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.

Appendix A Interim Report

ACRP 02-43: Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations Task 9 Interim Report Prepared for: Airport Cooperative Research Program (ACRP) Transportation Research Board of The National Academies Michael Kenney, Principal Investigator KB Environmental Sciences, Inc. 9500 Koger Blvd, Ste. 211 St Petersburg, FL 33702 May 19, 2015

A ‐ ii  Executive Summary  I. Purpose of the Report  The purpose of this Interim Report is to (i.) provide the ACRP 02-43 Project Panel with an update and overview of the Research Team’s findings in connection with Tasks 5 through 8 of the Amplified Work Program and (ii.) provide the Panel with the Team’s approach for commencing with Tasks 10 through 12. A meeting of the Panel and the Research Team is scheduled for June 30th in Washington D.C. to review and discuss the contents of this report as well as to reach consensus on the “go-forward” plan. II. Contents of the Report For ease of reference, a “refresher” on the outcomes from Tasks 1 through 3 previously presented in the Task 4 White Paper is provided first. This includes an overview of the literature search, the current knowledge and information gaps, and the available speciation methods as they pertain to the modeling of NO2 concentrations near airports. The remainder of the Report presents the following:  Task 5, Comparison of Emission Factors to Measured Values – This work involved a comparison of NO/NO2 emissions factors used in EDMS/AEDT to available published measurements of NO and NO2 in aircraft engine (and other airport sources of) emissions at airports.  Task 6, Comparison of Modeled & Measured NO2 Data – Under this task, a compressive comparison was undertaken of modeled and measured NO2 data at three “test-case” airports representing various fleets, activity levels, geography, and meteorological conditions. This included the use of the EDMS/AEDT/AERMOD modeling system and a range of NO2/NOx conversion methods.  Task  7,  Research  Plan – Based on information gathered and developed in support of Tasks 1 through 6, the Research Team has prepared a Research Plan to (i.) further evaluate existing methods for predicting NO2 concentrations from airport NOx emissions, (ii.) assess alternative method(s), and (iii.) develop a Preferred Method.  Task 8,  Independent Technical Review – Under this task, the Research Team prepared a plan for conducting an independent technical review (ITR) of the outcomes of Tasks 5, 6 and 7 with an emphasis on providing feedback on the Task 7 Research Plan. Task 10, Execute Research Plan, Task 11, ITR, Task 12, Future Research Needs, and Task 13, Final Report will be initiated and completed following the Interim Report Meeting pending the ACRP 02-43 Panel’s authorization. III. Observations and Recommendations Based upon the research that had been completed thus far, there are several “key” observations and recommendations that the Research Team will seek feedback on from the ACRP 02-43 and the ITR Panels. These include the following:  Predicted NO2 Concentrations ‐ The results of this assessment indicate that very poor correlation between monitored and modeled NO2 values has been achieved using the EDMS/AEDT/AERMOD modeling system and the various NO2/NOx conversion methods. Further assessment should be conducted on the NO2/NOx emission ratios for both aircraft and non-aircraft sources and aircraft engine thrust settings. The  contents  of  this  report  focus  mainly on  the outcomes of  Tasks 6  through 8.   Tasks 1  through 3 were  reported  upon  previously  in  the  Task  4  White  Paper  and  are  again  summarized  herein.  Tasks  10  through  13  will  be  completed  following  the  Interim  Report  Meeting. 

A ‐ iii   Background Concentrations - Time-varying background concentrations of NO2 and NOx have not been fully applied to modeling results. Further assessment of the non-airport sources of these compounds is warranted.  Source Apportionment - Although emissions from aircraft landing / takeoff cycles are the largest source of NOx emissions at each of the three test-case airports analyzed they are not necessarily the biggest contributor to predicted NOx and NO2 concentrations based on the modeling. It is recommended that these observations and issues be discussed by the Research Team and the Project Panel at the Interim Report Meeting with the goal of achieving consensus on how they can be remedied within the remaining schedule and funding available to the ACRP 02-43 Research Project.

A ‐ iv  ACRP 02‐43 Research Participants  Research Team: Michael Kenney, Principal Investigator, KB Environmental Sciences (KBE) Carrol Fowler, Administrative Officer, KBE Cristina Schoonard, Air Quality Specialist, KBE Paola Pringle, Air Quality Specialist, KBE Rick Miake-Lye, Principal Air Quality Scientist, Aerodyne Scott Herndon, Air Quality Scientist, Aerodyne Ezra Wood, Aircraft Engine Emissions Specialist, Univ. of Mass. - Amherst Brian Kim, Principal Model Architect, Wyle Roger Wayson, Air Quality Scientist, Wyle ACRP Manager: Joe Navarrete Topic Panel: Maria Pope, (Chair), John Wayne Airport William Brewer, City of Dallas Aviation Department Phil DeVita, Harris Miller Miller & Hanson Eric More, Mecklenburg County Air Quality Thomas Ryerson, NOAA Panel Liaisons: Peggy Wade, Federal Aviation Administration Chris Owen, Environmental Protection Agency Christine Gerencher, Transportation Research Board

A ‐ v  Table of Contents 1. Purpose of the Report ........................................................................................................................... 1 2. Background Information, Research Objectives & Research Plan ......................................................... 1 2.1  Background Information ............................................................................................................... 1  2.2  Research Objectives ...................................................................................................................... 2  2.3  Work Program ............................................................................................................................... 2  3. Tasks 1, 2 & 3 Refresher ........................................................................................................................ 4 3.1  Regulations & Standards (Task 1) ................................................................................................. 4  3.2  Current Understanding and Research Gaps (Task 2) .................................................................... 6  3.3  Emission/Dispersion Models & NOx Prediction Methods (Task 3) ............................................. 12  4. Comparison of Emission Factors to Measured Data (Task 5) ............................................................. 17 4.1  Task Objective ............................................................................................................................. 17  4.2  Aircraft Engine Emission Factors & Measurements .................................................................... 17  4.3  Non‐Aircraft Emission Factors .................................................................................................... 22  5. Comparison of of Modeled and Measured NO2 Data (Task 6) ........................................................... 24 5.1  Task Objective ............................................................................................................................. 24  5.2  Air Monitoring Data .................................................................................................................... 24  5.3  Air Modeling Data ....................................................................................................................... 27  5.4  Monitored Data: Description and Statistics ................................................................................ 32  5.5  Monitored Versus Modeled NO2 Concentrations Assessment .................................................. 39  5.6        Conclusions and Observations .................................................................................................... 60  6. Research Plan (Task 7) ........................................................................................................................ 62 6.1  Task Objectives ........................................................................................................................... 62  6.2  Available Models/Methods ......................................................................................................... 62  6.3  Research Plan Evaluation Methods & Criteria ............................................................................ 84  7. Independent Technical Review Plan (Task 8) ...................................................................................... 90 7.1  Task Objective ............................................................................................................................. 90  7.2  ITR Panel Candidates ................................................................................................................... 90  7.3  ITR Implementation Plan ............................................................................................................ 91  7.4  ITR Feedback Template ............................................................................................................... 91  8. Issues in Need of Resolution ............................................................................................................... 94

A ‐ 1  1. Purpose of the Report The principal objective of this Interim Report prepared in support of the Airport Cooperative Research Program (ACRP) 02-43 Research Project entitled Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations is essentially two-fold: Purpose of the Report   To provide the Project Panel with an update and overview of the Research Team’s findings in connection with Tasks 5 through 8 of the Work Program; and  To provide the Panel with an approach for completing the remainder of the Program and reach consensus with the Research Team for commencing with Tasks 10 through 13. For ease of reference, a “refresher” of the outcomes from Tasks 1 through 3 previously presented in the Task 4 White Paper is also provided.1 A meeting of the Panel and the Research Team is scheduled for June 30th in Washington D.C. to review and discuss the contents of this report as well as the “go-forward” plan. 2. Background Information, Research Objectives & Research Plan As a reminder to the Panel, this introductory section provides pertinent background information on the Principal Objectives and the Work Plan for addressing the problem that is the subject of the ACRP 02-43 Research Project. Additional and more detailed information on these topics is contained in the Amplified Work Plan, published separately.2 2.1 Background Information  In 2010, the U.S. Environmental Protection Agency (EPA) promulgated a new short-term (i.e., one-hour) National Ambient Air Quality Standard (NAAQS) for the pollutant nitrogen dioxide (NO2).3 As a means of assessing NO2 concentrations and compliance with this standard, computer models are often used. For airport-related applications, the available models comprise the Federal Aviation Administration (FAA) EDMS/AEDT models and the U.S. EPA AERMOD model.4 Simply stated, EDMS/AEDT is used to compute airport emissions of nitrogen oxides (NOx) from aircraft engines, auxiliary power units (APUs), ground support equipment (GSE), etc. and AERMOD is used to predict the resultant ambient (i.e., outdoor) NO2 concentrations. 1 ACRP 02-43 Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations Task 4 Working Paper, February 28, 2014. 2 ACRP 02-43 Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations Amplified Work Plan, August 6, 2013. 3 The one-hour National Ambient Air Quality Standard (NAAQS) for NO2 is 53 parts per billion (ppb). 4 EDMS/AEDT - Emissions and Dispersion Modeling System/Aviation Environmental Design Tool and AERMOD - American Meteorological Society/Environmental Protection Agency Regulatory Model. Report Purpose  This  Interim  Report  is  designed  to  inform the ACRP 02‐43 Panel on the  progress of the research prior to the  Panel/Research  Team  Meeting  and  to  help  achieve  consensus  for  progressing  with  the  remainder  of  the Work Plan.    

   Within A applying methods. include:  A  U  U  U Unfortuna (i.e., calib characteri result, the the availa pressing n 2.2 Resea Based upo computed Research  A FA N  D  Pr ai The produ Module fo computer utilizing underlyin alternative 2.3 Work Consisten Team Pro for this P tasks. For (the Tasks the tasks a                    5 EDMS/ Meteor 6 The Re electro 7 ACRP ERMOD, th both screen Arranged in ssuming com se of the Am se of the Ozo se of the Plum tely, neither rated) when stics than sta re is present ble means fo eed to remed rch Objecti n the need to , the ACRP Project: ssess the acc A’s EDMS O2 concentra evelop and a oduce a Pref rport emissio cts of the re r EDMS/AE model code the correct g mechanics s to these mo  Program  t with the ap posal, combi roject7, the ease of assi shown in bl re briefly sum                         AEDT - Emissi ological Society search Plan call nic copy of the m 02-43 Kick-Off M   e U.S. EPA ing and det increasing o plete convers bient Ratio M ne Limiting e Volume M the models n applied to tionary and o ly very limit r computing y this problem ves assess and a 02-43 Panel ACRP uracy of the /AEDT mode tions near air ssess alternat erred Method ns.6 search, a Fin DT to Predi , together wi inputs (e.g. and predictiv dels). proach conta ned with outc resultant Am milation, this ue are compl marized bel                     ons and Dispers /Environmental s for at least one odule code wil eeting Minutes has develo ailed NOx-to rder of com ion of all NO ethod (ARM Method (OLM olar Ratio M or these met airports wh ther mobile ed experienc NO2 concen . dvance how developed th  02‐43 Obje NOx emissi ls and the U ports5; ive methods; for predictin al Report ent ct NO2 Conc ll help ensur , NO2/NOx e accuracies ined in both omes of the plified Work Work Progr eted and thos ow. ion Modeling Sy Protection Age prediction met l be included in September 4, 2 A ‐ 2  ped guidan -NO2 specia plexity (i.e. x to NO2; /ARM-2); ); and ethod (PVM hods have be ich have va sources when e and confid trations at a NO2 concent e three follo ctives  ons and spec .S. EPA AER and g NO2 conce itled Develop entrations, a e that the m emission rat of EDMS/AE the ACRP s “Kick-Off” M Program c am is shown e shown in b stem/Aviation ncy Regulatory hod involving th the final deliver 013. ce for deve tion (i.e., c , Tiers), thes RM). en evaluated stly differen it comes to ence in the a irports and, t rations near wing objectiv iation metho MOD mode ntrations res ment of NOx Preferred M odels and m ios) and im DT and AE olicitation an eeting/Conf omprises 13 in the adjoi lack are rem Environmental D Model. e NO-NO2-ozo able. loping and onversion) e methods or verified t emission NO2. As a ccuracy of herefore, a airports are es for this ds used in l to predict ulting from Chemistry ethod and ethods are prove the RMOD (or d the KBE erence Call individual ning figure aining) and esign Tool. AE ne photo-station AC RMOD - Amer ary state and an RP 02‐43 Wor ican k Program

A ‐ 3  ACRP 02‐43 Amplified Work Program   Task  1,  Regulations  &  Standards – The primary aim of this task was to compile, evaluate and summarize what is presently known or currently under study on the combined topics of air quality and the new 1-hour NO2 NAAQS.  Task 2, Current Understanding & Research Gaps – This task described the current understanding and research gaps relative to measuring and predicting NO2 concentrations in airport environments.  Task 3, NOx Speciation Methods – Under this task, the Research Team analyzed and evaluated how emissions and chemistry of NOx are accounted for in various NOx speciation methods in EDMS/AEDT/AERMOD.  Task 4, White Paper –The results of Tasks 1 through 3 were summarized in a Working Paper8 and submitted to the Panel for review and a webinar was held with the Panel to discuss the findings.9  Task  5,  Compare  Emission  Factors  to  Measured  Values – This work involved a comparison of NO/NO2 emissions factors used in EDMS/AEDT to available published measurements of NO and NO2 in aircraft engine exhaust at airports.  Task 6, Compare Modeled & Measured NO2 Data – Under this task, a comparison was undertaken of modeled and measured NO2 data at a sample of airports representing various fleets, activity levels, geography, and meteorological conditions.  Task 7, Research Plan – Based on information gathered in Tasks 1 through 6, the Research Team prepared a Research Plan to develop alternative method(s) for predicting NO2 concentrations from airport NOx emissions, to evaluate existing and alternative methods and recommend a Preferred Method.  Task 8,  Independent Technical Review – Under this task, the Research Team prepared a plan for conducting an independent technical review (ITR) of the Task 7 Research Plan.  Task 9, Interim Report – The product of this task is this Interim Report and meeting of the Panel and the Research Team to discuss the findings, obtain feedback from the Panel, and resolve any issues before execution of the Research Plan.  Task 10, Execute Research Plan – The approved Task 7 Research Plan will be executed under this task.  Task  11,  Execute  Independent  Technical  Review – The approved Task 8 Independent Technical Review Plan will be executed under this task.  Task 12, Future Research Needs – This work will identify future related research need(s) and prepare ACRP problem statement(s) to address these need(s).  Task 13, Final Report – Under this task, the Research Team will prepare a final report documenting the research approach, methodologies and results. Importantly, the contents of this Interim Report focus mainly on the outcomes of Tasks 6 through 8. Tasks 1 through 3 were reported upon in the Task 4 Working Paper and Tasks 10 through 13 have not yet been completed. 8 ACRP 02-43, Task 4 Working Paper, February 28, 2014. 9 ACRP 02-45, Task 4 Webinar, April 10, 2014. Interim Report  The contents of this report focus  mainly on the outcomes of Tasks 6  through 8.  Tasks 1 through 3 were  reported upon in the Task 4 White  Paper and Tasks 10 through 13 have  not been completed. Interim Report 

    A ‐ 4    3. Tasks 1, 2 & 3 Refresher  The outcomes of Tasks 1, 2 and 3 were published previously in the Task 4 Working Paper and discussed during a subsequent webinar. For ease of reference, the essential elements of these materials are summarized below. 3.1 Regulations & Standards (Task 1)  With a focus on aviation, Task 1 identified and summarized pertinent regulations (e.g., acts, laws, policies, standards, etc.) related to air quality, in general, and airports, in particular. 3.1.1 Air Quality Standards   U.S. EPA promulgated the NAAQS10 to safeguard public health and environmental welfare against the detrimental effects of outdoor air pollution. NAAQS have been established for six “criteria” air pollutants: carbon monoxide (CO), lead (Pb), NO2, ozone (O3), particulate matter (PM10/2.5), and sulfur dioxide (SO2). EPA periodically reviews the NAAQS to determine if revisions or supplements are warranted. EPA also uses NO2 as the indicator for the larger group of NOx. EPA first set standards for NO2 in 1971, setting both a primary standard (to protect health) and a secondary standard (to protect the public welfare) at 53 parts per billion (ppb -100 µg/m3), averaged annually. The EPA has reviewed the standards twice since that time, but chose not to revise the annual standards at the conclusion of each review. However, in January 2010, EPA established an additional primary standard at 100 ppb (188 µg/m3), averaged over one hour.11 The form of the 1-hour standard is the 3-year average of the 98th percentile of the yearly distribution of 1-hour daily maximum NO2 concentrations. 3.1.2 Emission Standards  The regulations that are most relevant to the ACRP 02-43 research are those that mandate the maximum level of NOx emissions that can be emitted from the engine(s) on an aircraft or from other emission sources within the airport environs. The following provides a synopsis of these NOx emission regulations:  Aircraft  Engines – U.S EPA’s initial regulations for gaseous exhaust emissions from aircraft (including NOx) were promulgated in 1997 and are aligned with the emissions standards established by the International Civil Aeronautics Organization (ICAO) and generally referred to as the Committee on Aviation Environmental Protection (CAEP) standards. In summary, these emission standards (i.e., Tiers) have become progressively more stringent and are expected to continue as such into the future. Since 2012, new aircraft engine models must meet the CAEP/6 standards (the Tier 6 standards) and engines certified after 2014 must meet the CAEP/Tier 8 standards.  Motor Vehicle  Engines – U.S. EPA has also long established motor vehicle tailpipe emission standards that apply to automobiles, trucks, vans, buses, etc. For example, starting in 2017, Tier 3 would set new vehicle standards, including those for NOx. Compared to current requirements,                                                              10 Title 40 CFR Part 50 – National Primary and Secondary Ambient Air Quality Standards, http://www.ecfr.gov/cgi-bin/text- idx?c=ecfr&tpl=/ecfrbrowse/Title40/40cfr50_main_02.tpl. 11 Federal Register / Vol. 75, No. 26 / Tuesday, February 9, 2010 / Rules and Regulations, http://www.gpo.gov/fdsys/pkg/FR- 2010-02-09/pdf/2010-2554.pdf. Refresher Materials  The  materials  in  this  section  are  intended to provide the ACRP 02‐45  Panel with a broad overview and the  highlights  of  Tasks  1,  2  and  3.  The  Task 4 Working Paper contains more  detailed information.    

A ‐ 5  these new NOx tailpipe standards for light-duty vehicles represent an 80 percent reduction per fleet average.  Nonroad Engines – U.S. EPA has promulgated emissions standards for compression ignition (i.e., diesel powered) nonroad vehicles and equipment representing levels of emissions reductions that vary based on horsepower and whose phase-in (i.e., implementation schedule) varies by model year. Tier 4 is currently the most stringent, although EPA promulgated interim standards to allow extra time for industry and in-use fleets to become compliant. These nonroad standards apply to airport ground support equipment (GSE).  Stationary  Sources  – Under the federal Clean Air Act (CAA), there are several drivers for regulating and controlling NOx emissions associated with stationary sources. These include Title I: Provisions for Attainment and Maintenance of National Ambient Air Quality Standards, Title IV: Acid Deposition Control, and Title V: Permits. There are no emission standards that apply to NOx emissions from aircraft auxiliary power units (APUs). 3.1.3 Air Monitoring Data  Presently, all areas within the U.S. meet the annual NO2 NAAQS, with concentrations measured at area- wide monitors well below the level of the standard (i.e., 53 ppb). Therefore, there are no “nonattainment” areas for NO2, although the South Coast Air Basin in California is classified as an “attainment/maintenance” area. Since 1980, annual average ambient NO2 concentrations have decreased by more than 40 percent and currently range from approximately 10 to 20 ppb. There are also no violations of the one-hour NAAQS in the U.S., although assessing the monitoring data has just gotten underway and the standard is based on a three-year average. Notably, the national trend also shows a 29 percent decrease in 1-hour NO2 concentrations from 2000 through 2012. There are also no known violations of the short-term standard near any U.S. airports. However, violations of the standard have been measured near London Heathrow International Airport. Significantly, NO2 concentrations near roadways are appreciably higher than those measured at typical monitor networks. In fact, in-vehicle concentrations can be two to three times higher than measured at nearby area-wide monitors. Near-roadway (i.e., within 50 meters) concentrations of NO2 have been measured to be approximately 30 to 100 percent higher than concentrations away from roadways. 3.1.4 Monitoring & Measuring Methods  The analytical method for quantifying NO/NO2 in both the ambient monitoring and for aircraft exhaust is based on chemiluminescence. In this technique, sampled air is mixed with artificially produced O3 to produce, via chemical reaction, NO2 in an excited electronic state. The relaxation of the excited state NO2 to its "ground state" energy produces a photon that can be detected as light/signal. Since this chemiluminescence technique detects only NO, the sample is also passed through a hot molybdenum catalyst to convert NO2 to NO, so that the sum of NO and NO2 (i.e., "NOx") is detected. The NO2 concentration is then calculated by finding the difference between this NOx measurement (in which the sampled air passes through the catalyst) and NO measurements (in which the sampled air does not pass through the catalyst.) Airport NO2 Concentrations  There are no known violations of the  one‐hour NAAQS  for NO2  near U.S.  airports.   However, violations of the  standard  have  been  recorded  at  London Heathrow Airport. 

    A ‐ 6    Dedicated calibration results show that when properly operated the catalytic conversion of NO2 to NO can be near unity, but other compounds besides NO2 are also converted to NO via this process (e.g., nitric acid [HNO3], peroxyacetyl nitrate [PAN], and nitrous acid [HONO]). This can result in an overestimate of the true NO2 concentration. The extent to which NO2 is over reported depends on the concentrations of these other compounds, most of which are produced photochemically in the atmosphere. Recognizing that even small overestimates of true ambient NO2 concentrations could have serious regulatory ramifications, the U.S. EPA has field-tested several “new-technology” NO2 sensors that are immune to these interferences Finally, given the current use of the chemiluminescence technique in engine certification tests, it is likely that engine certification NOx emission indices actually include the contribution from HONO, which can account for several percent of the total NOx emission index. 3.2 Current Understanding and Research Gaps (Task 2)  The primary aim of Task 2 was to provide a synopsis of current understanding (i.e., What Is Known) and research gaps (i.e., What Is Not Known) regarding the topics of measuring and modeling of NOx emissions and NO2 concentrations in airport environments. The following subject areas were discussed and are summarized as follows:  3.2.1 Aircraft Performance Characteristics Aspects of an aircraft’s performance while in the taxi and takeoff modes that are relevant to NOx emissions include the amount of time an aircraft spends in each mode and the corresponding engine thrust settings. Presently, there are two prominent research projects underway by the ACRP that deal with aircraft NO2/NOx emissions in the taxi and takeoff mode. These include the following:  ACRP 02-45: Methodology to Improve EDMS/AEDT Quantification of Aircraft Taxi/Idle Emissions is being conducted to better quantify aircraft taxi/idle emissions through a better understanding of how aircraft operate (i.e., thrust setting, speed, operating time) within the taxi in and taxi out.  ACRP 02-41: Estimating Takeoff Thrust Settings for Airport Emissions Inventories is aimed at improving aircraft takeoff NOx emission estimates. From these and other studies, it is recognized that during taxi, aircraft thrust settings vary considerably and full thrust settings during takeoffs are seldom used. The taxi times also vary by airport, runway/taxiway system and aircraft type. 3.2.2 Factors Affecting Aircraft NOx Emissions Factors that affect aircraft NOx emissions include aircraft/engine combinations, ICAO emission indices/fuel flow rates, aircraft times-in-modes, single-engine taxi procedures, aircraft takeoff weight, thrust settings/reverse takeoff thrust, mixing height and altitude, and alternative fuels. Total emissions of NOx per engine per LTO mode are calculated by the product of the fuel-based emission index (grams of pollutant emitted per kilogram of fuel burned), the fuel flow rate (kilogram of fuel per second), and the total time-in-mode (in seconds). 3.2.3 Emission Indices  The calculation of aircraft NOx emissions requires emissions indices (i.e., rates) and corresponding fuel flow rates collected under the same conditions. Most of these quantities come primarily from the ICAO Aircraft Engine Emissions Databank and are based on measurements taken from engines running in test

A ‐ 7  beds and at conditions outlined in a certification standard. Importantly, the certification standard has been set for total NOx emissions, and not for the individual species of NO and NO2. Table  3.1 provides a summary of the NOx emission indices for aircraft engines currently within the EDMS/AEDT models. As shown, the NOx emission indices are higher for takeoff than for climbout, approach and idle. Secondly, the NOx emission indices tend to be higher for turbine engines than for turboprop and piston. Table 3.1. EDMS/AEDT Emission Indices for NOx  (grams per kilogram of fuel) Engine Type  Operating Mode  (Thrust Settings)  Maximum Minimum  Average  Turbine  Takeoff (100%) 65.8 2.09 26.8 Climbout (85%) 46.3 2.30 21.2 Approach (30%) 28.0 1.15 9.40 Idle (7%) 8.53 0.80 3.89 Turboprop  Takeoff (100%) 20.4 0.01 10.2 Climbout (85%) 17.9 0.01 9.53 Approach (30%) 10.7 0.89 6.99 Idle (7%) 7.47 0.45 3.03 Piston  Takeoff (100%) 5.88 0.36 2.69 Climbout (85%) 5.60 0.24 3.89 Approach (30%) 10.2 0.95 2.00 Idle (7%) 1.90 0.39 1.01 It is also noteworthy that the ICAO certification thrust settings for idle, takeoff, climb-out, and approach are 7, 100, 85, and 30 percent of full throttle, respectively. Since EDMS/AEDT use the corresponding NOx emission indices at these four thrust settings to calculate NOx emissions, deviations from these values during actual operation can lead to inaccuracies in NOx emissions inventories and predicted NO2 concentrations. Several studies have demonstrated that actual aircraft operations can be at thrust settings other than these ICAO settings. Most notably, 7 percent thrust usually overestimates actual ground-idle thrust and 100 percent thrust usually overestimates actual takeoff thrust. Lower thrusts reduce the amount of fuel burned during a takeoff and the combustion temperature inside the combustion chamber, which results in significant reductions in air emissions – particularly those of NOx. EDMS/AEDT also does not account for de-rated takeoff thrust which is part of the ACRP 02-41 Estimating Takeoff Thrust Settings for Airport Emissions Inventories research and may result in a modification to future applications of EDMS/AEDT. The impact of actual idle thrust settings on total NO2 and NOx emissions has not been extensively studied. However, it is noteworthy that although the idle component of a LTO cycle usually accounts for a small portion of total NOx emissions compared to climb-out and takeoff emissions, it can account for well over 50 percent of total NO2 emissions.

    A ‐ 8    3.2.4 NO2/NOx Emission Ratios Extensive NO, NO2 and total NOx emissions testing has been conducted on a wide range of aircraft engines in the last decade.12,13,14,15 From this, a few of the most important points regarding actual aircraft NO and NO2 emissions and how they are handled in EDMS/AEDT/AERMOD are listed as follows:  ICAO Values - Total NOx emission indices from aircraft engines agree reasonably well with the ICAO certification values, although the actual thrust values used can greatly affect NOx/NO2 values, as discussed above.  Thrust Settings  ‐ In contrast to most other NOx emission sources (e.g., gasoline vehicles, coal combustion), direct NO2 emissions account for a variable and at times large portion of total NOx emissions from aircraft engines. The NO2/NOx emission ratio decreases with power, from over 0.95 at the lowest power setting (four percent rated thrust or taxi/idle) to under 0.10 at higher power settings (65 to 100 percent rated thrust or climbout and takeoff).  Idle Versus Takeoff  ‐ As a result of the thrust dependence of the NO2/NOx emission ratio, the climb-out and takeoff portions account for the biggest portion of total NO (and NOx) emissions from the LTO cycle, but the idle mode accounts for the greatest portion of NO2 emissions.  Vertical Variation ‐ Because of the dependence of the NO2/NOx emission ratio on thrust setting, the NO2/NOx emission ratio is altitude-dependent. In other words, NO2 is emitted mainly at the surface, whereas NO emissions occur over a wider range of altitudes (surface and aloft). The extent to which the vertical variation in the NO2/NOx emission ratio affects downwind NO2 concentrations is not well understood.  Non‐Aircraft  Sources  ‐ NO2/NOx emission ratios from on-road vehicles can be summarized as less than 0.05 for gasoline-fueled vehicles, 0.05 to 0.2 for diesel vehicles not equipped with particulate filters or oxidation catalysts, and 0.15 to 0.8 for diesel-fueled vehicles equipped with particulate filters/oxidation catalysts. Taken from the Task 4 Interim Report, Table 3.2 provides a summary listing of the emission ratios for most airport-related sources of NOx/NO2 emissions.                                                              12 Aircraft Particulate Emissions eXperiment – APEX (2004), JETS-APEX2 (2005), and APEX3 (2005). 13 Additional sources include but are not limited to the F100 study, and the Frankfurt remote sensing measurements. 14 Miller, T., J. Ballenthin, D. Hunton, A. Viggiano, C. Wey, and B. Anderson. Nitric Acid Emission from the F100 Jet Engine. Journal of Geophysical Research, 2003. 15 Schäfer, Klaus, Carsten Jahn, Peter Sturm, Bernhard Lechner, and Michael Bacher. Aircraft Emission Measurements by Remote Sensing Methodologies at Airports, Atmospheric Environment, 2003.

A ‐ 9  Table 3.2. NO2/NOx Emission Ratios for Various Airport Emission Sources Source  NO2/NOx Ratio  Aircraft - Approach ~0.16 Aircraft - Idle 0.6 to 0.98 Aircraft - Takeoff ~0.08 Aircraft - Climbout ~0.09 APU 0.25 to 0.55 Gasoline Vehicles 0.01 to 0.05 Diesel Vehicles 0.05 to 0.20 Diesel Vehicles with DPM 0.15 to 0.80 Gasoline GSE 0.05 Diesel GSE 0.05 to 0.90 Stationary Engines (Diesel) 0.06 Stationary Engine (CNG) 0.13 3.2.5 Plume Dynamics and Measurements NO2 formation and its concentration are also highly dependent on plume dispersion characteristics. Specifically, plume growth and mixing with the surrounding air affects chemical transformations involving NOx/NO2/O3. In short, plume growth and mixing are dependent on various dispersive forces acting on the plume, which can include the following: atmospheric turbulence, atmospheric rise and wake turbulence. For example, all moving sources at an airport produce wakes with turbulent eddies that can impact pollutant dispersion. However, smaller and slower-moving equipment such as GSE generally produce less turbulent wakes than larger and faster moving equipment such as motor vehicles and aircraft. As illustrated to the right, during landing and takeoff operations, the wake vortices formed by the aircraft fuselage and wing tips can be relatively large and have significant impacts on dispersion. In addition, an aircraft also contributes to the wake effect through its engine blast (or wash) which includes the exhaust emissions themselves.   Finally, for reactive pollutants like NO2 (and its precursors), the interactions that occur when plumes merge also need to be accounted for in order to properly model the associated chemical transformations. 3.2.6 Plume Chemistry In simple terms, NOx (i.e., NO and NO2) is formed during fuel combustion mainly as a result of the thermal oxidation of atmospheric nitrogen (N2). From most combustion sources, NOx is emitted mostly in the form of NO and in the atmosphere this NO can be converted to NO2 by reaction with ozone (as follows): Wake Vortex Wake Vortex Jet Wash and  Exhaust Emissions Aircraft Wake Effects

    A ‐ 10     NO + O3 → NO2 + O2 NO can also be converted to NO2 by reaction with hydroperoxy radicals (HO2) and organic peroxy radicals (RO2, where “R” represents a carbon-based radical like CH3•). During the day (when sunlight is available), NO2 undergoes photolysis, reforming NO and O3 (as follows):  NO2 + light + O2 → NO + O3 Both of these reactions (i.e., NO to NO2 and NO2 to NO) should be quantified to accurately predict downwind NO2 concentrations. Notably, when the rates of these two reactions are equal, NO, NO2, and O3 are said to be in a “photostationary state.” The time required for the photostationary state to form is typically at least a few minutes, or significantly longer if mixing of the exhaust with ambient air is the limiting factor in attaining a photostationary state. As discussed above, mixing of different exhaust plumes (e.g., aircraft exhaust from the runway and exhaust from on-road vehicles at the terminal) can also disrupt the photostationary state. Meteorological conditions are also important factors because of their role in determining atmospheric mixing and plume evolution and also because of the dependence of the reaction NO + O3 on temperature and the dependence of the NO2 photolysis rate on solar radiation. Compounding this process, emissions of NO can rapidly lead to reductions in O3 concentrations and the associated increases in NO2 concentrations. In this way, the maximum concentration of NO2 that can be formed (i.e., “secondary NO2”) is limited by the ambient O3 concentration (and to a smaller extent, peroxy radicals) which is typically between 0 and 100 ppb. Further downwind, O3 levels can recover and even exceed background values via additional photochemical reactions involving volatile organic compounds (VOCs). This description of the impact of NOx emissions on NO2 and O3 values is fairly accurate for numerous types of NOx sources (e.g., power plants, on-road and off-road vehicles - all of which emit NOx mainly in the form of NO [i.e., when NO2/NOx emission ratios less than five percent]). By comparison, aircraft NOx emissions are different since at low engine thrust (i.e., during idle/taxi), the NO2/NOx emission ratio can exceed 90 percent. As a result, the maximum possible NO2 concentrations are not “limited” by the ambient O3 concentration since there is so much “primary” NO2 emitted. There is also an additional mechanism by which NO emitted from aircraft engines at low power (i.e., during taxi/idle) is converted to NO2. Although not well understood, this conversion was not from O3 chemistry (as discussed above) but presumably by the oxidation of NO resulting from the elevated peroxy radicals concentrations in the plume. 3.2.7 Dispersion Modeling at Airports  Although much progress has been made in the last 20 years to better characterize the dispersive nature of the atmosphere in modeling airport air quality, more research is necessary. The following factors can make airport-related dispersion modeling challenging:  Source location ‐ Since an airport can encompass a relatively large area (some major airports may be 20 square miles or more in size), it can be difficult to accurately characterize the location of Aircraft NO2/NOx Ratios  The  derivation  of  aircraft  engine  NO2/NOx ratios  is very complex and  measurably  different  from  most  emission  sources.  Although  significant progress has been made,  there  are  still  some  uncertainties  about  the  effects  of  time,  distance  and chemistry on these values.  

A ‐ 11  each source. Along with meteorological data, the proper characterization of each source’s position information can be critical. Oftentimes, the data for airport sources will only include schedule-type information of when certain activities occurred, such as when an aircraft departed or when a GAV was used, and may not indicate which taxiways or access roads were used. Incorrect location data can significantly affect concentration predictions at a receptor location since the receptor may experience the wrong part of a plume (may under- or over-predict based on the location of the receptor in relation to the emission point).  Multiple Sources ‐ Unlike a single, stationary source, the diversity of sources at an airport and the aforementioned spatial distribution over a large area provides a modeling challenge. The diversity of source types as well as the sheer number of sources requires the modeling of many plumes to represent all of the sources. Even with the aggregation of some sources into groups (e.g., GSE grouped into area sources, aircraft modeled as part of taxiway and runway area sources, etc.), the number of sources for an airport can number in the thousands for a full airport assessment (i.e., based on using the current modeling scheme implemented within EDMS/AERMOD). In addition, for reactive pollutants like NO2 (and its precursors), the interactions that occur when plumes merge need to be accounted for in order to properly model the associated chemical transformations.  Moving  sources  ‐ In addition to wake effects, moving sources also present a challenge in how to properly represent their activities. For example, aircraft may be modeled as part of segmented taxiway and runway area sources while GSE may be modeled as area sources around airport terminal gate areas. Since the emissions are uniform within each area, any specificity with regards to the actual location of the emissions within the area cannot be modeled. Moving sources also produce emissions that differ in position over time. This poses a challenge when using steady-state (static) models such as AERMOD to model dispersion. AERMOD is considered state-of-the-art, but it is mainly intended for application to stationary sources (e.g., power plant stacks). The Gaussian plume formulation used within AERMOD is typically used to model concentrations over a 1-hour period where the emissions, plume meandering, meteorology, etc. are all averaged over the hour. Although finer temporal resolutions are possible (e.g., 15- minutes), 1-hour is generally the minimum resolution used for such modeling efforts. As such, sub-hour impacts of airport operations and meteorology are difficult to determine. Models based on a time-varying Gaussian puff method can potentially improve the modeling of moving sources if each moving source is treated as a discrete source with temporally and spatially varying emissions release points. While the literature is plentiful in terms of using Gaussian plume models to assess air quality impacts from stationary (e.g., power plant stacks) and highway sources represented by line sources, there are relatively few studies that have assessed the use of AERMOD as well as Gaussian puff models with mobile sources and even fewer studies involving airport sources – thus not allowing for any substantive conclusion. 3.2.8 Background Concentrations  The “background” concentration of a pollutant is frequently defined as the concentration in the atmosphere in the absence of nearby sources representing the global or regional background levels. In the context of airports and air quality, background is defined as the concentration that would be measured in the absence of an airport contribution (i.e., affected by the global/regional background and contributions from non-airport sources like highways but not aircraft and GSE). Airport Sources  Unlike  stationary  sources,  the  large  number  and  diversity  of  NOx  emission  sources  at  an  airport  and  their spatial distribution over a large  area  provides  significant  modeling  demands.  The  merging  of  multiple  plumes  is  high  among  these  challenges 

    A ‐ 12    According to information reported in the Task 4 Working Paper, 1-hour averaged NO2 concentrations at all monitoring sites within “metropolitan statistical areas” between 2003, 2004 and 2005 were 5, 12, and 38 ppb, respectively. Among the U.S. EPA “SLAMS” monitoring sites, the 5th, 50th, and 95th percentile 1- hour daily maximum NO2 concentrations between 2009 and 2011 were 2, 17, and 45 ppb.  3.2.9 Health Risks  Current scientific evidence links short-term NO2 exposures, ranging from 30 minutes to 24 hours, with adverse respiratory effects, including airway inflammation in healthy people and increased respiratory symptoms in people with asthma. Also, studies show a connection between breathing elevated short-term NO2 concentrations and increased visits to emergency rooms and hospital admissions for respiratory issues, especially asthma. Human health risk assessments, also conducted by the U.S. EPA, concluded that the greatest exposures to short-term elevated NO2 values are adjacent to roadways and intersections with heavy motor vehicle traffic. Measurements have shown that these short-term levels can be more than twice as high as levels measured near residential areas and smaller roadways. In concert with the new one-hour standard, EPA is undertaking a phased deployment of near-road NO2 monitors that began in 2014 and will continue to 2017. Importantly, no additional monitors are proposed near airports as part of this initiative. 3.3 Emission/Dispersion Models & NOx Prediction Methods (Task 3)  Task 3 involved the assessment of tools and methods that are presently available to evaluate how emissions and chemistry of NOx are accounted for in in EDMS/AEDT and AERMOD to determine NO2 concentrations within the airport environment. The following subject areas are discussed in further detail within this section. 3.3.1 EDMS/AEDT and AERMOD The current version of the FAA’s EDMS is Version 5.1.4.1. As discussed above, EDMS is used to compute airport-related emissions (including emissions of NOx) and applies AERMOD to predict the resultant concentrations (such as NO2). Importantly, AEDT2b will fully replace (i.e., sunset) both the FAA’s Integrated Noise Model (INM) and EDMS in May 2015, incorporating dispersion modeling capabilities – again using AERMOD. Simply stated, dispersion modeling is conducted with the EDMS/AEDT based on the results of the emissions inventory and supplemented with meteorological data and the definition of receptor sites. Once the necessary data are input, EDMS/AEDT internally calls for AERMOD to perform the specified dispersion calculations. The resulting concentrations are displayed via maps of receptor sets and/or as tabular output. Developed by the U.S. EPA, AERMOD can simulate point, area, volume, and line sources and has the capability to include simple, intermediate, and complex terrains. It also predicts both short-term (1 to 24 hours) and long-term (quarterly or annual) average concentrations. The model can be executed by using the regulatory default options (e.g., stack-tip downwash, elevated terrain effects, calm wind speeds processing routine, missing data processing routine, buoyancy-induced dispersion, and final plume rise), default wind speed profile categories, default potential temperature gradients, and pollutant decay. Lastly, AERMOD has the capability to account for building downwash effects and to employ gas or particle EDMS/AEDT  The FAA will replace the EDMS with  the  AEDT  in  May  2015  but  the  products of the ACRP 02‐43 research  are  expected  to  be  applicable  to  both models.  

A ‐ 13  deposition or wet/dry depletion of the plume. AERMOD is commonly executed to yield 1-hour and season average concentrations (in µg/m3) at each receptor. 3.3.2 NO2 Prediction Methods  NO2 prediction methods have been implemented in various air quality models with varying degrees of complexity. These range from screening approaches based on NO2/NOx ambient ratio assumptions to O3 limiting approaches to detailed models that simulate NO-to-NO2 chemistry involving O3 and VOCs contributing to the formation of free nitrate radicals. The U.S EPA has provided three methods (i.e., tiers) of increasing complexity to compute NO2 concentrations. As indicated, the default option in AERMOD represents the Tier 1 method while the ARM/ARM-2 represents the Tier 2 method. The detailed methods of OLM and PVMRM represent Tier 3 methods. Algorithms implementing both Tier 3 methods are part of the standard AERMOD code, although neither method is considered a default regulatory method according to the GAQM (i.e., considered non-regulatory methods, meaning they cannot be used for normal regulatory modeling purposes).  Tier 1, Full Conversion - Assumes complete (i.e., 100 percent) conversion of all emitted NOx to NO2 based on the application of an appropriate refined modeling technique (such as AERMOD) to estimate ambient NOx concentrations.  Tier 2, ARM/ARM‐2 ‐ The Ambient Ratio Method is where model-predicted NOx concentrations are multiplied by a NO2/NOx ambient ratio derived from ambient monitoring data. There are two versions of this method:  ARM - The ARM multiplies the AERMOD results by empirically-derived ambient NO2/NOx ratios, with 0.75 as the default ratio for annual impacts and 0.80 as the default ratio for 1-hour impacts. According to U.S. EPA, site-specific ambient NO2/NOx ratios derived from appropriate ambient monitoring data may also be considered as detailed screening methods on a case-by-case basis and with proper justification.  ARM‐2 - This method incorporates a variable ambient ratio that is a function of AERMOD-predicted 1-hour NOx concentration and based on an analysis of hourly ambient NOx monitoring data from approximately 580 stations over a ten-year period. ARM-2 is a post‐process method that uses empirical data to define NO2/NOx ambient ratios based on the measured NOx concentration and distance from the source for each location rather than a single ratio. The ARM is the “default” Tier 2 method and the ARM-2 is a non-default “Beta” option.  Tier 3, OLM/PVMRM  ‐ Under this approach, detailed analyses are conducted of the NO2/NOx ratios taking into account the ambient background O3 concentrations. Two methods are available:  Ozone  Limiting  Method  (OLM) - The OLM involves an initial comparison of the estimated maximum NOx concentration and the ambient O3 concentration to determine which is the limiting factor to NO2 formation. If the O3 concentration is greater than the maximum NOx concentration, total conversion is assumed. If the maximum NOx concentration is greater than the O3 concentration, the formation of NO2 is limited by the ambient O3 concentration.  Plume Volume Molar Ratio Method  (PVMRM)  ‐ The PVMRM takes into account the plume size and models the reactions along the length of the plume. Rather than concentrations, moles of O3 and NOx are used to determine the NO2/NOx ratio at each receptor location. A number of moles are calculated based on the size of the plume at a

    A ‐ 14    receptor location calculated using the plume dispersion parameters. If the number of O3 moles is less than NOx moles, then the moles of NO2 is set equal to O3 moles plus the initial NO2 present in the exhaust (e.g., ten percent). If the number of O3 moles is greater, then the following photostationary reaction equation is used:  NO2/NOx = (K1/K3)O3 / [1+(K1/K3)O3] This equation represents the previously presented NO-to-NO2 and reverse reactions. K1 and K3 are the reaction rates where K3 is dependent on the zenith angle of the sun. Unlike OLM, PVMRM can handle multiple plumes where the dominant plume is enhanced based on distances to other plumes. Notably, neither the OLM nor PVMRM take into account other photochemical reactions (e.g., with volatile organic compounds). These Tier 3 methods require the most detailed level of analysis and produce the least conservative and, presumably, the most representative results. Key model inputs for both the OLM and PVMRM options are the “in-stack” ratios of NO2/NOx emissions and background O3 concentrations. U.S. EPA has provided some guidance for conducting NO2 modeling analysis using the Tier 3 methods.16,17 In general, the recommendations regarding the annual NO2 modeling are also applicable to modeling for the new 1-hour NO2 standard, but additional issues may need to be considered in the context of the latter. For example, certain input data requirements and assumptions for Tier 3 applications may be of greater importance for the 1-hour modeling given the more localized nature of peak hourly vs. annual impacts. 3.3.3 Limitations Associated with AERMOD, ARM, OLM, and PVMRM For the purposes of this ACRP 02-43 research, it is important to note that the ARM, OLM and PVMRM methods described above for computing NO2 using AERMOD are largely based on their applications to single or multiple “stack” stationary sources (i.e., power plants, incinerators, etc.). In contrast, there is comparatively less use of these methods at airports that are characterized by numerous moving sources; of differing sizes, shapes and altitudes; and with varying NO2/NOx emission indices and ratios. Notwithstanding this principal shortcoming, some of the more well-known limitations and gaps associated with these prediction methods include the following:  AERMOD - Plume models do not have the ability to handle recirculation of plumes from sea breeze effects or other diurnal/nocturnal phenomena. Moreover, plume models assume instantaneous mixing within the plume, whereas reactions (e.g., NO + O3) have finite reaction times. In addition, “pockets” of air within the plume can be either NO or O3 rich, resulting in less NO2 formation, and hence, over-predictions at the receptor location. The dispersion coefficients (sigma values) of plume models can also over- or under-predict the dispersion of all emitted pollutants (including NOx). Besides resulting in incorrect predictions of total NOx, this resulting concentration error also affects the NO-NO2 inter-conversion modeling. While AERMOD is generally considered a non-chemistry model, it offers two methods for dynamically modeling NO2 formation: the OLM and the PVMRM. Notably, neither of these                                                              16 EPA, Guidance Concerning the Implementation of the 1-hour NO2 NAAQS for the Prevention of Significant Deterioration Program, June 29, 2010. http://www.epa.gov/ttn/scram/ClarificationMemo_AppendixW_Hourly‐NO2‐NAAQS_FINAL_06‐28‐2010.pdf. 17 EPA, Additional Clarification Regarding Application of Appendix W Modeling Guidance for the 1-Hour NO2 National Ambient Air Quality Standard, March 1, 2011. http://www.epa.gov/ttn/scram/Additional_Clarifications_AppendixW_Hourly‐NO2‐NAAQS_FINAL_03‐01‐2011.pdf. Modeling Limitations  In general, the predictions (outputs)  from  computer  models  and  methods  involving  atmospheric  dispersion  and  chemistry  are  inherently  limited  and  are  only  as  good as the inputs. 

A ‐ 15  methods is accessible through the current version of EDMS and it is not clear when the methods will be available in AEDT.  ARM/ARM‐2 - The main limitations of the ARM/ARM-2 methods are that they are overly simplistic and utilize conservatively high ambient NO2/NOx ratios.  OLM  ‐ Among the limitations with OLM is that it cannot be used to simulate multiple plumes. However, the OLM allows for the specification of multiple sources to be combined (i.e., to allow the formation of a single plume). Also, this method does not take into account the dynamics associated with plume growth and changes in O3 along the length of the plume. However, the OLM is also reportedly likely to produce conservatively high results.  PVMRM ‐ The PVMRM may have a tendency to overestimate the conversion of NO to NO2 for low-level plumes by overstating the amount of O3 available for the conversion due to the manner in which the plume volume is calculated. The plume volume calculation in PVMRM also does not account for the vertical extent of the plume (based on the vertical dispersion coefficient) that may extend below ground for low-level plumes. This overestimation of the volume of the plume could contribute to overestimating conversion to NO2. The PVMRM method has further limitations for area source applications, especially for elongated area sources that may be used to simulate line segments. In these cases, the lateral extent of the plume used in calculating the plume volume depends on the projected width of the area source, even if only a portion of the area source actually impacts a nearby receptor. This again would tend to overestimate the volume of the plume for purposes of determining the amount of O3 available for conversion of NO to NO2, and would likely overestimate ambient NO2 concentrations. With the variety of sources at an airport, including the itemized modeling of taxiway and runway segments, the errors associated with modeling multiple plumes from such varying sources are not well understood. Although some methods are able to take into account multiple plumes, these methods either require uniformity of sources or use an aggregating technique to merge plumes. The use of different source types (i.e., point, area, and volume) also creates difficulties in properly modeling NOx chemistry. As discussed above, incorrect values for the aircraft engine thrust settings actually used at airports can lead to errors in NOx and NO2 concentration predictions. In addition, because airports have a mixture of stationary and mobile sources, not all NOx is emitted into the same chemical environment, which affects the partitioning of NO and NO2 and the rate of NO-NO2 inter-conversion. For example, VOCs are co-emitted with NOx from idling aircraft but not so much from NOx emitted from stationary sources or aircraft during takeoff. 3.3.4 Alternative Models and Methods  For the purposes of this research, four alternative methods for predicting NO2 concentrations apart from those used with AERMOD are also being evaluated. The following provides brief summaries of these models:   CALINE4  - Termed the California Line Source Model, this Gaussian “line source” model was developed by the California Department of Transportation (CalTrans) principally to predict air quality around roadways. The CALINE4 NO2 methodology employs a “discrete parcel method” similar in concept to the PVMRM, but assumes an initial mixing zone around the roadway where initial conditions are set. The same NO + O3 and NO2 photostationary state reactions are modeled taking into account plume size.

    A ‐ 16     CALPUFF  - The California Puff Model is also a Gaussian model with the capability to model point, line, area, and volume sources. Its domain of usage is similar to that of AERMOD, except that it offers time-varying concentrations and other capabilities (e.g., PM speciation predictions). NO2 modeling is based on the use of pseudo first-order chemistry with optional methods involving different species.  CMAQ - The Community Multiscale Air Quality model is the U.S. EPA’s grid-based, regional air quality model. Unlike the Gaussian models, CMAQ uses a mass conservation principle with large scale grids (e.g., smallest grids are usually 4 km x 4 km) to model pollutant concentrations. CMAQ incorporates a full complement of photochemical reactions including reactions with O3 and VOCs. These reactions are grouped into modules including the Carbon Bond 5 (CB5). Although these methods are incorporated in the grid model, they can potentially be implemented into a Gaussian plume environment.  SCICHEM - This is a “plume-in-grid” model, which starts with a chemically reactive plume model characterized by a reference frame moving according to the local wind vector. Emissions are dispersed perpendicular to the wind vector using a scheme that takes into account local turbulence and then allowed to exchange at the plume boundaries between the two frameworks. The objectives are to simulate plume dispersion more accurately than a relatively coarse-gridded model and to provide a means for more accurately simulating rates at which plume materials chemically react than when they are instantly diluted into a relatively large grid-cell volume.

A ‐ 17  4. Comparison of Emission Factors to Measured Data (Task 5) This section discusses the findings of Task 5 of the ACRP 02-43 Amplified Work Plan. 4.1 Task Objective  As discussed in Section 2, the purpose of this task is for the Research Team to address and accomplish the following: Task 5 Objective   Compare the NO/NO2 emissions factors used in EDMS/AEDT to available published measurements of NO and NO2 in aircraft engine exhaust at airports. The relative precision of NO2 concentration predictions by the combined use of EDMS/AEDT with AERMOD and a NO2 conversion method (e.g., PVMRM) depends in large part on the accuracy of the NOx emission factors (or “emission indices”) and, for the OLM and PVMRM methods, the NO2/NOx emission ratio. Similarly, because of their importance in determining total NOx emissions (and therefore resulting NO2 concentrations), engine fuel flow rates and times-in-mode are also assessed and discussed in this section. Finally, emission factors for non-aircraft emission sources common to airports (e.g., GSE, motor vehicles, etc.) are also briefly discussed. 4.2 Aircraft Engine Emission Factors & Measurements   4.2.1 EDMS/AEDT Emission Factors  The FAA’s EDMS/AEDT contains aircraft NOx emission indices (in grams of NOx per kilogram of fuel burned18) based on the International Civil Aviation Organization (ICAO) Aircraft Engine Exhaust Databank.19 These emission indices are based on four engine operating modes: takeoff, climbout, approach, and idle. Presently, the EDMS/AEDT contains 632 aircraft engines including jet turbine (488), piston (8), and turboprop engines (136). Table  4.1 provides a range of the NOx emission indices, by operating mode/thrust setting, for aircraft engines within the EDMS/AEDT. Importantly, EDMS/AEDT currently only provide a total NOx emission index and not separate emission indices for NO or NO2 or the NO2/NOX emission ratios. EDMS/AEDT also does not account for de-rated takeoff thrust20 which is part of the ACRP 02-41 Estimating Takeoff Thrust Settings for Airport Emissions Inventories research and may result in a modification to future applications EDMS/AEDT. 18 NOX emission indices (g of NOX per kg of fuel combusted) are conventionally calculated as if the NOX is 100 percent NO2, even though the true NO/NO2 speciation varies significantly (see later text). 19 Emission indices (grams per kg of fuel combusted) are a function of the thrust setting, aircraft weight, and ambient conditions (e.g., temperature). Total emissions per LTO are a function of the operating time within each mode and the number of engines per aircraft type. 20 For example, in some cases the adoption of a reduced thrust setting for an aircraft during takeoff can lower NOx emissions by up to 30 percent when compared to a full-thrust setting. In general, for takeoff plumes, the measured NOx emission index is lower (approximately 18 percent) than that predicted by engine certification data corrected for ambient conditions.

    A ‐ 18    Table 4.1. EDMS/AEDT NOx Emission Indices (grams per kilogram of fuel)  Engine Type  Operating Mode/ Thrust Settings  Maximum  Minimum  Average  Jet Engines    Takeoff (100%) 65.8 2.09 26.8 Climbout (85%) 46.3 2.30 21.2 Approach (30%) 28.0 1.15 9.40 Idle (7%) 8.53 0.80 3.89 Piston Engines    Takeoff (100%) 5.88 0.36 2.69 Climbout (85%) 5.60 0.24 3.89 Approach (30%) 10.2 0.95 2.00 Idle (7%) 1.90 0.39 1.01 Turboprop  Engines  Takeoff (100%) 20.4 0.01 10.2 Climbout (85%) 17.9 0.01 9.53 Approach (30%) 10.7 0.89 6.99 Idle (7%) 7.47 0.45 3.03 4.2.2 Published Measurements  The most comprehensive set of emission measurements from commercial aircraft are the JETS-APEX2 and APEX3 studies of 2005 at Oakland International Airport and Cleveland Hopkins International Airport, respectively.21,22,23 Emissions from six types of commercial aircraft engines were studied during these tests. NO and NO2 emission indexes and the NO2/NOx emission ratio were calculated at several thrust settings, nominally corresponding to the four phases of the landing takeoff cycle: idle/taxi, approach, climb-out, and takeoff. For safety reasons, several engines were not operated at full (i.e., 100%) thrust, which is the ICAO certification setting for takeoff, and instead tested at a maximum thrust of between 80% and 85%, depending on engine type. In addition to operating at the ICAO certification setting for the idle/taxi phase of 7% maximum thrust, most engines were also operated at “ground idle” which was more representative of the fuel flow rates used during actual idle/taxi operation. The thrust setting was approximately 4% thrust for those engines, with fuel flow rates approximately 20% lower than the ICAO idle/taxi fuel flow rate. It is important to note that ACRP Project 02-45 is researching the issue of actual thrust settings (and fuel flow rates) during the idle/taxi phase of an LTO cycle, and actual thrusts used during takeoff have been investigated as part of ACRP Project 02-41, Estimating Takeoff Thrust Settings for Airport Emissions Inventories. Table  4.2 and Figure  4.1 summarize and illustrate both the overall NOx emission indexes and the NO2/NOx emission ratios derived from the measured data. This data is discussed in greater detailed and portrayed graphically in several figures in Timko et al.24                                                              21 Aircraft Particulate Emissions eXperiment – APEX (2004), JETS-APEX2 (2005), and APEX3 (2005). 22 ACRP Project 02-03a focused on near-idle emissions of HAPs; NO and NO2 emissions were also quantified. Herndon et al., ACRP report 63. 23 Wood et al., Environmental Science and Technology, 42 (6) pp 1884-1891, 2008; Timko et al., Journal of Engineering for Gas Turbines and Power, 2010 132 pp 061504-1 to -14. 24 Timko, M., Herndon, S., Wood, E., Onasch, T.; Northway, M.; Jayne, J.; Canagaratna, M.; Miake-Lye, R; “Gas Turbine Engine Emissions Part 1. Hydrocarbons and Nitrogen Oxides”, Journal of Engineering for Gas Turbines and Power, 132 doi: 10.1115/1.4000131 (2010).

A ‐ 19  Table 4.2. NOx Emission Indexes and NO2/NOx Ratios Derived from Measured Data  With the exception of the CJ610 engine, which is the only turbojet engine (the others are all turbofan engines), the NO2/NOx emission ratio decreases from values between 70 and 95% at the lowest thrust setting studied (e.g., idle) to values between 5 and 10% at the highest thrust settings studied (e.g., climb- out/takeoff). While there is some inter-engine variability evident in the data, the overall trends are consistent across all engines. Furthermore, in almost all cases the total NOx emission indexes agreed with the ICAO certification values within 10%. As noted above, the actual thrust levels used by in-use aircraft do not always agree with the ICAO settings. Most notably, thrust settings/fuel flow rates appear to be significantly lower during actual idle/taxi and takeoff operation compared to the ICAO settings. For example, thrust settings for idle/taxi Engine  LTO Mode  NO2/NOx Emission Ratio  NOx EI  (APEX2/3)  Fuel Flow  Rate  (APEX2/3)  AE_3007  Idle (8%) 0.76 3.5, 3.4 0.05, 0.05 Approach 0.17 8.0, 7.1 0.10, 0.10 Climb-out 0.08 12.9, 11.3 0.30, 0.28 Takeoff 0.07 20.9, 13.1 0.35, 0.33 CFM56‐3B1  Idle (ground idle) 0.89 2.4 0.09 Idle (7%) 0.79 3.1 0.11 Approach 0.18 7.2 0.30 Climb-out 0.07 15.5 0.98 Takeoff CFM56‐7B22  Idle (ground idle) 0.95 3.0, 2.8 0.06, 0.09 Idle (7%) 0.56 4.0, 3.9 0.09, 0.11 Approach 0.10 8.7, 9.1 0.32, 0.32 Climb-out 0.07 17.7, 24.4 0.86, 1.14 Takeoff RB211‐535E4‐B  Idle (ground idle) 0.73 3.9, 3.7 0.13, 0.14 Idle (7%) 0.45 4.5, 5.0 0.21, 0.19 Approach 0.12 7.8, 9.8 0.61, 0.60 Climb-out 0.06 21.3, 22.9 1.81, 1.75 Takeoff 0.05 NA, 1.92 PW4158  Idle (ground idle) Idle (7%) 0.93 3.5 0.19 Approach 0.12 9.6 0.63 Climb-out (80%) 0.06 22.4 1.95 Takeoff CJ6108A  Idle (8%) 0.48 2.2, NA 0.05 Approach 0.71 2.6, 3.0 0.13 Climb-out 0.28 5.7, 4.2 0.29 Takeoff 0.22 5.4, 4.6 0.34

    A ‐ 20    appear to be closer to 4% thrust (approximately 20% lower fuel flow rate), and thrust settings for takeoff are closer to 85% than to 100% thrust as a result of “reduced thrust takeoff”. Notes: - Only 15, 30, and 43 m probe data used for NOx EIs (not 1 m data) - Dual entries separated by commas denote results from multiple engines - For the AE3007, the 1st entry is for the AE3007-A1E, the 2nd for the AE-3007A   Figure 4.1. Dependence of the NO2/NOx emission ratio (“ER”) as a function of thrust for the engines  studied during the JETS‐APEX2 and APEX3 studies. (Lines are added simply to guide the eye.)    Several other published datasets corroborate the JETS-APEX2 and APEX2 measurements regarding the total NOx EIs and actual thrusts used, including measurements conducted at Frankfurt airport25, Atlanta Hartsfield Jackson International Airport26, Brisbane Airport in Australia 27, and Roanoke Regional Airport in Virginia28. 4.2.3 Comparison of Emission Factors to Measured Data  For ease of comparison, Table 4.3 provides a summary discussion of the aircraft engine emission factors, emission ratios, fuel flow rates and engine times-in-mode currently contained in EDMS/AEDT in comparisons to published measurements.                                                              25 Schaefer, K. et al. (2003). Aircraft emission measurements by remote sensing methodologies at airports, Atmos. Environ. 37: 5261 - 5271. 26 Herndon, S. C. et al., (2008), “Commercial Aircraft Engine Emissions Characterization of in-use Aircraft at Hartsfield- Jackson Atlanta International Airport”. Environ. Sci. Technol. 42, 1877-1883. 27 Mazaheri et al., (2009), “Particle and Gaseous Emissions from Commercial Aircraft at Each Stage of the Landing and Takeoff Cycle”, Environ. Sci. Technol. 43, 441 – 446. 28 Klapmeyer and Marr (2012), “CO2, NOx, and Particle Emissions from Aircraft and Support Activities at a Regional Airport” Environ. Sci. Technol. 46, 10974 – 10981. 1.0 0.8 0.6 0.4 0.2 0.0 N O 2 /N O x E R 100806040200 % max thrust AE3007 CFM56_3 CFM56_7 RB211 PW4158 CJ610

A ‐ 21  Table 4.3. Comparison of Emission Factors to Measured Data  Parameter  Currently Used in EDMS/AEDT  Published Measurements  Aircraft NOx Emission  Indexes  ICAO values At the same thrust, measurements agree to within ~10% Aircraft NO2/NOx Emission  Ratio  For the Full Conversion and ARM methods, none is needed. For OLM and PVMRM, there is no default value. EPA guidance states that emission ratio used should be justified based on the specific application. In the absence of appropriate source-specific information, a default value of 0.5 may be used. Ranges from 70 to 95% at idle/taxi operation down to 5 to 10% at climb-out and takeoff thrust (see discussion below). Fuel Flow Rates   (and thrust settings)  Values in ICAO databank for thrust values of 7%, 30%, 85%, and 100% corresponding to idle/taxi, approach, climb-out, and takeoff.  Takeoff thrust is often significantly reduced from 100% (“reduced thrust takeoff”). ACRP Project 02-41 is investigating this issue.  Idle/taxi fuel flow rates are often lower than Time‐in‐Mode  Default values (26 min idle/taxi, 0.7 min takeoff, 2.2 min climb-out, 4 min approach) or Airport-specific input values by user or Values generated by Delay/Sequence module in EDMS. Times can vary significantly. Idle time-in-mode investigated as part of ACRP Project 02-45. A few of the most important points regarding actual aircraft NO and NO2 emissions and how they are handled in EDMS/AEDT/AERMOD are listed below:  Total NOx emission indices from aircraft engines agree reasonably well with the ICAO certification values, though the actual thrust values used can greatly affect NOx emissions (e.g., reduced thrust takeoff).  Aircraft engines emit NOx mainly as NO2 at low power (idle/taxi) and NO at high power (takeoff). The combined NO2/NOx emission ratio for an LTO cycle is very sensitive to the time spent idling, but is approximately in the 15 to 35 percent range.29 This is higher than the historical default value of 0.1 used by AERMOD when using the PVMRM method to model NO2 concentrations, and higher than the value of 0.05 used to model the impact of Paris airports on NO2 and ozone, but lower than the EPA “default” value of 0.5.  The NO2/NOx emission ratio is altitude-dependent. NO2 is emitted mainly at the surface, whereas NO emissions occur over a wider range of altitudes (surface and aloft). The extent to which the vertical variation in the NO2/NOx emission ratio affects downwind NO2 concentrations is not well understood. 29 Wood et al., (2008), Environmental Science and Technology, 42 (6) pp 1884-1891, and Stettler et al. (2011), “Air quality and public health impacts of UK airports. Part I: Emissions”, 42, 5415 – 5424.

    A ‐ 22    4.3 Non‐Aircraft Emission Factors  Although aircraft are the biggest source of NOx emissions at airports, other sources of NOx emissions are not insignificant. These include ground access vehicles (GAV), auxiliary power units (APU), ground service equipment (GSE), and stationary sources. Depending on the relative location of these sources and downwind receptor sites, these non-aircraft sources could have a disproportionately large impact on nearby NO2 concentrations. Some preliminary analysis reveals the following:  APU emissions were studied during the NASA-AAFEX and ACRP Project 02-17 and show NO2/NOx emission ratios in the range 0.25 to 0.55. 30  Diesel NOx emissions are changing rapidly in some areas (e.g., California) due to the implementation of diesel particulate filters and oxidation catalysts. These devices reduce PM emissions but increase the NO2/NOx emission ratio, to values exceeding 50 percent in some cases.31  Light-duty ground access vehicles (mostly gasoline-fueled) have low NO2/NOx emission ratios (less than 5 percent), but may increase if diesel-fueled engines becomes more widespread.  The relative importance of these sources on NO2 and NOx concentrations depends not only on the NOx emission rates but also their respective NO2/NOx emission ratios.  MOVES and NONROAD  ‐ Airport GAV emission factors are mostly derived from the U.S. EPA MOVES emission factor model. Conversion factors from the MOVES model applied to exhaust NOx emissions to derive NO/NO2 shows how these conversion rates vary by model year range and fuel type. Local MOVES parameters (i.e., fleet mix, fuel formulation, meteorology, I/M programs) also dictate how the model modifies its base emissions information to generate NOx emissions and emissions rates, based on local conditions. Airport GSE, another potentially significant source of NOx in airport environs, also have emission factors listed in the EPA NONROAD model.  NO2/NOx emission ratios from on-road vehicles can be summarized as less than 0.05 for gasoline- fueled vehicles, 0.05 to 0.2 for diesel vehicles not equipped with particulate filters or oxidation catalysts, and 0.15 to 0.8 for diesel-fueled vehicles equipped with particulate filters/oxidation catalysts.  APUs ‐ The NO2/NOx emission ratio from a single APU was studied as part of the NASA-AAFEX campaign (2009) and summarized in Kinsey et al.32 Emissions from a Garrett- AiResearch/Honeywell Model GTCP85–98CK, which is a relatively old APU and not representative of the current commercial fleet, were quantified when separately burning JP-8 fuel and a coal-derived Fischer Tropsch fuel. The NO2/NOx ratio decreased from 0.55 at an exhaust temperature of 350° C to 0.25 at an exhaust temperature of 620° C.    GSE ‐ Ground Support Equipment run on a variety of fuels, including gasoline, diesel, natural gas, and electricity. Overall pollutant emissions (CO, PM, etc.) from gasoline and diesel-fueled vehicles may differ from similar engines used by on-road vehicles due to differences in emission                                                              30 "Determination of the Emissions from an Aircraft Auxiliary Power Unit (APU) during the Alternative Aviation Fuels EXperiment (AAFEX)." J. Air & Waste Manage. Assoc. 62(4): 420 - 430. 31 Dallman et al (2012). "On-Road Measurement of Gas and Particle Phase Pollutant Emission Factors for Individual Heavy- Duty Diesel Trucks." Environ Sci Technol 46: 8511-8518. 32 Kinsey, J. S., M. T. Timko, S. C. Herndon, E. C. Wood, Z. Yu, R. C. Miake-Lye, P. Lobo, P. Whitefield, D. Hagen, C. Wey, B. E. Anderson, A. J. Beyersdorf, C. H. Hudgins, K. L. Thornhill, E. L. Winstead, R. Howard, D. I. Bulzan, K. B. Tacina, and W. B. Knighton. 2012. Determination of the Emissions from an Aircraft Auxiliary Power Unit (APU) during the Alternative Aviation Fuels Experiment.

A ‐ 23  regulations. Nevertheless, it is likely that NO2/NOx emission ratios exhibit the same overall trend as those from on-road vehicles (i.e., NO2/NOx < 0.05 for gasoline-fueled GSE and 0.05 to 0.9 for diesel-fueled GSE depending on the presence of post-combustion control technology).  Stationary Sources - Stationary sources of NOx at airports include (but are not limited to) power plants, boilers, generators and can use a range of fuels (diesel, fuel oil, natural gas, etc.). “In- stack” NO2/NOx emission ratios vary greatly depending on fuel type, emission source, and operational state (e.g., BTU/hr). Values range from 1-10% for natural gas fired boilers and turbines (CAPCOA), 5% for fuel oil-fired boilers (EPA document AP-42), and above 0.5 for natural gas fired reciprocating IC Engines. Post-combustion pollutant control technologies can increase the NO2/ NOx in-stack ratio – e.g., catalytic oxidizers. (EPA’s NO2/NOx In-Stack Ratio (ISR) Database http://www.epa.gov/scram001/no2_isr_database.htm). As noted in Table 4.3, the U.S. EPA recommends source-specific information be used when deciding on NO2/NOx emission ratios to be used as inputs into AERMOD when using OLM and PVMRM, and that in the absence of such information a default value of 0.5 may be used. Given that most (but certainly not all) non-aircraft NOx sources at airports have NO2/NOx emission ratios of less than 0.25, use of the default value of 0.5 in AERMOD will lead to overestimates of NO2 concentrations when using OLM or PVMRM in conjunction w/ AERMOD.  

   5.  Comp This secti Amplified 5.1. Task As discus following  C ac This obje comprised    These air based upo medium-t activity le ambient N modeling located in airports e Research necessary The follo and mode these airp 5.2 Air M For this a concurren collected note that for this support o These dat 5.2.1 A ADL is lo classified Mediterra mountains arison of M on discusses Work Plan.  Objective  sed in Sectio : ompare mod tivity levels, ctive was ach the followin Los Angel Adelaide ( Montreal ( Internation ports were n two “key o-large comm vels deemed O2/NOx con and (ii.) th their close xpressed w Team in ob data. wing subsec ling data obta orts in suppor onitoring D nalysis, the t NOx, NO2 at the test-ca these data w research pro f other effor a, and the con delaide Inter cated in sou as a “medium nean-like wit to the east o   odeled an the method n 2, the purpo eled and me geography, a ieved by ass g: es Internation Australia) Ai Canada) Pier al Airport (Y selected for ” criteria: (i ercial servi sufficient en centrations ey had air proximities. illingness to taining and/o tions describ ined from, a t of this rese ata  air monitori , O3 and me se airports. ere not coll ject but we ts and used ditions unde national Airp thern Austra -sized” com h moderate r f the airport i d Measur ologies and se of this tas Task asured NO2 nd meteorolo essing pertin al Airport (L rport (ADL), re Elliot Tru UL). this assessm .) they repre ce facilities ough to esti and ratios u quality mon In addition, t work with r developing e the monito nd developed arch. ng data incl teorological It is importa ected specifi re develope for this anal r which they ort (ADL)  lia, is the co mercial airpo ainfall and m nfluence loca A ‐ 24  ed NO2 Dat presents the k is for the R  6 Objective data at a sam gical conditi ent data from AX) and deau ent sent with mate sing itors hese the the ring for, uded data nt to cally d in ysis. were collecte untry’s fifth rt. Located o ild winters. L l air quality a (Task 6) outcomes o esearch Team   ple of airpo ons. three “case d, are describ busiest (104 ceanside, the ocal sea bre as air mass re Figure 5.1.  f Task 6 of to address a rts represent -study” airpo ed below, by ,000 annual climate is g eze effects in circulation e ADL Air Mon the ACRP nd accompli ing various f rts. These ai airport. operations) a enerally warm combination ffects do occu itoring Stati 02-43 sh the leets, rports nd is and with r. on  

The NO2/ located ad 5.2.2 Lo LAX is lo California (667,000 LAX is c the winds the nearby The NO2 analysis w monitorin Quality a and (ii.) (SPAS).34 The LAX monitorin airport - a  Up M sit the  Co loc  Co cen  Co cen Notably, monitor a 5.2.3 Mo YUL is lo the fourth climate an The NO2/ permanen 33 The N Apport accesse 34 The NO (SPAS) SPAS/A NOx monitor jacent to, and s Angeles In cated on the and is the f annual opera lassified as are predomi land-sea bre /NOx monit ere derived g studies con nd Source Ap the LAX Sp AQSAS da g stations lo s shown in Fi wind  North anagement uated 1.7 mi airport; mmunity  Ea ated approxi mmunity No ter of the air mmunity  So ter of the air the LAX SP s modeling d ntreal‐Pierr cated adjace busiest airpo d the winds a NOx monitor t air monitori O2 monitoring ionment Study ( d at http://www 2 monitoring d by LAWA bout_the_Stud ing data for A 1.2 miles ea ternational A Pacific coa ifth busiest a tions). The c Subtropical-M nately from w eze affect. oring datase from two s ducted at LA portionment ecific Plan taset was co cated in clos gure 5.2 and west (i.e., District - les northwes st (i.e., the mately 3.2 m rth (i.e., the port; and uth (i.e., the port. AS dataset w ata were only e Elliot Trude nt to the Sai rt in Canada re generally ing data used ng station loc data was deriv AQSAS) by Los .lawa.org/airQu ata was derived . Final R y_What_is_the. DL is colle st from the c irport (LAX) st in Los An irport in the limate surrou editerranea est-to-east d ts used for eparate air q X: (i.) the LA Study (AQS Amendment llected at fo e proximity described be the Air Q AQMD) m t of the cen Lennox) m iles east of th Westchester) El Segundo as singularly available for au Internati nt Lawrence (231,000 an from the east for this analy ated in the c ed from the L Angeles World alityStudy.aspx? from the Los A eport and aspx. A ‐ 25  cted by the A enter of, the a geles, world nding n and ue to this uality X Air AS)33 Study ur air to the low: uality onitor ter of onitor e center of th monitor loc ) monitor lo collected a this site. onal Airport  River in Mo nual operatio and northeas sis were coll enter of the a os Angeles In Airports (LAW id=7716.. ngeles Interna Materials can DL staff fro irport, as sho e airport; ated approxim cated approx t the Upwin (YUL)  ntreal in sou ns). The are t. ected and pr irport, as sho ternational Airp A), June 18, 20 tional Airport ( be acces Figure 5. m a permane wn in Figure ately 2 mil imately 1.2 m d Northwest thwest Quebe a has a semi- ovided by the wn in Figure  ort (LAX) Air 13. Final Repor LAX) Specific P sed at http 2. LAX Air Mo nt monitorin  5.1. es northeast o iles south o (i.e., the AQ c Province a humid contin YUL staff f 5.3. Quality and t and Materials lan Amendmen ://www.lawa.or nitoring Sta g site f the f the MD) nd is ental rom a Source can be t Study g/LAX tions 

   5.2.4 Tim The air m airports assessmen periods an is beca independe purposes. YUL wer air monit from LAX special st recently-a collected periods monitored Table 5.1 A Adelaide ( Los Inte ( Montreal ( 5.2.5 Qu This secti the case-s e Periods   onitoring da identified ab t were coll d with differ use these ntly from ea For example e collected a oring program which was udies. Howev vailable data for use in that were and mode . irport  Internationa ADL) Angeles rnational LAX) Internationa YUL) ality Assuran on describes tudy airports   ta obtained f ove and u ected over ing time reso data wer ch other and , the data fr s part of the s in contra collected as er, in all ca from each this analys examined f led data are Ta Purp l Perma stati Air Qua Sou Apportio Study (A Specifi Amend Study ( l Perma stati ce: Monitor quality assura . rom the thre sed for thi varying tim lutions. Thi e collected for differen om ADL and ir permanen st to the dat part of som ses, the mos program wa is. The tim or both th reported in ble 5.1. Air M ose  nent on Ju lity & rce nment QSAS) Fe J c Plan ment SPAS) J nent on Ja ed Data  nce methods A ‐ 26  Fi e s e s t t a e t s e e onitoring Ti Timef ly 1, 2011 - bruary 1 - M Season 1 uly 18 - Aug Season 2 ( anuary 1 - D 200 nuary 1, 200 31, 2 applied to th gure 5.3. YU me Periods rame  May 31, 201 arch 16, 201 (winter) ust 28, 2012 summer) ecember 31, 9 9 - Decembe 011 e air monitor L Air Monitor Time  Period 2 1 Year 2 6 weeks 6 weeks 1 Year r 3 Years ing data obta ing Station   Resoluti Hourly Minut Minut Hourly Hourly ined from ea on  e e ch of

A ‐ 27   Data Censoring ‐ A data censoring process was applied to the monitored NO2, NO, and NOx datasets. Essentially, this involved changing any negative values in the raw datasets to zero values. Blank, or missing, values were left as blank values.  Data Evaluation  ‐ The air monitored data were also assessed for normality and outliers to help ascertain how statistical results should be interpreted. The outcomes of this analysis are discussed below.  Normal Distribution Assessment ‐ These test parameters include the Pearson, Cramer- Von Mises, and Lilliefors (Kolmogorov-Smirnov)35 tests for normality. As expected, these results reveal that a normal distribution does not exist within any of the datasets. A normal Q-Q plot (i.e., quantile-quantile plot) is a probability plot that also aids in determining if a dataset’s distribution is normal. Quantiles of a given dataset are plotted against the quantiles of a normal distribution. If the given distribution follows the 1:1 (x = y) line (plotted for reference), then the given dataset follows a normal distribution. As expected, these plots support the findings of the Pearson, Cramer-Von Mises and Lilliefors (Kolmogorov-Smirnov) tests that a normal distribution is not present in any of the airport datasets.  Determination of Outliers  ‐ The Chi-squared and Grubbs tests were used to determine the existence of data outliers. These tests identified that outliers are present in the airport air monitoring datasets. However, it should be noted that Chi-squared and Grubbs tests for outliers assume a normal distribution, which, as discussed above, is not present. The distribution of a dataset highly effects whether an observed value is considered an outlier or not. Thus, providing multiple statistical tests and visual representations of the dataset was necessary in order to provide a complete examination of possible outliers. It is important to note that outliers were only identified, not removed. All outliers were included in the statistical analysis. 5.3   Air Modeling Data  The modeling of NO2/NOx emissions and concentrations for the three test-case airports was performed using the latest version of the Federal Aviation Administration’s (FAA) Emissions and Dispersion Modeling System (EDMS).36 The EDMS also contains the U.S. Environmental Protection Agency’s (EPA) AERMOD37 model. These models are described below. 5.3.1 EDMS/AERMOD Models  EDMS is designed to assess the air quality impacts of airport-related emission sources including aircraft, auxiliary power units (APUs), ground support equipment (GSE), ground access vehicles (GAV), fuel facilities and stationary sources. EDMS has both emissions inventory and dispersion modeling capabilities. Notably, EDMS will be replaced by the FAA’s Aviation Environmental Design Tool (AEDT) in 2015. For  dispersion  modeling  purposes,  EDMS  employs  EPA’s  AERMOD  atmospheric  dispersion  model which is listed as a Guideline model by the EPA in Appendix W of 40 CFR Part 51.38 AERMOD can 35 Geyer, Charles. Statistics 5601 Kolmogorov-Smirnov and Lilliefors Tests. University of Minnesota, School of Statistics. October 2, 2013. http://www.stat.umn.edu/geyer/5601/examp/kolmogorov.html. 36 FAA, Emissions and Dispersion Modeling System (EDMS), http://www.faa.gov/about/office_org/headquarters_offices/apl/research/models/edms_model/. 37 EPA, Preferred/Recommended Models, AERMOD Modeling System, http://www.epa.gov/ttn/scram/dispersion_prefrec.htm#aermod. 38 EPA, Appendix W of 40 CFR Part 51. July 1, 2011 http://www.gpo.gov/fdsys/pkg/CFR-2011-title40-vol2/pdf/CFR-2011- title40-vol2-part51-appW.pdf

    A ‐ 28    simulate point, area, volume, and line sources and has the capability to include simple, intermediate, and complex terrains.39,40 It also predicts both short-term (1 to 24 hours) and long-term (quarterly or annual) average concentrations. The model can be executed by using the regulatory default options (e.g., stack-tip downwash, elevated terrain effects, calm wind speeds processing routine, missing data processing routine, buoyancy-induced dispersion, and final plume rise), default wind speed profile categories, default potential temperature gradients, and pollutant decay. AERMOD also has the capability to account for building downwash effects and to employ gas or particle deposition or wet/dry depletion of the plume. Notably, EDMS does not have the capability to execute this downwash feature, thus it was not accounted for in the modeling. AERMOD is commonly executed to yield 1-hour and season average concentrations (in micrograms per cubic meter [µg/m3]) at each analyzed receptor. These concentrations may be presented as plot files and receptor files showing the results at each receptor for tabular and graphical display. Existing methods within AERMOD to convert NOx to NO2 include the Ambient Ratio Method (ARM/ARM2),41 the Ozone Limiting Method (OLM), and the Plume Volume Molar Ratio Method (PVMRM). All three of these methods, in additional to full conversion, have been referenced in EPA’s Appendix W of 40 CFR Part 51. The three different methods (called Tiers) are described further in Section 5.3.4. A simple schematic of the inputs and outputs of using EDMS and AERMOD to predict NO2 concentrations is shown below as Figure 5.4.                                                              39 EPA Preferred/Recommended Models, AERMOD Modeling System, http://www.epa.gov/ttn/scram/dispersion_prefrec.htm#aermod. 40 Title 40 CFR Part 51, Revision to the Guideline on Air Quality Models: Adoption of a Preferred General Purpose (Flat and Complex Terrain) Dispersion Model and Other Revisions; Final Rule, http://www.epa.gov/ttn/scram/guidance/guide/appw_05.pdf. 41 In January 2014, an updated version of ARM, referred to as “ARM-2” was incorporated into the latest AERMOD (Version 13350). Presentation of Data  It  should  be  noted  that  the  data  presented  and  discussed  in  this  Interim Report is only a small subset  of the entire database developed in  support  of  the  ACRP  02‐43  Research.  

5.3.2 So Using ED based upo and time p For disper Point sour sources w activity); out and a sources. These mo produced U.S. EPA OLM and EPA’s rec for the Tie Receptors meters ab 5.3.3 Me AERMOD data, uppe 42 RTP E Septem 43 EPA, U http://w urce Input Pa MS, emissio n source-spe eriod analyz sion modelin ces were use ere used to aircraft taxiin pproach mo del input da for this resea recommend PVMRM) a ommended m r 2 (i.e., ARM were placed ove ground le teorological contains a r air soundin nvironmental As ber 20, 2013. ser’s Guide for ww.epa.gov/ttn Figu rameters  ns from airp cific activity ed. g, airport-rel d to represen represent air g, queuing a des. Airport ta were deri rch project fo ed default eq nd the defau inimum NO 2) method.4 at the air mo vel.  Data Input  meteorologic gs, and data sociates. Ambie the AERMOD M /scram/metobsd re 5.4. Schem ort-related so data, emissi ated sources t stacks from craft gate ap nd accelerati roadways an ved from pr r YUL. uilibrium an lt equilibriu 2/NOx ratio 2 nitoring stat al data prepr from on-site nt Ratio Method eteorological P ata_procaccprog A ‐ 29  atic of EDM urces (i.e., on rates and modeling we stationary s rons (i.e., ai ng on the tax d parking f eviously-prep d in-stack ra m ratio was value of 0.2 ion(s) describ ocessor, AER instrument to Version 2 (ARM reprocessor (A s.htm#aermet. S/AERMOD. aircraft, GSE temporal val re simulated ources such rcraft at star iway/runway acilities were ared EDMS tios were use 0.9 and the and maximu ed above fo MET43 that wers such as 2) for use with ERMET), Nove   , APU’s, etc ues for each as either are as boilers an tup, GSE op system; and modeled a files for LA d for the Ti default in-st m ratio value r each airport accepts surfa a sonic Dete AERMOD for 1 mber 2004, .) were com case-study a a or point sou d generators. erations and aircraft in c s a series of X and ADL er 3 methods ack ratio wa of 0.9 were at a height o ce meteorolo ction and Ra -hour NO2 Mod puted irport rces. Area APU limb- area and (i.e., s 0.5. used f 1.8 gical nging eling.

    A ‐ 30    (SODAR) system. Atmospheric turbulence characteristics, mixing heights, friction velocity, Monin- Obukov length and surface heat flux are calculated in this manner. Airport-specific meteorological data was obtained from a variety of sources, based on availability. These include the following:  ADL - these data were obtained from a previous air quality assessment conducted at ADL.  LAX AQSAS - these data were based on the analysis completed for the LAX Air Quality and Source Apportionment Study (AQSAS). This study included SODAR, surface, and upper air data for both seasons analyzed for this research project.  LAX  SPAS - these data were obtained from the Specific Plan Amendment Study (SPAS) Environmental Impact Report.  YUL - these data were obtained from Lakes Environmental44 and then developed for the AERMOD input files. 5.3.4 Conversion Methods  As discussed previously, AERMOD allows a three-tiered approach in deriving NOx to NO2 conversions for application in atmospheric dispersion modeling. Tiers 1 and 2 represent the default regulatory methods used under most cases. Under Tier 3, the OLM and PVMRM are non-regulatory methods that can be used on a case-by-case basis with EPA’s approval. The Tier 1 approach is the most conservative and assumes complete conversion of NOx emissions within an air plume to NO2. The Tier 2 approach pertains to the ARM and ARM2 methods. The EPA recommends that a default ratio of 0.80 be used for ARM without additional justification. A variable ambient ratio based on the NOx concentration is used for ARM2. Tier 3 options rely upon the OLM and PVMRM. Unfortunately, none of these methods have been evaluated or verified (i.e., calibrated) when applied to airports which have vastly different emission characteristics than stationary and other mobile sources when it comes to NO2. Specifically, the NO2/NOx emission ratio for aircraft differs markedly from most other NOx sources. The modeling analysis for the “test-case” airports included all three conversion methods. Additionally, the OLM and PVMRM methods were completed for two scenarios: (i.) using default values for the emission ratios and (ii.) with a variation of the emissions ratios for aircraft, depending on operating mode. These methods are labeled as “OLM with Variable” and “PVMRM with Variable,” respectively. All of the NO2/NOx conversion methods are further discussed below.  Full Conversion – This Tier 1 conversion method assumes full conversion of NOX to NO2.  ARM/ARM2 – Although AERMOD does not explicitly use the ARM, the default (i.e., regulatory) option in AERMOD can be used to predict NOx concentrations with measured representative NO2/NOx ratios. The ARM represents the Tier 2 method as specified in Appendix W of 40 CFR Part 51. Previous studies suggest that 1-hr NO2 concentrations in the near field are overestimated using a fixed ARM ratio of 0.8, coupled with a plume model’s unbiased estimate of NOx concentrations. As a result, a subsequent and more refined version of ARM has been made available, known as “ARM2”, which incorporates a variable ambient NO2/NOX ratio.45 The ARM2 conversion method was developed using data from 580 ambient monitoring sites over a                                                              44 Lakes Environmental at http://www.weblakes.com/. 45 Podrez, Mark. “Ambient Ratio Method Version 2 (ARM2) for use with AERMOD 1-hour NO2 Modeling”.

A ‐ 31  span of 10 years. The relationship is empirically derived between the upper limits of the observed NO2/NOX ambient ratio vs. the ambient NOx concentration.46  OLM – As previously mentioned in the Task 4 Working Paper,47 this method compares NOx concentrations to ambient O3 concentrations. If the O3 concentration is greater than the NOx concentration, complete conversion to NO2 is assumed. In the alternative case (i.e., NOx concentrations are greater); the NO2 concentration is set equal to the O3 concentration plus the initial NO2. The most well-known limitation with OLM is that it cannot be used with multiple plumes. However, sources can be combined to create a single plume. Also, the simplistic modeling does not take into account the dynamics associated with plume growth and changes in O3 along the length of the plume. But the simplicity provides an easy-to-understand framework that tends to produce conservative results.  PVMRM – Unlike the OLM, the PVMRM takes into account the plume size and models the reactions along the length of the plume. Rather than concentrations, moles of O3 and NOx are used to determine the NO2/NOx ratio at each receptor location. Number of moles are calculated based on the size of the plume (using a “slice” of the plume) at a receptor location calculated using the plume dispersion parameters (i.e., σy and σz). If the number of O3 moles is less than NOx moles, then the moles of NO2 is set equal to O3 moles plus the initial NO2 present in the exhaust (e.g., 10 percent). If the number of O3 moles is greater, then the following photostationary reaction equation is used:  NO2/NOx = (K1/K3) O3 / [1 + (K1/K3) O3]1  This equation represents the previously presented NO-to-NO2 and reverse reactions. K1 and K3 are the reaction rates where K3 is dependent on the zenith angle of the sun. Unlike OLM, PVMRM can handle multiple plumes where the dominant plume is enlarged (“enhanced”) based on distances to other plumes. It should be noted that neither the OLM nor PVMRM take into account other photochemical reactions (e.g., with volatile organic compounds). Finally, it should be noted that the following restrictions and options apply to OLM and/or PVMRM methods:48  Only one method can be used at a time.  Since both methods are non-regulatory, the default regulatory option cannot be used with either of these methods.  Both methods require a background O3 concentration (average or hourly).  Both methods require the NO2/NOx ambient ratio (if none is specified, the default value of 0.9 is used).  Both methods require the exhaust NO2/NOx ratio (if none is specified, the default value of 0.1 is often used).  The OLM allows for the specification of multiple sources to be combined (i.e., to allow the formation of a single plume). 46 RTP Environmental Associates, Inc. Ambient Ratio Method Version 2 (ARM2) for use with AERMOD for 1-hr NO2 Modeling: Development and Evaluation Report. Published September 20, 2013. Retrieved December 8, 2014. http://www.epa.gov/scram001/models/aermod/ARM2_Development_and_Evaluation_Report-September_20_2013.pdf. 47 KB Environmental, Inc. ACRP 02-43: Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations. Task 4 Working Paper: Review of Regulations/Ambient standards, Current Understanding/Research Gaps, and NOx Speciation Methods in EDMS/AEDT/AERMOD. February 28, 2014. 48 KB Environmental, Inc. ACRP 02-43: Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations. Task 4 Working Paper: Review of Regulations/Ambient standards, Current Understanding/Research Gaps, and NOx Speciation Methods in EDMS/AEDT/AERMOD. February 28, 2014.

    A ‐ 32     The PVMRM allows a “Beta” option to implement Prevention of Significant Deterioration (PSD) increments for any source(s). 5.4   Monitored Data: Description and Statistics   In this section, descriptive statistics such as measures of central tendency and scatterplots are provided for monitored concentrations to demonstrate that valid monitored concentrations were ingested into AERMOD. The results of these descriptive statistics are provided and described below and support that these concentrations are typical concentrations found at airports. 5.4.1 Measures of Central Tendency  Measures of central tendency such as mean, standard deviation, median, minimum, and maximum concentrations were examined on monitored concentrations to provide an overview of the monitored NO2, NOX, O3 and NO2/NOx Ratio concentrations. These measurements are listed in tables and visualized with boxplots and histograms provided below. 5.4.2 Mean, Standard Deviation, Median, Minimum, and Maximum  Mean, standard deviation, median, minimum and maximum monitored values for each dataset for NO2, NOX, O3 and NO2/NOX Ratios are useful in generally describing a dataset. An example from the YUL dataset is shown below for the YUL dataset as shown in Table 5.2. Table 5.2.  Mean, Standard Deviation, Median, Minimum and Maximum Values: YUL: 2009‐2011   Parameter      Values  Mean (ppb)  SD (ppb)  Median (ppb)  Minimum (ppb)  Maximum (ppb)  NO2  10.86 9.21 7.96 0.01 71.47 NOX  15.15 20.04 9.23 0.01 336.18 Ozone  23.32 13.15 23.11 0.01 76.56 NO2:NOX Ratio  0.83 0.16 0.87 0.05 1.00 5.4.3 Quartiles/Interquartile Ranges  Reporting quartiles and IQRs describes the statistical dispersion of a given dataset and the ranges are used for creating boxplots. To examine a dataset’s quartiles, a dataset is broken up into four quartiles based on the spread of the data. The IQR represents the middle 50 percent of the data, where 25 percent of the data falls on either side of the median value. An assessment of quartiles and IQRs of NO2, NOX, O3 and the NO2/NOX Ratio values were computed and an example of these data is shown below for the YUL dataset as Table 5.

5.4.4 B Box and depictions block of t notated w (Q3). The Outliers a for YUL a Pa NO2 ox and Whisk whisker plot of a dataset he boxplot, w ithin this box outer reache re plotted ou re provided b T rameter  NO2  NOX  Ozone  :NOX Ratio  er Plots  s are useful ’s quartiles. A ith the main . The “whisk s of these “w tside of the “ elow for the Figu able 5.3. Inte Q1  0 0 0 0.05 in visualizin s discussed portion of ers” on eith hiskers” rep whiskers” on YUL dataset re 5.5. YUL N A ‐ 33  rquartile Ra Interqua Q2  4.27 4.86 13.39 0.74 g a given d previously, th the box repre er ends of th resent the m either end o in Figure 5.5 O2, NOx and  nges: YUL rtile Values Q3  7.96 9.23 23.11 0.87 ataset’s rang e IQR (the m senting the e IQR box re aximum (and f the IQR. E and 5.6. O3 Boxplots. Q4  14.57 17.83 32.40 0.96 e and outlie iddle 50 per IQR. The me present Quar minimum) n xample box   rs and are v cent) is a bu dian value i tiles 1 (Q1) on-outlier v and whisker isual ilding s also and 3 alues. plots

   5.4.5 H Histogram range of v each test- through 5 istograms   plots are u alues) within case airport w .10, respectiv   seful in dete a given dat ere prepared ely.  Figure 5.6. YU rmining a di aset. Histogra and exampl Figure 5.7. H A ‐ 34  L NO2/NOx  stribution of ms of NO2, es for the YU istogram of Y Boxplot.  frequency o NOX, O3 and L dataset ar UL NO2.  f occurrence the NO2/NO e provided b of each valu X Ratio valu elow in Figur e (or es for e 5.7 

Figure 5.8. H Figure 5.9. H A ‐ 35  istogram of Y istogram of  UL NOx.  YUL O3. 

   5.4.6 Scat Comparis inter-varia  A  A  A  In  N  N An examp ter Plots: Int ons of monito ble influence s NOX increa s O3 increase s O3 increase most cases,  In som be rela o clear relatio o clear relatio le scatterplot   Figure er‐Variable I red NO2, NO s between th ses  NO2 in s  NO2 dec s NOX dec the NO2/NOX e cases at LA ted to roundi nship betwe nship betwe matrix for th  5.10. Histog nfluence  X, and O3 da ese paramete creases reases reases Ratio decrea X AQSAS, ng/quality as en NO2 and N en O3 and NO e YUL datas A ‐ 36  ram of YUL N ta at the test- rs. In general ses as NOx in NO2/NOX Ra surance issue O2/NOX Rati 2/NOX Ratio et is provided O2/NOx Rati case airports , these scatte creases. tio increased s in the moni os s below in Fig o.  were also as rplots reveale while NOX d tored dataset ure 5.11. sessed to exa d the followi ecreased, bu s. mine ng: t may

5.4.7 Tim Example shown, so Novembe several ho this time p evident by However, periods, concentra the NO2 vehicles) the form o of O3, w concentra e Series Data time series d me of the hig r 2010. These urs during th eriod is not comparison these data a and demons tions. Althou concentration have a low N f NO and no hich was us tions (30 to 5 Figure 5 Top Row: H 2nd Row: N 3rd Row: O    ata from the hest NOx (an high concen e night (on 1 representativ with the mea re fairly repr trate some gh NO (and rarely exce O2/NOx emis t NO2. As di ually less th 0 ppb) were n .11. Example istogram of Ox vs. NO2, N 3 vs. NOx, O3 YUL air mo d NO2) conc trations are 1/12/2010 to e of typical N n concentrat esentative of important a therefore NO eds 50 ppb. sion ratio (le scussed abov an 30 ppb ot able to gr A ‐ 37   Scatterplot NO2 concent Ox histogra vs. NO2, O3 nitoring stat entrations at likely of non 11/13/2010) O, NO2, and ions describe the “behavi spects of N x) has reache This is bec ss than 0.1), e, conversion – a typical eatly exceed  Matrix for Y rations, NO2 m, NOx vs. O histogram ion are prov YUL were o -airport origin when flight NOx concen d earlier. or” of NO2 a Ox chemis d very high ause most u meaning that of NO to N background the ozone con UL.  vs. NOx, NO 3 ided below i bserved duri given they activity is lo trations attri nd NOx dur try as obse levels – grea rban NOx so the NOx was O2 is limited concentration centration (a 2 vs. O3 n Figure 5.1 ng several da were sustaine w. In other w butable to YU ing high pol rved in am ter than 300 urces (i.e., m emitted mos by the availa . Thus, the t most 30 pp 2. As ys in d for ords, L as lutant bient ppb – otor tly in bility NO2 b).

    A ‐ 38    Much higher concentrations of NO2 (e.g., greater than the 1-hour air quality standard of 100 ppb) can only form with higher primary NO2 emissions (i.e., from an emission source with a high NO2/NOx emission ratio, such as idling aircraft), or if there had been much higher levels of photochemical activity which would have allowed for a much higher portion of emitted NO to be converted to NO2. Figure 5.12. Time Series Data of Measured NO, NO2, NOx, and O3 at YUL.  Similar high monitored NOx concentrations at ADL are shown in Figure 5.13, along with the modeled NOx concentrations (i.e., modeled NO2 using the full conversion method). In this case, the winds were light – less than 3 m/s – and during the high NOx episodes from the southeast and not from the airport.                           Figure 5.13. Time Series Data of Wind Speed and Direction and Measured and Modeled NOx at ADL.    200 150 100 50 0 5/25/2012 5/27/2012 5/29/2012 5/31/2012 6/2/2012 6/4/2012 6/6/2012 300 200 100 0 W in d di re ct io n 8 6 4 2 0 W in d sp ee d Measured NOx Modeled_NOx 350 300 250 200 150 100 50 0 11/13/2010 11/15/2010 11/17/2010 dat NOx NO NO2 O3

A ‐ 39  5.5 Monitored Versus Modeled NO2 Concentrations Assessment   This section discusses the assessment of the monitored and modeled NO2 values for the three “test-case” airports derived using the NO2/NOx conversion methods discussed above. For the purposes of this assessment, the ten highest modeled 1-hour NO2 concentrations were examined. 5.5.1 Monitored Data  For reference, the top ten monitored NO2 concentrations for each airport and dataset are first presented below in Tables  5.4 and 5.5. With respect to this assessment, the following is potentially noteworthy concerning these data:  Background  Contributions – It is likely that very few of the highest monitored NO2 concentrations occurred with winds from the airport and thus were mostly urban “background” in origin. As such they were not accounted for in the modeling results since the background NOx concentrations used were zero.  No Violations – None of these monitored data exceed the one-hour NAAQS for NO2 of 100 ppb with the highest value of 77 ppb at the LAX SPAS station. Table 5.4. Top 10 Monitored NO2 Concentrations (ppb) at ADL, YUL, & LAX SPAS  Rank  Airport  ADL  YUL  LAX SPAS  1  56 71 77 2  42 68 76 3  42 66 71 4  40 66 69 5  40 66 69 6  38 65 67 7  38 62 65 8  38 61 65 9  38 61 64 10  38 61 63

    A ‐ 40    Table 5.5. Top 10 Monitored NO2 Concentrations (ppb) at LAX AQSAS Sites      Rank  Monitoring Station  AQ  CE  CN  CS  AQ   CE   CN   CS   Season 1  Season 2  1  62 64 68 69 32 43 46 29 2  59 63 65 62 27 43 41 28 3  56 61 63 60 27 41 41 28 4  54 61 61 60 27 40 40 26 5  54 61 61 59 27 39 40 26 6  54 59 60 59 26 38 39 26 7  53 58 60 58 26 38 39 24 8  53 58 60 56 26 37 37 24 9  52 58 59 55 25 37 37 24 10  52 58 59 55 25 37 37 23 5.5.2 Modeling Data  The ten highest modeled NO2 concentrations for the six NOx conversion methods are presented below in Tables 5.6 through 5.9. Among the most notable aspects of these data are that the ten highest results for the AERMOD Tier 1 (i.e., Full Conversion) and Tier 3(i.e., OLM and PVMRM) methods produced unrealistically high results. The ARM2 method’s results were also high in comparison to the monitored values, but not nearly as much as the others. Importantly, it should also be noted that the ARM2 method does not depend on the input NO2/NOx emission ratio and is only based on the modeled NOx concentration. In contrast, the OLM and PVMRM outputs are a function of the NO2/NOx emission ratio, which was set to 0.5 for non-aircraft sources. Although this likely reflects common use of PVMRM or OLM in conjunction with EDMS/AERMOD, this value is likely too high as discussed in Section 4.

A ‐ 41  Table 5.6. Top 10 Modeled NO2 Concentrations (ppb) for ADL  Rank  Conversion Method  Full  Conv.  ARM2  OLM OLM w/ Variable  PVMRM PVMRM w/ Variable  1  454 91 454 454 454 454 2  310 84 186 182 279 279 3  278 81 163 147 237 237 4  263 80 147 145 225 225 5  250 79 147 143 212 212 6  236 78 146 139 195 195 7  216 77 143 139 184 184 8  204 76 143 131 181 181 9  202 76 132 127 156 155 10  172 76 131 120 155 143 Highest monitored concentration was 56 ppb. Table 5.7. Top 10 Modeled NO2 Concentrations (ppb) for LAX AQSAS (Season 1)  Rank  Conversion   Method  Full Conv.  ARM2  OLM  OLM w/  Variable  PVMRM  PVMRM w/  Variable  AQ Station 1  1584 317 1565 747 1426 1426 2  940 188 959 427 846 846 3  787 157 815 365 708 708 4  780 156 808 335 702 702 5  733 147 765 332 660 660 6  652 130 688 330 587 587 7  647 129 683 312 582 582 8  627 125 664 310 564 564 9  605 121 644 256 544 544 10  547 109 590 250 493 493 CE Station  1  5484 1097 2782 2229 4054 3085 2  2871 574 1463 1141 1828 1426 3  2471 494 1271 1055 1685 1420 4  1838 368 960 785 1654 1354 5  1736 347 908 511 1563 1121 6  1347 269 713 509 1212 992 7  1301 260 691 488 1171 902 8  1292 258 681 470 1111 878

    A ‐ 42    Rank  Conversion   Method (Continued)  Full Conv.  ARM2  OLM  OLM w/  Variable  PVMRM  PVMRM w/  Variable  9  1026 205 553 453 802 802 10  946 189 513 433 774 679 CN Station  1  1854 371 967 996 1478 1480 2  1479 296 779 812 1331 1331 3  1414 283 747 714 1273 1273 4  1341 268 710 712 1207 1207 5  1335 267 704 691 1202 1202 6  1278 256 679 659 1122 1122 7  1247 249 664 636 1024 983 8  1092 218 589 615 983 982 9  1091 218 585 613 982 970 10  1045 209 562 571 941 941 Highest monitored concentrations ranged from 52 - 69 ppb. Table 5.8. Top 10 Maximum NO2 Concentrations (ppb) for LAX SPAS   Rank  Conversion Method  Full  Conv.   ARM2  OLM  OLM w/  Variable  PVMRM  PVMRM  w/  Variable  1  331 86 298 298 298 298 2  313 84 229 229 298 298 3  309 84 221 221 282 282 4  304 83 205 194 278 278 5  298 83 196 192 273 273 6  288 82 194 190 259 259 7  282 81 194 178 254 254 8  278 81 192 164 251 251 9  276 81 192 164 249 249 10  244 78 190 159 229 229 Highest monitored concentrations ranged from 63 - 77 ppb.            

A ‐ 43  Table 5.9. Top 10 Maximum NO2 Concentrations (ppb) for YUL  Rank  Conversion Method  Full  Conv.  ARM2  OLM  OLM w/  Variable  PVMRM  PVMRM w/  Variable  1  1100 220 578 595 885 990 2  987 197 530 560 824 889 3  984 197 518 532 797 885 4  916 183 498 530 793 824 5  842 168 455 476 733 733 6  814 163 447 468 655 723 7  728 146 394 408 644 655 8  725 145 390 400 640 652 9  648 130 355 390 583 583 10  581 116 310 333 523 523 Highest monitored concentrations ranged from 61 - 71 ppb. 5.5.3 Validation: Monitored NO2 vs. Modeled NO2  This section focuses on validating how well each model conversion method corresponds to monitored concentrations and how they compare to one another. After reviewing the literature presented in the Task 4 Working Paper49 of this project, it was determined that a majority of studies comparing modeled to monitored data presented comparisons using multiple techniques. Based on the comparison techniques of previous literature, modeled to monitored model accuracy evaluations were conducted in six statistical techniques:  Q‐Q  Plots  –  For this analysis, NO2 modeled values are compared to NO2 monitored values (unpaired in time). U.S. EPA typically conducts model evaluations by comparing modeled values to monitored values unpaired in time in this manner.50  Scatterplots  (Paired  in  Time)  – In this case, paired in time monitored vs. modeled NO2/NOX ratios are compared.  Scatterplots (Unpaired  in Time) – For this analysis, NOX concentrations compared to NO2/NOX ratio concentrations are plotted. Modeled and monitored values are independently plotted on the same plot.  RHC  – The RHC aids in reducing the effect of extreme values on model comparisons. It is determined from a tail exponential fit to the high end of the frequency distribution of observed and predicted values. 49 KB Environmental, Inc. ACRP 02-43: Development of a NOx Chemistry Module for EDMS/AEDT to Predict NO2 Concentrations. Task 4 Working Paper: Review of Regulations/Ambient standards, Current Understanding/Research Gaps, and NOx Speciation Methods in EDMS/AEDT/AERMOD. February 28, 2014. 50 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).

    A ‐ 44     Mean Squared Error (MSE) – The MSE calculation is an important statistical test that is used to examine the performance of a model and is a measure of the squares of the departure from monitored values.  Maximum Values Reported – The maximum values reported by each model were examined to compare the upper limits and concentrations of each conversion method. This assists in determining if the models are reporting realistic upper limits and concentrations. These results were reported above in Tables 5.6 through 5.9. These overall findings of comparing modeled values to monitored values show that there is a significant need for improvement for estimating NOx by AERMOD. As discussed below, there is significant over- predicting of NO2 values by the methods. Scatterplots, regression analysis, and correlation coefficients computed for modeled vs. monitored values also showed very poor agreement (i.e., low r2 values). The six techniques used to validate the monitored NO2 to modeled NO2 data are described in detail below. 5.5.3.1 Q‐Q Plots  The Q-Q plots comparing the distribution of modeled and measured NO2 values unpaired in time are presented in Figure  5.20. Note that these Q-Q plots are not an assessment of model accuracy, only a comparison between distributions and ranges. Q-Q plots are simple ranked pairings of modeled and monitored concentrations that are useful when comparing the frequency distributions of two datasets. A given quantile of the modeled concentration is plotted against the same quantile of the monitored concentration. If the distributions are similar, they will fall on the 1:1 (x = y) line (plotted for reference). Over-predictions are plotted above the 1:1 (x = y) line, and under-predictions are plotted below the 1:1 (x = y) line. In general, at low monitored NO2 concentrations, most of the methods under-predict the modeled values and at high concentrations, methods over-predict the modeled values. The following brief discussions and figures provide a synopsis of these findings.  ADL (Figure 5.14) – As shown, at low monitored NO2 concentrations, the methods under-predict, and at high concentrations, methods over-predict.

 LA Figure X AQSAS – F  AQ  (Fi predic concen  CE  (Fig predic  CN (Fig over-p concen concen other m  CS  (Fig monito  5.14. Q‐Q P or this case, gure 5.15): A t, with the trations. At ure  5.16): A t, and at high ure 5.17): A redict and trations see trations, the odels at high ure  5.18): T red concentr lot of Monito the Season 1 t low moni exception of high concent t low NO2 concentratio t very low m then begin m to have methods ten er concentra he methods ations levels. A ‐ 45  red Vs. Mod results are sh tored NO2 co the OLM rations, the m monitored c ns, the metho onitored NO2 to under-p a fair agree d to over-pre tions. tend to over eled NO2 Co own and des ncentrations which slight ethods tend oncentration ds tend to ov concentratio redict at lo ment with dict. ARM2 -predict NO2 ncentrations cribed as foll , most metho ly over–pred to over-predi s, the metho er-predict. ns, the metho w concentr the monitor tends to ove concentratio : ADL.  ows: ds tend to u icts at very ct. ds tend to u ds tend to sl ations. Mid ed data. At r-predict less ns througho nder- low nder- ightly -level high than ut all

   Figur Figu e 5.15.  Q‐Q  re 5.16.  Q‐Q   Plot of Moni  Plot of Mon tored Vs. Mo itored Vs. Mo A ‐ 46  deled NO2 C deled NO2 C oncentration oncentratio s: LAX AQSA ns: LAX AQSA S: Season 1 – S Season 1 –  AQ.   CE. 

Figu Figu  LA pr re 5.17.  Q‐Q re 5.18.  Q‐Q X  SPAS (Fig edict, and at  Plot of Mon  Plot of Mon ure  5.19):  A high concent itored Vs. Mo itored Vs. Mo t low moni rations, they A ‐ 47  deled NO2 C deled NO2 C tored NO2 c tend to over-p oncentration oncentratio oncentration redict. s: LAX AQSA ns: LAX AQSA s, the metho S Season 1 – S Season 1 – ds tend to u  CN.   CS.  nder-

    Y hi The gener                    51 RTP En Modeli http://w Figure 5. UL (Figure 5 gh concentra Figure al distributio                         vironmental As ng: Developmen ww.epa.gov/scr   19.  Q‐Q Plot .20): At low tions, they te  5.20.  Q‐Q P n results gene                     sociates, Inc. Am t and Evaluatio am001/models/  of Monitore monitored co nd to over-pr lot of Monito rated by thes bient Ratio Me n Report. Publis aermod/ARM2_ A ‐ 48  d Vs. Modele ncentrations edict. red Vs. Mod e Q-Q plots thod Version 2 hed September Development_a d NO2 Conce , the method eled NO2 Co are supported (ARM2) for use 20, 2013. Retrie nd_Evaluation_ ntrations: LA s tend to un ncentrations by previous with AERMOD f ved December Report-Septem X SPAS.  der-predict, a : YUL.  research. 51,52 or 1-hr NO2 8, 2014. ber_20_2013.pd nd at ,53,54 f.

5.5.3.2 Sc For this an time mon model con under-pre presented  A A O to e  LA 52 Paine, R Evalua 53 Paine, R Evalua 54 Podrez atterplots: P alysis, mode itored ratios. version meth dictions occu and briefly d DL  (Figure 5 DL dataset. LM w/ varia have the mo stimating the Figure 5.21. X AQSAS - F  CE  Se round while amou variab gener conse high m .J., R. F. Lee, R tion Results For .J., R. F. Lee, R tion Results For , Mark. “Ambien aired in Time led NO2/NO A 1:1 (x = od over- or r below. Th escribed belo .21)  ‐ A larg A large amou ble, PVMRM st points bel NO2/NOx ra  Scatterplots or this case, ason  1 (Fig ing errors o modeled ra nt of scatter le, PVMRM ally below th rvative than onitored/low . Brode, R. B. W AERMOD: Dra . Brode, R. B. W AERMOD: Dra t Ratio Method : NO2/NOX R x ratios were y) reference under-predic ese scatterpl w. e range of b nt of scatter , and PVMR ow the 1:1 lin tios than othe  (Paired In Ti the Season 1 ure  5.22): M r QA issues tios seemed is also prese , and PVMR e 1:1 line, in other conver modeled qu ilson, A.J. Cim ft Document. 1 ilson, A.J. Cim ft Document. 1 Version 2 (ARM A ‐ 49  atios (Monit compiled and line has also ts the values ots for ADL oth monitore is also prese M w/ variab e, indicating r methods (i. me): Monito and 2 results onitored rat in monitored to be more nt in all con M w/ variab dicating that sion method adrant. orelli, S.G. Per 998. orelli, S.G. Per 998. 2) for use with ored to Mod plotted agai been added . Over-predic and two of d and mode nt in all conv le). The OL that this met e., ratios are red Vs. Mod are shown an ios range fr data, mostl tightly cluste version met le). This scat it under-pred s. ARM2 al ry, J.C. Weil, A ry, J.C. Weil, A AERMOD 1-h eled)  nst the corre as an indica tions occur a the LAXAQ led ratios is ersion metho M w/ variab hod is a bit l under-predic eled NO2/NO d described om ~0.13 - y in very lo red within 0 hods (ARM2 terplot indic icts more fre so has a clu . Venkatram, an . Venkatram, an our NO2 Model sponding pai tion of whet bove this lin SAS datase present withi ds (ARM2, O le method ap ess conservat ted). X Ratio: ADL as follows: 1.1 (indicati w concentra .2 – 0.9. A , OLM, OL ates that ARM quently and i stering aroun d W. D. Peters. d W. D. Peters. ing”. red in her a e and ts are n the LM, pears ive in   ve of tions) large M w/ 2 is s less d the Model Model

   Fig Fig  CN Se almos As a were 1:1 li other ure 5.22. Sca ure 5.23. Sca   ason 2 (Figu t 0 to 6.35 (n result, no val found in this ne, indicating conversion m tterplots (Pa tterplots (Pa re 5.23): Mo ote: modeled ues were exc dataset. Thi that it unde ethods. ired In Time) ired In Time) A ‐ 50  nitored range values were luded on this s scatterplot r-predicts m : Monitored  Season 1 CE. : Monitored  Season 2 CN. s have a wid at zero at th plot). Note indicates that ore frequentl Vs. Modeled   Vs. Modeled   e spread of ra is high monit that no mode ARM2 is g y and is less  NO2/NOX Ra  NO2/NOX Ra tios, ranging ored ratio of led ratios ab enerally belo conservative tio: LAX AQS tio: LAX AQS from 6.35. ove 1 w the than AS  AS 

Again, the cited prev investigat 5.5.3.3 Sc For this a concentra they plot and descri  A m th e p F  LA 55 RTP En Modeli http://w general resu iously in th ion. atterplots: U nalysis, mod tions were pl monitored da bed below: DL (Figure 5 onitored con e monitored quation repro roduces the h igure 5.24. S X AQSAS – F  CE  (Fi genera conce higher NO2/N  CN (F variab vironmental As ng: Developmen ww.epa.gov/scr lts generated is section, a npaired in T eled and mon otted. Note th ta and mode .24) – At low centrations b concentratio duced by sim ighest estima catterplots ( or this case, gure  5.25):  A lly match ntrations, the NOx conce Ox Ratio. igure 5.26): A le generally sociates, Inc. Am t and Evaluatio am001/models/ by these sc lthough the ime: NOx to N itored NO2/N at these mon led data ind er NOx conc est, and at m ns the best. ilar studies.5 te of the NO Unpaired In T the Season 1 t lower NO the monitor ARM2 met ntrations, th t lower NO match the m bient Ratio Me n Report. Publis aermod/ARM2_ A ‐ 51  atterplots (pa near-constan O2/NOX Rat Ox ratios as itored and m ependent of e entrations, th id-to-higher The ARM2 5 At higher N 2/NOx Ratio. ime): NOX V and 2 results x concentrat ed concentr hod seems to e PVMRM x concentrati onitored con thod Version 2 hed September Development_a ired in time) t value of 0 io  a function o odeled data p ach other. T e PVMRM NOx concen method rati Ox concentr s. NO2/NOX R are shown a ions, the res ations best, fit the mon w/variable i ons, the PVM centrations b (ARM2) for use 20, 2013. Retrie nd_Evaluation_ are supporte .9 for PVM f modeled a oints are no hese scatterp w/variable g trations, the os also fit th ations, the PV atio Concen nd described ults from PV and at m itored conce s the highes RM w/ vari est, and at m with AERMOD f ved December Report-Septem d by past res RM warran nd monitored t paired in tim lots are pres enerally matc ARM2 meth e “ARM2 c MRM w/va trations: ADL as follows: MRM w/va id-to-higher ntrations be t estimate o able and OL id to higher or 1-hr NO2 8, 2014. ber_20_2013.pd earch ts re- NOx e, as ented h the od fit urve” riable .  riable NOx st. At f the M w/ NOx f.

   Figu Figu conce higher NO2/N re 5.25.  Scat re 5.26. Scat   ntrations, the NOx conce Ox Ratio. terplots (Unp terplots (Unp ARM2 meth ntrations, th aired In Tim aired In Tim A ‐ 52  od seems to e PVMRM w e): NOX Vs. N Season 1 CE. e): NOX Vs. N Season 1 CN. fit the monito / variable O2/NOX Rati O2/NOX Rati   red concentr is the highe o Concentra o Concentrat ations the be st estimate o tions: LAX AQ ions: LAX AQ st. At f the SAS  SAS 

 L v N h R Figu  Y m N h N AX SPAS (Fig ariable are th Ox concentr igher NOx c atio. re 5.27.  Sca UL (Figure 5. ethods are t Ox concentr igher NOx c O2/NOx Rati ure 5.27) – A e most clos ations, the A oncentrations tterplots (Un 28) – At low he most clos ations, the A oncentration o. t lower NO ely aligned w RM2 method , the PVMR paired In Tim er NOx conce ely aligned w RM2 method s, the PVM A ‐ 53  x concentrati ith the mon seems to fi M w/ variab e): NOX Vs.  ntrations, the ith the mon seems to fi RM w/varia ons, the PVM itored concen t the monito le is the hig NO2/NOX Rat PVMRM w itored conce t the monito ble produces RM w/ vari trations, and red concentra hest estimate io Concentra / variable and ntrations and red concentra the highes able and OL at mid to h tions the be of the NO2 tions: LAX SP OLM w/ va , at mid to h tions the be t estimate o M w/ igher st. At /NOx AS.  riable igher st. At f the

   Fig The result earlier in findings o ambient ra 5.5.3.4 R The RHC observed  For demo represents column re divided b NOx to NO                    56 RTP En Modeli http://w ure 5.28.  Sca s generated b this section. f previous re tio curve inh HC  statistic is a NO2 concentr RHC = X(N Where N = th N = 26 X(N) = X = th nstration pu the 26th high presents the y the Full co 2), and the f                         vironmental As ng: Developmen ww.epa.gov/scr   tterplots (Un y these scatt The curve as search56 and erent in the m measure of ations. The R ) + [X ‐ X(N)] x : e number of v based on EP the Nth larg e average of t rposes, the r est value, th RHC score, t nversion RH ifth column r                     sociates, Inc. Am t and Evaluatio am001/models/ paired In Tim erplots (unpa sociated with is explained t ethod. model perfor HC value is  ln[(3N ‐ 1)/2 alues at the h A guidance ( est value, in he N-1 large esults for A e second colu he fourth co C score, sinc epresents the bient Ratio Me n Report. Publis aermod/ARM2_ A ‐ 54  e): NOX Vs.  ired in time) the ARM2 o be the resu mance and i calculated us ] in ug/m3  igh end of th Protocol for ppb st values, ppb DL are sum mn represen lumn represe e the Full Co highest NO2 thod Version 2 hed September Development_a NO2/NOX Ra are supported conversion m lt of the pow s based on th ing the follow e distributio Determining marized in ts the averag nts the RHC nversions as concentratio (ARM2) for use 20, 2013. Retrie nd_Evaluation_ tio Concentr by the prev ethod is also er function s e top 26 hig ing equation n of values the Best Perf Table  5.10. e of the top 2 ratio (i.e., m sumes comp n. with AERMOD f ved December Report-Septem ations: YUL. ious research supported b hape of the A hest modele : orming Mod The first co 5 values, the odeled RHC lete conversi or 1-hr NO2 8, 2014. ber_20_2013.pd cited y the RM2 d and el) lumn third score on of f.

A ‐ 55  As shown, the RHC results indicate that the ARM2 has the lowest over-predicted ratio concentration when compared to monitored ratios concentrations. PVMRM is greatly over-predicting ratios when compared to monitored ratio concentrations. All other conversion methods over-predict ratios when compared to the monitored ratio concentrations. Table 5.10. RHC: Monitored NO2 vs. Modeled NO2 (ppb): ADL Method Values  X(N)  X  RHC  RHC Ratio  Highest NO2Concentration  Full Conversion  88 169 386 454 ARM2  66 75 98 0.25 91 OLM  70 122 258 0.67 454 OLM w/Variable  60 109 240 0.62 454 PVMRM  73 148 348 0.90 454 PVMRM w/Variable  70 141 328 0.85 454 Monitored (NO2)  31 35 47 0.12 56 5.5.3.5 Mean Squared Error (MSE)  The mean squared error is a measure of the squares of the departure from monitored values. The lower the MSE, the closer the modeled concentrations are to the monitored concentrations. A sample of these values for ADL are shown in Table 5.11. The results indicate that the ARM2 conversion method has the lowest MSE, indicating that the difference associated with the ARM2 when compared to the monitored values is smaller than for other conversion methods. The PVMRM and OLM methods’ MSE values are likely very sensitive to the input NO2/NOx emission ratios. Table 5.11. 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 5.5.4 Meteorological Data Assessment  A sensitivity analysis was conducted to examine the influence meteorological factors have on the NO2/NOx conversion methods. For this analysis, YUL was chosen as it represents a middle spread of the modeling data. The meteorological conditions that were considered are as follows: (i.) wind speed, (ii.) wind direction, (iii.) temperature, and (iv.) relative humidity.

   For this a reported a 5.5.4.1 W Figure 5.2 plot show with prev Previous r Figure 5 5.5.4.2 W The effec higher ov the termin conversio                    57 ACI Eu 58 Carslaw Contrib 59 Langw Manag nalysis, an u nd therefore ind Speed   9 provides a s that as win ious studies eports have a .29.  YUL Mo ind Direction t of wind dir erestimated v al, as the te n methods as                         rope. Effects of , David, Sean B utions to Ambie orthy, Lucinda ement, 2012.    pper NO2 li NO2 values a scatterplot o d speed incre which have a lso illustrate deled Conce    ection indica alues than ot rminal is sou shown in Fig                     Air Traffic on A eevers, Karl Ro nt Nitrogen Ox Minton. New 1-H mit was also bove four up f the monitor ases, the rela lso reported d that AERM ntrations (p tes that win her wind dire thwest of th ure 5.30.  ir Quality in the pkins, and Mar ides in the Vicin r Air Quality S A ‐ 56  set since m per limit thre ed and mode tive accuracy an increase OD performs pb) as a Func d directions ctions. This e monitor. T Vicinity of Eur garet Bell. Detec ity of a Large In tandard Pose Im any extreme sholds were r led NO2 valu of the mode in accuracy poorly at lo tion of Wind coming from could be ind his overesti opean Airports. ting and Quant ternational Airp plementation C ly high mod emoved. es by wind s ls increase. T with increase w wind speed  Speed (all w the southwe icative of N mation is pre 2010. ifying Aircraft a ort, Atmospher hallenges. Env eling values peed at YUL hese results d wind spee s.59 ind direction st tended to O2 emissions sent in all m nd Other On-Ai ic Environment. ironmental were . The agree d.57,58 s).  have from odel rport 2006.

5.5.4.3 Te As shown temperatu Figure 5 mperature   in Figure 5. re on the NO .30. Boxplots 31, the asses 2/NOx conver  of Modeled sment of thi sion method A ‐ 57   NO2 ‐ Monit s variable ind s. ored NO2 by icated that t  Wind Direct here was no ion.  clear influen ce of

   5.5.4.4 R The analy influence 5.5.5 Sou Although of the th concentra For exam mode), “g each of th Noticeabl  A 75  W ai Figu elative Humi sis of relati from this var rce Apportio emissions fro ree test-case tions based o ple, from the ates” (compr e three airpor e outcomes o t ADL and Y %), whereas ithin the airc rports: 18% a   re 5.31. YUL dity  ve humidity iable. nment   m aircraft la airports th n the modelin EDMS/AER ising APUs ts. These resu f this analysi UL, the Ga at LAX aircr raft category t ADL, 19%  Monitored N on NO2/NO nding / takeo ey are not g. MOD outpu and GSE), r lts for NO2 ( s including th tes category aft are the la , the taxi/idle at LAX and A ‐ 58  O2 ‐ Modele x conversion ff cycles are necessarily t t files, Table oadways, etc i.e., full conv e following: is the larges rgest contribu phase accou 87% at YUL d NO2 and Te methods in the largest so he biggest  5.12 shows . to NO2 con ersion metho t contributor tor to NO2 ( nts for a vas . mperature.  dicate that th urce of NOx contributor t the contribut centrations d d) and ARM to both NO 70%). tly different p ere was no emissions at o NOx and ion of aircra uring one ho 2 are shown. x and NO2 ( ortion at the clear each NO2 ft (by ur at 71 to three

A ‐ 59  Table 5.12. Modeled NO2 Concentrations Apportioned by Sources  Airport  Sources  Full Conversion  ARM2  ug/m3 % ug/m3  % ADL  (July 1 ‐ Dec  31 2011)  Aircraft  Approach 5E-03 0 3E-03 0 Takeoff 40.7 17 23.1 17 Landing 1.4 1 0.8 1 Taxi 10.4 4 5.9 4 Gates (GSE & APUs) 175.7 71 99.8 71 Parking 0.8 0 0.5 0 Roadways 16.9 7 9.6 7 Stat. Sources 5E-08 0 3E-08 0 Training Fires 0 - 0 - All 246.0 100 139.7 100 LAXSPAS  (2009)  Aircraft  Approach 1.4 0 0.4 0 Takeoff 295.1 56 85.9 56 Landing 8.7 2 2.5 2 Taxi Q 75.9 14 22.1 14 Gates (GSE & APUs) 77.4 15 22.5 15 Parking 40.0 8 11.6 8 Roadways 25.2 5 7.3 5 Stat. Sources 0.0 - 0.0 - Training Fires 0.0 - 0.0 - All 523.7 100 152.4 100 YUL (2009‐ 2011)  Aircraft  Approach 4E-17 0 9E-18 0 Takeoff 22.1 3 5.9 4 Landing 1.0 0 0.3 0 Taxi Q 129.1 20 31.8 20 Gates (GSE & APUs) 490.0 75 120.2 75 Parking 0.5 0 0.1 0 Roadways 9.1 1 2.5 2 Stat. Sources 5E-04 0 1E-04 0 Training Fires 0.0 - 0 - All 651.8 100.0 160.8 100.0 These results suggest that the EDMS/AEDT/AERMOD system may be attributing disproportional impacts for different airport sources under certain conditions.

    A ‐ 60    5.6 Conclusions and Observations   As demonstrated in this section, the results of Task 6 reveal that modeling NO2 concentrations near airports using existing models and NOx/NO2 conversion methods is characterized by a number of significant complications and shortcomings. The following highlights some of the most noteworthy of these along with some potential remedies.  Poor Correlations - The comparison of modeled vs. measured NO2 values at the three test-case airports does not lead to obvious conclusions regarding the accuracy of the various NOx chemistry modules (e.g. ARM vs. OLM vs. PVMRM). Previous studies have supported this finding in both NO2 concentrations as well as other parameters. In particular, ACRP Report 7160 stated that low r2 values were reported when comparing modeled to monitored values for carbon monoxide (CO), NOx, carbon dioxide (CO2), and other parameters. Other studies also support that modeled vs. monitored CO concentrations have low r2 values.61  NO2/NOx  Ratios  ‐ As reported, 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). In short, the sensitivity of the modeled NO2 values should be re-assessed with a wider range of NO2/NOx emission ratios.    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. This limitation is not the fault of AERMOD or any of the NOx chemistry options. Due to the uncertainties in connection with background concentrations, this analysis is subject to some uncertainty.62  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.  AERMOD  Limitations  ‐ Predicted concentrations of NOx by AERMOD can be greatly overestimated at low wind speeds. Inaccuracies in predicting NOx (rather than NO2) indicate that there are problems with EDMS and/or the dispersion modeling within AERMOD itself, or that the inputs for the model were insufficient or inaccurate (e.g., inputs of NOx emissions, meteorological inputs, etc.).  Variable Results ‐ According to scatterplots (unpaired), generally speaking, 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 ARM2 model seems to fit the monitored concentrations the best. The ARM2 modeled ratios also fit the “ARM2 curve” equation reproduced by similar studies cited within this section. At higher NOx concentrations, the PVMRM w/ variable is the highest estimate of the NO2/NOx Ratio. These results could differ greatly with different NO2/NOx emission ratios used as inputs – the OLM and PVMRM would predict lower NO2 concentrations. Using the NO2/NOx emission ratio of 0.5, the OLM and PVMRM methods all produced maximum NO2 concentrations that were much higher than any of the measured NO2 concentrations, whereas the maximum NO2 concentrations produced by the ARM2 method were significantly lower for all datasets. Note that the ARM2 method is insensitive to the NO2/NOx emission ratio.                                                              60 Transportation Research Board. ACRP, Report 71, Guidance for Quantifying the Contribution of Airport Emissions to Local Air Quality. 2012. 61 Martin, Anjoli. Verification of FAA’s Emissions and Dispersion Modeling System. 2006. 62 Tran, Khanh. AERMOD NO2 Modeling of Barto Compressor Station. January 24, 2013.

A ‐ 61   Multiple Plumes – By its very nature, this assessment incorporated multiple emissions sources and plumes in the modeled. The merging of multiple plumes in the PVMRM method has been found to lead to discontinuities in the model predictions and this effect should be further considered when examining results presented in this study.  Source Apportionment – The modeling results suggest that the EDMS/AEDT/AERMOD system may be attributing disproportional impacts for different airport sources under certain conditions.

    A ‐ 62    6.  Research Plan (Task 7)  This section presents and discusses the outcomes of Task 7 of the ACRP 02-43 Amplified Work Plan. The materials include an assessment of existing models/methods for predicting NO2 in the vicinities of airports and the formulation of a Research Plan for developing a Preferred Method. 6.1 Task Objectives  As initially discussed in Section  2, the purpose of this task is for the Research Team to prepare a Research Plan designed to accomplish the following two primary objectives: Task 7 Research Plan Objectives   Evaluate and develop an alternative method(s) for predicting NO2 concentrations from airport- related emission sources, and  Recommend a Preferred Method. As a means to achieve these objectives, the ACRP 02-43 Work Plan calls for the Research Team to undertake the following tasks:  Assessment and evaluation of existing models/methods, including:  Comparison of models/methods using qualitative criteria, and  Comparison of models/methods using quantitative criteria;  Assessment and evaluation of an alternative models/method(s); and  Develop a plan for incorporating a Preferred Method(s) within EDMS/AEDT. Importantly, the full implementation of the Task 7 Research Plan is among the principal aims of Task 10, Execute Research Plan – which will be undertaken following the Interim Report Meeting and contingent upon the endorsement of the Panel. 6.2 Available Models/Methods  Presently, there are a number of available models that have been devised, or can be used, for computing NO2 concentrations in the ambient air, although none of them are specifically designed for airport applications. Most notable among these is the U.S. EPA AERMOD model with the alternative NO2 prediction methods (i.e., ARM/ARM-2, OLM and PVMRM). Other relevant models also include EPA’s CALINE4, CALPUF, CMAQ and SCICHEM. The Task 4, Working Paper contains a full description of these models and Section 3 of this Interim Report provides brief overviews. As discussed, there are both important similarities and distinctions as well as individual attributes and disadvantages that characterize these models – especially when it comes to predicting NO2 concentrations, in general, and associated with airports, in particular. 6.2.1 Model Properties & Characteristics  For ease in reviewing and assimilating the main properties and characteristics of the available models, Table  6.1 provides a “high-level” overview and simple comparison of some of their most important features. This appraisal is based upon the following criteria:     Research Plan Execution Importantly,  the  Task  7  Research  Plan  for  developing  a  Preferred  Method  for  modeling  NO2  in  the  vicinities  of  airports  is  to  be  completed in support of Task 10.   

A ‐ 63  Model Properties & Characteristics   U.S. EPA Classification – This refers to EPA’s Preferred/Recommended and Alternative Models (i.e., those included on the EPA’s Support Center for Regulatory Atmospheric Modeling) – a potentially important aspect when evaluating models that have undergone the agency’s “vetting” process.  Common  Uses  – This provides a generalized description of the models’ most common applications.  Underlying  Method  – This describes the computational means and processes by which the models function and perform.  Pollutants  – This provides a listing of the types and characteristics of air pollutants that are analyzed by the models.  NOx Chemistry Application – This describes the method(s) used by the models to account for chemical reactions and transformations – particularly as they apply to NOx.  Source  Types  – Using dispersion modeling terminology, this describes the types of sources represented by the models.  Terrain  Modeling  – This indicates whether or not a model takes into account changing topography.  Spatial Range – This generally describes the area or distances over which the models are most commonly used.  Temporal Resolution – This describes the time periods over which the models’ predictions are typically made.

A ‐ 64  Table 6.1. Overview & Comparison of AERMOD and Alternative Model Properties and Characteristics  Model  EPA Classification  Common Uses  Underlying  Method  Pollutants  NOx Chemistry  Application  Source  Types  Terrain  Modeling  Spatial Range  Temporal  Resolution  AERMOD  On U.S. EPA's Preferred / Recommended List Used for most regulatory point source assessments. Contained within EDMS Steady-State Gaussian Plume Any relatively stable, primary pollutant (e.g., CO, TSP, PM2.5, PM10), SO2, NO2 Full conversion, ARM/ARM2, OLM, and PVMRM Point, area, volume Yes 10 m to 20 km 1 hr to 1 year (multi- year possible) averaging times CALPUFF  On U.S. EPA's Preferred / Recommended List Used as an alternative to AERMOD, especially for "long range" transport Gaussian puff Any relatively stable, primary pollutant (e.g., CO, TSP, PM2.5, PM10, SO2, NO2, etc.) Pseudo first- order Point, line, area, volume Yes 10 m to >100 km 1 hr to 1 year averaging times CALINE4  Not on U.S. EPA's Preferred / Recommended List. Only applicable in California Gaussian line source model used for highway sources Steady-state Gaussian plume from line sources Any relatively stable, primary pollutant such as CO and NO2 First-order Discrete Parcel Method Line No Gaussian short range 1 hr averaging time CMAQ  On U.S. EPA's List of Photochemical Grid Models Used for local, state, and regional air quality modeling Nested Eulerian grid and plume- in-grid Criteria gases including O3, pri. & sec. PM, PM, PM components, and HAP species Full gas phase and aerosol chemistry (Carbon Bond Module) Grid and point Yes 1 km to >100 km grids (local to continental coverage with multiple grids) Few minutes to annual averaging times SCIPUFF/  SCICHEM  On U.S. EPA's Alternative Models List Alternative general model Second- order closure integrated Gaussian puffs Criteria gases including O3, pri. & sec. PM, PM, PM components, and HAP species Simplified NOx chemistry and full chemistry in SCICHEM (Carbon Bond Module) Point, area, volume Yes 10 m to >100 km 10 min to 1 year averaging times

A ‐ 65  6.2.2 Discussion of Model Properties and Characteristics   6.2.2.1 Overview  As shown above in Table  6.1, AERMOD and CALPUF are both listed among the EPA’s Preferred/Recommended Models63 – a potentially important aspect when evaluating models that have undergone the agency’s “vetting” process and have been deemed acceptable for regulatory applications. CMAQ and SCICHEM are also models sanctioned by EPA, but not as highly or appropriately. CALINE4 is used nationwide but is not among the EPA-approved models. Most of the models share common applications but differ from one another with respect to most common uses which range from regulatory purposes to area-wide and long range simulations. Notably, AERMOD is already incorporated into the EDMS/AEDT modeling architecture. All of the models also share a common list of pollutants (i.e., CO, PM, etc.) but some also include highly reactive compounds such as O3 (i.e., CMAQ and SCICEM). The source types mostly include point, area, line and volume sources; most models address varying terrain, the spatial ranges extend from meters to kilometers and the time-based predictions are commonly from 1 hour to a year – although shorter time periods are possible with SCICHEM. Because the methods by which the models address NOx/NO2 chemistry and transformation is “central” to the ACRP 02-43 Research Project, this criterion is discussed separately in the next section. 6.2.2.2 NOx/NO2 Conversion Methods  The following provides a focused discussion on the alternative methods used by the various models for addressing the NOx/NO2 chemistry and transformation processes.  AERMOD - The NOx/NO2 conversion methods employed in AERMOD were previously described in Section 3 and include the following:  Tier 1: Full conversion (“Default” method)  Tier 2: Fixed ratio and empirical methods (ARM and ARM2)  Tier 3: Detailed NOx conversion (OLM and PVMRM) The Tier 1 method is the simplest and is based upon full conversion. The Tier 1 ARM and ARM2 methods are based on fixed and empirical conversation rations, respectively. And the Tier 3 methods include the OLM and PVMRM which directly incorporate the photostationary state using aggregated NO2 formation equations with simplifying atmospheric conditions and applications. Notably, although AERMOD is a U.S. EPA recommended model, it should be noted that since PVMRM, OLM, and ARM2 are not “default” options, these methods are not included within this designation.  CALINE4 (Simple First‐Order Reactions) ‐ CALINE4 is the latest version of the “CALINE-series” of highway/roadway air quality models. CALINE3 is the U.S. EPA recommended model for free- flow roadway conditions while CALINE4 is generally only used in California to model interrupted highway (i.e., intersection) air quality. The Discrete Parcel Method employed in the CALINE4 uses simple first-order reactions and is similar in complexity to the OLM and PVMRM. Simply stated, the “parcel” method uses the concept of a parcel to model changing concentration gradients between the parcel and the 63 Preferred models” as defined in 40 CFR Part 51, Appendix W (Guideline on Air Quality Models).

    A ‐ 66    surrounding atmosphere. The method is fully integrated with the mixing zone concept used in all of the CALINE-based models where pollutants are assumed to be fully mixed within the mixing zone (the immediate area surrounding the roadway) to a height of 3.5 m. With CALINE4, NO2 concentrations are based on the initial concentration within the mixing zone and the travel time of a parcel from each finite line element to a receptor. The following simplified set of first-order rate equations are used to describe the NOx conversion reactions:  NO2 + hv  NO + O  O + O2  O3  NO + O3  NO2 + O2 For simplification, these reactions are assumed to occur independently within each parcel during the dispersion process. Also, the reactions are assumed to occur as isolated processes within each parcel, thereby avoiding complications from overlapping plumes.  CALPUF  (Pseudo  First‐Order  Reactions)  ‐  CALPUFF incorporates two methods that employ “pseudo first-order” chemical reaction rates: (i.) the methodology from the MESOPUFF II model and (ii.) the RIVAD/ARM3 method. In addition to the conversion of NO to NO2, the methods also treat the conversion of NO2 to nitrate (NO3-) aerosols and the formation of sulfate (SO4-2) aerosols from SO2. Ozone concentrations can be specified by hour or as a constant background value.   MESOPUFF - The MESOPUFF II chemical transformation methods simulate the formation of nitrate and sulfate aerosols with NOx photochemistry playing an important part of the gas phase reactions. This photochemistry includes reactive organic gases (ROGs) with diurnal changes where the free radical chemistry is much more active during the day and the oxidation rates are lower during night hours. The reaction rates for the formation of aerosols are highly dependent on temperature and relative humidity.  RIVAD/ARM3 - This method is similar in modeling the formation of aerosols, but uses a condensed set of simple, first-order rate equations to model the photostationary state to obtain pseudo steady-state concentrations of NO, NO2, and O3. The diurnal cycle is modeled with the formation of NO and O3 balanced with the formation of NO2 during daylight hours. During nighttime, only the NO and O3 titration reaction is modeled. The O3 concentration within each puff does evolve based on distance from the source, but rather is replenished during each time-step. This may cause errors if the sources are high NOx emitters resulting in a depletion of O3 close to the source, resulting in low OH radical formation and low formation of aerosols from NO2 and SO2. Also, because this method assumes low background VOC concentrations, it is not well suited to modeling urban environments.  CMAQ (Carbon Bond [CB] Full Atmospheric Chemistry)  ‐ As previously discussed, grid models assume uniform mixing within relatively large grids resulting in a single average concentration for each pollutant and grid cell. The mass balance methods employed under this Eulerian grid framework is conducive to modeling a comprehensive set of chemical reactions that describe the chemistry of both inorganic and organic species.  Perhaps the most well-known and well-used grid-based, gas phase chemistry module is the Carbon Bond (CB) mechanism, currently at version 6.64 This module is used in the leading grid                                                              64 Yarwood, Greg, Jaegun Jung, Greg Z. Whitten, Gookyung Heo, Jocelyn Mellberg, and Mark Estes. “Updates to the Carbon Bond Mechanism for Version 6 (CB6).” Presented at the 9th Annual CMAS Conference. Chapel Hill, NC. October 11-13, 2010.

A ‐ 67  models, CMAQ65 and the Comprehensive Air Quality Model with extensions (CAMx),66 which have typically been used to predict regional O3 formation. The modeled species include various nitrogen species (NO, NO2, HONO, etc.), oxidants (O3 and H2O2), hydrocarbons (paraffin carbon bond, olefinic carbon bond, ethane, toluene, etc.), carbonyls and phenols (formaldehyde, acetaldehyde, etc.), etc. Although the chemistry represented in the CB mechanism is considered to be comprehensive, it does not explicitly include every organic chemical reaction since it would be computationally burdensome to do so. Rather, simplifications are made to group organic reactions based on the carbon bond type associated with individual carbon atoms. Another simplification involves the treatment of each reaction as elementary – meaning each reaction involves no more than three reactants and all stoichiometric coefficients for these reactants have a value of one. As such, the derivation of reaction rates is simplified and the use of reaction rate constants (K) is straightforward. A small sample of the nitrogen species-related, photochemical reactions is exemplified below:  NO2 + hv  NO + O  O + O2  O3  O3 + NO  NO2 + O2  NO + NO3 --> NO2 + NO2  O3 + NO2  NO3 + O2  O3 + hv  O2 + O The corresponding reaction rate constants are based on various empirical expressions as well as the Arrhenius equation: K = Ae-(E/T) where, A = Pre-exponential factor E = Activation energy divided by the gas constant T = Temperature Although the CB mechanism is state-of-the-art, continuous improvements can be made to increase the accuracy of O3 and other species formations. Improvements may include “tightening” certain reaction rates and interactions between NOx species and other chemicals. So while improvements may continue, the historical evolution of this module appears to indicate a level of maturity where the core mechanisms appear to be stable.  SCICHEM  (Staged  Chemistry)  ‐  The overall set of chemical reactions used under a grid-based system described above can also be applied to a Gaussian plume-based framework. This is the case with the SCICHEM model where essentially the CB suite of reactions are used.67 SCICHEM is the reactive pollutant version of the Second-Order Closure Integrated Puff (SCIPUFF) model. As the name implies, both SCICHEM and SCIPUFF are based on a Gaussian puff modeling framework that allows the modeling of both static plumes and instantaneous puffs. Using time- steps, the puffs allow for modeling of detailed time-varying pollutant concentrations. The overall concentration predictions within SCICHEM are based on the use of the basic, three-dimensional 65 Byun, D. W. and J. K. S. Ching. “Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System.” Office of Research and Development, United States Environmental Protection Agency. EPA/600/R- 99/030. Washington, DC. March 1999. 66 Environ. “User’s Guide, Comprehensive Air Quality Model with Extensions, Version 6.0.” May (2013). 67 Electric Power Research Institute (EPRI). “SCICHEM Version 1.2: Technical Documentation.” Final Report. December 2000.

   G tu T be ac co In se st th ac oc tr O da sa aussian puff rbulence clos A + ζB where, A and Pi = Pr ζ and K = R he chemistry tween reacti count result mputationall order to si parated into ate with NO, e plume whe ids. Stage 3 curs requirin ansition crite H for the Sta y or night c mple set of r Source: EPRI   equation, an ure theory. T K --> ΣγiPi B = Reactant oduct γi = Stoichio eaction rate modeling ta ve species, ing in incre y optimized mplify the o three stages. NO2, and O re radical co is the furthe g full chem ria between ge 1 to Stage onditions as eactions for t Figure 6.1. Sa 2000. d the turbul he basic chem s metric coef constant kes into acc thus reducing ased species by allocating verall chem Stage 1 is th 3 assuming n ncentrations st part of the istry. The fu stages are ty 2 transition) well as var he first two s mple set of  A ‐ 68  ent diffusion ical reaction ficients ount plume/ reaction ra concentratio puffs to three istry and red e early part egligible imp start to beco plume wher ll CB mech pically based . Transition ious other fa tages. SCICHEM Sta parameters s in SCICHE puff turbulen tes. Overlap ns for react -dimensiona uce the com of the plume acts from ra me significan e sufficient m anism is app on concentr criteria betwe ctors. Figure ge 1 and 2 C are based on M take the f ce which re ping puffs a ions. This p l grids to ide putational e representing dicals. Stage t with forma ixing with t lied only un ations of ce en the stage   6.1 present hemistry  the second- ollowing form duces the co re also taken uff interacti ntify nearby p ffort, a plum a photostati 2 is a mid-p tion of seco he backgroun der Stage 3 rtain species s are depende s a facsimile order :  ntact into on is uffs. e is onary art of ndary d air . The (e.g., nt on of a

A ‐ 69  In addition to the use of the full CB chemistry, SCICHEM also allows the use of an optimized set of simple first-order NOx reaction equations that simplifies the input needs for modeling NO2 concentrations as well as runtime. The use of this option is generally suitable for near-source scenarios with simplifying assumptions such as less mixing with background air. The reaction equations used under this option include:  [NO2]  [NO] + [O]  [O]  [O3]  [O3] + [NO]  [NO2]  [NO] + [NO]  2 [NO2] In recent years, SCICHEM has been used as the plume-in-grid solution for grid modeling, including within CMAQ and CAMx. SCICHEM improves upon earlier plume-in-grid models through advancements such as improved mixing with the background air, wind shear, plume overlaps, and plume turbulence impacts on reaction rates.68 The implementation within CMAQ is referred to as CMAQ with an Advanced Plume Treatment (CMAQ-APT). 6.2.3 Qualitative Assessment of Existing Methods and Models  This section provides a comparative qualitative assessment of the existing models and methods that are used to compute NO2 concentrations. Importantly, the emphasis is on the potential application of these models/methods using EDMS/AEDT to predict NO2 levels in the vicinities of airports. 6.2.3.1 Approach & Criteria  As with most qualitative assessments, this review is made using criteria ratings that should be viewed as “relative” among the models/methods. That is, they do not represent characteristics of the models/methods in an absolute sense and should not be taken as such. In other words, the ratings of “low,” “medium,” and “high” designations only apply within the specified criteria. For example, the medium rating for CALPUFF could have been considered high if CMAQ and SCICHEM were not included in these comparisons. Even though the methods used in AERMOD (i.e., full conversion, ARM/ARM2, OLM, and PVMRM) are included as baselines for comparison purposes (i.e., not as an “alternative” method), they can also serve as a basis for a new method depending on the results of the overall evaluations. The default, full conversion method is included to serve as the baseline among the AERMOD methods since it does not include any NOx conversion methods (only physical dispersion). As such, the full conversion NO2 concentrations will be highest at all receptor locations. Although ARM and ARM2 are different methods, they are grouped together since the overall scopes of the two methods are similar. It should also be noted that in most cases, the comparisons are necessarily based on a combination of the NOx conversion methods and the corresponding models. In most cases, the methods are fully integrated into the dispersion modeling methodology, making it difficult to evaluate just the method alone. As such, while not perfect, the comparisons should be considered a best effort at focusing on the methods. 68 Karamchandani, P., R. Morris, B. Brashers, G. Yarwood, L. Parker, E. Knipping, N. Kumar, B. Chowdhury, and I. Sykes. “Application of SCICHEM for Near-Field and Far-Field Single Source Impacts.” Presented at the AWMA Guideline on Air Quality Models: The Path Forward Conference, Raleigh, NC, March 2013.

    A ‐ 70      Evaluation Criteria  AERMOD  CALINE4  CALPUFF  CMAQ  SCICHEM    Tier 1  Full  Conversion    Tier 2  ARM/ARM2  Tier 3  Simple 1st  Order  Reactions  Pseudo 1st  Order  Reactions  Full  Chemistry  (Carbon  Bond)  Staged  Chemistry OLM  PVMRM  Use of photostationary state reactions  No No (but ARM2 does indirectly) Yes Yes Yes Yes Yes Yes Accounts for chemical species beyond  NO/NO2/O3  No No No No No Yes (as part of the aerosol chemistry) Yes (full atmospheric chemistry) Yes (full atmospheric chemistry) Model/method acceptability  Good, EPA “Default “ model (Tier 1) Good (EPA Tier 2) Good (EPA Tier 2) Good (EPA Tier 2) Fair (California required model) Good (EPA recommended model) Good (EPA recommended regional model) Good (EPA Alternative model)  Input data complexity   Low Low Medium Medium Medium Medium High Medium to High Computational complexity & runtime  Low Low Medium Medium Medium High High High Plume growth & changes in O3  concentrations  No No but ARM2 accounts for these effects No Yes Yes Yes N/A for grids, but in PinG Yes  Simulates multiple plumes  No No, but ARM2 inherently accounts for multiple plumes No, but approximate by a source grouping feature Yes No Yes, but limited by number of sources allowed (200) N/A for grids, but in PinG Yes User‐supplied ambient O3  concentrations  No No, but ARM2 inherently accounts for background conditions Yes Yes Yes Yes Yes Yes       Table 6.2. Qualitative Assessment of Models & Methods 

A ‐ 71  Evaluation Criteria  AERMOD CALINE4 CALPUFF CMAQ SCICHEM  Tier 1  Full  Conversion  Tier 2  ARM/ARM2  Tier 3  Simple 1st  Order  Reactions  Pseudo 1st  Order  Reactions  Full  Chemistry  (Carbon  Bond)  Staged  Chemistry  OLM  PVMRM  Plume segmentation and/or region  modeling  No No, but ARM2 inherently accounts for plume evolution No No, but plume changes No No N/A for grids, but in PinG Yes User‐supplied NO2/NOx ratios by  source and operational mode  No No (only NOx concentration for ARM2) Yes Yes No Yes Yes, but number of sources limitations Yes Use of monitoring data to improve  background air characteristics (e.g.,  temperature, humidity, non‐O3  concentrations, etc.)  No No, but ARM2 inherently accounts for background conditions Yes, but only MET data Yes, but only MET data Yes Yes Yes Yes Feasibility of implementation within  AEDT/AERMOD  Already in AERMOD Already in AERMOD Already in AERMOD Already in AERMOD Difficult due to reliance on mixing zone Difficult due to time-varying puff method Difficult due to Eulerian grid method Difficult due to time-varying puff method Source‐specific modeling  Yes Yes Yes Yes No (sources grouped as line sources and within mixing zone) Yes No (sources grouped within grids unless PinG used) Yes Modeling – Receptor Limitations  None None None None None None N/A as grids are used, but PinG limits on number of receptors None Modeling – Source Limitations  None None None None None 200 source limit N/A because sources are grouped into grids None Finest time‐varying concentrations  Coarse (1-hr) Coarse (1-hr) Coarse (1-hr) Coarse (1-hr) Coarse (1-hr) Fine (1-sec) Coarse (1-hr) Fine (1-sec) Spatial domain  Local Local Local Local Local Local Regional Local

    A ‐ 72    6.2.3.2 Key Findings  A number of important findings are evident from the qualitative evaluation and comparison of the existing models/methods that are available for computing airport-related NO2 concentrations. The most relevant of these are discussed below:  Photostationary state  reactions - With the exceptions of AERMOD Tiers I (Full Conversion) and II (ARM/ARM2) the evaluated models/methods take into account the photostationary reactions between NOx and O3 through direct modeling of NO/NO2/O3 conversion equations. Notably, while the ARM2 method does not explicitly model this condition, the method’s underlying regression algorithm accounts for the state.  Chemical  species  beyond NO/NO2/O3  – The methods within AERMOD are specific towards modeling NOx conversion and currently do not allow modeling of other reactive species. In contrast, the methods in some of the other models, allow for modeling various other reactive chemicals. CMAQ and SCICHEM are the most robust in allowing for full chemistry of the atmosphere. The ability to model other reactive species may be beneficial but issues of complexity and runtime would need to be considered.  Model/method  acceptability  ‐ Each of the non-full conversion methods can be considered “good” in terms of model acceptability because each have been vetted through the U.S. EPA or state model evaluation processes. The AERMOD methods are a part of the U.S. EPA tiered methods for NO2 while the others are EPA recommended/alternative or state-approved methods. This indicates an established history and support for the methods/models.  Input  data  complexity  ‐ Input data complexity refers to the types and amount of input data required to use the methods/models. While simpler input requirements may be preferred, input data requirements may also be viewed as an indication of the level of modeling robustness (i.e., ability to take into account various factors). Except for the full conversion method and ARM/ARM2, the methods have similar input data complexities with options to use default or representative data versus more detailed, user-supplied information (e.g., a single ambient O3 concentration or hourly values). SCICHEM and CMAQ can have the most complex requirements if the user chooses to supply detailed atmospheric data under the full chemistry option.  Full Chemistry Challenges  ‐ While the full chemistry capabilities within SCICHEM and CMAQ may allow for greater fidelity, the increased degree of control can also make the use of these capabilities more difficult as it requires a deeper understanding of the impacts caused by each input parameter. Although region-specific, default background concentration data is available for the full chemistry option, modifying this default data for greater accuracy may be burdensome due to obtaining the data as well as understanding the impacts on modeled NO2.  Plume  Growth  &  Multiple  Plumes  ‐ Plume growth and accounting for changes in O3 concentrations within evolving plumes, as well as the ability to model the interactions of multiple plumes, are effects that improve the overall NO2 predictions. Methods (e.g., OLM) that do not explicitly model these effects may rely on assumptions that the impacts are small. While not explicitly taking into account these effects may not produce inaccurate results, it reduces the capability of the method. Regarding OLM, the methodology as implemented in AERMOD does allow the use of a grouping option where sources with overlapping plumes can be merged (grouped) so that the aggregate NO in the merged plume competes for available O3. This generally results in lower, potentially more accurate NO2 concentrations.  User‐supplied ambient O3 concentrations ‐ Although ARM2 also does not explicitly model the effects of evolving plumes, the regressed NO2 prediction equation inherently takes into account these effects. This is also true for ambient O3 concentrations. Although ARM2 does not allow

A ‐ 73  user-specified O3 concentrations, the regression methodology inherently accounts for background conditions.  Plume segmentation ‐ Plume segmentation and plume region modeling is similar to accounting for changes in plume characteristics (e.g., plume growth, O3 concentrations, etc.), but may be seen as further refinement. This allows for more explicitly modeling of evolving chemical characteristics of plumes and the conversions that occur at different locations along the plume.  User‐supplied NO2/NOx ratios by source and operational mode – All of the methods except for the full conversion and the method used in CALINE4 allow for the use of exhaust NO2/NOx ratios by source mode of operation. For those that do allow this, it is necessarily tied to the encompassing model and therefore reflects the flexibility of each model to allow modeling by mode.  Use  of  monitoring  data  to  improve  background  air  characteristics  – This criteria refers to improving the overall fidelity of the modeling work through the use of “additional” air data. With the exception of some meteorological data in OLM and PVMRM, the methods in AERMOD are mainly focused on using a NO2/NOx and O3 concentration data. In contrast, the other methods can potentially use other meteorological data and background pollutant concentrations to improve the modeling work, providing more robust modeling environments.  Feasibility  of  implementation  within  AEDT/AERMOD  – Due to various reasons, the non- AERMOD methods would be difficult or impossible (at least not without alterations to the methods) to implement or work with AERMOD. Reasons for this include CALINE4’s reliance on the use of mixing zones, the time-varying framework used by puff models, and the Eulerian gridding scheme employed in CMAQ. In general, this can be summarized as an incompatibility of methods and modeling frameworks – an important consideration when selecting methods for potential implementation within AERMOD.  Source‐specific modeling  – Most of the models allow for the flexibility in modeling specific, individual sources except for CALINE4 (models sources as part of line sources) and CMAQ (models sources as gridded areas although plume-in-grid is available to model some specific sources). This is an indication of the modeling framework rather than the capabilities of the methods as each method should be applicable to individual sources which could allow greater degrees of flexibility in modeling local air quality.  Modeling  (Receptor  &  Source  Limitations)  –  In considering the implementation of methods within models, it should be recognized that some models and modeling frameworks may have limitations, some of which are tied to computer resources and runtime issues. Although it can potentially be modified, CALPUFF currently has a limit of 200 sources, and CMAQ limits the number of receptors that can be used in plume-in-grid modeling. While it is not a reflection of the chemistry module’s capabilities, CALPUFF’s limitations on only allowing 200 sources makes it difficult to properly test the NO2 modeling capabilities. This impacts criteria such as computational complexity and source specifications of NO2/NOx ratios by mode of operation. In general, all of the core NOx conversion methods can take into account different source modes of operation to account for differing NO2/NOx ratios, but it depends on the current implementation within each model. Such limitations are part of the modeling framework rather than the methods and should be considered when considering methods for implementation in models.  Finest  time‐varying  concentrations  –  The time period (averaging period) for each modeled concentration depends on the combination of methods and models. But in general, the Gaussian plume (AERMOD) and Eulrian grid (e.g., CMAQ) models will be limited to a 1-hour timeframe whereas Gaussian puff models have the potential to generate concentrations at very fine time resolutions (e.g., 1-min, 1-sec, etc.). However, the usefulness and fidelity of such concentrations

    A ‐ 74    depends on whether the input source characteristics (e.g., emission factors, location, etc.), weather data, background concentrations, etc. also provide such resolution. The overall consideration is that finer time-varying frameworks may provide a more robust modeling environment.  AEDT/AERMOD  Compatibility  – An important consideration in assessing candidate models/methods for potential incorporation into EDMS/AEDT/AERMOD are their compatibilities. For example, CALINE4 is difficult to implement in other models because the method is dependent on the characteristics of the mixing zone around each line source. While the method could be modified to be more generic to allow for use with other dispersion methodologies, the current framework will not allow such use. The methods used in CALPUFF and SCICHEM are also difficult to implement within AERMOD due to a difference in the modeling framework. In particular, it would be very difficult to implement the time-varying methods in CALPUFF and SCICHEM into AERMOD’s static plume framework. While simplifications and assumptions can be made, it would require significant changes to the methods. The above limitations are also applicable for the method in CMAQ which uses an Eulerian grid framework. However, since CMAQ can also employ a plume-in-grid (PinG) module, the implementation scheme used with the PinG can potentially be adapted for use in AERMOD, but would still require modifications to work with the plume-only AERMOD framework. It should be clarified that although AERMOD is characterized as a static model, it can still produce time- varying results, but on a much coarser level (e.g., 1-hr) than the puff models (CALPUFF and SCICHEM) which can produce results at much finer levels (e.g., 1-sec). 6.2.4 Quantitative Assessment  This section provides a quantitative assessment of the existing models and methods that are used to compute NO2 concentrations. Again, the emphasis is on the potential application of these models/methods using EDMS/AEDT to predict NO2 levels in the vicinities of airports. 6.2.4.1 Approach  For the purposes of this assessment, “modeled-to-modeled” comparisons were made using the AERMOD methods (i.e., Tiers 1, 2, 3) as a basis for comparison since the AERMOD methods were previously evaluated as part of the modeled-to-measured assessments conducted in Support of Task 5. As such, this section provides quantitative evaluations of the following methods/models:  AERMOD (Full Conversion)  AERMOD (ARM/ARM2)  AERMOD (OLM)  AERMOD (PVMRM)  CALPUFF (with CALMET)  SCICHEM (full chemistry)  SCICHEM (simple NOx conversion) CALINE4 was not included in the quantitative assessment as its reliance on the highway source mixing zones would have made it difficult to accurately model airport sources. CMAQ was also not included in part because of the grid-based nature of the model which also would have made comparisons difficult. Moreover, full chemistry modeling in CMAQ through the CB mechanism is replicated in SCICHEM with

reportedly options w 6.2.4.2 M For this an airport co The opera  T  O  A  W  W From EDM resulting OLM, and modeling the defaul ambient e For the ai types wer weighted 69 Kelly, 6 July similar resu ere exercised odeling Scen alysis, an ai nsisting of tw Figu tional and m ime Periods: perations: 20 ircraft: Mixtu ind speed: 2 ind direction S, the AER in area sourc PVMRM m these method t ambient NO quilibrium N rcraft engine e correlated ratios by mod James T. and K 1999.” EPA Off lts.69 For S . arios and Lim rport layout w o runways, s re 6.2. Simp eteorological Morning (7-8 departure an re of small ( .0 m/s : 0 deg. MOD input es with emi ethods were s. For the A 2/NOx max/ O2/NOx ratio exhaust (i.e to a large s e: irk R. Baker. “P ice of Air Quali CICHEM, b itations  as devised a everal taxiwa lified LAX Air data used for am) and afte d arrival flig e.g., ER145) files (.inp file ssion rates. set manually RM option, t min limits of of 0.9 was us ., in-stack) N et of North lume Chemistry ty Planning & S A ‐ 75  oth the full nd input into ys, and an ai port Layout a the analysis rnoon (3-4 p hts per hour and larger (e s) and the ho The AERMO outside of ED he default am 0.9/0.2 were ed. O2/NOx ratio American f Modeling with tandards. 10th C chemistry an EDMS base rcraft gate ap s Represent are listed as m) (January .g., B757) jet urly emissio D specifica MS as EDM bient NO2/N used. For O s, a set of ra light operati SCICHEM and onference on Ai d the simpl d on a simpli ron as illustra ed in EDMS.  follows: 1, 2010) s n files (.hre tions to run S does not p Ox ratio of 0 LM and PVM tios for diffe ons resulting CMAQ: Cumb r Quality Mode e NOx conve fied setup of ted in Figure files) were cr the ARM/A rovide optio .8 and for A RM, the de rent sized ai in the follo erland Power P ling. March 14, rsion LAX  6.2. eated RM2, ns for RM2, fault, rcraft wing lant on 2012.

    A ‐ 76     Idle: 0.768  Approach: 0.147  Climbout: 0.070  Takeoff: 0.061 The mode of each area source was identified based on the placement (runway or taxiway) and height of the each source. For simplicity, the ambient O3 concentrations were kept constant at the following nominal values for all methods/models:  7-8 am: 30 ppb  3-4 pm: 60 ppb Both AERMOD and SCICHEM allowed the use and allocation of different exhaust NO2/NOx ratios by source, allowing a single run of all sources to account for these differences. But for CALPUFF, a run was made for each mode (i.e., idle, approach, etc.) and the resulting four sets of concentrations were added at each receptor. This was necessary since CALPUFF currently has a limit of 200 sources. Although this prevents the interaction of pollutants from different modes, it allowed the inclusion of mode-specific ratios. It may have been possible to recompile the CALPUFF Fortran code to bypass the 200-source limit, but this was not accomplished due to the difficulty of recompiling a specific compiler (or development environment). Since the CALPUFF area sources were fragmented, the test did not allow the interaction of puffs from one run to those of the other three. As such, the modeled results from CALPUFF are approximations of what may be generated by the underlying method if the limitation did not exist. Unlike the area sources used in AERMOD and CALPUFF, point sources were used in SCICHEM because the current version of SCICHEM does not allow the use of the hourly emission file from AERMOD and also does not allow multi-component pollutant modeling using area sources. Therefore, a point source was placed at the centroid of each area source and the emission rate was derived accordingly from the area source. The point source “stack” diameter was made equivalent to the width of the shortest side of the area source and a nominal value for exhaust temperature (366 K) was used to represent the average jet exhaust conditions. Finally, because the jet exhaust moves horizontally (no vertical momentum), the exhaust speed was set to zero so that plume rise only occurs from a temperature gradient (i.e., buoyancy). The receptor locations used for the analysis are shown in Figure 6.3 and were placed perpendicular to the runway/taxiway system and vertically aligned with a constant wind direction of 0 degrees. Receptors R1 through R8 were located close to the sources and in-between the two runways to capture any potential “near field” effects. R9 is positioned 800 ft directly south of Runway 25L; R10, 11 and 12 are incrementally further south by 1,600 ft; and the remaining receptors are placed 3,200 ft from each other - with the furthest receptor (R18) approximately 5 miles away. All receptors have “flag-pole” heights of 5.9 ft (1.8 m) above ground level (agl). In contrast, all of the sources have release heights of 126 ft (38.4 m) agl – the default, average height coded in EDMS/AEDT to account for aircraft engine height and plume rise.

6.2.4.3 Di The result  N co m th te S F th by A th di ch A co th scussion of R s of the quan O2  Concentr ncentrations ethod results e other meth nd to produ CICHEM me or the PM per e AERMOD the OLM ERMOD me e CALPUFF fferences in emistry. long with t ncentrations e total NOx Figure 6.3. R esults  titative asses ations  - Figu for the AM in the highe ods tested. N ce the high thods produc iod, CALPU full convers and PVMRM thods. The S drop-off f concentration he AERMOD for SCICHE accounted fo eceptor Loca sment discuss res  6.4  and and PM st st NO2 valu otably, for t est concentr ed the lowest FF produced ion, but with (very simi CICHEM m alls below t s near the s full conv M and CAL r at each rec A ‐ 77  tions Simplif ed above are 6.5 provide udy periods. es among the he AM perio ations follow . the highest c a sharp decl lar concentra ethods produ he SCICHEM ources are la ersion metho PUFF. These eptor. For th ied Airport S summarized modeled-to- As expected AERMOD- d, the OLM ed by CAL oncentration ine at further tions for bo ced the lowe results st rgely due to d, the figu illustrate th e morning h cenario.  as follows: modeled com , the ARM based metho , PVMRM, a PUFF and at Receptor 8 distances. T th), ARM an st concentra arting at Re dispersion res also sho e amount of our, the PV parisons of “full conver ds but also a nd ARM me ARM2 whil , even highe hese are foll d ARM2 fo tions except ceptor 10. T effects rather w the total NO2 in relati MRM, OLM NO2 sion” mong thods e the r than owed r the when hese than NOx on to , and

    A ‐ 78    ARM indicate most of the NOx is in the form of NO2 while the other methods generally indicate a much smaller portion being due to NO2 for the receptors that are closer to the source. For the afternoon hour, the NO2 from the AERMOD methods account for a relatively large proportion of NOx similar to the morning hour. This is true even though the NO2 levels appear to be generally lower than the morning concentrations. The afternoon NO2 levels for SCICHEM and CALPUFF appear to be proportionally similar to those of the morning hour, but the NO2 levels appear to fall off quicker in the afternoon.            Figure 6.4. Modeled‐to‐Modeled Comparisons of NO2 Concentrations (AM Period).  0 50 100 150 200 250 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 N O 2 Co nc en tr at io n (p pb ) Receptor AERMOD‐PVMRM AERMOD‐OLM AERMOD‐ARM AERMOD‐ARM2 SCICHEM‐Full Chem SCICHEM‐Simple CALPUFF AERMOD‐Full Conversion SCICHEM‐Full Chem, Total NOx SCICHEM‐Simple, Total NOx CALPUFF, Total NOx

A ‐ 79  Figure 6.5.  Modeled‐to‐Modeled Comparisons of NO2 Concentrations (PM Period).   NO2/NOx Ratios - Figures 6.6 and 6.7 present the ambient NO2/NOx ratios for R9 through 18. The other receptors (i.e., 1 through 8) are not shown in order to more clearly show the ratio with distance away from the source. Also, the ratios were plotted against receptors rather than distance because since there are many sources (i.e., runway segments, taxiway segments, etc.), it is difficult to define a single distance for each receptor (i.e., distance from the sources). The convergence of the plots at a ratio of 0.9 for the AERMOD methods is due to the fact that this is the default equilibrium ratio that was used. Figure 6.6. Modeled Ambient NO2/NOx Ratio (AM Period).  0 0.2 0.4 0.6 0.8 1 1.2 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 M od el ed  Am bi en t N O 2/ N O x Ra tio Receptor AERMOD‐OLM AERMOD‐PVMRM AERMOD‐ARM AERMOD‐ARM2 SCICHEM‐Full Chem SCICHEM‐Simple CALPUFF 0 50 100 150 200 250 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 N O 2 Co nc en tr at io n ( pp b) Receptor AERMOD‐PVMRM AERMOD‐OLM AERMOD‐ARM AERMOD‐ARM2 SCICHEM‐Full Chem SCICHEM‐Simple CALPUFF AERMOD‐Full Conversion SCICHEM‐Full Chem, Total NOx SCICHEM‐Simple, Total NOx CALPUFF, Total NOx

    A ‐ 80    Figure 6.7.  Modeled Ambient NO2/NOx Ratios (PM Period).  For ease of review, Table 6.3 provides a direct comparison of average modeled-to-modeled ratios of (NO2 to NO2 ratios) for each method. Similar to the NO2/NOx ratios, only Receptors 9 through 18 were included for clarity to focus on concentrations away from the sources. PVMRM was used as the basis for comparison as the literature currently appears to promote PVMRM over other methods (e.g., OLM) in terms of methodology and accuracy (MACTEC 2004 and Hendrick 2012). However, any of the methods could have been used as the basis for comparisons since these are all relative comparisons. Table 6.3. Average Ratio of Each Method to PVMRM Concentrations    OLM Source Grouping - Figures 6.8 and 6.9 illustrate the differences in using the OLM with and without the source grouping option (i.e., “OLMGROUP”) within AERMOD. In concept, when two or more sources are grouped due to expected merging of plumes, this option allows for more accurate titration of NO by O3 to form NO2 by allowing the NO within the mixture (from Average Ratio (PVMRM as Basis) Method/Model  AM Period  PM Period  AERMOD‐PVMRM  1 1 AERMOD‐OLM  0.96 1 AERMOD‐ARM  1.02 0.89 AERMOD‐ARM2  0.74 0.97 SCICHEM‐Full Chem  0.42 0.20 SCICHEM‐Simple  0.37 0.21 CALPUFF  0.12 0.13 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 M od el ed  Am bi en t N O 2/ N O x Ra tio Receptor AERMOD‐OLM AERMOD‐PVMRM AERMOD‐ARM AERMOD‐ARM2 SCICHEM‐Full Chem SCICHEM‐Simple CALPUFF

A ‐ 81  different sources) to compete for the available O3. As shown, the results reveal significant differences at receptor locations close to the southern-most runway, and the concentrations tend to converge with distance (further south) of the runway. Figure 6.8.  OLM versus OLM Grouped Source NO2 Predictions (AM Period).  Figure 6.9.  OLM versus OLM Grouped Source NO2 Predictions (PM Period).  0 50 100 150 200 250 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 N O 2 Co nc en tr at io n ( pp b) Receptor AERMOD‐Full Conversion AERMOD‐OLM AERMOD‐OLM w/OLMGROUP 0 20 40 60 80 100 120 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 N O 2 Co nc en tr at io n ( pp b) Receptor AERMOD‐Full Conversion AERMOD‐OLM AERMOD‐OLM w/OLMGROUP

    A ‐ 82    6.2.4.4 Key Findings  A number of important findings are evident from the quantitative assessment and comparison of the existing models/methods that are available for computing airport-related NO2 concentrations. The most relevant of these are discussed below:  Distance Versus Concentrations – In general, the greatest differences between the model results are observed at closer distances to the sources (e.g., runways and taxiways). At distances of about 2 – 2.5 miles from the closest runway, the differences in NO2 concentrations start to become far less. This is indicative of how the models treat near field dispersion and chemistry differently and how these differences decrease with receptor distance. For the scenario modeled, the receptors (e.g., 1, 2, 3, etc.) that are further upwind of the lower runway showed relatively low concentrations, in part because they did not experience the plumes from the lower runway, but also because they were too close to some of the other sources (e.g., the other runway and taxiways). Since the height of the receptors (5.9 ft) above ground level is much lower than the height of the source release points (126 ft), receptors too close to the sources may not experience much of the plumes. Also, the NO2/NOx ratios are low in the early part of the plumes generated by jet aircraft, and can result in relatively low NO2 concentrations at close receptors. For both the morning and afternoon hour, the concentrations for these close receptors appear to noticeably start increasing at about Receptor 5 which is approximately 300 ft south of the northern runway.  Morning versus Afternoon Concentrations – The generally lower concentrations of NO2 in the afternoon hour is largely due to the lower overall NOx emissions during that hour (as compared to the morning hour). The higher ambient O3 concentrations in the afternoon hour (60 ppb versus 30 ppb for the morning) is reflected in the generally higher NO2/NOx ratios for the afternoon. The equilibrium level is quickly reached at about the Receptor 11 location. Also, the higher O3 concentration in the afternoon hour appear to, in part, result in less diversity between the different methods.  Confirmed CALPUFF and SCICHEM are Less Conservative – The NO2 concentration plots show that the PVMRM, OLM, and ARM are more conservative than the other methods. This is in-line with past studies that have indicated the conservative nature of these AERMOD methods.70 The lower concentrations generated by the more refined methods in CALPUFF and SCICHEM support these findings as well.71, 72, 73, 74 SCICHEM produced significantly lower results while CALPUFF was the lowest. As these ratios include results from Receptors 9 through 18, they will reflect the larger differences from the receptors (e.g., 9) that are closer to the sources and potential effects of early plume characteristics. As such, they do not necessarily reflect the decreasing differences of more mature plumes.                                                              70 Environmental Protection Agency (EPA). Memorandum. Clarification on the Use of AERMOD Dispersion Modeling for Demonstrating Compliance with the NO2 National Ambient Air Quality Standard. EPA Office of Air Quality Planning and Standards. 2010. 71 Karamchandanib, Prakash, et al. “Application of SCICHEM-2012 for 1-Hour NO2 Concentration Assessments.” Presented at the 11th Annual CMAS Conference, Chapel Hill, NC. October 2012. 72 Karamchandani, P., R. Morris, B. Brashers, G. Yarwood, L. Parker, E. Knipping, N. Kumar, B. Chowdhury, and I. Sykes. “Application of SCICHEM for Near-Field and Far-Field Single Source Impacts.” Presented at the AWMA Guideline on Air Quality Models: The Path Forward Conference, Raleigh, NC, March 2013. 73 Chowdhury, B., I. Sykes, D. Henn, N. Kumar, E. Knipping, and P. Karamchandani. “Comparative Studies Using SCICHEM.” Presented at the AWMA Guideline on Air Quality Models: The Path Forward Conference, Raleigh, NC, March 2013. 74 Environ. “Evaluation of Chemical Dispersion Models using Atmospheric Plume Measurements from Field Experiments.” Prepared for EPA Office of Air Quality Planning and Standards. EPA Contract No. EP-D-07-102. September 2012.

A ‐ 83   ARM2  is Less Conservative  than other AERMOD Methods Close  to Sources – The differences are shown by the average modeled-to-modeled ratios of NO2/NO2 which indicate that PVMRM, OLM, and ARM are comparable whereas ARM2 appears to be a little less conservative than these methods. The ARM2 concentrations were noticeably lower than the other AERMOD methods for receptors close to the runways. This appears to reflect the origins of the ARM2 methodology which is based on regressing data from EPA monitoring sites representative of background levels. Although past studies have indicated that ARM2 may be more conservative than PVMRM and OLM that understanding has mainly been based on comparing percentile-type values rather than concentrations paired in space and time as well as sensitivity studies.75, 76, 77 However, the ARM2 method shows similar concentration levels at receptor locations further away from the runways (further south).  Plume Merging/Grouping – Exercising the OLM grouping options showed consistency with the understanding that the grouping of sources (plume merging) results in less O3 being available and lower NO2 concentrations. Similar to the other methods, the predictions from the OLM and OLM-grouped option tend to merge quickly with distance from the source, especially when more O3 is available (i.e., as indicated by the afternoon hour).  Comparison  of Dispersion without  Chemistry – The “Total NOx” results for CALPUFF and SCICHEM are strictly based on the effects of atmospheric dispersion (no chemistry). The comparisons for the morning hour indicate that total NOx concentrations generated by AERMOD (full conversion) are similar in maximum concentration to those generated by SCICHEM and CALPUFF. This appears to indicate that in terms of near field peaks, the dispersion modeling capabilities of each model is comparable for the scenario modeled. However, for the afternoon hour, the AERMOD total NOx is noticeably different than those of the other models. The difference is likely due to the formulations between plume and puff models (i.e., the CALPUFF and SCICHEM dispersion show similar characteristics during the afternoon hour) and the atmospheric characteristics of afternoon hour. This is also evident by how the concentrations predicted by the puff models fall off noticeably quicker than AERMOD with distance.  Full Chemistry versus Simplified NOx in SCICHEM – The full chemistry (CB mechanism) and the simplified NOx conversion methods in SCICHEM produced very similar results. This indicates that at least for the airport scenario that was modeled including the spatial domain (e.g., up to 5 miles), the simplified method appears to be adequate.  Impact  of  CALPUFF  Modeling  Limitations – The resulting CALPUFF NO2 predictions are noticeably different than those from other methods/models. In addition to the aforementioned differences in model formulations, these differences are also likely due to the use of four separate runs (one for each aircraft mode) in order to accommodate the 200-source limitation. As such, interactions of the puffs across the runs were not possible. The quicker drop-off of NO2 concentrations indicate potentially greater degree of dispersion and sensitivity to distance. 75 MACTEC. Sensitivity Analysis of PVMRM and OLM in AERMOD. Final Report, Alaska DEC Contract No. 18-8018-04. MACTEC Federal Programs, Inc., Research Triangle Park, NC. 2004. 76 RTP Envrionmental Associates, Inc. Ambient Ratio Method Version 2 (ARM2) for use with AERMOD for 1-hr NO2 Modeling, Development and Evaluation Report. Prepared for the American Petroleum Institute. September 20, 2013. 77 Environmental Protection Agency (EPA). Memorandum. Clarification on the Use of AERMOD Dispersion Modeling for Demonstrating Compliance with the NO2 National Ambient Air Quality Standard. EPA Office of Air Quality Planning and Standards. 2010.

    A ‐ 84    6.3 Research Plan Evaluation Methods & Criteria  This section provides the Research Team’s recommended approach to evaluating and developing an alternative method(s) for predicting NO2 concentrations from airport-related emission sources and recommending a Preferred Method. As discussed above, this Research Plan will be completed under Task 10 and it is expected that the ACRP 02-43 Panel will provide comments on this approach before it is implemented. 6.3.1 Target Characteristics  Based upon the outcomes and findings derived from the qualitative and quantitative assessments of the existing models/methods, the overall approach and evaluation criteria for developing an alternative(s), should encompass several “key” targets. These attributes are identified and described as follows:  Reasonable  Data  Requirements  - Reasonable input data requirements address two important factors in modeling: (i.) it allows for easier use of the model by modelers and (ii.) thus more frequent and consistent use of the method. It should be noted that simpler data requirements does not necessarily imply a corresponding lack of accuracy.  Computational Requirements - Preferably, the model/method should be comparatively fast so that computer requirements and runtimes are appropriate by most users. In general, dispersion modeling can require extensive computer resources which are magnified when modeling airport- related sources which are comprised of numerous (i.e., >1,000) individual sources.  Reasonable  Accuracy - The model/method should be reasonably accurate and comparable to other available methods. For dispersion modeling, reasonable accuracy is generally considered to be able to be accurately predict concentrations to within a factor of two. As the data in Section 6 shows, it is very difficult for dispersion models to show close agreement when comparing modeled-to-measured results.  Technically Defensible Methodology - The overall methodology should be technically defensible and supported with back-up data demonstrating its applicability and accuracy.  Compatible with EDMS/AEDT/AERMOD - Because the FAA follows the U.S. EPA’s strategy of using AERMOD for most dispersion modeling needs, the method should be either adoptable within, or work in conjunction with, AERMOD. This is particularly relevant to use of AERMOD in the EDMS/AEDT models. 6.3.2 Alternative Method Candidate  Again, based upon the findings derived from the qualitative and quantitative assessments of the existing models/methods combined with the outcomes of Tasks 2, 3 and 6, the Research Team initially recommends that a new method be developed using the ARM2 methodology as the basis. In short, this alternative offers an appropriate balance of data and computational requirements, reasonably accuracy and compatibility with EDMS/AEDT/AERMOD framework. Although past studies involving stationary sources have indicated that ARM2 tends to produce results similar or more conservative than PVMRM and OLM (RTP 2013 and EPA 2010), the quantitative comparisons conducted in support of the ACRP 02-43 research with aircraft sources have shown ARM2 producing NO2 concentrations lower than that of PVMRM and OLM (and comparable to OLM-grouped concentrations) for receptors relatively close to the sources. For receptors further away, ARM2 performs similar to PVMRM and OLM. This provides some confidence that when modeling aircraft sources,

A ‐ 85  ARM2 performs similar to the Tier 3 methods and for close receptor locations, may perform similar to the more refined CALPUFF and SCICHEM methods. Furthermore, while the methods in CALPUFF and SCICHEM may be more detailed and scientifically acceptable than ARM2, the CALPUFF and SCICHEM methods are closely tied (dependent) on the time- varying Gaussian puff framework. It would be very difficult (i.e., forcing simplifications causing potential inaccuracies) to implement the methods within AERMOD’s static plume environment. Simpler methods such as OLM and PVMRM would require modifications to the AERMOD software code and recompiling to create a new executable version. Although ARM2 is also currently coded within Using AERMOD, a new method based on ARM2’s regression methodology could be developed as an independent module such that AERMOD would not need to be modified. While ARM2 only directly accounts for NOx concentrations as the dependent variable, the methodology could be expanded to potentially include parameters such as the following:  Atmospheric conditions (e.g., temperature, humidity, stability, etc.)  Ambient O3 concentrations  Exhaust NO2/NOx ratios A multivariable regression model developed using airport-specific data may also help make the ARM2 method more appropriate and specific to airport applications (e.g., “ARM2-Airports” method). Depending on the assessment of the available datasets, the resulting regression strategy could potentially use weighted average input parameters, categories based on receptor locations, types of sources and operations, etc. While the regression work would need to be further evaluated, the overall goal is to include additional parameters to make the regression more airport-specific. It is further recommended that the ARM2-Airports method be developed from regressing the airport monitoring data used previously to evaluate (e.g., compare modeled-to-measured) the AERMOD methods. Depending on the viability of the datasets, approximately 75% of the data can be used for the regression work. This should allow thousands of data points to provide good statistical analyses. The remaining data could be augmented with additional collected from the ACRP Project 02-08 Airport Air Quality Contributions Project, Report 71 (Kim 2012) to help validate the new method. The following steps would also need to be undertaken in the development of the new method:  Review available monitoring data to identify available parameters.  Design preliminary regression scheme with specifications for dependent variables.  Process monitoring data to get it into a suitable form for statistical evaluations.  Group the data according to the regression scheme.  Conduct the regression work.  Specify the final regressed equations with rules and categories as necessary. 6.3.3 Evaluation Plan  Once the new method is developed, the above-mentioned airport datasets will be used to evaluate the method. The evaluations will be conducted to establish an initial understanding of the applicability and accuracy of the new method which will include sensitivity analyses as well as model validation-type assessments.

   The sensi variable ( regression  S  Id in  R de  Pr id The plan evaluation measured good corr possible i statistics sense. As PVMRM) Specifical that are p will be vis  R  S  y- tivity analyse i.e., NO2/NO will be inclu et up a suitab entify the ma crement. un the meth velop a matr ocess and an entify the mo for the va s already co comparisons elations betw n an absolut among differ such, in addi will also be ly, the mode aired in spac ualized using 2 lope Intercept Figure    s will involv x ratio). It is ded in the an le airport sce x and min ra od while var ix of output r alyze the re st impactful lidation-type nducted und that are pai een modeled e sense (e.g ent methods tion to the ne exercised for led-to-measu e and time. T scatter plots 6.10. Compa e varying on expected tha alyses. More nario (likely nge for each ying one pa esponses. sults to ident variables). assessments er Task 6. red in space and measur ., low R2 va should be ab w method, th the same sce red comparis he following with a 1:1 a rison Scheme A ‐ 86  e parameter a t each of the specifically, the same as th independent rameter at a ify the relati is consiste The assessm and time. W ed concentra lues are exp le to show h e AERMOD narios. ons will be c statistics wi greement line          for Modele t a time to id dependent v the work wil at use for th parameter as time while ve contributi nt (e.g., sam ents will in hile past stud tions (or NO ected), comp ow the new methods (fu onducted us ll be used to (illustrated i Bias Mean Squ d‐to‐Measur entify its im ariables inclu l involve the e validation w well as a su holding all ons from ea e statistics) volve the us ies have cle 2/NOx ratios) arisons of t method perf ll conversion ing 1-hour N assess mode n Figure 6.10 ared Error (M ed Evaluatio pact on the o ded as part following ste ork). itable perturb others consta ch parameter with the m e of modele arly indicate will likely n he goodness- orms in a re , ARM, OLM O2 concentra l performanc ): SE) ns.  utput of the ps: ation nt to (i.e., odel d-to- d that ot be of-fit lative , and tions e and

A ‐ 87  In addition to the paired modeled-versus-measured comparisons, various other statistics will be used including those representing non-paired datasets, such as:  Summary Statistics - Summary statistics are the typical statistics used to describe the population of concentrations such as the mean, standard deviation, min, max (highest), 2nd highest, etc. Since the statistics are based on the population as a whole, no pairing is involved. These simple statistics serve as a starting point for understanding the distribution of each population, providing indications of model performance on concentration ranges.  Quantile‐Quantile  (Q‐Q)  Plots - Q-Q plots are used to compare distributions of modeled and monitored datasets with no pairing. A quantile is essentially a fraction (or percent) of data points that fall below a specified value. As such, a Q-Q plot represents a comparison of two frequency distributions. These plots provide an indication of the overall performance of the models in predicting concentration ranges based on the scenario(s) modeled.  Robust High Concentration (RHC) - The RHC is an aggregation statistic representing the highest concentrations from modeled or measured datasets. Similar to a geometric mean, the RHC helps to smooth out the effects of extreme values. RHC is calculated through a tail exponential fit to the high end of the frequency distribution.  Scatter Plots of NO2/NOx Ratios Versus NOx Concentrations - Scatter plots of NO2/NOx ratios versus total NOx concentrations will also be developed to further understand model behavior and sensitivities to NOx concentrations. In addition to comparisons between the different methods, the plots will also provide an indication of the impacts of including multiple variables in the new method (i.e., beyond the ARM2 functional dependence on just NOx concentrations). 6.3.4 Plans for Implementation within AEDT  6.3.4.1 Module Development  The FAA’s AEDT is built on a modular architecture where discreet computational modules work in conjunction with a central database to predict airport noise and air quality impacts. As such, any new methods with the potential to work within this system must be developed as a self-contained module that can be accessed by the system taskmaster software. Figure 6.11 provides an overview of how the new method could be used in conjunction with AERMOD to predict NO2 concentrations. As indicated, the new method would be a separate computational module that uses inputs from AERMOD (i.e., NOx concentration) and the user (e.g., O3 concentrations) to predict NO2 concentrations. As this is a conceptual diagram, there would be various data-handling and module management software (e.g., taskmaster, data handler, etc.) that would control the flow of data. In essence, the new module would be in the form of a computational library.

   In accord managed (VS.NET) module w programm 6.3.4.2 AE The docu computati  A m co (o  In do of ex Figure 6.11 ance with FA .NET Dyna developmen ith clear spe ing language DT Module D mentation re onal modules lgorithm Des ethodology i ncepts the i nly develope terface  Con cument is sp the module ample shown   . High‐Level  A’s AEDT mic Link L t environme cifications fo can be used ocumentati quirements : ign Documen n a concise mplemented rs) and serve trol  Docume ecially geare ’s input para in Table 6.4 Data Flow Di architecture ibrary (DLL nt. Either Vi r input and o . on  for AEDT s t (ADD) – Ex manner so t calculation r s to also docu nt  (ICD) – d towards to meters and o . Similar to t A ‐ 88  agram of the requirements ) developed sual Basic.N utput data. A pecifies the plains the ba hat AEDT m outines. This ment the me Also intende software im utput results he ADD, the  Airport‐Spe , the new m under the ET or C#.NE s long as th need for tw sic theory, a odel develo document is thodology ve d only for A plementation . This is acco ICD also has cific ARM2 M odule must b Microsoft V T will be us e product is o specific d lgorithms, eq pers can un not intende rsioning as w EDT mode . The docum mplished us strict versio odule.  e developed isual Studio ed to develo a .NET DLL ocuments fo uations, etc. derstand the d for model ell. l developers ent specifies ing tables lik ning control. as a .NET p the , any r all of the basic users , this each e the

A ‐ 89  Table 6.4. Example Input Parameter List  Parameter  Precision Type  Units  Min  Max  Description  Temp  Double Kelvin ----- ----- Temperature O3  Double ppb ----- ----- Ozone Concentration NOx  Double ppb ----- ----- State/condition NO2NOx  Double ----- ----- ----- Exhaust NO2/NOx ratio Each computational module is essentially a programmable object once it is referenced into the development environment (VS.NET). As such, the input and output parameters are the objects properties. If there are any object methods, those must be clearly specified as well using tables and figures as necessary. For complex modules with multiple methods, arrays, properties, etc., an object diagram is usually included in the ICD. Both of these documents will be developed and included as part of the documentation for the new method. They will be developed so that they can readily be implemented within the AEDT system.

    A ‐ 90    7.  Independent Technical Review Plan (Task 8)  This section provides an overview and summary discussion of the outcomes of Task 8. 7.1 Task Objective  The purpose of Task 8 is for the Research Team to prepare a plan for having an independent technical review (ITR) conducted of the ACRP 02-43 Tasks 5, 6, and 7 research approaches and findings. For ease of reference, these tasks are restated below:  Task 5, Compare Emission Factors to Measured Values – This work involved a comparison of NO/NO2 emissions factors used in EDMS/AEDT to available published measurements of NO and NO2 in aircraft engine exhaust at airports.  Task  6,  Compare Modeled & Measured NO2 Data – Under this task, a comparison was undertaken of modeled and measured NO2 data at a three “test-case” airports representing various fleets, activity levels, geography, and meteorological conditions.  Task 7, Research Plan – Based on information gathered in Tasks 1 through 6, the Research Team prepared a Research Plan to develop alternative method(s) for predicting NO2 concentrations from airport NOx emissions, to evaluate existing and alternative methods and recommend a Preferred Method. 7.2 ITR Panel Candidates  In order to accomplish this task, the Research Team has obtained expressions of interest from several individuals that have demonstrated skills and interests in advancing what is known about predicting the effects of airport-related emissions on air quality. These ITR candidates are listed in Table  7.1 and represent a unique combination of expertise and experience pertaining to air quality, in general, and in airport environs, in particular. Their specialized understanding of the pertinent issues should add greatly to the success of the research and the end-products of this project. Importantly, the ITR Panel will serve on a voluntary basis. The following provides a list of review panel candidates, along with a rationale for their consideration, submitted for review and approval.78   Table 7.1. Independent Technical Review Panel Candidates  Name and Organization  Qualifications  Dr. Prem Lobo, Center of Excellence for Aerospace Particulate Emissions Reduction Research, Missouri University of Science and Technology Extensive research experience on the development and use of aircraft engine emission factors, air quality monitoring and impacts of aviation on the environment. Dr. Sarav Arunachalam, Research Associate Professor, Center for Environmental Modeling for Policy Development, UNC-Chapel Hill Extensive research experience on the development and use of computer modeling of aviation air quality. Eric Lu, ENVIRON International Meteorologist with extensive experience on the development and application of NO2/NOx ratios and the use of EDMS/AEDT.                                                              78 At the request of the Project Panel, five Independent Technical Review candidates are offered. ITR Plan  The  ITR  Plan  is  designed  and  in  a  way  that  is  intended  to achieve  the  objectives  and  deliver  meaningful  results  while  making  it  as  easy  as  possible for the ITR Panel to provide  their feedback.    It  should  also  be  noted  that  the  implementation  of  the  ITR  is  to  be  accomplished  as  Task  11  of  the  ACRP 02‐43 Amplified Work Plan. 

A ‐ 91  John Pehrson, CDM Smith, Irvine, CA Engineer with extensive experience on the development and application of NO2/NOx ratios and the use of EDMS/AEDT. Vincent Tito, Epsilon Associates, Maynard, MA Meteorologist with extensive experience on the development and application of NO2/NOx ratios and the use of EDMS/AEDT. Upon approval by the ACRP 02-43 Panel, these individuals will be re-contacted to confirm their interest and availability to support the ITR and to provide them with the relevant materials for this review in accordance with the ITR Implementation Plan described below. 7.3 ITR Implementation Plan  This section describes the Research Team’s four-step Implementation Plan for completing the ITR.  Obtain Panel Approval  –  This step involves obtaining the ACRP 02-43 Panel’s feedback and approval for enlisting the ITR candidates identified above. This will be accomplished at the Interim Report Meeting. If recommended, alternative candidates will be considered.  Refine  ITR  Feedback  Template  – Again, this step involves obtaining the ACRP 02-43 Panel’s feedback and approval on the ITR Feedback Template discussed below. This too will be accomplished at the Interim Report Meeting and alternative formats/ questions will be considered.  Conduct  the  ITR  –  This step will be undertaken and completed as Task 11 of the Amplified Work Plan. This includes the preparation of the ITR package, distribution to the ITR Panel and obtaining the feedback.    Adopt the ITR Recommendations – This step will involve the review of the ITR feedback and the implementation of the recommendations wherever appropriate and feasible. To the extent necessary, coordination will be undertaken with the ACRP 02-43 Panel on any ITR recommendations that require their input. The ITR Implementation Plan can be initiated immediately following the Interim Report Meeting and can be completed in one month’s time. 7.4 ITR Feedback Template  Because the extensive volume of materials collected and developed in support of Tasks 5, 6 and 7, the ITR of this information and data should be as comprehensive, yet streamlined, as possible. It is also important that the ITR Panel provide their feedback to the Research Team in a clear and efficient manner. In this way, the objectives and benefits of the ITR are more likely to be realized. For these reasons, the Research Team has developed, and is proposing to use, an ITR Feedback Template as a means of aiding the ITR Panel in providing their input. A draft version of this template is provided below as Table7.2. As shown, the form is designed to solicit feedback from the Panel on both the overall accomplishments of the research as well as on the specific elements of each task. The form will be provided to the Panel members in an MSWord format for ease of input, editing and expanding, if necessary. The electronic format will also aid the Research Team in organizing and assimilating the feedback. ITR Materials  The  ITR Panel will be provided with  a  consolidated  and  condensed  version  of  relevant  materials  from  the Interim Report, the ITR Response  Form,  a  set  of  instructions  and  a  timetable  for  submitting  their  feedback.    

    A ‐ 92    Table 7.2. ACRP 02‐43 Independent Technical Review Response Form Template  ACRP 02‐43 Independent Technical Review Response Form  Reviewer Information  Name: Affiliation: Contact Information: Date: General Feedback  Overall Summary  Provide comments highlighting the overall achievements and shortcomings of the research and its contribution to improving methods for predicting NO2 concentrations from airport NOx emissions. Response: Overall Progress   Provide feedback on the overall progress of the research (check one): Excellent progress (the Research has fully achieved its objectives and technical goals for the tasks).  Good progress (the Research has achieved most of its objectives and technical goals for the tasks with relatively minor deviations).  Unsatisfactory progress (the Research has failed to achieve critical objectives and/or is not on course).  Overall  Recommendations   Provide overall recommendations (e.g. overall modifications, corrective actions or re-tuning the objectives to optimize the remainder of the Research). Response: Task 5, Compare Emission Factors to Measured Values  This  task  involved  a  comparison  of  NO/NO2  emissions  factors  used  in  EDMS/AEDT  to  available  published measurements of NO and NO2 in aircraft engine exhaust at airports.   Overall Assessment  Has the research achieved the Task objective(s)? Yes Partially No Detailed Response: Compliments/Criticisms   Provide specific feedback on the reported findings pertaining to NO/NO2 emission factors contained in EDMS/AEDT versus other published data. Detailed Response: Recommendations  Provide specific guidance and/or recommendations on how the task findings can be expanded and/or approved: (e.g., data/information sources, alternative interpretations, etc.). Detailed Response:

A ‐ 93  Task 6, Compare Modeled & Measured NO2 Data  Under this task, a comparison was undertaken of modeled and measured NO2 data at a sample of airports  representing various fleets, activity levels, geography, and meteorological conditions.   Overall Assessment  Has the research achieved the Task objective(s)? Yes Partially No Detailed Response: Compliments/Criticisms   Provide specific feedback on the reported findings pertaining to modeled and measured NO2 data at the test-case airports. Detailed Response: Recommendations  Provide specific guidance and/or recommendations on how the task findings can be expanded and/or approved upon: (e.g., data/information sources, alternative interpretations, etc.). Detailed Response: Task 7, Research Plan  This task involved the preparation of a Research Plan to evaluate existing and alternative methods for  predicting NO2 concentrations from airport NOx and recommend a Preferred Method.   Overall Assessment  Has the research achieved the Task objective(s)? Yes Partially No Detailed Response: Compliments/Criticisms   Provide specific feedback on the adequacy of the Research Plan developed in Task 7 to evaluate existing and alternative methods for predicting NO2 concentrations from airport NOx and identify a Preferred Method. Detailed Response: Recommendations  Provide specific guidance and/or recommendations on how the Research Plan developed in Task 7 can be modified or improved upon to better achieve the research objectives. Detailed Response: Additional Comments

    A ‐ 94    As discussed above, the ACRP 02-43 Panel’s review comments on the ITR Feedback Form Template will be obtained at the Interim Report Meeting and any necessary modifications will be made before it is distributed to the ITR Panel.   8. Issues in Need of Resolution  The preceding sections of this Report have identified some issues associated with the ACRP 02-43 research conducted thus far that are in need of resolution. For example, as described in Section 5 poor agreement has been reached between the monitored and modeled NO2 values at the three “test-case” airports. In particular, high NOx concentrations have been predicted by EDMS/AEDT/AERMOD when compared to monitored values. Notably, the highest modeled concentrations occurred mainly under low wind speeds suggesting that improvements to the dispersion modeling itself or relevant input parameters are necessary. Potential deficiencies in other aspects of the modeling are also apparent. Some of the issues can be addressed by the Research Team while others should be discussed with the Project Panel at the Interim Report Meeting. These issues are identified and described below along with suggested remedies – wherever possible.  Issue No. 1: NO2/NOx Emission Ratio - The ARM, OLM and PVMRM NO2/NOx conversion methods resulted in two sets of NO2/NOx emission ratios: (i.) with all sources set to 0.5 and (ii.) with non-aircraft sources set to 0.5 but aircraft emissions set equal to those observed during the JETS-APEX2 and APEX3 studies. This may result in overly conservative (i.e., high) modeling results for non-aircraft sources. Proposed Resolution - Additional model runs should be conducted with the NO2/NOx emission ratio for non-aircraft sources set at a range of lower, more realistic, values. This would test the sensitivity of the model to this key input.  Issue No. 2: Thrust Settings - Total NOx emissions are sensitive to the application of reduced thrust takeoff since it reduces both the NOx emission index (g per kg of fuel) and the fuel flow rate (kg/s). Additionally, the NO2/NOx emission ratio is sensitive to the actual thrusts used – both at low and high thrust settings. Proposed Resolution  - Additional computed model runs with a range of thrust settings values should be conducted to determine the sensitivity of predicted NO2 and NOx concentrations to this input.  Issue No. 3: Dispersion Coefficients - The horizontal and vertical dispersion coefficients used in Gaussian dispersion modeling such as AERMOD are a function of atmospheric stability and wind speed. Proposed Resolution  - The sensitivity of the modeled NOx concentrations (ignoring the NO – NOx chemistry initially) to these values should be further investigated - especially under the low wind speed conditions.  Issue No. 4: Background Concentrations –Time-varying background concentrations of NO2 and NOx have not been fully applied to modeling results. Proposed  Resolution  - The method(s) of determining appropriate background NO2 and NOx concentrations needs to be further assessed. Issue Resolution  It is recommended that these Issues  in Need of Resolution be discussed  at  the  Interim Report Meeting with  the goal of achieving consensus on  how  they  can  be  remedied  within  the remaining schedule and funding  available  to  the  ACRP  02‐43  research.  

A ‐ 95   Issue No. 5: Source Apportionment - Although emissions from aircraft landing / takeoff cycles are the largest source of NOx emissions at each of the three test-case airports they are not necessarily the biggest contributor to predicted NOx and NO2 concentrations based on the modeling. Proposed Resolution – (None presently proposed). Again, it is recommended that these issues be discussed by the Research Team and the Project Panel at the Interim Report Meeting with the goal of achieving consensus on how they can be remedied within the remaining schedule and funding available to the ACRP 02-43 research.

    A ‐ 96    REFERENCES Chowdhury, B., I. Sykes, D. Henn, N. Kumar, E. Knipping, and P. Karamchandani. “Comparative Studies Using SCICHEM.” Presented at the AWMA Guideline on Air Quality Models: The Path Forward Conference, Raleigh, NC, March 2013. Dickerson, R.R.; Stedman, D.H.; Delany, A.C. “Direct measurements of O3 and nitrogen dioxide photolysis rates in the troposphere,” J. Geophys. Resh. 1982, 87(C7), 4933-4946. Environ. “Evaluation of Chemical Dispersion Models using Atmospheric Plume Measurements from Field Experiments.” Prepared for EPA Office of Air Quality Planning and Standards. EPA Contract No. EP-D-07-102. September 2012. Environmental Protection Agency (EPA). Memorandum. Clarification on the Use of AERMOD Dispersion Modeling for Demonstrating Compliance with the NO2 National Ambient Air Quality Standard. EPA Office of Air Quality Planning and Standards. 2010. Environmental Protection Agency (EPA). Memorandum. Clarification on the Use of AERMOD Dispersion Modeling for Demonstrating Compliance with the NO2 National Ambient Air Quality Standard. EPA Office of Air Quality Planning and Standards. 2010. EPA, Additional Clarification Regarding Application of Appendix W Modeling Guidance for the 1-Hour NO2 National Ambient Air Quality Standard, March 1, 2011. http://www.epa.gov/ttn/scram/Additional_Clarifications_AppendixW_Hourly‐NO2‐NAAQS_FINAL_03‐ 01‐2011.pdf. EPA, Guidance Concerning the Implementation of the 1-hour NO2 NAAQS for the Prevention of Significant Deterioration Program, June 29, 2010. http://www.epa.gov/ttn/scram/ClarificationMemo_AppendixW_Hourly‐NO2‐NAAQS_FINAL_06‐28‐ 2010.pdf. Hamson, R.F., Jr.; Garvin, D. Reaction Rate and Photochemical Data for Atmospheric Chemistry 1977; NBS Technical Note 513; U.S. Department of Commerce: Washington, DC, 1978. Karamchandani, P., R. Morris, B. Brashers, G. Yarwood, L. Parker, E. Knipping, N. Kumar, B. Chowdhury, and I. Sykes. “Application of SCICHEM for Near-Field and Far-Field Single Source Impacts.” Presented at the AWMA Guideline on Air Quality Models: The Path Forward Conference, Raleigh, NC, March 2013. Karamchandanib, Prakash, et al. “Application of SCICHEM-2012 for 1-Hour NO2 Concentration Assessments.” Presented at the 11th Annual CMAS Conference, Chapel Hill, NC. October 2012. Kelly, James T. and Kirk R. Baker. “Plume Chemistry Modeling with SCICHEM and CMAQ: Cumberland Power Plant on 6 July 1999.” EPA Office of Air Quality Planning & Standards. 10th Conference on Air Quality Modeling. March 14, 2012. Kim, B., J. Rachami, D. Robinson, B. Robinette, K. Nakada, S. Arunachalam, N. Davis, B. H. Baek, U. Shankar, K. Talgo, D. Yang, A. F. Hanna, R. L. Wayson, G. Noel, S. S. Cliff, Y. Zhao, P. K. Hopke, and P. Kumar. ACRP Report 71: Guidance for Quantifying the Contribution of Airport Emissions to Local Air Quality. Transportation Research Board of the National Academies, Washington, D.C., 2012.

A ‐ 97  MACTEC. Sensitivity Analysis of PVMRM and OLM in AERMOD. Final Report, Alaska DEC Contract No. 18-8018-04. MACTEC Federal Programs, Inc., Research Triangle Park, NC. 2004. Martin, Anjoli. Verification of FAA’s Emissions and Dispersion Modeling System. 2006. RTP Environmental Associates, Inc. Ambient Ratio Method Version 2 (ARM2) for use with AERMOD for 1-hr NO2 Modeling: Development and Evaluation Report. Published September 20, 2013. Retrieved December 8, 2014. RTP Envrionmental Associates, Inc. Ambient Ratio Method Version 2 (ARM2) for use with AERMOD for 1-hr NO2 Modeling, Development and Evaluation Report. Prepared for the American Petroleum Institute. September 20, 2013. http://www.epa.gov/scram001/models/aermod/ARM2_Development_and_Evaluation_Report- September_20_2013.pdf. Zhang, L., Jacob, D. J., Yue, X., Downey, N. V., Wood, D. A., & Blewitt, D. (2014). Sources contributing to background surface ozone in the US Intermountain West. Atmospheric Chemistry and Physics, 14(11), 5295-5309.