Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

co  PV ov pl th pl ac ve gr vo th T ar ca po ov co Based up AERMOD the highes 4.0 Mod As a mea current predicting measured assessed a 4.1 Case‐ For this a three airpo    These air three imp large com sufficient concentra airport em Furthermo developed layout an periods fo nservatively MRM  ‐ Th erestimate th umes by ove e manner in ume volume count for the rtical disper ound for low lume of the e conversion he PVMRM ea sources us lculating the rtion of the erestimate th nversion of N on the outco software w t potential fo eled vs. Me ns of better EDMS/AEDT NO2 valu NO2 data at nd compared Study Airpo nalysis, mon rts were eva Los Angel Adelaide ( Montreal Internation ports were ortant criteri mercial serv enough t tions and rati ission sourc re, and in the and utilized d monitor lo r the study. high results. e PVMRM e conversion rstating the a which the plu calculation vertical exte sion coeffici -level plume plume could to NO2.  method has fu ed to simula plume volum area sourc e volume of O to NO2, a mes of this ith the “built r meeting the asured NO quantifying /AERMOD es near airp three “case- . rts  itoring data luated: es Internation Australia) Ai (Canada) Pi al Airport (Y specifically s a: (i.) they r ice facilities o model os; (ii.) they e activity le case of LAX for the purpo cations are s may have of NO to N mount of O3 me volume i in PVMRM nt of the plu ent) that ma s. This over contribute t rther limitat te line segme e depends o e actually im the plume fo nd would lik qualitativel -in” Tiers 1, Preferred M 2 Data  the accuracy configurati orts, model study” airpo from the fo al Airport (L rport (ADL), erre Elliot UL). elected base epresent med with activit ambient N had air qual vels could b and ADL, s se of predict hown in the 16  a tendency O2 for low-l available du s calculated. also does me (based on y extend be estimation of o overestima ions for area nts. In these n the projec pacts a ne r purposes of ely overestim y-based asse 2, and/or 3 f ethod Key T of the ons in ed and rts were llowing AX) and Trudeau d upon ium-to- y levels O2/NOx ity monitors e computed ome modelin ing annual N adjoining fig to evel e to The not the low the ting source applic cases, the lat ted width of arby recepto determining ate ambient N ssment, the or predicting argets (see Se located in th concurrent g data or mo O2 concentra ure. Table  Qu The AER in”  pre pos me Tar LAX Air Mo ations, espec eral extent of the area sou r. This aga the amount O2 concentr application short-term N ctions 1.3 an eir close pro with the air del input file tions. For illu 5 presents th alitative Ass  application  MOD softwa Tiers 1, 2, an dicting short sess the high eting the Pref gets  nitoring Sta ially for elon the plume u rce, even if o in would ten of O3 availab ations. of EDMS/A O2 levels po d 3.3). ximities; and monitoring s had already stration, the e air quality essment Out of EDMS/AED re with the “ d/or 3 for  ‐term NO2 lev est potential erred Metho tions (Examp gated sed in nly a d to le for EDT- ssess (iii.) data. been LAX time come  T‐ built‐ els   for  d Key  le) 

17  Table 5.  Air Monitoring Time Periods  Airport  Purpose  Timeframe  Time  Period  Resolution  Adelaide International (ADL) Permanent station July 1, 2011 - May 31, 2012 1 Year Hourly Los Angeles International (LAX) Air Quality & Source Apportionment Study (AQSAS) February 1 - March 16, 2012 Season 1 (winter) 6 weeks Minute July 18 - August 28, 2012 Season 2 (summer) 6 weeks Minute Specific Plan Amendment Study (SPAS) January 1 - December 31, 2009 1 Year Hourly Montreal International (YUL) Permanent station January 1, 2009 - December 31, 2011 3 Years Hourly 4.2 Modeling Methodology  Using EDMS, emissions from airport-related sources (i.e., aircraft, GSE, APU’s, etc.) were computed based upon source-specific activity data (e.g., landing/take-offs, operating times, etc.), emission rates, and temporal values (hours, quarter hours, minutes) for each case-study airport and time period analyzed. Within AERMOD (and following modeling convention), airport-related sources were simulated as either “area” or “point” sources. For example, area sources were used to represent aircraft gate aprons (i.e., aircraft at startup, GSE operations, and APU activity); aircraft taxiing, queuing, and accelerating on the taxiway/runway system; and aircraft in climbout and approach modes. Airport roadways and parking facilities were also modeled as a series of area sources. Point sources were used to represent stacks from stationary sources such as boilers and generators.  4.3 Statistical Analyses  Descriptive statistics were used to assess the accuracies of the modeled-to-monitored comparisons for the case-study airports. These tests comprised (but were not limited to) the following:  R-Squared (R2) - For this analysis, R2 statistics measured how close the modeled vs.-monitored data fitted 1:1 regression line.  Q‐Q  Plots  -  For this analysis, NO2 modeled values were compared to NO2 monitored values (unpaired in time). Notably, U.S. EPA typically conducts model evaluations using this method of comparison.3  Scatterplots  (Paired  in  Time)  -  For this application, paired-in-time monitored vs. modeled NO2/NOX ratios were compared. 3 This approach is the only approach EPA has used and is widely accepted as a comparison of modeled to monitored values (Blewitt and Wood, 2014).

 Sc ra gr  R de an  M ex m  M co de Presented summariz 4.3.1 Q‐Q The Q-Q against ea ADL: As s concentra high conc atterplots  (U tio concentra aph. HC  - The RH termined fro d predicted v ean Squared amine the p onitored valu aximum Val mpare the termining if below in ed above:  Plots  plot is a gra ch other. As hown, at low tions, the met entrations, m npaired  in T tions were p C aids in r m a tail exp alues.  Error (MSE erformance es. ues Reporte upper limits the models a the followin phical metho noted above, monitored N hods under-p ethods over-p ime) - For t lotted. Mode educing the onential fit to ) - The MSE of a model a d  - The max and concen re reporting r g subsection d for compa the U.S. EPA O2 redict, and a redict.  18  his analysis, led and mon effect of ex the high en calculation i nd is a mea imum values trations of ealistic upper s is expand ring probabi typically em t LAX AQ most m excepti at very the met NOX concen itored values treme values d of the freq s an importa sure of the reported by each conver limits and c ed informat lity distributi ploys this m SAS: At low ethods tend t on of the OLM low concentr hods tend to trations comp were also p on model uency distrib nt statistical squares of th each model sion method oncentrations ion from th ons by plott ethod in eva monitored N o under-pred which slig ations. At hi over-predict. ared to NO2 lotted on the comparisons. ution of obs test that is us e departure were examin . This assis . e statistical ing their qua luating mode O2 concentra ict, with the htly over–pre gh concentra   /NOX same It is erved ed to from ed to ts in tests ntiles ls. tions, dicts tions,

LAX AQSA concentra and at hig over-pred LAX AQSA NO2 conc concentra S – CE: At lo tions, the met h concentrati ict.  S – CS: The m entrations thr tions levels.  w NO2 monit hods tend to ons, the meth ethods tend oughout all m ored under-predic ods tend to to over-pred onitored 19  t, LAX AQ concent predict concent have a f high co predict. models ict LAX SPA the met concent SAS – CN: At rations, the m and then beg rations. Mid- air agreemen ncentrations, ARM2 tend at higher con S: At low m hods tend to rations, they very low mo ethods tend in to under-p level concen t with the mo the methods s to over-pred centrations.  onitored NO under-predict tend to over- nitored NO2 to slightly ov redict at low trations seem nitored data tend to over- ict less than 2 concentratio , and at high predict. er- to . At other ns,

YUL: At lo they tend 4.3.2 Sca For this an time mon model con under-pre presented ADL: A la amount of and PVM 1:1 line, in w monitored to over-predi tter Plots (P alysis, mode itored ratios. version meth dictions occu and briefly d rge range of scatter is als RM w/ varia dicating that concentratio ct.  aired in Tim led NO2/NO A 1:1 (x = od over- or r below. Th escribed belo both monito o present in a ble). The OL this method ns, the metho e)  x ratios were y) reference under-predic ese scatterplo w. red and mod ll conversion M w/ variab is a bit less c 20  ds tend to un compiled and line has also ts the values ts for ADL eled ratios is methods (A le method a onservative i der-predict, plotted agai been added . Over-predic and two of present with RM2, OLM, ppears to hav n estimating and at high c nst the corre as an indica tions occur a the LAX AQ in the ADL OLM w/ var e the most p the NO2/NOx oncentrations sponding pai tion of whet bove this lin SAS datase dataset. A la iable, PVMR oints below ratios. , red in her a e and ts are rge M, the

LAX AQSA issues in m tightly clu (ARM2, O ARM2 is conservat monitored LAX AQSA 6.35 (note excluded indicates t and is less S‐ CE Season onitored da stered within LM, OLM w generally be ive than oth /low modele S ‐ CN Seaso : modeled va on this plot). hat ARM2 is conservative  1: Monitore ta, mostly in 0.2 – 0.9. A / variable, P low the 1:1 er conversi d quadrant. n 2: Monitor lues were at Note that no generally be than other c d ratios range very low con large amou VMRM, and line, indicati on methods ed ranges hav zero at this hi modeled ratio low the 1:1 l onversion me 21  from ~0.13 centrations) nt of scatter PVMRM w ng that it un . ARM2 al e a wide spre gh monitored s above 1 w ine, indicatin thods. - 1.1 (indicat while model is also prese / variable). T der-predicts so has a c ad of ratios, ratio of 6.35 ere found in t g that it unde ive of roundi ed ratios see nt in all conv his scatterpl more freque lustering ar ranging from . As a result his dataset. T r-predicts mo ng errors or Q med to be m ersion metho ot indicates t ntly and is l ound the h almost 0 to , no values w his scatterplo re frequently A ore ds hat ess igh ere t

4.3.3 Sca Modeled plotted an they plot m ADL:  At concentra concentra similar stu of the NO 4 RTP En Modeli http://w tter Plots (U and monitore d described b onitored da lower NOx tions best, an tions the best dies.4 At hig 2/NOx Ratio. vironmental Ass ng: Developmen ww.epa.gov/scr npaired in  d NO2/NOx r elow. Note t ta and model concentratio d at mid-to- . The ARM2 her NOx con ociates, Inc. Am t and Evaluatio am001/models/ Time)  atios as a fun hat these mon ed data indep ns, the PVM higher NOx method rati centrations, t bient Ratio Met n Report. Publis aermod/ARM2_ 22  ction of mod itored and m endent of eac RM w/vari concentration os also fit th he PVMRM hod Version 2 (A hed September Development_a eled and mo odeled data h other. able genera s, the ARM e “ARM2 cu w/variable p RM2) for use w 20, 2013. Retrie nd_Evaluation_ nitored NOx points are no lly matched 2 method fi rve” equation roduces the h ith AERMOD f ved December Report-Septem concentration t paired in tim the monito t the monito reproduced ighest estim or 1-hr NO2 8, 2014. ber_20_2013.pd s are e, as red red by ate f.

 LAX AQSA the monit to fit the m highest es LAX  AQSA generally ARM2 m PVMRM LAX SPAS closely al method se w/ variabl S – CE: At lo ored concentr onitored co timate of the S  –  CN:  At match the m ethod seems w/ variable is : At lower N igned with th ems to fit the e is the highe wer NOx co ations best, a ncentrations NO2/NOx Ra lower NOx onitored co to fit the mo the highest Ox concentra e monitored c monitored c st estimate o ncentrations, nd at mid-to best. At high tio.  concentratio ncentrations nitored conc estimate of th tions, the PV oncentration oncentration f the NO2/NO 23  the results fr -higher NOx er NOx conc ns, the PVM best, and at entrations the e NO2/NOx R MRM w/ va s, and at mid s the best. At x Ratio.  om PVMRM concentration entrations, th RM w/ var mid to high best. At hig atio.  riable and O to higher NO higher NOx w/variable s, the ARM2 e PVMRM w iable and OL er NOx con her NOx con LM w/ varia x concentrat concentration generally ma method see /variable is   M w/ varia centrations, centrations,   ble are the m ions, the ARM s, the PVMR tch ms the ble the the ost 2 M

YUL: At l most clos ARM2 m PVMRM 4.3.4  RH As shown concentra ratios whe when com M Full C A OLM w PV PVMRM Monit 4.3.5 Me The mean MSE, the values for The result associated ower NOx co ely aligned w ethod seems w/variable pr C in Table 6 b tion when co n compared pared to the ethod onversion  RM2  OLM  /Variable  MRM   w/Variable  ored (NO2)  an Squared squared erro closer the m ADL is show s indicate tha with the AR ncentrations ith the moni to fit the mo oduces the h elow, the RH mpared to m to monitored monitored rat Table 6. RHC X(N)  88 66 70 60 73 70 31  Error (MSE r is a measur odeled con n in Table 7 t the ARM2 M2 when co , the PVMRM tored concen nitored conc ighest estima C results ind onitored rat ratio concen io concentrat : Monitored X  169 75 122 109 148 141 35 )  e of the squar centrations a . conversion m mpared to th 24  w/ variabl trations and, entrations the te of the NO2 icate that th ios concentr trations. All o ions.  NO2 vs. Mod Valu RHC 386 98 258 240 348 328 47 es of the dep re to the mo ethod has th e monitored e and OLM at mid to hig best. At hig /NOx Ratio. e ARM2 has ations. PVM ther convers eled NO2 (pp es  RHC Ra - 0.25 0.67 0.62 0.90 0.85 0.12 arture from m nitored conc e lowest MS values is sm w/ variable m her NOx con her NOx con the lowest o RM is great ion methods b): ADL  tio  Highe Conce 4 4 4 4 4 onitored val entrations. A E, indicating aller than fo ethods are centrations, centrations, ver-predicted ly over-pred over-predict st NO2 ntration  54 91 54 54 54 54 56 ues. The low sample of that the diffe r other conve the the the ratio icting ratios er the these rence rsion

25  methods. The PVMRM and OLM methods’ MSE values are likely very sensitive to the input NO2/NOx emission ratios. Table 7. Mean Squared Error (ppb): ADL  Method  MSE (ppb)  Full 252.47 ARM2 144.36 OLM 190.31 OLM w/variable 174.51 PVMRM 221.15 PVMRM w/ variable 216.12 4.4 Case‐Study Results   Based on the above, the results of the statistical analyses revealed that estimated NO2 concentrations near the case-study airports derived from existing models and using existing conventional NO2/NOx conversion methods have a number of potential shortcomings. The following highlights some of the most noteworthy of these findings:  Low Correlations - The comparison of modeled vs. measured NO2 values at the three test-case airports demonstrated poor overall correlation and does not lead to positive conclusions regarding the accuracy of the various NOx conversion methods. Previous studies supported this same finding relative to NO2 concentrations. In particular, ACRP Report 715 stated that low R2 values were reported when comparing modeled-to- monitored values for carbon monoxide (CO), NOx, carbon dioxide (CO2), and other parameters.6 Other studies also support that modeled vs. monitored CO concentrations have low R2 values.7  NO2/NOx  Ratios  - ARM2 had the least amount of greatly overestimated NO2 concentrations. However, this observation could differ if lower non-aircraft NO2/NOx emission ratios were used (0.5 was used for all non-aircraft sources).  Background  Values  - Including proper “background” concentrations to use as inputs for the modeling would be required for more accurate output concentrations but is a difficult task given the time-varying nature of these concentrations. For this study, no background values were used. Comparison of modeled and measured NO2 at more secluded, non-urban airports where the non- airport contribution to ambient concentrations is much smaller should be considered in future research.  AERMOD  Performance  - Predicted concentrations of NOx by AERMOD can be greatly overestimated at low wind speeds. Inaccuracies in predicting NOx (rather than NO2) indicate that 5 Transportation Research Board. ACRP, Report 71, Guidance for Quantifying the Contribution of Airport Emissions to Local Air Quality. 2012. 6 R-square can take on any value between 0 and 1, with a value closer to 1 indicating that a greater proportion of variance is accounted for by the model. For example, an R-square value of 0.8234 means that the fit explains 82.34% of the total variation in the data about the average. 7 Martin, Anjoli. Verification of FAA’s Emissions and Dispersion Modeling System. 2006. Case‐Study Airports Modeled‐to‐ Monitored Comparisons  In general, at lower NOx values, the  PVMRM‐w/variable and OLM‐ w/variable correlate with the  monitored concentrations best and,  at mid‐to‐higher NOx  concentrations, the AERMOD ARM2  method correlates best. However,  in all cases, there was relatively  poor statistical correlation.