMOD (American Meteorological Society–AMS/ adjustments. Environmental Protection Agency–EPA Regulatory MODel), improvestheaccuracyofpredictions,whichtranslatesdirectlyinto 2.Methodology a better understanding of the risks at involved receptors and an improved assessment of compliance with air quality standards, The AERMOD modelling system includes the AERMOD enablingmoreinformeddecisions.However,datarequirementfor dispersionmodel(EPA,2004a)andtwoinputdataprocessorsthat high–resolution models is higher than for screening models, are regulatory components: AERMET (EPA, 2004b), a thereforetheiruseislimitedindevelopingcountries. meteorological data processor, and AERMAP (EPA, 2004c), a terrain data processor. Other non–regulatory components of this TheU.S.EPAestablishedAERMODastheregulatorymodelin system are AERSURFACE (EPA, 2008), a surface characteristics 2005 (EPA, 2005), to replace ISCST3 (Industrial Source Complex processor,andBPIP–PRIME(EPA,2004d)forprocessingdatafrom modelforShortTerms,version3).AERMODisanadvancedplume buildings and obstacles near emission points to determine their model that incorporates updated treatments of turbulence and Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111103 interferencewithplumeriseandtoestimatethevariablesneeded For Cuba, the following adjustments to the default values byAERMODtoevaluatebuildingdownwasheffect. definedbyAERMET’suserguidefordifferentseasons(definedby defaultformid–latitudecontinentalareas)areproposed: 2.1.AERMAP •ValuesinsummerequaltoAERMET’sdefaultsinsummer, One of the main limitations for the use of AERMAP in Cuba •ValuesinwinterequaltoAERMET’sdefaultsinautumn, and other developing countries is the availability of a digital • Values in autumn equal to the average of the AERMET’s elevation model (DEM) containing topographical data of the defaultvaluesforsummerandautumn, modelling domain with an adequate resolution that can be • Values in spring equal to the average of the AERMET’s acquired quickly and inexpensively. There are online free DEM defaultvaluesforspringandsummer. sourcesthatcanbeusedtorunAERMAPwhenalocaloneisnot available.Thefollowingdatasetswereevaluated: In all cases, the default values of the Bowen ratio must be usedforaveragemoistureconditions.Insomeareasofthecountry GTOPO30.ADEMwithsamplespacingof30arc–seconds(~900m) astheverydryregionsofGuantanamo,theseconsiderationsmay (GTOPO30,1996),usingtheLatitude/LongitudeWGS84projection. vary. InthecaseofCuba,CentralAmerica,MexicoandtheCaribbean, thefilesw100n40.demandw140n40.demarerequired. InCuba,asinmanyothercountries,thereisalackofupdated digital land use layers. In this case, the layer contained in the SRTM (Shuttle Radar Topography Mission). A DEM with sample International North America land cover database could be used. spacing of 3 arc–seconds (~90 m), using the Latitude/Longitude ThislayerispartoftheGlobalLandCoverCharacteristicsDatabase projection. Each file corresponds to one degree of latitude and (availableinhttp://LPDAAC.usgs.gov/glcc/glcc.asp)fromtheUSGS longitude. The largest local domain (100x100km) could require and includes all continents. Data share the same projections uptoninefiles.Thenamesofthefiles,e.g.N23W075.hgt,contain (Interrupted Goode Homolosine and Lambert Azimuthal Equal the latitude (North or South) and longitude (West or East) Areas)ataspatialresolutionof~1000m.Thedecisionwasmade correspondingtothelowerleftcornerofthegridsystem. tousethedataintheLambertAzimuthalEqualAreasprojection, whichissupporteddirectlyinmostGISapplications. SRTMisaninternationalprojectspearheadedbytheNational Geospatial–Intelligence Agency (NGA) and the National The data is presented as a raster image with the spatial AeronauticsandSpaceAdministration(NASA).Elevationdataona resolution of 1000m. A land use value corresponds to a 24– near–global scale was obtained to generate the most complete categorylanduseclassificationgiveninthefirstcolumninTable1. high–resolutiondigitaltopographicaldatabaseofEarth(Rodriguez Inaddition,Table1showsthecorrespondenceproposedbetween etal.,2005;Farretal.,2007). thecategoriesusedinthisdatabase,inAERMET,andAERMOD. The resolution of this data set is 10 times higher than Ifanothersourceofdataisused,thecorrespondencetothe GTOPO30.Recently,version2oftheSRTMwasreleased.Version2 categories established for AERMET and AERMOD should be istheresultofasubstantialeditingeffortbytheNGAandexhibits verified. well–defined water bodies and coastlines and the absence of spikes and wells (single pixel errors), although some areas of Table1.LandusecategoriesinUSGSandAERMET–AERMOD missing data (“voids”) are still present. Both are available using anonymousftpate0srp01u.ecs.nasa.gov/srtm. USGScategories AERMET AERMOD Urbanland, 2.2.Landuse UrbanandBuiltͲUpLand Urban no vegetation AERSURFACE, designed to aid in obtaining realistic and DryͲland,CroplandandPasture reproducible surface characteristic values for the AERMOD Agricultural modellingsystem,isavailablefromearly2008(version08009)and IrrigatedCroplandandPasture Cultivated land requires the input of land cover data from the U.S. Geological MixedDryͲland/IrrigatedCroplandandPasture Survey(USGS)NationalLandCoverData1992archives(NLCD92), Land Cropland/GrasslandMosaic ataspatialresolutionof30meters.Thisinformationisavailableat nochargeonlyforusersfromtheUnitedStates.Thismethodology Cropland/WoodlandMosaic Rangeland proposes to replace AERSURFACE combining the use of the Grassland Grassland followingtools: Desert Barrenland, ShrubͲland Shrubland mostlydesert 1)AGeographicalInformationSystem(GIS),fortheprocessing MixedShrubͲland/Grassland oftheavailablelanduselayerisusedtointegrateitwiththelayer Grassland Rangeland that represents the modelling domain in order to estimate the Savannas percentage corresponding to each land use category for each DeciduousBroadleafForest Deciduous sector. In order to satisfy both AERMET and AERMOD requireͲ DeciduousNeedleͲleafForest Forest ments, the modelling domain must be composed of 72 radial sectorsoffivedegreeseach. EvergreenBroadleafForest Forest Coniferous EvergreenNeedleͲleafForest Forest 2)TheMSExcelapplication,LandUse.xls,isusedtocalculatea MixedForest weightedaverageoftheAlbedoandBowenratioatmiddayandan Bodiesof WaterBodies Water arithmetic average for the surface roughness length as inputs to water AERMET(seeSupportingMaterial).Theestimationisbasedonthe Non–forested land use cover of each category for each sector and the default HerbaceousWetland wetlands valuesofthesethreesurfacecharacteristicsforeachseason/land Swamp usecategoryareadjusteddependingonthestudyarea. WoodedWetland Forest Desert Barrenland, BarrenorSparselyVegetated Shrubland mostlydesert 104 Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111
Figure1.Exampleoflandusemapprocessedbyradialsectors. Figure 1 shows an example of a land use category map dZ H 3Tu processedbyradialsectors,asrequiredbyAERMOD–AERMET,for C * 21 A 2B 2 (1) amodellingdomainof100x60km. dt JU T ZC CP JT Zg C 2.3.AERMET 2.TheequationoriginallyproposedbyDriedonks(1982),used in conjunction with the equation for the variation of the AERMET requires surface and upper air meteorological data. temperature jump across the top of the boundary layer. This Sinceaversionthatsupportsafreeformatforsurfacedataisstill modelwasadaptedforuseinADMS3.1(CERC,2001),asdescribed pending, the program SD_Aermet was developed to convert the by Thomson (1992), Thomson (2000) and also in Lakes EnvironͲ nationalsurfacedataintoaformatsupportedbyAERMET. mentalAERMOD–ISCView(Theetal.,2001). AERMET estimates the mixing height in the Convective Boundary Layer, taking into account its dependence on both dZ S C D (2) mechanical and convective processes. The mixing height is dt 'T calculatedbasedonthefollowingcriteria: d'T JT S H S • During the day, when the Monin–Obukhov Length is D D (3) dt ' UT ZC Z negative, it is estimated as the larger of the convective or the CP C mechanicalmixingheight. 3 • During the night, when the Monin–Obukhov Length is H *Tu D AS B (4) positive,itisequaltothemechanicalmixingheight. UC gZCP A problem emerges when trying to estimate the convective where u* is the friction velocity (m/s), H is the surface heat flux mixingheightbecauseupperairmeteorologicaldataarerequired. 2 In Cuba and in other countries, upper air soundings are not (Joule/sm ),Zicistheconvectivemixingheightforhour“i”(m),tis the time of calculation (s), ZC is the convective mixing height for available with the required frequency (twice daily) or they are 3 never observed. Under these conditions, an upper air estimator time“t”(m),AandBareconstants,ʌisthedensityofair(kg/m ), CPisthespecificheatofairatconstantpressure(Joule/kgK),gis (UAE)thatcanestimatetheconvectivemixingheightsisrequired. 2 ThesimplestUAEcouldbebasedonsurfacemeteoͲrologicaldata. acceleration of gravity (m/s ), JT is the potential temperature Other more complicated estimators could be implemented using gradient above the mixing height (K/m), T is the reference the results of 3D meteorological meso–scale models as MM5 temperature(ongroundsurface)(K),'Tisthetemperaturejump (PSU/NCAR, 2005) or WRF–ARW (Advanced Research WRF) across the boundary layer top (K). The potential temperature (PSU/NCAR,2010). gradient above the mixing height JTwas estimated using the expressionproposedbyGill(1982): To develop the required UAE two algorithms were added to the MPPBL module of AERMET and a new version was obtained 2 uTN (AERMET+).Itdoesnotrequireupperairmeteorologicaldataand JT (5) g estimatestheverticaldata(specificallytheheightoftheconvective mixed layer, the convective velocity scale, and the potential temperaturegradientabovethemixingheight)fromsurfacedata. whereNuistheBruntͲVaisalafrequencyabovetheboundarylayer in (1/s). The default values are 0.013, as suggested by Thomson Thealgorithmsmentionedaboveare: (2000), and 0.011, as implied by the U.S. Standard Atmosphere (Gill,1982). 1.TheequationproposedbyBatchvarovaandGryning(1991), Both algorithms are given in Seibert et al. (2000). A combinationofalgorithms1and2wassatisfactorilyimplemented Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111105 asdescribedinTurtosetal.(2009a).Thedifferentialequation[1] satisfactorily the convective mixing heights based on surface wassolvednumerically,takingintoaccountthatforthesmalltime meteorological data, with relative independence of the values intervals't=360secondsatwhichcalculationswereconducted, selectedforthealgorithmparametersNu,AandB,providedthat the rate of change with time of the dependent variable ZC, theyaresufficientlyrepresentative. estimated as the finite difference C / '' tZ , can be satisfactorily approximated as the variable derivative C /dtdZ in the •Theresultsofalgorithms1and2areverysimilar(Figure2a) correspondingtimeinterval. and the algorithms could be used interchangeably without significant changes in simulated air pollutant concentrations. When this calculation method based on a finite differences Algorithm2generatedresultsthatareslightlymoreaccurateifthe scheme was used, the estimated values for all convective mixing parametersassociatedwiththetrendlinesarecompared,butthe heightsatthedifferenthoursofthedaywerefoundtobestrongly Batchvarova and Gryning algorithm, composed by just one dependent on the accuracy of the convective mixing height differential equation, is very suitable for the present application estimatedforthefirstconvectivehour. duetoitssimplicityandaccuracy. t0=0wasconsideredtobethespecificmomentofdawnfor •Verygoodcorrelationswereachievedinalltestcases.The which the surface heat flux, and consequently the convective slopes of the regression equation and the coefficients of mixing height, began to grow, starting from the initial condition: determinationR2werecloseto1.0(seeforexampleinFigure2b, 2 ZC(0)=0;'T(0)=0;H(0)=0. y=1.071x,R =0.921).Ontheaverage,theproposedmethodology overestimated concentrations by approximately 5–7% for Themainproblemwastoestimateanaccuratefirstnon–zero Nu=0.011 and underestimated concentrations by 2–4% for value 1 CC ' tZZ toinitializethecalculation.Itwasdemonstrated Nu=0.013 with respect to the values derived from upper air thatthealgorithmtodeterminethefirstZCvalueisthesameasfor soundings. algorithms 1 and 2. It included an iterative scheme and the determination of the upper value limit in order to avoid nonͲ 2.4.AERMOD convergenceofthesolution. The methodology for the implementation of the dispersion Although AERMOD does not evaluate the concentration at modelAERMOD,inadditiontotheprevioussteps,involved: calm wind hours, the implementations of algorithms 1 and 2 in AERMETconsideredtheestimationoftheincrementofconvective 1) Proposalofthecontroloptions: mixingheightduringthesehoursbecausethemixingheightalways depends on the increment in preceding hours, even if the a) Estimation of concentrations and wet and dry conditions were calm. The same analysis leads to the linear depositionofpollutants(gasesandparticulates) interpolationforestimatingthemissingmeteorologicaldata. b) Selection of the averaging periods (in Cuba, hourly, daily and annual) according to the pollutant and Sensitivity analysis and assessment of results. The results of establishedstandardsforanalyzingresults severalsensitivityanalysesconductedwiththeimplementationof c) Useofthefollowingdefaultoutputoptions: these algorithms were presented in Turtos et al. (2009a). They wereperformedtoevaluatetheimpactofcertainparametersand • Concentrations and depositions at each receptor theuseofalgorithms1or2.ValuesconsideredforNuwereeither due to emissions from each group of sources: the 0.011 (as suggested by EPA) or 0.013 (as suggested by UK average during long periods (annual or whole period) authorities), and two possible values were considered for the and the maximum in each short averaging period constant B in Equations (1) (Batchvarova and Gryning algorithm) (PLOTFILEoption) and[4](Driedonksalgorithm). • Concentrations higher than the threshold values due to emissions from each group of sources in each The comparison of the convective mixing heights estimated shortaveragingperiod(MAXIFILEoption) from upper air data for all AERMET–EPA test cases (EX01–EX05), •All concentrations at each receptor in each short andtheresultsofalgorithms1and2fordifferentvaluesofNuand averagingperiod(POSTFILEoption);onlyforvalidation B,showsthat: studies 2) Establishment of methodologies to generate the data •Themodelisnotsignificantlysensitivetothevariationof required by the high–resolution models for characterization of the analyzed parameters within their usual ranges. Therefore, it emissionsources,whichisunavailableinthecountry: can be concluded that the proposed methodology can predict
Figure2.Comparisonofconvectivemixingheightsestimatedusingalgorithms1and2.EPAtestcaseEX03,Nu=0.013andB=5. 106 Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111
a) Particulate emissions and size distribution of particles thesimulationsD1,D2andDnwerecalculatedusingthefollowing (for each category, the mass fractions and density) expression: from available international databases such as AP–42 (EPA, 1998). If the concentrations are half of the ݉ referencevalues,sensitivitystudiesmustbeconducted ݀ݎ ൌ ቆͳ െ ቇ ͳͲͲǡ ሺΨሻ (6) fortherangeofprobablevalues. ݉ b) Simulation of flares as point sources with the corresponding adjustment in equivalent emission wheremrepresentstheconcentrationordepositionvaluesandi,j heightanddiameter. indicatethesimulationsbeingcompared. 3) Approaches and algorithms for using variable emission Themaximum,minimumandmeanvaluesofdrforonehour, patterns,evenforrealͲtimemodelling. 24hoursandthewholeperiodshowthattherelativedifferenceis small,onaverage5.6%for1hour,4.7%for24hours,and1.4%for 4) Defining urban and rural conditions based on the Auer the whole period. Consequently, the deposition increases on the methodorthepopulationdensityprocedure. average by 19% for the whole period and 28% for 1 hour. In general, the value selected for cuticular resistance does not As in ISCST3, the modelling of each pollutant in AERMOD is significantlyinfluencetheresults. independent and therefore necessary to build a control file for eachpollutant.Inordertoavoidrenamingthecontrolandoutput 3.CaseStudies files for each pollutant, a “.bat” file is used (Aermodb.bat). It is further recommended that a “.bat” file is created for each case 3.1.CASE1ͲValidationofAERMET+ studyinordertoconsecutivelysimulatetheconsideredpollutants. A study was conducted in order to compare the surface Data for deposition algorithms. The deposition algorithms in concentrations calculated by AERMOD, using mixing heights AERMOD(EPA,2004e;EPA,2004f)aredifferentfromthoseusedin obtained from upper air meteorological soundings and mixing ISCST3 (Wesely et al., 2002). The new deposition algorithms heightsderivedfromtheUAEusingalgorithms1and2(C1andC2, require the previously mentioned land use categories as well as respectively). This case study included hourly, daily, and annual somegasdepositionresistancetermsforfiveseasonalcategories: estimates of SO2 concentrations and deposition rates, with emissionsfrom6pointsourceslocatedinthedomainoftheEPA’s 1) Midsummerwithlushvegetation, Test Case 3 for AERMET (EPA, 2007). The concentrations were 2) Autumnwithunharvestedcropland, estimated at three discrete receptors and at 720 receptors 3) Late autumn after frost and harvest, or winter with no included in two Cartesian grids centred on the sources. The first snow, gridincludes240receptors,hasahorizontalresolutionof5kmand 4) Winterwithsnowonground,and coversanareaof60x100km,fromsouthtonorthandwestto 5) Transitional spring with partial green coverage or short east,respectively.Thesecondgridincludes480receptorsandhas annuals. a horizontal resolution of 0.5 km on a rectangular domain of 12x10km. ForCuba,Category2isappropriatefromNovembertoMarch andCategory1isapplicablefortherestoftheyear. Table3showsthemaximumhourlyanddailyconcentrations andthehighestaverageconcentrationsatallreceptors,withinitial Table 2 lists the chosen values of parameters needed for mixingheightscalculatedwithdifferentalgorithms.Themaximum modellingsulphurandnitrogenoxides,H2SO4andH2SinAERMOD values were used in the comparison because the maximum at25°C.TheEPAproposesvaluestouseforalonglistofpollutants concentrations are the focus of regulations. The changes in the (Weselyetal.,2002). concentrations with respect to C are calculated as a percentage 0 (C1–C0)/C0x100)andasafractionalbias(FB)whenUAEareused. Table2.ConstantsforgasdepositionalgorithmsinAERMOD The fractional bias was selected as the basic measure of Parameters SO2 NOx H2S H2SO4 performance in this evaluation because it is symmetrical and a a b c bounded.ItsvaluesrangebetweenͲ2.0(extremeover–prediction) Diffusivityinair,Da 0.1509 0.1656 0.1623 0.986 (cm2/s) and+2.0(extremeunder–prediction).Thefractionalbiasbetween Ͳ5 d Ͳ5 d Ͳ5 e Ͳ5 c C1andC0isestimatedas2(C1–C0)/(C1+C0). Diffusivityinwater,Dw 1.83x10 1.4x10 1.36x10 1.28x10 (cm2/s) f f As can be observed from Table 3, maximum concentrations Cuticularresistance,rc 80 200 ͲͲ (s/cm) are not highly sensitive to the algorithm used in mixing height g 3 g 3 g Ͳ10 c estimation, because the differences are less than 3% for all Henry'sLawconstant,HC 72.37 84.43x10 1.01x10 5.08x10 (PaͲm3/mol) averagingperiods(hourly,daily,andannual).Themaximumhourly and daily concentrations were predicted at the same receptor a(Scireetal.,2000),b(GasSim2,2005),c(Dortchetal.,2005),d(Boerboometal.,1969), independently of which mixing height estimation algorithm was e(Himmelblau,1964),f(CurrieandBass,2005),g(Sander,1999) used: at a receptor 2237m from the central source point for hourlyconcentrationsand1275mfordailyconcentrations. Anadequatevaluewasnotfoundforthecuticularresistance toH SO andH S,therefore,sensitivitystudieswereconductedto 2 4 2 Thehighestaverageannualconcentrationwasestimatedata evaluate the influence of the chosen values. Three AERMOD receptor located 313m away from the central source when the simulationswereconductedforH S;thefirstsimulation,identified 2 mixing heights were obtained from the upper air meteorological as D1, used the default proposed (rc=80 s/cm), the second (D2) Ͳ5 soundings and the UAE using algorithm 2. When algorithm 1 is usedthevaluesselectedforSO (r =1.36x10 s/cm),andthethird 3 2 c used, the maximum concentration of 146μg/m was found at a (D )didnotconsiderdeposition. n receptorlocated829mfromthesourcecentre,slightlyhigherthan 3 a concentration of 145μg/m calculated at a receptor located Tocomparetheresults,therelativedifferences(dr)between 313mfromthesource. thecalculatedconcentrationsanddepositionsateachreceptorin Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111107
Table3.ComparisonofsurfaceconcentrationscalculatedbyAERMODusingmixingheightsobtainedfromupperair
meteorologicalsoundings(C0)andderivedfromtheUAEusingalgorithms1and2(C1andC2respectively)
ActualUAData(C0) UAE,algorithm1(C1)ȴC1_0,% FB1_0 UAE,algorithm2(C2)ȴC2_0,% FB2_0 Max. 148 145Ͳ2.04Ͳ.0205 149 0.92 0.0092 AnnualAverage Average 12.7 12.4Ͳ6.01Ͳ.0640 12.6Ͳ2.86 0.0292 Max. 15458 15469 0.07 0.0007 15469 0.07 0.0007 Highest1Ͳhour Average 848 812 4.16Ͳ.0292 824Ͳ2.75 0.0279 Maximum 2610 2603Ͳ0.25Ͳ.0025 2600Ͳ0.36 0.0036 Highest1Ͳday Average 159 158Ͳ2.50Ͳ.0304 158Ͳ1.49 0.0164 The deviations in average concentrations are bigger than stations (identified as MS–A and MS–B) in order to compare deviations in maximum concentrations, but less than 6% for measurementsandmodelresults. algorithm 1 and less than 3% for algorithm 2. These results corͲ roboratethestatementoftheslightsuperiorityofalgorithm2over The comparison was mainly qualitative. It was not based on algorithm1. statisticaltechniques,sincethestudyassumedaverageoperating conditionsforthewholeassessmentperiodanddidnottakeinto The criterion that the performance of a model can be account variations in the operating management for any reason regarded acceptable if FB is betweenͲ0.5 and 0.5 (Kumar et al., (e.g. maintenance periods or partial cleaning system failures). 2006) was used. Therefore, it can be concluded that the UAE is Additionally, fugitive emissions were not considered, which were sufficiently accurate and is a viable approach to provide mixing expectedtobeimportant,butcannotbeestimated. heightsformodelssuchasAERMODwhenupperairsoundingdata arenotavailable. A period of almost 4 years was simulated including the emissionsofSO2,H2S,PM10andPM2.5fromallpointsourcesanda 3.2. CASE 2 – Comparing modelling results with measurement river basin which was modelled as a complex polygonal area data source. The developed methodology was successfully applied to a Figure3showsthelocationoftheMSrelativetothesources nationalcasestudy(Turtosetal.,2007b)wherelocaltopography, and the wind rose for the 36 sectors in the region, which was landuseandupperairmeteorologicaldataarenotavailable.The obtainedfromthestatisticalprocessingofthehourlydatausedin AERMODmodelwasrunwiththePOSTFILEoutputoptionfortwo thecalculations. receptorsfixedattheexactlocationoftwoairqualitymonitoring
Figure3.WindroseforthestudyzoneandlocationofsourcesandMS. 108 Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111
Predicted and measured concentrations were compared. In thepollutantconcentrationmeasurementsindirectlyvalidatesthe general,measuredconcentrationsexceededpredictedvalues.For proposedmethodology. PM10andPM2.5,theaveragepredictedconcentrationislessthan 1% of the measured values at both MS, suggesting a low 3.3.CASE3–Assessingtheinfluenceofmeteorologicalconditions contribution of the modelled sources to particulate air pollution. tosupportoperationalairpollutionmitigation This is in agreement with the analysis of existing sources in the modelling domain, and therefore a detailed comparison for this The aim of this case study is to evaluate the influence of pollutantisnotnecessary. meteorological conditions on air quality deterioration in order to support operational mitigation procedures (Turtos et al., 2008; ForH2SandSO2,themostinterestingfindingisthefollowing Turtosetal.,2009b). apparentcontradictionintheresults:thesimulatedandmeasured SO2concentrationsatMS–AarelargerthanatMS–B,whereasthe In the framework of the distributed power generation oppositeoccursforH2S.Theisolinesofaverageconcentrationsat program,groupsofgeneratorsets(GS,internalcombustionengine all modelled receptors as shown in Figure 4 support this distinct for electric generation) are being installed in urban areas of the behaviour, despite having been obtained by using the same country to provide base and emergency power by burning fossil meteorological,topographicalandlandusedata. fuel. The operation of a GS could cause significant local air pollutionduetoemissionsofsulphurdioxide,nitrogenoxidesand ForSO2,theconcentrationisolinesfollowtheprevailingwind particulatematter. direction (see the wind rose and the relative positions of the sources and the MS shown in Figure 3). For H2S, although the As air pollution depends on many factors, various mitigation prevailingwindscomefromtheeast–northeastandtheeast,the strategies can be applied including proper selection of a GS highest concentrations were found in the east–northeast, caused location, design or operational alternatives. The selected byanimportantcomponentofthewindsfromthewest–southwest mitigationtypedependsuponthestatusoftheparticularfacility. at very low speeds (<2m/s). These winds cause very high Where GSs are already installed, design and location options are concentrations because of the evaporative emissions from the difficulttoimplement. river modelled as an area source. These results agreed with the measurements. Itisextremelydifficulttoassessalleffectsofmeteorological conditionsonairquality.Amongthelocalmeteorologicalvariables, For both pollutants, the measured values exceeded the thewindpattern(primarilyhorizontalspeedanddirection)hasthe simulatedvalues.ThisisanexpectedresultforSO2becausenotall mostpronouncedinfluenceonpollutantconcentrations.Common of its sources in the domain were modelled. For H2S, the practice is to locate facilities so that the prevailing winds do not differencesinthehourlyanddailyvaluesaccountforthefactthat transport contaminants to populated areas. However, the the study assumed average operating conditions for the whole incremental concentrations of pollutants depend on the energy period. This does not take into account the variations in the balance in the atmosphere, which determines the stability and operating regimen; i.e., the total emissions are uniformly turbulence (both thermal and mechanical), and the thickness of distributed, but in reality there are emissions exceeding those the mixing layer. The incremental concentrations are negatively considered. Differences in mean values were assumed to come correlated with wind speed and the mixing height. Other from fugitive emissions, which were not considered in the meteorological variables are cloudiness, ceiling height (because modellingstudies. they alter the atmospheric energy balance and consequently the mixingheight),temperature,precipitation,andrelativehumidity. Itisconcludedthatthereisareasonablematchbetweenthe predicted and measured concentrations taking into account the This case study uses the AERMOD modelling system to simplificationsintroducedinthemodelling,theabsenceofupper evaluatetheincrementalconcentrationsduetotheemissionsofa air meteorological data and other uncertainties. Typically, this hypothetical power plant consisting of 4 groups of 4 engines of correspondencecanonlybeachievedbytheuseofhigh–resolution 1.7MW each in a grid of polar receptors. The calculations were models. A simplified model could never explain the apparent carried out assuming flat terrain to remove the influence of contradictionbetweentheconcentrationsatbothMSforSO2and topographyontheresults. H2S. The consistency of the model estimates in comparison with
Figure4.Contoursofaverageconcentrationsof(a)SO2and(b)H2S. Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111109
Figure 5 shows hourly incremental concentrations as a winds in Cuba is NE (northeast), i.e. winds blowing from the functionofthedistancefromthesource[upto30kmin(a)andup northeasttothesouthwest. to5kmin(b)]forJanuary1,2006,assumingthatduringthewhole Operationalmeasurescanbeappliedtoreducetheimpactof day the wind blew to the receptors. This assumption allows GSonairquality.Itisproposedtoadapttheoperatingregimenfor excludingtheeffectofwinddirectionontheresults.Fortheactual eachsiteaccordingtothedistancetoareaswithhighpopulation meteorological data file for 2006, the correlation between density: incrementalconcentrationsandwinddirectionisaround0.2atany distancefromthesource. 1.Duringdaytimehours,itisrecommendedthatGSslocated indenselypopulatedareasoperateatlowcapacityortheydonot operateatall. 2.Atnight,itisencouragedthatGSlocatedinurbanareasbut morethan2kmfromdenselypopulatedareasdonotoperateor operate at low capacity, since peak concentrations are encountered relatively far from the source and then decrease graduallywithdistance. A real–time monitoring of both emissions and air quality throughmodellingcouldsupportdispatchingofmitigationoptions. This is recommended even when air quality monitoring systems areoperating,duetotheirinabilitytoidentifythecontributionofa specificsource. 4.Conclusions The present work confirms the possibility of implementing high–resolutionmodelssuchasAERMODtosimulatedispersionof localpollutantsinCubaandothercountries,despitetheabsence of some required input data and different meteorological and climatological conditions with respect to the region where the modelwasdeveloped.Thisassertionisbasedonthevalidationof the developed methodological solutions and their corresponding computertools.Theimplementationallowsustoconcludethat: 1. The methods proposed to complete the data required by the high–resolution models for the emission sources characterization,whichareunavailableinthecountry,areatthe levelofinternationalpractices. Figure5.Dependenceofincrementalconcentrationsatdifferenthourson 2. The lack of required upper air meteorological data was thedistancefromthesource(a)upto30kmand(b)upto5km,assuminga resolved. The major efforts were concentrated on developing of constantwinddirectionrelativetothereceptors. alternative methodologies for estimating the mixing height from surface data only. Regarding these methodologies, it was concludedthat: A distinctive behaviour was observed during daytime and nighttime hours, although in both cases the maximum a)TheinfluenceoftheparametersNu,AandBusedinthe concentrationandthelocationwhereitispredictedwereshown algorithmsisnotsignificantintheresultsintherange to depend crucially on the effective emission height. During the ofassumedvaluesfortheseparameters,althoughthe day,theconcentrationsareveryhighatthereceptorsclosetothe bestresultsareachievedwithNu=0.013,A=2.5and sourceanddecaysharply(at2kmfromthesource,concentrations B=5. have been reduced by an order of magnitude). At night, when b)Theresultsofbothalgorithmsareverysimilarandcan stableconditionsprevail,thepeakconcentrationswerereachedat be used with no significant variation in the concenͲ receptorsfartherfromthesource,butthedecayissofter,sothat trationslatercalculatedbyAERMOD. significant concentrations are maintained away from the source. c)Theimplementationwasvalidatedthrough: Although the graph corresponds to one day of the year, this •The correlation of the mixing heights calculated behaviourisrepresentativeofalldays. with AERMET+ using surface and upper air meteorological data, for all test cases of the original This distinctive behaviour allows making some suggestions model. regardingtheoperationalmanagementofthegeneratorsets,and •Thecomparisonoftheincrementalconcentrations even about their location, i.e. to locate the generator sets at a estimatedbyAERMODusingAERMET(withsurfaceand greater distance from residential areas than the distance where upperairdata)andAERMET+(usingonlysurfacedata). themaximumconcentrationisreachedatnight.Thisexceedsthe distanceatwhichthemaximumconcentrationispredictedduring 3.ThelackofaDEMwiththeappropriateresolutionforthe daytimehoursandconsequentlyitisguaranteedthatthedaytime model domain does not prevent the use of high–resolution concentrationswillbesignificantlyattenuated. models, as there are free online data sets that fulfil these requirements. It is proposed to combine the SRTM2 data set Theplantsshouldneverbelocateddownwindofotherplants (spatialresolutionof3'')withGTOPO30(spatialresolution30''). inthedirectionofprevailingwinds,becausetheoverlappingofthe pollutantplumesisgreatestinthiscase.Thedirectionofprevailing 4. The availability of updated land use digital layers of the modeldomainwiththeproperresolutionisalsonotalimitation. 110 Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111
The "North America land cover database" in Lambert Azimuthal Driedonks, A.G.M., 1982. Sensitivity analysis of the equations for a projection of equal areas can be used. It is proposed to replace convectivemixedlayer.BoundaryLayerMeteorology22,475Ͳ480. AERSURFACEbythecombineduseofGISandalanduselayerof EPA,2008.AERSURFACEUser’sGuide.EPAͲ454/BͲ08Ͳ001. the model domain processed with the application LandUse.xls, developedspecificallyforthispurpose. EPA, 2007. AERMET Test Cases, aermet_testcase.zip available in http://www.epa.gov/scram001/metobsdata_procaccprogs.htm#aermet. 5. A comparison of AERMOD results using the proposed EPA,2005.RevisiontotheGuidelineonAirQualityModels:Adoptionofa methodologywithvaluesmeasuredattwocontinuousmonitoring Preferred General Purpose (Flat and Complex Terrain) Dispersion stations, although primarily qualitative because statistical Model and Other Revisions. 40 CFR Part 51, [AH–FRL–7990–9] RIN techniques were not used, showed good agreement and it was 2060–AK60,FederalRegister/Vol.70,No.216/RulesandRegulations. presentedasanindirectvalidationofthemethodology,reinforcing EPA,2004a.User'sGuidefortheAMS/EPARegulatoryModel–AERMOD. theneedtointroducehigh–resolutionmodelling. EPAͲ454/BͲ03Ͳ001. EPA,2004b.User’sGuidefortheAERMODMeteorologicalPreͲprocessor– ThisvalidationallowsustoproposetheinclusionofAERMOD AERMET.EPAͲ454/BͲ03Ͳ002. inthenationalmodellingframeworkatthelocalscale,despiteits relative complexity in data and large computation time when EPA, 2004c. User's Guide for the AERMOD Terrain PreͲprocessor – multiplesourcesandreceptorsaremodelled. AERMAP.EPAͲ454/BͲ03Ͳ003. EPA,2004d.User'sGuidetotheBuildingProfileInputProgram.EPAͲ454/RͲ Thisproposalisthebaseforsolvingtheproblemofinsufficient 93Ͳ038,RevisedApril21,2004. capacity of modelling tools in the country for assessing and EPA,2004e.ADDENDUMtoUser'sguidefortheAMS/EPAregulatorymodel monitoring local air pollution. Such models are an indispensable –AERMODEPAͲ454/BͲ03Ͳ001. complement to measurements in any context, but most importantlyinCubaandotherdevelopingcountries,duetolackof EPA,2004f.AERMODDepositionAlgorithms–ScienceDocument(Revised resources for installation of an adequate air quality monitoring Draft).http://www.epa.gov/scram001/7thconf/aermod/aer_scid.pdf. network. EPA, 1998. Compilation of Air Pollutant Emission Factors, APͲ42, Fifth Edition, Volume I: Stationary Point and Area Sources. Acknowledgements http://www.epa.gov/ttn/chief/ap42/ch01. Farr,T.G.,Rosen,P.A.,Caro,E.,Crippen,R.,Duren,R.,Hensley,S.,Kobrick, ThankstoFernandoCuevas.Heencouragedustoinitiatethe M.,Paller,M.,Rodriguez,E.,Roth,L.,Seal,D.,Shaffer,S.,Shimada,J., workrelatedtotheAERMODimplementationintheframeworkof Umland,J.,Werner,M.,Oskin,M.,Burbank,D.,Alsdorf,D.E.,2007.The the Collaboration Agreement between the Mexican Environment ShuttleRadarTopographyMission.ReviewsofGeophysics,45,Article andNaturalResourcesSecretariat(SEMARNATfromitsacronymin NumberRG2004.(Availableat:http://www2.jpl.nasa.gov/srtm/SRTM_ Spanish)andtheEconomicCommissionforLatinAmericaandthe paper.pdf) Caribbean(ECLAC)(CEPAL/SEMARNAT,2007).Healsoencouraged ustopursuemanyotherscientificandtechnicalchallenges. GasSim 2, 2005. User Manual. UK Environment Agency. http://www. gassim.co.uk/graphics/Documents/gassim%202%20manual.pdf The critical comments of two anonymous referees and the Gill,A.E.,1982.AtmosphereͲOceanDynamics,AcademicPress. writingimprovementsofMeganMarieBela,onearlierversionsof GTOPO30,1996.Global30ArcͲSecondElevationDataSet.U.S.Geological thispaperaregratefullyacknowledged. Survey Centre for Earth Resources Observation and Science (EROS). http://www.webgis.com/terr_world.html, http://eros.usgs.gov/#/ SupportingMaterialAvailable Find_Data/Products_and_Data_Available/GTOPO30 Himmelblau, D.M., 1964. Diffusion of dissolved gases in liquids. Chemical The MSExcel application (LandUse.xls) that was used to Reviews64,527Ͳ550. calculate a weighted average of the Albedo and Bowen ratio at middayandanarithmeticaverageforthesurfaceroughnesslength Kumar, A., Dixit, S., Varadarajan, C., Vijayan, A., Masuraha, A., 2006. asinputstoAERMET.Thisinformationisavailablefreeofcharge Evaluation of the AERMOD dispersion model as a function of viatheInternetathttp://www.atmospolres.com. atmospheric stability for an urban area. Environmental Progress 25, 141Ͳ151. References PSU/NCAR,2005.TutorialClassNotesandUser’sGuide:MM5Modelling System Version 3. Mesoscale and Microscale Meteorology Division; Batchvarova,E.,Gryning,S.E.,1991.Appliedmodelforthegrowthofthe NationalCentreforAtmosphericResearch. daytimemixedlayer.BoundaryLayerMeteorology56,261Ͳ274. PSU/NCAR,2010.ARWVersion3Modellingsystemuser’sguide.Mesoscale Boerboom,A.J.H.,Kleyn,G.,1969.Diffusioncoefficientsofnoblegasesin andMicroscaleMeteorologyDivision;NationalCentreforAtmospheric water.JournalofChemicalPhysics50,1086Ͳ1088. Research. Cambridge Environmental Research Consultants (CERC), 2001. ADMS 3.1 Rodriguez,E.,Morris,C.S.,Belz,J.E.,Chapin,E.C.,Martin,J.M.,Daffer,W., UsersGuide. Hensley, S., 2005. An assessment of the SRTM topographic products. Technical Report JPL DͲ31639, Jet Propulsion Laboratory, Pasadena, CEPAL/SEMARNAT,2007.Assessmentofenvironmentalexternalitiesfrom California, 143 pp. http://www2.jpl.nasa.gov/srtm energy sector in critical areas of Tula and Salamanca. /srtmBibliography.html,linktoSRTM_D31639.pdf LC/MEX/L.788/Rev.1. www.cepal.cl/mexico/publicaciones/xml/7 /33647/L788_Rev_1__vf.pdf Sander, R., 1999. Compilation of Henry’s law constants for inorganic and organic species of potential importance in environmental chemistry. Currie,B.A.,Bass,B.,2005.Estimatesofairpollutionmitigationwithgreen Air Chemistry Department, MaxͲPlanck Institute of Chemistry, plants and green roofs using the UFORE model. Conference in 6th http//www.mpchͲmainz.mpg.de/sander/res/henry.html. biennialofCanadianSocietyforEcologicalEconomics,CANSEE,27Ͳ29, October,Toronto. Scire, J.S., Strimaitis, D.G., Yamartino, R.J., 2000. User’s Guide for the CALPUFFDispersionModel(Version5.0).EarthTech,Inc.,Concord,MA. Dortch,M.,Zakikhani,M.,Furey,J.,Meyer,R.,Fant,S.,Gerald,J.,Qasim, M., Fredrickson, H., Honea, P., Bausum, H., Walker, K., Johnson, M. Seibert,P.,Beyrich,F.,Gryning,S.E.,Joffre,S.,Rasmussen,A.,Tercier,P., 2005. Data Gap Analysis and Database Expansion of Parameters for 2000. Review and intercomparison of operational methods for the Munitions Constituents. ERDC/EL TRͲ05Ͳ16, US Army Corps of determination of the mixing height. Atmospheric Environment 34, Engineers,EngineerResearchDevelopmentCentre,Washington. 1001Ͳ1027. Turtosetal.–AtmosphericPollutionResearch1(2010)102Ͳ111111
The,J.L.,Lee,R.,Brode,R.W.,2001.WorldwideDataQualityEffectsonPBL Turtos,L.,Meneses,E.,Sanchez,M.,Diaz,N.,Rivero,J.,Paz,E.,Pena,A., ShortͲRange Regulatory Air Dispersion Models. Proceedings 7th 2008. Air pollution mitigation strategies for generator sets which International Conference on Harmonization within Atmospheric provide base and emergency power by burning fuel oil, and for the DispersionModellingforRegulatoryPurposes,Belgirate,Italy,pp.202Ͳ refurbishing process in thermal power plants. Technical Report to 206. Contract SCT No.06/2008, CUBAENERGÍA – Cuban National Thomson, D.J., 2000. The met input module, ADMS 3 Technical EnvironmentalandNuclearRegulatoryOffice,ORASEN. Specification.CERC. TurtosL.,Meneses,E.,Sanchez,M.,Rivero,J.,Diaz,N.,2007a.Assessment Thomson,D.J.,1992.AnanalyticalsolutionofTennekes’equationsforthe oftheimpactsonhealthduetotheemissionsofCubanpowerplants growthofboundarylayerdepth.BoundaryLayerMeteorology59,227Ͳ that use fossil fuel oils with high content of sulphur. Estimation of 229. externalcosts.AtmosphericEnvironment41,2202–2213. Turtos,L.,Rivero,J.,Curbelo,L.,Sanchez,M.,Meneses,E.,Diaz,N.,2009a. Turtos, L., Meneses, E., Sanchez, M., Molina, E., Diaz, N., Rivero, J., Method for the estimation of the convective mixing height aimed to Fernandez,M.,Paz,E.,2007b.Studyofgaseslocaldispersionemitted atmospheric local dispersion modelling. Environmental Impact byMoaNickelS.A.ͲPedroSotoAlba.TechnicalReporttoContractSCT Assessments,NovaSciencePublishers,Inc. No.03/2007,CUBAENERGÍA–INHEMͲMoaNickelS.A. Turtos, L., Sanchez, M., Meneses, E., Rivero, J., Pire, S., Diaz, N., Paz, E., Wesely, M.L., Doskey, P.V., Shannon, J.D., 2002. Deposition 2009b. The air quality modelling to support operational atmospheric Parameterizations for the Industrial Source Complex (ISC3) Model. pollution mitigation in generator sets. Environmental Capacity ANL/ER/TRͲ01/003withAppendixBtoF. EnhancementProject(ECEPͲC)FinalWorkshop,Havana.