Guidance for Quantifying the Contribution of Airport Emissions to Local Air Quality (2012)

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.

48 Findings, Recommendations, and Suggested Future Research This chapter provides findings and recommendations resulting from both the reviews of the various measurement and modeling techniques and the airport case study assessments presented in the appendices. Based on those findings and recommendations, this chapter suggests areas for future research that might be considered. 6.1 Summary of Findings Regarding Measurement Equipment and Data The following list summarizes the findings from the measurement campaigns and analysis of the measured data: • In general, all of the equipment used during the measurement campaigns was demonstrated to provide adequate measurement capabilities for either typical airport use (e.g., regulatory purposes) or for research projects. Appendixes A and D provide details on measurement issues and associated guidance. All of the proposed equipment worked adequately (i.e., met expectations) in collecting data. One of the lessons learned from the first campaign was to use on-site line power whenever possible. The use of battery power for short-term measurements is recommended when possible, since portable power may not be reliable and can be disrup- tive when it needs to be refuelled. Also, enclosures to protect the equipment are necessary. • All of the criteria pollutants were adequately measured and assessed. While NOx can be rela- tively high, it should be expected from the makeup of modern jet engine technology and the relatively low content of sulfur in jet fuel that CO and SOx concentrations at airports will be low (i.e., much lower than the NAAQS). • The concentrations of a few VOCs (including HAPs) were quantified, but most VOCs were largely undetectable using the current standard sampling methods from USEPA. Therefore, more complex equipment with greater sensitivity could be used to measure these pollutants but would require support such as line power and a shelter/enclosure. Due to the relatively small concentrations for VOC, the sampling equipment (Summa canisters and DNPH car- tridges) should be strategically placed to obtain the highest concentrations possible. Smaller concentrations may also be an indication of health impacts but this will need to be further studied with receptors appropriately placed closer to typical human activity areas. • Similar to emissions, the concentration correlations between formaldehyde and other VOCs appear to suggest that some general relationships could be established (perhaps with error bars) to relate formaldehyde concentrations to those of other VOCs. This assumes uniform mixing in the atmosphere among the various VOC species. • The concentrations of PM2.5 and PM10 were more prevalent compared with the other pollut- ants in relation to the NAAQS. However, the concentration levels of these pollutants were C h a p t e R 6

Findings, Recommendations, and Suggested Future Research 49 similar, indicating that, due to the much smaller size of PM from aircraft exhaust, virtually all of the mass is due to PM with an aerodynamic diameter smaller than 2.5 µm. • While Minivols allow cost-effect spatial coverage, their lower flow rates restrict their use to longer sample periods. It should be noted that even the USEPA reference methods must impact sufficient mass on the filters to be quantified accurately. This is not a serious issue for comparisons to the NAAQS given that 24-hour standards are available. The 24-hour sampling by the Minivols revealed adequate mass collection. • Based on the methodology developed under this project for determining background con- centrations, there appears to be significant variations in background concentrations on an hourly and daily basis. Comparisons with the USEPA AQS data also revealed that airport levels are lower than some background sites far away that may or may not be located next to a major traffic road or power plant. This seems to indicate the relatively low contribution of airport sources to local air quality. However, this may be specific to the airport and region studied under this project and will need to be corroborated with other airports. Overall, the important finding is that background concentrations need to be as specific in time and loca- tion as possible. • While a sufficient amount of background concentrations for some receptor locations was developed, the base site in particular was limited. As such, future projects should make more concerted efforts to strategically place receptors to obtain adequate upwind concentrations. This will depend on the resources available to a project and will involve juggling the needs to obtain higher concentrations (close to sources) and the availability of equipment and person- nel to cover multiple locations. • As with the determination of background concentrations, it would be helpful to co-locate more sampling equipment whenever resources permit so that comparisons can be made to better understand uncertainties in the data. Again, this will depend on resources available to each project. • As expected, the trend assessments showed that there appear to be correlations between con- centrations and seasons. For example, CO concentrations appeared to be higher in the winter and spring, while O3 concentrations were higher in the summer. This type of information can be helpful in better understanding air quality trends, possibly providing information for air quality mitigation strategies. • NO2/NOx concentration ratios were developed to better understand the conversion capa- bilities of the atmosphere. The assessments showed that these ratios can vary significantly by location, distance from source, and by time of day. Therefore, assessments to calculate these ratios should include the location and relatively large sets to obtain representative averages. More work will be necessary to further understand this ratio. • The overall findings seem to show some impacts of aircraft operations, and wind directions on concentrations showed some correlations that confirm our understanding that these fac- tors affect airport air quality. Although this is intuitive, further assessments can potentially be conducted to show better correlations. • To further evaluate atmospheric impacts, time-series comparisons were made to the Monin- Obukhov Length (L) and the Sensible Heat Flux (H), two indicators of atmospheric stability. While there did not seem to be a correlation with CO and NOx, the other pollutants (SOx, O3, PM10, and PM2.5) appeared to generally show an increase in concentrations under an unstable atmosphere, possibly indicating that the spread of the plume is increased, thus making more of the plume visible to receptors that are close to the ground. • Comparisons with background concentrations generally showed that CO concentrations gen- erated by the airport were very low and may not even be worth measuring in future projects (i.e., can just be modeled). • The assessments in Appendix D show that, although noise event data can be useful in identi- fying events, the data are limited by the fact that such data cannot be feasibly used to identify

50 Guidance for Quantifying the Contribution of airport emissions to Local air Quality specific aircraft types. As such, these data can only be used to help corroborate incidences of flights in a flight plan or schedule. The placement of sound level meters is critical since, ideally, only aircraft noise should be measured. The use of more than one meter (e.g., on opposite ends of a runway) would be helpful in corroborating aircraft operations. Also, it is recom- mended that, if possible, either videocameras alone or in combination with sound level meters be used to both record incidences and identify aircraft types. In any case, it is further recom- mended that a computer program be used to automate the processing of the data because the research team’s assessments have shown that it can be very tedious to process the data. 6.2 Summary of Findings Regarding EDMS/AERMOD Modeling Capabilities The following summarizes findings, including limitations/gaps from the various EDMS/ AERMOD assessments and improvements/recommendations that can either address the limita- tions/gaps or help to improve the modeling work: • The overall findings from the modeled-versus-measured assessments is that although there appears to be much room for improvement in accuracy, the combination of EDMS and AERMOD seems to be appropriate or adequate for modeling emissions and concentrations of the various criteria pollutants studied. It should be noted that although these assessments may have similar elements, the purpose (and the scope of this project) was not to conduct a formal validation study, but to evaluate the models’ outputs and capabilities to determine their appropriateness for assessing airport air quality contributions. With that understanding, our assessments have generally shown that while overall (for all receptor locations as a whole) the modeled and measured concentrations showed moderate to little agreement (e.g., low R2 values), the correlations were generally all positive (positive slope of the regressed lines) or could be explained due to deficiencies in the comparison data (e.g., background concentra- tions and potential outliers). Also, the y-intercepts were generally all positive, providing a level of desired conservatism for smaller values. Overall, our evaluations showed that there do not appear to be significant gaps that would prevent improvements to these models and/or improvements to the application thereof. As previously indicated, the research team leveraged its understanding of concentration levels and source locations from the previous FAA study (Wayson 2003). Our modeled-to-measured comparisons generally show an improvement over the previous study’s findings. It should be noted that the results from the previous study have not been made public. • Since aircraft were indeed confirmed to be the biggest contributor to both emissions and concentrations (based on the chosen receptor locations) and wind and atmospheric stability are the most significant meteorological factors influencing dispersion, the general agreement between these components and concentrations provided good evidence of proper responses by the model. • Although both modeled emissions and concentrations can be improved through the use of better quality source activity data, the potential for improvement with modeled concentra- tions appears to be greater because the location of moving sources can have a significant impact on concentrations. • The assessments revealed the importance of developing accurate background concentrations. If at all possible, hourly background concentrations that are specific to each receptor location should be developed. Our results showed that using approximations like daily averages or interpolations between long periods (several hours) could result in significant errors in the prediction of absolute concentration values. This is especially important because background concentrations can be significantly greater than the modeled values and, therefore, can have a significant influence as they did in our evaluations.

Findings, Recommendations, and Suggested Future Research 51 • As exemplified in our study, EDMS/AERMOD can be used to apportion the concentrations at each receptor location by source. Unfortunately, this can only be conducted by manually choosing one source at a time in the EDMS interface. Although the current version of EDMS allows selecting one source type (e.g., aircrafts) at a time to run in AERMOD, there is no batch mode to automatically generate source-apportioned concentrations. Each source type must be selected one at a time. Implementation of a batching method in AEDT might be considered as a possible enhancement. • Although it should be determined on a case-by-case basis, most airports can probably be considered to have flat terrain. Our assessments seemed to indicate little difference in con- centrations for receptors placed within the relatively flat terrain of an airport environment when assuming a flat terrain versus using DEM data. However, since the process of obtaining and implementing the freely available DEM data from USGS data is not difficult or time con- suming, it should be included to help reduce any uncertainties in a study. Ultimately, it will depend on the terrain surrounding an airport and where receptors are located. • While EDMS/AERMOD currently allows the emissions and concentration modeling of vari- ous aggregated hydrocarbon pollutants (e.g., VOC and NMHC), the concentrations of indi- vidual VOC and HAPs species (e.g., formaldehyde and benzene) cannot be modeled using the publicly available version of EDMS. The EDMS interface for creating AERMOD interface provides no provisions for modeling individual species. The modeling also cannot be per- formed directly within AERMOD because there is no feasible way to edit the HRE files. To do so would require a developer’s knowledge of EDMS to determine how the hydrocarbon pol- lutants were allocated to each source listing in the HRE file, and a processing tool or method would need to be developed to modify the HRE file accordingly to reflect the emissions of each species. It is possible that this capability can be implemented within AEDT. A post-processing method could be applied to proportionally modify concentrations according to emissions ratios, but it would need to be accomplished using one source (e.g., aircraft) at a time since it would be difficult (or impossible) to do so with multiple sources resulting in a mixture of different source pollutant profiles. • Based on our limited comparisons of modeled-versus-measured NO2 concentrations, it appears that the PVMRM option in AERMOD may be adequate for predicting NO2 (i.e., as good as pre- dicting NOx concentrations) but further studies will be necessary. Further sensitivity assessments modifying the input values for background ozone concentrations, ambient NO2/NOx equilib- rium ratio, and the in-stack NO2/NOx ratio should be conducted. The PVMRM option must be specified directly through the AERMOD input file since the EDMS interface does not provide this option. Similarly, various other options in AERMOD can be used as long as they do not require modifications to the HRE files (which are difficult or impossible to do without an EDMS devel- oper’s knowledge of how the HRE files were constructed). • Exercising the SO2 decay option using the nominal 4-hour half-life value showed very little difference in concentrations. Only a few higher level concentration points were affected. There was very little difference in R2 values. • The default regulatory modeling option in AERMOD was also exercised as part of the sensitiv- ity assessments. The resulting concentrations showed virtually no change, producing similar or identical R2 values. • As our assessments have shown, the emissions and air quality contributions from GAVs on roadways surrounding the airport can be significant. The inclusion of these sources and how much of them (how much of the roadways) to include will depend on the purpose of the study (e.g., NEPA project-level assessment or air quality health study). • For model evaluations, and for better application of the models, choosing receptor locations that are closest to airport sources is advantageous for characterizing the sources. As our com- parisons of results between the base station and the other receptor sites have shown, the agree- ment between modeled and measured concentrations is better when closer to the sources

52 Guidance for Quantifying the Contribution of airport emissions to Local air Quality (e.g., near runways). Although this is intuitive, it should be one of the most important con- siderations when planning an air quality study because smaller concentrations can be much harder to work with, especially since airport concentration contributions generally appear to be either comparable to or less than background concentrations. 6.3 Summary of Findings Regarding CMAQ Modeling Capabilities The following summarizes the conclusions, including limitations/gaps, from the various CMAQ assessments and insights for improving the modeling work: • The model performance for the WRF model was better at 4 km than at 12 km, especially in simulating the temperature and humidity fields, while the 12-km simulated the wind fields better. The differences in model performance between the 24-hour and 48-hour forecasts were negligible. Initializing the meteorological fields using NCEP’s North American Model (NAM) was crucial in obtaining such favorable model performance, when the models are run in near real-time. • The WRF model performance was better for the April 2009 and January 2010 periods than for the July 2010 period. This is in large part due to the poor temperature performance, which was observed in previous studies and is especially magnified during the summer months due to the higher average temperatures. • A methodology to process EPA’s National Emissions Inventories (NEI-2005) was imple- mented to create emissions for the 2009–2010 time period. Incorporating growth and con- trols for this period for the various anthropogenic source sectors was critical to obtain good model performance for the CMAQ model for the period modeled for this study. • Using the EDMS2Inv tool was critical for smooth transfer of information of EDMS outputs to the SMOKE-CMAQ modeling system and to model aircraft sources in a more realistic representation with 4-D variability. Aircraft sources at Dulles contribute from 45 to 80% of total airport-level emissions for the various pollutants. • Even though the release version of EDMS does not provide explicit estimates of HAPs from aircraft sources for use in dispersion modeling, the FAA-EPA chemical speciation profile for TOG was used to estimate HAPs, which were modeled with CMAQ to further estimate incre- mental ambient concentrations. • Using initial and boundary conditions from NCEP’s Eta-CMAQ forecast system was criti- cal in obtaining realistic flow of upwind concentrations to the Dulles region and to further obtain desirable model performance in the study region, when compared with various surface networks in the region. • Airport operators desiring to set up CMAQ model applications are recommended to use the NAM and CMAQ model outputs from NOAA’s NCEP to drive their local-scale model appli- cations. In the near future, the model resolution from NOAA will likely be a high resolution of 4 km, which may be a valuable resource for local-scale studies. • When compared to routine EPA monitors, the CMAQ model performance was statistically insignificant between the 12-km and 4-km modeling outputs. For PM2.5, the model under- estimated high concentrations (warm season in July 2010) and overestimated lower concen- trations (cool seasons, April 2009 and January 2010). In general, the predictions over the 12-km domain are higher than the predictions over the 4-km domain. Model performance was best in predicting daily maximum 1-h and 8-h O3 concentrations and reasonable for the other gas-phase species and within the ranges of performance metrics found in other comparable modeling studies to date. • In the evaluation of CMAQ against the Dulles study measurements, the model reproduced the observations very well for April 2009, overestimated for January 2010, and underestimated

Findings, Recommendations, and Suggested Future Research 53 for July 2010. Overall, predictions in the 12-km domain are higher than those in the 4-km domain. The daily variability observed by the field study measurements is captured well by CMAQ at both resolutions, which is a key indicator of model performance to be used for performing subsequent sensitivity studies for policy assessments. • Across the three seasons, the incremental airport contributions varied from 0.30 (April 2009) to 0.41 µg/m3 (January 2010) to ambient PM2.5 in the Dulles airport grid-cell. On a percent basis, the range was from 4.8% (April 2009) to 7.8% (July 2010). • In all three seasons, AEC (non-volatile soot PM) was the highest contributor, ranging from 49% (July 2010) to 96% (April 2009). Organic Carbonaceous Aerosol ranged from 10% (July 2010) to 20% (January 2010). Of the inorganic PM2.5 components, nitrate aerosol (ANO3) was the key contributor, ranging from 27.6% in April 2009 to 154.5% in July 2010. This large change in ANO3 is explained by a large change in a very low value of ANO3 in the background scenario. However, locally, ANO3 was reduced in the vicinity of the airport. Notably, the inorganic PM2.5 contributions were higher at downwind distances and varied for each season depending on the meteorology. These results are consistent with prior airport modeling stud- ies, which showed that secondary inorganic aerosols from airport emissions have larger effects at downwind distances of 200 to 300 km from an airport. Similar trends of local decreases and downwind increases are also seen for the two other inorganic PM2.5 components (sulfate and ammonium). However, these increases and decreases are both less than 1%. • Dulles airport contributes from up to 13.3% (January 2010) to 110% (July 2010) increase in NO2 concentrations. However, due to high incremental emissions of NOx from airport aircraft activity, decreases in ozone levels in all three seasons can be seen. These decreases range from -6.0% (July 2010) to -30.1% (April 2009). During the summer campaign alone, Dulles airport contributes about 0.98 ppm or 1.8% increase in daily maximum 8-h O3 slightly northeast of the airport. • An approach to evaluate CMAQ’s modal treatment of PM2.5 against the 8-section distribution from the RDI measurements was developed. The measurements show a much more pro- nounced bimodal distribution than the model. The RDI data are point measurements, each better able to identify near-source concentrations than otherwise allowed by the 4-km model resolution. This is especially true of the smallest size ranges, which are not represented ade- quately in the emissions inputs to the model (due to lack of information in emissions invento- ries). However, this evaluation could provide a basis for improving size fraction assumptions in aircraft emissions used in models. • The total PM (TM) is uniformly under-predicted across the size distribution at all sites in warmer months, except in the size range representing accumulation mode aerosol. This indi- cates that the model is more representative of aged aerosol and of mixing within the grid of locally produced PM with that transported long range. • Total PM is generally under-predicted in the smallest size bins, indicative of the inability of the model to capture near-source primarily emitted (or nucleating) particles, both due to the coarse resolution compared to point measurements and due to not using a plume-in-grid model. • Soil is generally under-predicted in the largest size ranges and over-predicted in the fine size cuts, pointing to the need to re-examine the soil emissions magnitudes and mean diameters in which emitted. • The main contributor to the under-bias is components other than SO4 and soil, as these are present in much smaller magnitudes relative to TM; however, the contributing sources of these other PM constituents are not obvious in the data provided because the breakdown into other PM components is not available in the RDI measurements. This makes it difficult to conclude where the model representation of PM-Other needs to be improved. • The winter-time measurements also show a significant over-prediction of TM, due to over- predictions in every constituent. This indicates that secondary species are a more significant component of the modeled size distributions relative to the measured.

54 Guidance for Quantifying the Contribution of airport emissions to Local air Quality • The impact of aircraft emissions is seen to decrease with increasing particle size on an event- average basis, with the largest size cut showing little or no impact, and in some cases a very slight increase in concentrations when the aircraft emissions are removed. • The evaluation of the hybrid model for each of PM2.5, CO, NOx, and SOx showed good model performance for PM2.5 when compared with observations for all three periods. It had a slight positive bias during the April and January periods and a slight under-bias during the July period. In both the April and July periods, the hybrid model using the 12-km CMAQ simula- tion performed better than the 4-km simulation as seen in the scatter plots. This, when com- bined with the results from the model evaluation of both WRF and CMAQ earlier, indicates that 12-km resolution is adequate for regional-scale model application. • Airport operators desiring to develop hybrid applications for performing airport air qual- ity assessments are recommended to use WRF-SMOKE-CMAQ at a resolution of 12 km, and instrument AERMOD with a finely resolved grid of receptors uniformly spaced at 500 m within the area of interest, and perform hybrid calculations using the methodologies dis- cussed in Appendix F.2. • Use of a hybrid model allows blending of regional contributions from background sources with highly resolved local-scale impacts to obtain variability at local scales and provides a powerful technique for estimating local-scale air quality impacts of airport emissions. • The hybrid model performance was not as good for CO, not showing much correlation with observed values during the April period, while dramatically over-predicting concentrations dur- ing both the January and July periods. The modeling data also showed some extreme outlier values which have been seen in previous hybrid models, and has been introduced through high concentrations in AERMOD. Additional investigation is needed to look at specific conditions under which AERMOD predicts these high concentrations while modeling aircraft sources. • The model performance is fairly good for NOx in that the model picks up the changes in dis- tribution across the seasons and fairly closely matches in all three seasons. However the scatter plot shows that, when paired in time and space, the model does not tend to agree with the observations all the time. This is similar to other AERMOD results in other studies. As with CO, the model contains some extraneous outliers, and this further reinforces the need for additional investigation of AERMOD behavior. • The SOx model performance is the poorest that has been seen, especially for January and July where the model dramatically overestimates the observed values. The model performed relatively well during the April campaign, but both the January and July campaigns showed much lower observed SOx values and consequently poorer model performance. Further study is recommended. • From analyses of the spatial distribution of PM2.5, the primary airport impact was near the gates, and secondly, the January period had much higher concentrations than either the April or July period. It was also found that the impact of the airport did not extend much beyond the airport property. In the 4-km CMAQ modeling, the secondary impact was found to be isolated to the western edge of the airport, while the 12-km modeling resulted in secondary impacts which were more widespread but of lower magnitude. • The spatial distribution of SOx is similar to that of NOx with peaks near the gates and sec- ondary maximum values near the runways; however, the concentrations are much smaller compared to the background values for SOx. As with CO and PM2.5, the impact of the airport on SOx concentrations does not extend much beyond the airport itself. • Using CMAQ, it was possible to estimate incremental contributions of aircraft emissions to various air toxics. Based on the analyses of differences in primary versus secondary constitu- ents of key air toxics, up to 96% of formaldehyde and 95% of acetaldehyde (maximum value in a single 4-km grid-cell) are from secondary production, rather than primarily emitted by airport sources. On a monthly average basis, the average secondary to total formaldehyde and acetaldehyde for the 3 months range from 92.4 - 95.6% and 86.4 - 88.4%, respectively. On a daily

Findings, Recommendations, and Suggested Future Research 55 average basis, the corresponding numbers are 87.5 - 99.2% (formaldehyde) and 72.1 - 93.6% (acetaldehyde), showing the importance of secondary production due to atmospheric chem- istry, rather than primary emissions of these two HAPs. The emissions analyses further show that formaldehyde and acetaldehyde contribute 8.7% and 2.8% to the total VOCs from IAD. When compared to all emissions sources in Loudoun county (where Dulles airport is mostly located), formaldehyde and acetaldehyde from Dulles airport contribute 9.7% and 4.0% to the total countywide emissions. 6.4 Summary of Findings Regarding Receptor Modeling The major conclusions of the PM-focused receptor modeling work conducted under this study are • Based on the use of the elemental data from the RDI, the receptor modeling work showed that useful insights on overall contributions from different sources can be accounted for. This includes the potential for identification of sources/processes through signature-type species identified in the measurements as well as comparisons to source-modeled, apportioned ambi- ent concentrations. • A source not included in the EDMS emissions modeling work—aircraft landings—was iden- tified as a potential source of PM emissions. The higher concentrations of Zn appeared to sup- port this conclusion as well as an analysis involving a conditional probability function (CPF) using wind direction data that pointed to the runways as significant sources. • Landings appeared to produce significant amounts of airborne particulate matter, with most of the mass being PM2.5 as would be expected from the ablation of tires and brake parts. Landing emissions are characterized by tire and brake wear particles as well as vaporization/ condensation of lubricating grease. Similar impacts of this source were observed during the three sampling campaigns. • High concentrations of transported secondary sulfate were observed, with highest concentra- tions observed during the July campaign. High concentrations of deicing salt were observed during January. Suspended soil impacts were highest in the April campaign and lowest during the winter. These findings are consistent with expectations by seasons and provide indica- tions of receptor modeling methods to help support the identification and relative amounts of certain species by season. • Soil and salt contributed mostly to particle sizes greater than 1 µm. • The component concentration comparisons against the source-oriented models showed that significant PM mass from the EDMS/AERMOD-predicted concentrations appeared to be missing. This appeared to corroborate the understanding that there are various missing PM components within the modeled aircraft emissions inventory, including gaps in the EDMS emission factors database as well as certain processes that are currently not included, such as emissions from landings (tire and brake wear). • While currently not considered a main technique for analyzing air quality near airports, the receptor modeling work has demonstrated potential in serving as a supplemental tool for corroborating and investigating gaps in the overall source-oriented modeling work. 6.5 Suggestions for Improving EDMS/AERMOD Modeling Capabilities Suggestions for future research based on the reviews and assessments conducted under this project are presented below. These suggestions are loosely ranked based on a combination of their expected impact and ease of investigation. That is, the top-ranked suggestions would tend to have the most significant impact on air quality modeling and would tend to be easier

56 Guidance for Quantifying the Contribution of airport emissions to Local air Quality to implement. As such, the higher ranked suggestions are expected to provide a better return on resource investment. 1. Aircraft emissions under idling and taxiing conditions are modeled assuming a power set- ting with corresponding fuel flow at the ICAO standard 7% level. It is widely understood that aircraft under idling conditions (and possibly under taxiing/queuing conditions) expe- rience power settings that are generally lower than the ICAO standard (Kim 2008). As such, emissions for CO and hydrocarbon species may be significantly higher than predicted, since these emissions increase exponentially below the 7% power setting. There is some dispute by the engine manufacturers as to whether this increase in emissions actually occurs because there are other operational changes in the engine below 7% that tend to result in a capping effect of the emissions at the 7% level. Measured data from ACRP Project 02-03A (TRB 2012); however, appear to confirm that there is significant increase in emissions below 7%. Therefore, a new power setting for idle/taxi to more accurately model emissions should be considered for incorporation into AEDT. 2. It is suggested that both GSE and APU operating times as well as the default fleet mix for GSEs be further investigated to determine if more accurate (more current) data are available. It is further recommended that videocameras be used to facilitate the data collection work. 3. For aircraft sources, EDMS currently elevates the area sources to account for the height of the aircraft engine exhaust as well as plume rise due to thermal buoyancy. An average height is used for all aircraft types. If it is found that an average height is not representative, then it is suggested that different heights be developed by either aircraft type or by aircraft size categories. This would help to improve the overall air quality modeling work, since some airports may only service certain types/categories of aircraft that may not correlate well with the average height of the elevated area sources. 4. Along with improving taxi emissions, the accuracy of modeled aircraft engine start-up emis- sions should also be investigated to determine their impacts on the overall ambient con- centration modeling process. For most jet engine aircraft, the start-up emissions for THC, NMHC, VOC, and TOG are the largest aircraft modal emissions for those pollutants (larger than the other modes). 5. Further studies should be conducted to verify the aircraft PM-related emissions from EDMS. Most of these emissions are apparently based on the use of the FOA methodology. Until measurement-based emissions indices can be made available, this methodology will likely be used for the foreseeable future. Therefore, the accuracy of the FOA should be further assessed. 6. While some sensitivity assessments were conducted to help gauge the behavior and responses by EDMS/AERMOD, the assessments are by no means comprehensive. Therefore, it is sug- gested that a formal sensitivity assessment be conducted to exercise most or all of the pertinent options in both EDMS and AERMOD. While numerous studies regarding airport emissions have been conducted, very few studies have been performed regarding air quality (concentra- tion) modeling. Therefore, such a study could help to better understand the behavior and accuracy of the models. 7. EDMS has no method to predict PM emissions from aircraft tire and brake wear during landing events. As such, a methodology should be developed and implemented into EDMS to account for these emissions. Such emissions could contribute significantly to PM con- centrations near an airport, especially near runways. 8. OG-speciation profiles for non-aircraft sources (e.g., GSEs) are not considered to be as accurate as those for aircraft. Therefore, the improvement of these non-aircraft, ground- level source profiles for OG-speciation may provide more confidence in the speciated emissions inventories. Exhaust emissions would need to be measured to develop new profiles.

Findings, Recommendations, and Suggested Future Research 57 9. With limited understanding of the evolution of plumes, especially involving chemically reactive pollutants, it would be beneficial to conduct a study to evaluate and/or confirm the conversion, decay, settling, and so forth of various pollutants, including VOC/HAPs spe- cies in the airport environment which represents different reaction conditions than other environments, including those that may have been simulated previously in a laboratory. Potentially, monitors could be located at varying locations away from a runway to record the concentration fall-off characteristics. 10. While EDMS allows the use of volume sources for stationary sources and gate-related sources (i.e., parked aircraft and GSEs), GAVs and aircraft are modeled as part of area sources rep- resenting roadways, taxiways, and runways. Because of the three-dimensional nature of plumes and the elevated aircraft exhausts, the use of volume sources should be investigated for aircraft and possibly GAVs as well. The current source specifications in HRE files cannot be modified without a developer’s knowledge of how the files are constructed. 11. Although the use of AERMOD has unified the modeling work in EDMS (i.e., only one dispersion model), it is suggested that the use of a model such as CAL3QHC be revisited to allow for the possibility of more accurately modeling the impacts of vehicle queues. CAL3QHC may provide better micro-scale modeling of atmospheric dispersion from interrupted roadway vehicle activities. 12. With the increasing availability of specific aircraft operational information and finer reso- lution (time-wise) weather data at airports, it is suggested that time-varying models (e.g., Gaussian puff models) be investigated to determine their utility in providing finer resolu- tion (e.g., per second and per minute) concentrations and the potential for more accurate results. Using a flight schedule, the dispersion of aircraft emissions can be modeled more accurately since each aircraft’s movements can be simulated and, therefore, treated as a discrete moving source (i.e., rather than as part of a stationary homogenous area source). 13. Whether it involves the use of static plumes or time-varying Gaussian puffs, it is suggested that the impacts of an aircraft’s wake be investigated. Unlike stationary sources, the wake of moving sources such as an aircraft can both drag and disperse pollutants. Although some of the dispersive impacts of an aircraft’s wake may be accounted for in the initial dispersion parameters of the area sources representing taxiways and runways, the effects of the wake on plumes/puffs generated by other sources (including other aircraft) are not clear.