Comparative evaluation of the impact of WRF/NMM and WRF/ARW

16 Pages·2012·4.08 MB·English

Atmos. Chem. Phys.,12,4091–4106,2012
Atmospheric
www.atmos-chem-phys.net/12/4091/2012/
Chemistry
doi:10.5194/acp-12-4091-2012
©Author(s)2012. CCAttribution3.0License. and Physics
Comparative evaluation of the impact of WRF/NMM and
WRF/ARW meteorology on CMAQ simulations for PM and its
2.5
related precursors during the 2006 TexAQS/GoMACCS study
S.Yu,R.Mathur,J.Pleim,G.Pouliot,D.Wong,B.Eder,K.Schere,R.Gilliam,andS.T.Rao
AtmosphericModelingandAnalysisDivision,NationalExposureResearchLaboratory,USEnvironmentalProtection
Agency,ResearchTrianglePark,NC27711,USA
Correspondenceto:S.Yu([email protected])
Received:17March2011–PublishedinAtmos.Chem.Phys.Discuss.:7December2011
Revised:5March2012–Accepted:12April2012 –Published:9May2012
Abstract. This study presents a comparative evaluation of gaseous species (HNO , SO , NH , isoprene, toluene, ter-
3 2 3
the impact of WRF-NMM and WRF-ARW meteorology penes)as bothmodelsused the samechemicalmechanisms
on CMAQ simulations of PM , its composition and re- and emissions. The results of ship along the coast of south-
2.5
lated precursors over the eastern United States with the in- easternTexasovertheGulfofMexicoshowthatbothmodels
tensive observations obtained by aircraft (NOAA WP-3), capturedthetemporalvariationsandbroadsynopticchange
shipandsurfacemonitoringnetworks(AIRNow,IMPROVE, seenintheobservedHCHOandacetaldehydewiththemeans
CASTNet and STN) during the 2006 TexAQS/GoMACCS NMB<30%mostofthetimebuttheyconsistentlyunderes-
study. The results at the AIRNow surface sites show that timatedterpenes,isoprene,tolueneandSO .
2
both ARW-CMAQ and NMM-CMAQ reproduced day-to-
day variations of observed PM and captured the majority
2.5
of observed PM within a factor of 2 with a NMB value
2.5
of −0.4% for ARW-CMAQ and −18% for NMM-CMAQ. 1 Introduction
Both models performed much better at the urban sites than
at the rural sites, with greater underpredictions at the rural Fine particulate matter (PM2.5, particles with aerodynamic
sites.Bothmodelsconsistentlyunderestimatedtheobserved diameterslessthan2.5µm)resultsfromprimarydirectemis-
PM at the rural IMPROVE sites by −1% for the ARW- sions and secondary formation through atmospheric oxida-
2.5
CMAQ and −19% for the NMM-CMAQ. The greater un- tionofgaseousprecursorssuchassulfuroxides(SOx),nitro-
derestimations of SO24−, OC and EC by the NMM-CMAQ gen oxides (NOx) and volatile organic compounds (VOCs),
contributedtoincreasedunderestimationofPM attheIM- and subsequent gas-to-particle conversion processes. To re-
2.5
PROVEsites.TheNMBvaluesforPM attheSTNurban flectmorerecenthealtheffectstudiesandprovideincreased
2.5
sites are 15% and −16% for the ARW-CMAQ and NMM- protectionofpublichealthandwelfare,thelevelofthe24-h
CMAQ, respectively. The underestimation of PM2.5 at the PM2.5 National Ambient Air Quality Standards (NAAQS)
STNsitesbytheNMM-CMAQmainlyresultsfromtheun- has been revised from 65µgm−3 to 35µgm−3, effective on
derestimations of the SO2−, NH+ and TCM components, 18 December 2006 (Federal Register, 2006). The rationale
4 4
whereas the overestimation of PM at the STN sites by for this revision includes consideration of: (1) Evidence of
2.5
theARW-CMAQresultsfromtheoverestimationsofSO2−, health effects related to short- and long-term exposures to
4
NO−, and NH+. The Comparison with WP-3 aircraft mea- fineparticles;(2)insightsgainedfromaquantitativeriskas-
3 4
surementsrevealsthatbothARW-CMAQandNMM-CMAQ sessment;and(3)specificconclusionsregardingtheneedfor
have very similar model performance for vertical profiles revisionstothecurrentstandardsandtheelementsofPM2.5
for PM chemical components (SO2−, NH+) and related standards (i.e., indicator, averaging time, form, and level)
2.5 4 4
that,takentogether,arerequisitetoprotectpublichealthwith
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
4092 S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology
an adequate margin of safety (Federal Register, 2006). Un- periment,especially,forthreetypesofplumes(powerplant
likeO pollutionwhichoccurstypicallyduringthehighpres- plumes,HoustonandDallasurbanplumesandShipChannel
3
sure, hot, sunny and stagnant atmospheric conditions at the plumes) over the Houston-Galveston-Brazoria and Dallas-
locationswithsubstantialVOCandNO concentrations,el- FortWorth(DFW)metropolitanareas.Theinfluenceofthese
x
evated PM concentrations occur throughout the year be- two different meteorological fields on spatial and temporal
2.5
causePM iscomposedofavarietyofparticlesdifferingin variations of PM , and its chemical composition over the
2.5 2.5
sizeandchemicalcompositionandalsobecausesourceemis- easternUSisalsoevaluatedagainsttheobservationsfromthe
sions of each component of the atmospheric particles vary surfacemonitoringnetworks(AIRNOW,IMPROVE,CAST-
differently and seasonally. For example, sulfate is produced NetandSTN)duringthe2006TexAQS/GoMACCSstudy.
frombothprimaryandsecondarysourcesbutelementalcar-
bon (EC) is emitted from the primary sources. Differences
inthecompositionofparticlesproducedbydifferentsources 2 Descriptionofthemodelingsystemand
leadtospatialandtemporalheterogeneityinthecomposition observationaldatabases
oftheatmosphericaerosols.
The relationship between PM2.5 and meteorological con- 2.1 Descriptionofthemodelingsystem
ditions has been examined by several studies (Whiteaker et
al., 2002; Wehner and Wiedensohler, 2003; Wise and Com- The detailed description of the modeling system and con-
rie, 2005; Dawson et al., 2007). The meteorological con- figurations is given by Yu et al. (2012). Here a brief sum-
ditions can have complex effects on the concentrations of mary relevant to the present study is presented. The WRF
PM due to the fact that PM is comprised of many dif- modelisastate-of-sciencemesoscalemodelframeworkwith
2.5 2.5
ferent species and the meteorological impacts on individual twoavailabledynamiccores:theNon-hydrostaticMesoscale
speciesaredifferent.Forexample,inthestudyofsensitivity Model (NMM) developed by NCEP (Janjic, 2003) and the
ofPM tovariousmeteorologicalparametersintheeastern AdvancedResearchWRF(ARW)developedbyNCAR(Ska-
2.5
US,Dawsonetal.(2007)showedthatthestrongesteffectsof marock et al., 2005). These two dynamic cores cannot be
changesinmeteorologyonPM concentrationswerefrom merged because each dynamic core corresponds to a set of
2.5
temperature, wind speed, absolute humidity, mixing height dynamicsolversthatoperatesonaparticulargridprojection,
andprecipitationeffects,whereascloudliquidwatercontent, grid staggering and vertical coordinate (Skamarock, 2005).
optical depth and cloudy area can lead to small changes in AssummarizedbySkamarock(2005),operationalresultsin-
PM onaveragewithappreciableresponsesinsomeareas. dicatedthatthesignificantdifferencesbetweenthesetwody-
2.5
The changes in concentrations of PM caused by changes namiccoreforecastsaremoretheresultofdifferentphysics
2.5
inmeteorologyshouldbetakenintoaccountinlong-termair but not dynamical core designs. The NMM core is a fully
qualitymanagementasconcludedbythem. compressiblehydrostaticNWP(NumericalWeatherPredic-
The 2006 Texas Air Quality Study/Gulf of Mex- tion)modelusingmassbasedverticalcoordinate,whichhas
ico Atmospheric Composition and Climate Study (Tex- beenextendedtoincludethenon-hydrostaticmotions(Janjic´,
AQS/GoMACCS)wasajointregionalairqualityandclimate 2003),whereastheARWcoreisafullycompressible,Eule-
changestudyconductedduringthelatesummer(1Augustto rian nonhydrostatic model with a run-time hydrostatic op-
15 October 2006). The objective of the program is to pro- tion available. The NMM core uses a terrain-following hy-
vide a better understanding of the sources and atmospheric brid (sigma-pressure) vertical coordinate and Arakawa E-
processes responsible for the formation and distribution of grid staggering for horizontal grid, whereas the ARW core
ozoneandaerosolsintheatmosphere,theirimpactonhuman uses a terrain-following hydrostatic-pressure vertical coor-
healthandregionalhazeaswellastheinfluenceontheradia- dinate with vertical grid stretching permitted and Arakawa
tiveforcingofclimateoverTexasandthenorthwesternGulf C-grid staggering for horizontal grid. As summarized in
of Mexico. The comprehensive observational data from the Yu et al. (2012), the physics package of the NMM (ARW)
2006 TexAQS/GoMACCS can be used to examine in detail includes the Betts-Miller-Janjic (Kain-Fritsch (KF2)) con-
theperformanceofairqualitymodelsfromamultipollutant vective mixing scheme, Mellor-Yamada-Janjic (Asymmet-
perspective, in terms of their surface concentrations as well ric Convective Model (ACM2)) planetary boundary layer
asverticaldistributions.Inthisstudy,weexaminetheimpact (PBL) scheme, Lacis-Hansen (Dudhia) shortwave and Fels-
ofthesetwodifferentmeteorologicalfields(WRF-ARWand Schwartzkopf (RRTM) longwave radiation scheme, Ferrier
WRF-NMM)ontheCMAQsimulationsforPM ,itschem- (Thompson) cloud microphysics, and NOAH (Pleim-Xiu
2.5
ical composition and precursors. The purpose of this paper (PX)) land-surface scheme. In this study, both WRF-ARW
is to comparatively examine the impact of these two differ- and WRF-NMM are employed to provide meteorological
entmeteorologicalfieldsonCMAQsimulationsforvertical fields for CMAQ (the notations ARW-CMAQ and NMM-
profiles of PM , its chemical composition and precursors CMAQ will be used hereafter to represent these two con-
2.5
onthebasisoftheextensivemeasurementsobtainedbyair- figurations). NMM-CMAQ uses the lowest 22 layered ver-
craftandshipduringthe2006TexAQS/GoMACCSfieldex- tical grid structure of the 60 hybrid layers in WRF-NMM
Atmos. Chem. Phys.,12,4091–4106,2012 www.atmos-chem-phys.net/12/4091/2012/
S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology 4093
Fig. 1. Comparison of the modeled (ARW-CMAQ, NMM-CMAQ) and observed daily PM2.5 concentrations at the AIRNow monitoring
sites (a) scatterplot (ppbv); (d) The NMB values of each model as a function of the observed daily PM2.5 concentration ranges; spatial
distributionsofNMBfor(c)ARW-CMAQand(d)NMM-CMAQduringtheperiod5Augustand7October2006.
meteorological fields directly without vertical interpolation smaller ARW domain sizes, and is able to use the emis-
through the use of a common vertical coordinate system. sion data from the NMM-CMAQ forecast model. Second,
On the other hand, the WRF-ARW model has been em- thepointsourceemissionswereredistributedtothe34layers
ployedtogeneratemeteorologicalfieldsforCMAQbecause according to the ARW meteorological fields on the basis of
the WRF-ARW meteorological model is compatible with thosefromtheNMM-CMAQmodel.Inaddition,theARW-
CMAQlikemm5before.FortheNMM-CMAQrun,there- CMAQ uses the same area sources such as the mobile and
sultsfromthetargetforecastperiod(04:00UTCtonextday’s biogenic sources as those in NMM-CMAQ. Therefore, the
03:00UTC)basedonthe12:00UTCNMM-CMAQsimula- totalemissionbudgetsforbothmodelsarethesame.Inboth
tion cycle over the domain of the continental United States ARW-CMAQ and NMM-CMAQ, the lateral boundary con-
(seeFig.1aofYuetal.,2012)areused,whereastheARW- ditions are horizontally constant and are specified by conti-
CMAQmodelwith34verticallayerswasappliedoverado- nental “clean” profile for O and other trace gases; the ver-
3
mainencompassingtheeasternUnitedStates(seeFig.1bof ticalvariationsarebasedonclimatology(ByunandSchere,
Yuetal.,2012)andwasrunfromthebeginningtoendwith 2006).Forbothmodels,thethicknessoflayer1isabout38m
firstthreedaysasmodelspin-upoverthewholeperiod. and the vertical coordinate system resolves the atmosphere
Giventhefactthatbothmodelsusedifferentmapprojec- between the surface and 50hPa although each model uses
tions and grid staggering, it is difficult to make the WRF- differentnumberofverticallevels.
ARW grid coverage identical to the WRF-NMM coverage. The Carbon Bond chemical mechanism (version 4.2) has
Several steps are taken to ensure that both the models are been used to represent photochemical reaction pathways in
set up as consistently as possible so that the comparison bothNMM-CMAQandARW-CMAQ.Theareasourceemis-
of the two models is meaningful. First, the meteorological sions are based on the 2001 National Emission Inventory.
fields of ARW were padded by 5 cells in both x and y di- The point source emissions are based on the 2001 Contin-
rections around the original meteorological domain when uous Emission Monitoring estimates of SO and NO pro-
2 x
themeteorologicalfieldswereprocessedusingMeteorology- jected to 2006 on a regional basis using the Department of
Chemistry Interface Program (MCIP) to create the CMAQ- Energy’s 2006 Annual Energy Outlook issued in January
ready files. This helps match the larger NMM domain and of 2006 (DOE, 2006). The mobile source emissions were
www.atmos-chem-phys.net/12/4091/2012/ Atmos. Chem. Phys.,12,4091–4106,2012
4094 S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology
Table1.ComparisonofARW-CMAQandNMM-CMAQmodelsforoperationalevaluationofdailyPM2.5 concentrationsonthebasisof
theAQSdataovertheeasternUnitedStates.
DomainMean,µgm−3 RMSE,µgm−3 MB,µgm−3 NMB(%) NME(%) R
Number Obs ARW NMM ARW NMM ARW NMM ARW NMM ARW NMM ARW NMM
Rural 4103 12.8 10.0 8.1 6.9 7.9 −2.8 −4.7 −21.9 −36.9 38.8 45.5 0.63 0.60
Suburban 6554 13.6 13.6 11.2 7.7 7.7 0.0 −2.3 0.2 −17.2 39.4 40.9 0.56 0.52
Urban 5299 13.2 13.5 11.2 8.1 7.8 0.4 −2.0 2.8 −15.4 41.7 42.2 0.53 0.50
Alldata 19168 12.3 12.2 10.0 7.9 7.6 −0.1 −2.3 −0.4 −18.4 43.7 44.3 0.53 0.51
generated by EPA’S MOBILE6 model using 1999 Vehicle served PM , SO2−, NO−, and NH+ data are available
2.5 4 3 4
MilesTraveleddataandafleetyearof2006. at 178STN sites within the model domain. The CAST-
TheaerosolmoduleinCMAQisdescribedbyBinkowski Net (http://www.epa.gov/castnet/) collected the concentra-
and Roselle (2003) and updates are described by Bhave et tion data at predominately rural sites using filter packs that
al. (2004) and Yu et al. (2007). The size distribution of areexposedfor1-weekintervals(i.e.,TuesdaytoTuesday).
aerosols in tropospheric air quality models can be repre- The aerosol species at the 34 CASTNet sites used in this
sented by the sectional approach (Zhang et al., 2004), the evaluationinclude:SO2−,NO−,andNH+.Thehourlynear
4 3 4
momentapproach(Yuetal.,2003),andthemodalapproach real-timePM dataat309sitesintheeasternUnitedStates
2.5
(Binkowski and Roselle, 2003). In the aerosol module of are measured by tapered element oscillating microbalance
CMAQ, the aerosol distribution is modeled as a superpo- (TEOM) instruments at the US EPA’s Air Quality System
sition of three lognormal modes that correspond nominally (AQS) network sites. In addition, measurements of verti-
to the ultrafine (diameter (Dp) <0.1mm), fine (0.1<Dp< cal profiles of PM , its related chemical composition and
2.5
2.5mm), and coarse (Dp>2.5mm) particle size ranges. gas species (CO, NO, NO , HNO , PAN, ethylene), and
2 3
Eachlognormalmodeischaracterizedbytotalnumbercon- meteorological parameters (liquid water content, water va-
centration,geometricmeandiameterandgeometricstandard por, temperature, wind speed and direction, and pressure)
deviation. The model results for PM concentrations are were carried out by instrumented aircraft (NOAA WP-3)
2.5
obtained by summing aerosol species concentrations over and a research ship deployed as part of the 2006 Tex-
the first two modes. Generally speaking, the modal ap- AQS/GoMACCS field experiment. The detailed instrumen-
proach offers the advantage of being computationally effi- tationandprotocolsformeasurementsaredescribedathttp:
cient,whereasthesectionalrepresentationprovidesmoreac- //esrl.noaa.gov/csd/2006/fieldops/mobileplatforms.html.The
curacy at the expense of computational cost. The CMAQ overviewofdataqualityandtheprincipalfindingsfromthe
modelisabletosimulatetheintegralpropertiesoffinepar- 2006TexAQS/GoMACCSfieldexperimentisgivenbyPar-
ticles such as PM mass and visible aerosol optical depth ris et al. (2009). The flight tracks of the WP-3 aircraft, and
2.5
well but it cannot resolve PM size distributions accurately ship movements are presented in Fig. 2 of Yu et al. (2012).
(Yu et al., 2008). In this study, we only present the model Theresultsforcomparisonoftheimpactoftwometeorolog-
performanceforPM massbutnotsizedistributions. icalmodelsonCMAQsimulationsovertheeasternUS(e.g.,
2.5
ARWdomainasshowninFig.1bofYuetal.,2012)during
2.2 Observationaldatabases theperiodof6Augustand6October2006arepresentedin
thisstudy.
Four surface monitoring networks for PM measurements
2.5
were employed in this evaluation (Interagency Monitoring
of Protected Visual Environments (IMPROVE), Speciated 3 Resultsanddiscussions
Trends Network (STN), Clean Air Status Trends Network
(CASTNet) and Air Quality System (AQS)), each with its 3.1 Impactofmeteorologyonspatialandtemporal
ownandoftendisparatesamplingprotocolandstandardop- variations of PM over the eastern US domain at
2.5
eratingprocedures.IntheIMPROVEnetwork,two24-hsam- theAQSsites
ples are collected on quartz filters each week, on Wednes-
day and Saturday, beginning at midnight local time (Sisler Table 1 summarizes the comparison results of the ARW-
and Malm, 2000). The observed PM , SO2−, NO−, EC CMAQandNMM-CMAQforthedaily(24-h)averagePM
2.5 4 3 2.5
and OC data are available at 71 rural sites across the east- concentrations.FollowingtheprotocoloftheIMPROVEnet-
ern United States. The STN network (http://www.epa.gov/ work, the daily (24-h) PM concentrations at the AQS
2.5
air/data/aqsdb.html) follows the protocol of the IMPROVE sites were calculated from midnight to midnight local time
network (i.e., every third day collection) with the excep- of the next day on the basis of hourly PM observations.
2.5
tion that most of the sites are in urban areas. The ob- The evaluation results at the urban and rural sites are also
Atmos. Chem. Phys.,12,4091–4106,2012 www.atmos-chem-phys.net/12/4091/2012/
S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology 4095
25
-3m) 20 MMeeaann ((AORbsW))
g Mean (NMM)
μM ( 2.5 1105
P
5
4 MB (ARW)
-3m) 2 MB (NMM)
g 0
μB ( --42
M -6
-8
-3m) 1101 RRMMSSEE ((ANRMWM))
g 9
μE ( 78
S 6
M
R 5
4
0.4 NMB (ARW)
0.2 NMB (NMM)
B
M 0
N -0.2
-0.4
-0.6
NME (ARW)
0.6
NME (NMM)
E 0.5
M
N 0.4
0.3
0.2
r (ARW)
0.8
r (NMM)
0.6
r 0.4
0.2
0
8/11 8/17 8/23 8/29 9/4 9/10 9/16 9/22 9/28 10/4
Time (2006)

Fig.2.ComparisonofdailyvariatiFoingsuroef 2t.h eCovmalpuaersisoofn doofm daaiinly- wviadreiatmioenasn o,fM thBe, vRaMluSesE o,fN dMomBa,iNn-MwiEdea nmdeacno,r rMelBat,i oncoefficient(r)forthe
dailyPM2.5massconcentrationsattheRMAISREN, oNwMmBo, nNitMorEin gansdit ecsofrroerlaAtiRonW -cCoeMffAicQienatn (dr)N fMorM th-Ce MdaAilQy sPiMmu2.5la tmioansss .
concentrations at the AIRNow monitoring sites for ARW-CMAQ and NMM-
CMAQ simulations.
summarized in Table 1. The domain wide mean values of ledtomoreunderestimationofPM attheIMPROVErural
2.5
mean bias (MB) and root mean square error (RMSE) (Yu sites.SinceTEOMmeasurementsforPM attheAQSsites
2.5
et al., 2006) for all daily PM at the AQS sites during the shouldbeconsideredaslowerlimitsbecauseofvolatilization
2.5 1
2006 TexAQS/GoMACCS period are −0.1 (−2.3) and 7.9 ofsolubleorganiccarbonspeciesinthedryingstagesofthe
(7.6)µgm−3forARW-CMAQ(NMM-CMAQ),respectively, measurement (Grover et al., 2005), the underprediction by
andthosefornormalizedmeanbias(NMB)andnormalized themodelislikelymoreseverethanthisevaluationsuggests.
meanerror(NME)are−0.4(−18.4)%and43.7(44.3)%for Additionalinsightintothenegativebias(underestimation)
ARW-CMAQ (NMM-CMAQ), respectively. It is of interest anderrors(scatter)ofbothmodelscanbegainedfromFig.1a
to note that both models performed much better at the ur- forthescatterplotandFig.1bfortheNMBvaluesasafunc-
bansitesthanattheruralsites,withgreaterunderpredictions tion of the different observed PM concentration ranges.
2.5
at the rural sites. As shown in section 3.2, the underestima- Table 1 and Fig. 1 depict that the model performance for
tion of PM at the STN urban sites by the NMM-CMAQ ARW-CMAQ and NMM-CMAQ is similar and reasonable
2.5
mainlyresultsfromtheunderestimationsoftheSO2−,NH+ forthePM concentrationwithveryclosevaluesofRMSE,
4 4 2.5
andTCMcomponents,whereastheoverestimationofPM NME, and correlation coefficient for both models although
2.5
at the STN sites by the ARW-CMAQ results from the over- theARW-CMAQhastheslightlybetterperformanceonthe
estimationsofSO2−,NO−,NH+,andOTHER.Thegreater basis of values of MB and NMB. Figure 1a and b clearly
4 3 4
underestimationsofSO2−,OCandECbytheNMM-CMAQ indicate that both ARW-CMAQ and NMM-CMAQ models
4
www.atmos-chem-phys.net/12/4091/2012/ Atmos. Chem. Phys.,12,4091–4106,2012
4096 S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology

100 40
PM2.5 (IMPROVE-arw) SO4 (IMPROVE-arw)
PM2.5 (IMPROVE-nmm) 35 SO4 (IMPROVE-nmm)
80 PM2.5 (STN-arw) SO4 (STN-arw)
PM2.5 (STN-nmm) 30 SO4 (STN-nmm)
SO4 (CASTNet-arw)
60 PM 25 SO4 (CASTNet-nmm)
2.5
( μ g m-3) 20
40 15
10 SO42-
20
5 ( μ g m-3)
0 0
0 20 40 60 80 100 0 5 10 15 20 25 30 35 40
15 15
NH4 (STN-arw) NO3 (IMPROVE-arw)
NH4 (STN-nmm) NO3 (IMPROVE-nmm)
NH4 (CASTNet-arw) NO3 (STN-arw)
NH4 (CASTNet-nmm) NO3 (STN-nmm)
10 10 NO3 (CASTNet-arw)
NO3 (CASTNet-nmm)
el
d
o
M
5 5
NO-
NH+ 3
4
( μ g m-3) ( μ g m-3)
0 0
0 5 10 15 0 5 10 15
15 25
TC (IMPROVE-arw)
OC (IMPROVE-arw)
TC (IMPROVE-nmm)
OC (IMPROVE-nmm)
20 TC (STN-arw)
TC (STN-nmm)
10
OC
15
( μ gC m-3) TC
10 ( μ gC m-3)
5
5
0 0
0 5 10 15 0 5 10 15 20 25
Observation
Figure 3a. Comparison of observed and modeled (ARW-CMAQ and NMM-CMAQ)
Fig.3a.ComparisonofobsePrvMe2d.5 aanndd mitso dcheelemdic(Aal RcWom-CpoMsiAtiQon aant dthNe MIMMP-RCOMVAEQ, S)TPNM a2n.5d aCnAdSiTtsNcehte smiteicsa dlucroinmgp ositionattheIMPROVE,
STNandCASTNetsitesduritnhge t2h0e062 0T0e6xATeQxSA/QGSoM/GAoMCCASC pCeSriopde riod.

reproducedthemajority(78%)oftheobserveddailyPM served daily PM concentrations in the southeast, espe-
2.5 2.5
concentrations within a factor of 2, especially for the con- cially for the NMM-CMAQ. To investigate the model per-
centration range of 10 to 35µgm−3. However, both models formanceovertime,thevaluesofmea1n,MB,RMSE,NMB,
overestimated the observations in the low PM concentra- NME and correlation coefficient (r) were calculated (do-
2.5
tionrange(<10µgm−3)withNMBvaluesof37.8%(ARW- mainwideaverages)andplottedasdailytimeseriesforthe
CMAQ) and 15.6% (NMM-CMAQ), respectively, but un- daily PM concentrations as shown in Fig. 2. The NMB
2.5
derestimates the observations in the high PM concentra- values range from −50.4% (23 September) to 18.9% (25
2.5
tionrange(>10µgm−3)consistently.ThesmallNMBvalue September)forNMM-CMAQandfrom−36.8%(7August)
(−0.4%)fortheARW-CMAQmodelresultsfromthecom- to 41.1% (2 October) for the ARW-CMAQ. Both models
pensationerrorbetweenlargePM overestimationforlow hadconsistentlyslightunderestimationsofPM forthefirst
2.5 2.5
PM concentrationportion(<10µgm−3)andunderestima- period from 6 August to 3 September but general overesti-
2.5
tion of high PM concentration portion (>10µgm−3) as mations after 3 September. The domain daily mean PM
2.5 2.5
indicated in Fig. 1b. The spatial distributions of NMB val- concentrations for the ARW-CMAQ are consistently about
ues for ARW-CMAQ (Fig. 1c) and NMM-CMAQ (Fig. 1d) 17%higherthanthosefortheNNM-CMAQduringthe2006
showthatbothmodelshadlargeunderestimationoftheob- TexAQS/GoMACCS period although the RMSE, NME and
Atmos. Chem. Phys.,12,4091–4106,2012 www.atmos-chem-phys.net/12/4091/2012/
S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology 4097
Table2.ComparisonofARW-CMAQandNMM-CMAQmodelsforPM2.5 anditscomponentsforeachnetworkovertheeasternUnited
Statesduringthe2006TexAQS/GoMACCSperiod.
CASTNet IMPROVE STN
SO24− NH+4 NO−3 SO2 TotS PM2.5 SO24− NO−3 OC EC TC PM2.5 SO24− NH+4 NO−3 TC
ARW-CMAQ
Mean(Obs) 4.16 1.26 0.32 0.72 2.42 7.19 2.48 0.22 1.24 0.31 1.55 13.49 3.86 1.32 0.56 4.32
Mean(Model) 3.74 1.23 0.47 1.45 3.31 7.14 2.88 0.38 1.11 0.28 1.38 15.53 4.90 1.91 1.12 3.23
Number 500 500 500 500 500 1648 1169 1169 1628 1628 1628 1816 1945 1945 1854 1971
correlation 0.88 0.82 0.30 0.73 0.81 0.52 0.64 0.35 0.30 0.48 0.37 0.30 0.57 0.45 0.36 0.29
MB −0.41 −0.03 0.15 0.72 0.89 −0.05 0.40 0.16 −0.13 −0.04 −0.17 2.04 1.04 0.59 0.56 −1.09
RMSE 1.46 0.52 0.60 1.09 1.58 6.25 2.62 0.92 1.27 0.58 1.71 11.10 3.38 1.49 1.54 2.74
NMB(%) −9.9 −2.5 46.7 99.4 36.8 −0.7 16.3 71.4 −10.7 −11.8 −10.9 15.1 27.0 100.6 44.8 −25.1
NME(%) 24.5 29.9 115.7 105.6 45.3 56.1 64.2 165.0 65.5 67.4 63.1 57.2 62.2 159.4 80.0 47.8
NMM-CMAQ
Mean(Obs) 4.16 1.26 0.32 0.72 2.42 7.19 2.48 0.22 1.24 0.31 1.55 13.49 3.86 1.32 0.56 4.32
Mean(Model) 2.94 0.99 0.43 1.36 2.93 5.85 2.04 0.37 1.00 0.22 1.22 11.32 3.33 1.33 0.74 2.52
Number 500 500 500 500 500 1648 1169 1169 1628 1628 1628 1816 1945 1945 1854 1971
correlation 0.89 0.83 0.23 0.77 0.83 0.61 0.77 0.39 0.42 0.53 0.46 0.35 0.66 0.54 0.36 0.37
MB −1.22 −0.27 0.10 0.64 0.50 −1.34 −0.44 0.15 −0.24 −0.09 −0.33 −2.17 −0.53 0.01 0.19 −1.80
RMSE 1.98 0.60 0.52 0.93 1.20 5.02 1.93 0.77 1.06 0.44 1.38 9.53 2.50 1.03 1.00 2.75

NMB(%) −29.3 −21.6 32.2 87.9 20.8 −18.6 −17.7 69.5 −19.7 −28.4 −21.5 −16.1 −13.7 0.4 33.3 −41.8
NME(%) 33.0 32.9 103.3 94.5 34.1 45.7 43.9 157.9 56.5 60.2 54.5 46.4 43.6 53.0 106.5 52.9

*TheunitofMean,MB,RMSEisµgm−3,andTotSistotalsulfur(SO24−+SO2)concentrations(µgSm−3).
3.2 Influenceofmeteorologyonspatialand
15
temporalevaluationforPM anditschemical
ARW-CMAQ 2.5
componentsattheCASTNet,IMPROVE,STN
NMM-CMAQ
sitesovertheeasternUS
The scatter plots of Fig. 3a indicate that at the IM-
10
PROVE, CASTNet and STN sites, both ARW-CMAQ
and NMM-CMAQ captured a majority of observed SO2−
4
+
(65% (ARW-CMAQ), 74% (NMM-CMAQ)), NH (60%
4
del (ARW-CMAQ),69%(NMM-CMAQ)),PM2.5(66%(ARW-
o CMAQ), 72% (NMM-CMAQ)) concentrations within a
M
5 factor of 2. The examination of the domain-wide bias
and errors (Table 2) for different networks reveals that
theNMM-CMAQconsistentlyunderestimatedtheobserved
mean SO2− by 29%, 18% and 14% at the CASTNet, IM-
Total Sulfur ( μ g S m-3) 4
PROVE and STN sites, respectively, whereas the ARW-
0 CMAQoverestimatedtheobservedmeanSO2−by16%and
4
27% at the IMPROVE and STN sites, respectively, with
0 5 10 15
Obs slight underestimation of 10% at the CASTNet site. Both
+
models overestimated the observed NH at the STN sites
4
Fig.3b.Comparisonofobservedandmodeled(ARW-CMAQand (by45 %forARW-CMAQand33%forNMM-CMAQ)but

Figure 3b. CompNaMriMso-CnM oAf Qob)steortavledsu alfnurd (mSOo24d−el+edSO (2A)RcWonc-CenMtraAtioQns anatd tNheMM-uCnMdeAreQsti)m ated at the CASTNet sites (by −3% for ARW-
total sulfur (SO 2C- A+S TSNOet)s citoesndcuernintrgatthieon20s0 a6t TtehxeA CQAS/GSToMNAeCt sCiStepse rdioudr.ing the 200C6M AQand−22%forNMM-CMAQ).Bothmodelsoveres-
4 2
timated the observed SO by more than 80% at the CAST-
TexAQS/GoMACCS period. 2
Net sites. The comparison of the modeled and observed to-

correlationcoefficientvaluesarecloseforthesetwomodels tal sulfur (SO2−+SO ) at the CASTNet sites in Fig. 3b re-
4 2
asshowninFig.2. vealsthatbothmodelsoverestimatedtheobservedtotalsul-

fursymmetricallyandthemodeledmeantotalsulfurvalues

arehigherthantheobservationsby37%and21%forARW-

CMAQ and NMM-CMAQ, respectively. This indicates too

muchSO emissionintheemissioninventory.
2

www.atmos-chem-phys.net/12/4091/2012/ Atmos. Chem. Phys.,12,4091–4106,2012

1
4098 S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology
Table3.FlightObservationSummaryforWP-3aircraftforPMduringthe2006TexAQS/GoMACCSstudy.
Date ObservationsummaryforWP-3∗
9/13 Dallasemissioncharacterizationandchemicalprocessing,meanflowwindspeedis5ms−1withnortherlywind,takeoffat10:45
andlandingat16:45(LST)
9/15 HoustonUrban,Parishpowerplant,Isolatedrefineries,lightwinds,Emissioncharacterization,chemicalprocessing,meanflow
windspeedis4.5ms−1withsoutheasterlywind,takeoffat09:50:andlandingat16:20(LST)
9/16 Houstonemissioncharacterizationandchemicalprocessing,NETexapowerplantsandagedHoustonplume,takeoffat9:55and
landingat16:30(LST)
9/19 HoustonUrban,Parishpowerplant,Isolatedrefineries,flowfromtheNE,Emissioncharacterization,chemicalprocessing,mean
flowwindspeedis7.5ms−1withnortheasterlywind,takeoffat09:50andlandingat16:20(LST)
9/20 Beaumont Port Arthur, Houston Urban, Parish power plant, Isolated refineries, emission characterization, chemical processing,
meanflowwindspeedis5.0ms−1witheasterlywind,takeoffat09:55andlandingat16:15(LST)
9/21 HoustonUrbanandIndustrial,Parishpowerplant,emissioncharacterization,chemicalprocessing,meanflowwindspeedis9.5
ms−1withsoutherlywind,takeoffat09:50andlandingat16:25(LST)
9/25 Dallas,Parishpowerplant,BigBrownandLimestonepowerplants,GMDtower,emissioncharacterization,chemicalprocessing,
meanflowwindspeedis3.5ms−1withnortherlywind,takeoffat09:45andlandingat16:25(LST)
9/27 HoustonUrban&Industrial,Parishpowerplant,Beaumont-PortArthur,emissioncharacterization,chemicalprocessing,meanflow
windspeedis3.5ms−1withsoutherlywind,takeoffat12:45andlandingat17:55(LST)
9/29 HoustonUrban&Industrial,Parishpowerplant,Emissioncharacterization,chemicalprocessingintothenight,meanflowwind
speedis7.0ms−1withsoutheasterlywind,takeoffat13:45andlandingat20:10(LST)
10/5 HoustonUrban&Industrial,Parishpowerplant,Chemicalprocessingandtransport,lightwindsfromthenortheasterly,takeoffat
9:50andlandingat16:20(LST)
10/6 HoustonUrban&Industrial,Parishpowerplant,VictoriaandSeadrift,chemicalprocessingandtransport,windsfromthenorth-
easterly,takeoffat09:50andlandingat16:00(LST)
*Basedonflightinformationpresentedathttp://esrl.noaa.gov/csd/tropchem/2006TexAQS/P3/index.htmlandMcKeenetal.(2009).
−
The poor model performance for NO (see scatter plot underestimation of secondary OC (SOA) formation such as
3
in Fig. 3a and correlation <0.40 except that at the STN sources from the oxidation of isoprene and sesquiterpenes
sites for the NMM-CMAQ in Table 2) is related in part to (Edney et al., 2005) and an aqueous-phase mechanism for
−
volatilityissuesofmeasurementsassociatedwithNO ,and SOA formation from the oxidation of VOCs (Carlton et al.,
3
their exacerbation because of uncertainties associated with 2006)thatwerenotyetincludedintheversionoftheCMAQ
SO2− and total NH+ simulations in the model (Yu, et al., modelusedhere.Morrisetal.(2006)foundthatincludingthe
4 4
2005).Table2indicatesthatbothmodelsunderestimatedthe SOA formation from sesquiterpene and isoprene improved
observed mean OC, EC and TC concentrations at the IM- theCMAQmodelperformanceforOC.
PROVE sites by −11%, −12% and −11% for the ARW- Figure 4 shows comparisons of stacked bar-plots for
CMAQ,respectively,andby−20%,−28%and−21%for observed and modeled concentrations for each chemical
theNMM-CMAQ,respectively.NotethatsincetheSTNnet- constituent of PM at the STN sites. Note that “OTHER”
2.5
work used the thermo-optical transmittance (TOT) method species in Fig. 4 refers to unspecified anthropogenic mass
todefinethesplitbetweenOCandECwhiletheIMPROVE which comes from the emission inventory of PM , i.e.,
2.5
andthemodelemissioninventoryusethethermo-opticalre- [PM ]=[SO2−]+[NH+]+[NO−]+[TCM]+[OTHER].
2.5 4 4 3
flectance(TOR)method,onlythedeterminationoftotalcar- Since organic compounds comprising ambient particulate
bon(TC=OC+EC)iscomparablebetweenthesetwoanal- organic mass are largely unknown, an average multiplier is
ysisprotocols(Yuetal.,2004).Therefore,Table2onlylists frequently used to convert measurements of OC (typically
the performance results for TC comparisons from the STN reportedasµgCm−3)toorganiccarbonaceousaerosolmass
sites.Bothmodelsconsistentlyunderestimatedtheobserved (OCM). The value of 1.4 has been widely used to estimate
TC concentrations at the STN sites by −25% for ARW- particulate organic mass (e.g., Turpin and Lim, 2001) from
CMAQ and −42% for NMM-CMAQ. As pointed out by measured OC and is also used in our analysis. The ARW-
Yu et al. (2007), factors contributing to this underestima- CMAQ overestimated the observed PM at the STN sites
2.5
tionofthemodeledOCinclude:(1)missingsourcesofpri- (mostofthemarelocatedinurbanareas)by15%,whereas
mary OC in emission inventory used for the summer, (2) the NMM-CMAQ underestimated by −16% as listed in
Atmos. Chem. Phys.,12,4091–4106,2012 www.atmos-chem-phys.net/12/4091/2012/
S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology 4099
3.3 Influence of meteorology on vertical profiles for
20 STN sites SO4 PM2.5 chemical components (SO24−, NH+4), and its
NO3 relatedgasspeciesfrom2006TexAQS/GoMACCS
NH4
TCM
15 Tocomparethemodeled(ARW-CMAQ,NMM-CMAQ)and
Other
-3m) 32% observed vertical profiles, following Yu et al. (2012), the
Concentration ( g μ10 1420%9%% 173%% 30% mtTthehroepedaoeihlrlaoceturdeardflrytetstooruetlthsthsoeelwvcmoeedorrredemeeslpxogotdrrneailddceitidnendgdoiboucybtepssmue(tracsvtoacwlthuieoimrnnegnaa,ltlrhtsoieomwpelaoisnns.eidTtailrhoalenyyseoirbon)f--.
7%
servedandmodeleddatapairsweregroupedaccordingtothe
12%
27% modellayerforeachdayandeachflight.Theverticalprofiles
5 43%
28% from both models and observations obtained in this manner
canberegardedtorepresentaverageconditionsencountered
15% 21% 23% over the study domain. We refer to these average regional
0 vertical variations as composite vertical distributions in the
Obs ARW-CMAQ NMM-CMAQ
subsequentdiscussions.Table3summarizesthespecificmis-
Fig.4.Comparisonofstackedbar-plotsforobservedandmodeled sionsandweatherconditionsencounteredforeachflightused

(AFiRguWre 4-.C CMomApaQris,onN oMf stMack-eCd bMar-AplQots) foPr Mobs2er.v5edc ahned mmoidcealeldc (AoRmWp-CoMsiAtiQo, nNMatMt-he in this study. WP-3 conducted most of its measurements
STNTCeMxsAAitQQe)Ss /PGMdou2M.5r AcihnCeCgmSic tpahel recioodm2. p0 To0shi6eti opneTr acete xnthtAaeg SQeTs SNre/ psGirteesose Mdnut rtAhinegC f rthaCcet S2io0n0ps6 oe fr eiaocdh . The per- during the daytime (∼09:40 to ∼17:00LST) except on 29
c“eOnTtcwaHohgmiE cephosR csio”trmieosenpsp f rfoeerroc sPmieMe ntsh2.te5. r e“temOhfiTeessHriosfEnrR taio”nc vstpeuineonctoinserypss roeefocf ePfirMfis et2eo.5ad .u c nahspnetcchiofrimeod ppanootghsreiotpniooigcnenmicf oamrsasssP wMh2i.c5h. Sinetpotenmigbhetr(1w3h:4e5ntoth2e0W:10P-L3STm)e.aAsusrseummemntasriwzeedrebycoMndcuKceteedn
comes fromtheemissioninventoryofPM2.5. et al. (2009), the WP-3 spent a significant fraction of its al-

locatedflighttimebetween300and700mabovetheground

and had 10 daytime flights between 13 and 29 September

Table2.Thestackedbar-plotsofFig.4showthattheunder- 2006 which consisted of upwind and downwind transects
estimation of PM2.5 at the STN sites by the NMM-CMAQ of the Houston and Dallas urban areas. Figure 5 presents
mainly results from the underestimations of the SO2−, modeledandobserveddailycompositeverticaldistributions
4
NH +4 and TCM components, whereas the overestimation1of for PM2.5 chemical components (SO24−, NH+4) and related
PM2.5 attheSTNsitesbytheARW-CMAQresultsfromthe gaseousspecies(HNO3,SO2,NH3,VOC(isoprene,toluene,
overestimationsofSO2−,NO−,NH+,andOTHERalthough terpene))duringthe2006TexAQS/GoMACCSperiod.Mean
4 3 4
the ARW-CMAQ still underestimated the observed TCM. compositeverticaldistributionsaccordingtothemodellayer
Ontheotherhand,bothmodelsconsistentlyunderestimated forthemodels(ARW-CMAQandNMM-CMAQ)andobser-
theobservedPM attheIMPROVEsites(mostofthemare vationsforthewholeperiodaresummarizedinTable4.
2.5
located in rural areas) by −1% for the ARW-CMAQ and
−19%fortheNMM-CMAQ.Thenotableunderestimations 3.3.1 VerticalprofilesofSO2−,andNH+
4 4
of SO2−, OC and EC by the NMM-CMAQ led to the
4
underestimationofPM attheIMPROVEsitesasshownin As shown in Fig. 5 and Table 4, both ARW-CMAQ and
2.5
Table2.Theseresultssuggestaneedtoimproveaccuracyof NMM-CMAQgenerallyestimatedSO2− wellonmostdays
4
TCM at both rural and urban sites. On the basis of analysis except on 9/16 and 9/21 in which the NMM-CMAQ had
of the diurnal cycles from the AQS PM monitors and consistentlyhighSO2−.NMM-CMAQalsohasconsistently
2.5 4
+
comparison with model median diurnal cycles over the high NH on 9/16 and 9/21 relative to both observation
4
northeasternUSduringthe2004ICARTTstudy,McKeenet andARW-CMAQ.AsanalyzedinMcKeenetal.(2009),on
al.(2007)foundsomeinconsistencieswithcertainprocesses both9/15and9/21,theairmassesoriginatingfromwestern
within the models and the observations. They found very Louisiana merging with the Houston plume with high CO,
little diurnal variation in the median observed diurnal organic aerosol and EC but relative reduced enhancements
cycles at urban and suburban monitor locations. However, ofNOy,SO2 andtolueneweresampledbytheWP-3.There
significantdiurnalvariabilitywasexhibitedbysomemodels, was an additional influence of an aged continental air mass
suchastheEta-CMAQ,thatdoesnotcapturethedecreaseof fromtheeastorsoutheastaffectingthenortheasternHouston
observedPM from01:00to06:00LT,indicatingareduced with a possible biomass burning signature (McKeen et al.,
2.5
roleforaerosollossduringthelatenightandearlymorning 2009). These characteristics of air masses may make some
hours(McKeenetal.,2007).ThelargescatterinFig.3afor contribution to the poor performance of NMM-CMAQ for
PM canalsoariseduetoinadequaterepresentationofthe SO2−andNH+on9/21.Figure5andTable4revealthatboth
2.5 4 4
+
diurnal evolution of observed PM by both ARW-CMAQ models often overestimated NH for all altitudes except at
2.5 4
andNMM-CMAQ. layer1,whereasbothmodelssystematicallyunderestimated
www.atmos-chem-phys.net/12/4091/2012/ Atmos. Chem. Phys.,12,4091–4106,2012
4100 S.Yuetal.: EvaluationoftheimpactofWRF/NMMandWRF/ARWmeteorology
104
103 9/13 HNO3 (Obs)
HNO3 (ARW)
ht (m) 102 9/15 9/16 9/19 9/20 HNO3 (NMM)
eig 101
H104
103
102
9/21 9/25 9/27 9/29 10/5 10/6
101
0 4 8 4 8 4 8 4 8 4 8 4 8
HNO (ppbv)
3
104
NH3 (Obs)
103 9/20 9/21 NH3 (ARW)
m) NH3 (NMM)
ght (102 9/16 9/19
ei101
H104
103 10/5
9/29
102
9/25 9/27 10/6
101
0 2 4 2 4 2 4 2 4 2 4
NH (ppbv)
3
104
9/13 9/15 9/16 9/19 9/20
103 SO2 (Obs)
SO2 (ARW)
m)102 SO2 (NMM)
ht (101
eig104
H 9/21 9/25 9/27 9/29 10/5 10/6
103
102
101
0 4 8 4 8 4 8 4 8 4 8 4 8
SO (ppbv)
2
104
9/13 9/15 9/16 9/19 9/20 Isoprene (Obs)
103 Isoprene (ARW)
Isoprene (NMM)
m)102
ht (101
eig104
H 9/25 9/27 9/29 10/5 10/6
103 9/21
102
101
0 400 1000 400 1000 400 1000 400 1000 400 1000 400 1000
Isoprene (pptv)

Figure 5. Comparison of composite vertical distributions of observed and modeled
Fig. 5. Comparison of composite vertical distributions of observed and modeled (ARW-CMAQ and NMM-CMAQ) HNO3, NH3, SO2,
Isoprene,toluenS(eA,OteRrp2W-e na-enCs,dMP NMA2H.Q5S+ a Oan24l−do naNngdM tNhMHe +4-aCiarlMconragAftQth et)ra aiHrncNrsaefOtctt3rsa, noNsefHc Wts3,oP fS-WO3P 2d-,3 uIdrsiuonrpignrg ethtnheee ,22 0t0o006lu6Te enxAeQ, tSe/GrpoeMnAeCsC, SP.M2.5
4 4
TexAQS/GoMACCS.
Atmos. Chem. Phys.,12,4091–4106,2012 www.atmos-chem-phys.net/12/4091/2012/
1