Dispersion Modeling Guidance for Airports Addressing Local Air Quality Health Concerns (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.

16 4.1 Input Data Requirements On May 29, 2015, FAA revised its policy on air quality modeling procedures to identify the Aviation Environmental Design Tool (AEDT) Version 2b as the required model to perform air quality analyses for aviation, replacing EDMS. At the time ACRP Project 02-58 was being com- pleted, AEDT 2b had not been publicly released; therefore, after coordination with the ACRP project panel, the researchers decided to use the latest version of EDMS (5.1.4.1) to generate the airport emissions inventory. AEDT uses the same algorithms for estimating emissions as EDMS, and has the same cou- pling of dispersion to the AERMOD model. AEDT 2b implements the key functionalities from Integrated Noise Model (INM), EDMS, and AEDT 2a applications. Although mostly legacy functionalities were incorporated, differences in the models resulted in new overall capabilities, such as emissions dispersion from curved flight tracks (resulting from INM legacy capability) and noise using real weather (resulting from EDMS legacy capability). The ability to estimate emissions from curved flight tracks and then disperse them is a significant improvement over EDMS which assumes “straight in” and “straight out” tracks. Our general objective was to use equivalent input data in all four models to keep the inputs consistent so that any differences in output are the result of differences in the dispersion models. However, the researchers identified a dispersion modeling limitation where some of the disper- sion models (e.g., CALPUFF and SCIPUFF) are able to only model area sources with four edges. In the EDMS study used for the LAWA report, some EDMS sources such as parking structures and gates were modeled with more than four edges. To ensure consistency among all dispersion software models, the researchers converted the parking structures to have four edges. AEDT also has the ability to estimate full-flight emissions of multiple pollutants for global aircraft activity, as described in Wilkerson et al. (2010). Ongoing enhancements include the abil- ity to estimate both non-volatile PM mass and number, and will be available in a future release. 4.1.1 AERMOD AERMOD is included in EDMS/AEDT; hence, dispersion modeling for local-scale air quality and health using AERMOD can be made straight from the models and is relatively straight- forward, as long as the focus is solely on primary pollutants. For AERMOD to be used with one of the other models (e.g., CALPUFF, SCICHEM or ADMS-Airport), additional work needs to be done. 4.1.2 CALPUFF CALPUFF uses the same area-source treatment for aircraft sources as AERMOD, and it retains the same aspect ratio. To elaborate, EDMS represents most aircraft sources during landing and C h a p t e r 4 Models versus Data Inputs

Models versus Data Inputs 17 take-off with a dimension of 200 m × 20 m, thus yielding an aspect ratio of 10.0, which is retained in both AERMOD and CALPUFF modeling. Ideally, all area sources in CALPUFF should have an aspect ratio of less than 2.0. Sources with an aspect ratio greater than 2.0 will lead to some performance problems in the near field (at distances of 1–5 km), but not in the far field (at dis- tances greater than 5 km). In the research for ACRP Project 02-58, non-aircraft sources that are usually estimated on an annual basis with temporal profiles were converted to hourly emissions to be compatible with the CALPUFF model. 4.1.3 SCICHEM The version of SCICHEM used in this study also models aircraft sources as area sources. However, SCICHEM’s algorithm for modeling area sources involves splitting the area sources into smaller point sources, and then modeling them as individual Gaussian points. Thus, area sources that were elongated in one direction [or with aspect ratio (length/width) > 1.0] were split into individual point/stack sources. Even though AERMOD and CALPUFF maintain the same dimension when modeling aircraft sources, SCICHEM splits the source into multiple smaller Gaussian points. By following this procedure for the LAX study, the researchers found that the total number of sources increases by about a factor of 10. This is the key reason behind the exces- sive run times associated when using SCICHEM to model aircraft emissions as area sources. Like CALPUFF, SCICHEM does not support hour-of-day or static emission sources; hence, pollutant-specific hourly emissions for a single day in each season were generated for these sources and combined with the AERMOD’S hourly emissions files (labeled “.hre” files). 4.1.4 ADMS-Airport ADMS-Airport allows two methods for inputting aircraft emissions data to compute dispersion. 1. ADMS-VOL: The “Volume” option uses volume sources from aircraft sources. EDMS/AERMOD outputs the aircraft airborne emissions in area sources 20 m apart (vertically). For ACRP Project 02-58, these area sources were converted to volume sources in ADMS-Airport with a height of 20 m, thus making a very straightforward conversion process. 2. ADMS-AIR: The “Air File” option uses the model’s performance and chemistry capabilities to model aircraft sources as “jets.” This method uses an aircraft performance “engine” that uses specific aircraft and track positions to disperse the emissions. Unfortunately, specific aircraft and track position data are not available in the EDMS outputs, because EDMS out- puts aircraft emissions in area sources without any information on which aircraft contributed to these emissions. Therefore, the research team made several approximations when using the Air File (ADMS-AIR) option. For each hour, EDMS provides the area sources that had emissions (i.e., an aircraft passed through them) and the sources that did not (i.e., no aircraft passed through them). The active area sources were used to determine an average hourly air- craft track for a specific runway, and the researchers used that track for all aircraft in ADMS- Airport. The complexity of this ADMS-Airport modeling method required a reduction in the different aircraft types used; the researchers therefore mapped all the EDMS aircraft to 11 ADMS aircraft that could be used in the ADMS-Airport modeling. The emissions along each hourly track were distributed to the aircraft based on the number of flights by each air- craft and the aircraft emission indexes (EIs). The research team encountered no problems converting non-aircraft sources from EDMS to ADMS-Airport. EDMS gate and parking sources are already area sources and were eas- ily converted to volume sources. EDMS roadways are line sources with an EDMS-specified dispersion width, which was used to construct the volume sources. In EDMS, stationary sources can be area, point, or volume. Point sources in EDMS are specified with one set of coordinates (for the point); the coordinates are used, along with the source diameter, and

18 Dispersion Modeling Guidance for airports addressing Local air Quality health Concerns height (provided in EDMS), to construct volume sources. However, ADMS-Airport has an option to model roadways as road sources that include the treatment of street canyons, tunnels, and traffic-induced turbulence. ADMS-Airport also has a treatment for point sources that incor- porates buoyancy and momentum, and effects of nearby buildings. The ACRP Project 02-58 researchers did not use either of these options in this study. Rather, the team converted non-aircraft sources that were estimated on an annual basis with temporal profiles to hourly emissions to be compatible with the ADMS-Airport model. This is the first attempt to adapt the EDMS emissions for ADMS-Airport dispersion in the United States. The ADMS-Airport inputs are not directly compatible with EDMS, and the researchers had to make several assumptions and create conversion algorithms. EDMS and AEDT both utilize AERMOD and have the same dispersion output format. However, AEDT has additional output persist options, allowing for more details in its performance output files (e.g., results by flight rather than aggregate emissions results). In addition, AEDT has the ability to specify a detailed track for each aircraft in AEDT (including curved tracks) using aircraft radar data or other similar information. These AEDT features can be used in future studies to minimize the conversion assumptions, which in turn will increase the fidelity in creating ADMS-Airport inputs using the Air File (ADMS-AIR) option. 4.2 Modeling Systems AERMOD, CALPUFF, and SCICHEM models can be run on the Windows or Linux operating platforms. Commercial versions of AERMOD are available with GUI front-ends on Windows for ease of use, but the researchers compiled and installed all models on a Linux server. Doing this enabled the research team to run multiple instances for the two seasons and for various pol- lutants and receptor combinations. ADMS-Airport only runs through a Windows-based GUI, so for this model the researchers were limited to that environment. The LAX AQSAS provided access to an unprecedented dataset of ambient measurements at a large airport, but the data are only useful for validating the models at discrete locations. To develop dispersion model applications for local-scale health assessments, the researchers recommend that a gridded set of receptors be created in and around the airport, similar to what the researchers did for this study. This approach requires a Cartesian grid with receptors every 500 meters going up to 5 km from the airport, and a Polar grid with receptors every 5 km extending to 50 km from the airport. For airports near densely populated urban areas, the number of receptors may be increased (e.g., spacing one receptor every 100 meters), and near roadways increased even more (e.g., spacing one receptor every 10 meters) to capture the spatial gradients of traffic-related air pollutants in the near-road environment. The model then predicts concentrations at all these receptors at the ground level. This level of detail will be helpful to understand the spatial field of concentrations from the airport that will be used to study health impacts. To obtain seasonality, it is suggested that a future modeling study look at capturing estimates from at least two different seasons (e.g., summer versus winter) to examine the effects of meteo- rological conditions on the model predictions.

Models versus Data Inputs 19 Table 3 presents computational times in CPU-hours taken by the four modeling systems to perform the modeling for the LAX AQSAS for the winter and summer seasons. As seen in Table 3, SCICHEM modeling run times are prohibitively long. Unless additional work is done to improve source characterization, SCICHEM is not likely to be a viable option to perform local air quality modeling of airport sources. Another key distinction between steady-state models (such as AERMOD and ADMS-Airport) and non-steady-state models (such as CALPUFF and SCICHEM) is the generation of input meteorological data. When processing input meteorological data from National Weather Service (NWS) sites for AERMOD, some data gaps are normal and do not create an issue for steady-state models, aside from missing concentration fields. In the case of non-steady-state models, how- ever, missing hours of meteorological data need to be addressed before performing the model simulations. For this study the research team created a complete dataset by filling missing hours of data with observations from nearby NWS sites and by computing certain meteorological variables from first principles. (A description of the procedures used appears in the contractor’s final report, which can be obtained using a link from the ACRP Project 02-58 webpage.) LAWA Receptors, 2m Heights, 7 Pollutants* Winter 2012 Summer 2012 # CPUs Airc Airp Bg All # CPUs Airc Airp Bg All AERMOD 7 n/a n/a n/a 3.4 7 n/a n/a n/a 3.4 CALPUFF 1 1.4 1.7 0.3 2.0 1 1.0 1.1 0.2 1.3 SCICHEM 7 1167.9 1742.1 1439.1 3181.1 7 1652.0 2084.1 1459.6 3543.7 ADMS-AIR 7 27.3 27.7 0.5 28.3 7 22.0 22.4 0.5 22.9 ADMS-VOL 7 1.6 2.0 0.5 2.6 7 1.2 1.6 0.5 2.1 # CPUs = The number of CPUs used when running the model; Airc = aircraft sources; Airp = airport sources, including Airc; Bg = background sources (sources outside the airport but within the study region); All = the sum of Airp+Bg. AERMOD was run for all sources in one execution; for this model no data are broken out under Airc, Airp, and Bg. All models except ADMS-Airport (ADMS-AIR and ADMS-VOL) were run on Redhat Linux OS with Intel5866 x86_64 processor with 48 GB memory; ADMS-AIR and ADMS-VOL were run on Windows 7 OS using the same processor. * Particulate matter (PM) counted separately for PM2.5 and PM10. Table 3. CPU-hours used by the four models.