Enhanced AEDT Modeling of Aircraft Arrival and Departure Profiles, Volume 2: Research Report (2018)

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.

11 3 Research Results Results from each task conducted during this research effort are included in the following sections. Industry Review Findings An Industry Review was conducted to assess current practice regarding how aircraft arrival and departure profiles are modeled in the Integrated Noise Model (INM) and Aviation Environmental Design Tool (AEDT). The review also involved listing the profile options currently in the models and indicating the methods used to customize profiles. In addition, it included describing the process for Federal Aviation Administration (FAA) approval of customized profiles. The overall approach to the Industry Review included:  Document the FAA approval process for INM and AEDT profiles, by meeting with key FAA staff and reviewing the recent guidance updates released with FAA Order 1050.1F, Environmental Impacts: Policies and Procedures.  Document the current approaches to INM and AEDT profile modeling, by interviewing practitioners experienced with airport noise modeling.  Develop conclusions which will help the research team to refine and advance the project. Profile Customization Profile customization options and methods available in both the legacy INM model and the new AEDT model (which replaced INM in May 2015) were investigated. The research team’s interviews and discussions about current practices were in reference to INM, whereas recommendations for future modeling consider the capabilities of AEDT. Therefore, this section discusses both models Customization Options The profile customization options in INM include the creation of user-defined procedural or fixed-point profiles, and/or the creation of user-defined aircraft performance coefficients. When creating user-defined profiles, modelers typically alter profiles by modifying existing procedure steps, replacing existing procedure steps, or adding new procedure steps. Procedure steps include, for example, takeoff, climb, descend, land, and decelerate. It is uncommon for users to create user-defined fixed-point profiles because, while the points of the profile are easy to modify to match a desired trajectory, there is no underlying performance data that can be used to determine what changes in thrust would be caused by the changes in the trajectory. The creation of user-defined aircraft, with modified flight performance data, can also be used in conjunction with existing standard or new user-defined profiles to alter modeled trajectories. For example, a copy of an existing aircraft can be created with modified thrust coefficients and be assigned to fly a copy of an existing standard procedural profile. The thrust coefficients can be modified in such a way that, for example, they simulate a reduced-thrust takeoff version of the assigned flight profile. AEDT provides the additional ability to explicitly define aircraft altitude profiles – for any points along a flight path with an altitude of greater than 500 feet Above Field Elevation (AFE). This capability was originally used for airspace procedure noise modeling in the FAA’s Noise

12 Integrated Routing System (NIRS) model; with AEDT, it is now available for use in airport noise studies as well. Profiles are defined as a set of 3-dimensional points (latitude, longitude, and altitude) with “altitude controls” specifying the altitude and a constraint setting (At, At-Or-Above, At-Or-Below). AEDT then calculates flight paths (for civilian jet, turboprop, and piston aircraft) within the limits of a given aircraft’s performance characteristics based on the altitude controls. One advantage of this method is that, if modelers use AEDT to develop profiles, then FAA AEE approval is not required. AEDT also provides a “Sensor Path” functionality, allowing users to input a set of 4-dimensional points (latitude, longitude, altitude, and speed) for which AEDT will follow given the flight performance constraints defined in the model. This gives users the ability to explicitly define aircraft altitude and speed profiles, with the current limitations that the flights must be runway-to- runway and input speed values are ignored at altitudes below 10,000 ft. AFE. The models, methods, and variables currently available for custom profile and trajectory-based modeling are summarized in Table 3-1. Table 3-1 Custom Profile Options in INM and AEDT Current Methods Variables INM & AEDT – Procedural Profiles Takeoff, Climb, Cruise-Climb, Accelerate, Accel-Percent, Descend, Descend-Decel, Descend-Idle, Level, Level-Decel, Level-Idle, Land, and Decelerate INM & AEDT – Fixed-Point Profiles Ground Track Distance, Altitude, Speed, Thrust Setting, and Operational Mode INM & AEDT – Performance Coefficients Jet Thrust Coefficients, Propeller Thrust Coefficients, and Flap Coefficients AEDT – Altitude Controls User-Defined 3-D Trajectory, with At, At-Or-Above, or At-Or-Below Altitude Controls (for altitudes > 500 ft. AFE) AEDT – Sensor Path User-Defined 4-D Trajectory from Latitude, Longitude, Altitude, and Speed (runway-to-runway only, input speed ignored at altitudes < 10,000 ft. AFE) A future version of AEDT, at FAA’s discretion, will include support for speed controls in conjunction with altitude controls. Speed controls will consist of target speeds and speed restrictions (At, At-Or-Above, At-Or-Below) that are assigned to geographic coordinates along ground tracks. Speed controls will enhance user control over aircraft speeds in the analysis of terminal area operations, and will be leveraged in the analysis of flights driven by Sensor Path data, such that input speed values are taken into account for the analysis of the terminal area portion of runway-to-runway operations. It is also anticipated that future versions of AEDT will support greater flexibility in analyzing sensor path data, such as the ability to process partial flights (instead of full runway-to-runway analysis). Customization Methods Model users can exercise the AEDT and INM models using the profile customization options shown in the table above. In order to define the profiles, several methods are used to collect and process data and develop either procedure-based or trajectory-based profiles. Procedure-based profiles consist of procedure step assignments for each segment of the profile (such as takeoff, climb, accelerate, etc.). Source data can come from:  Operator or Air Traffic Controller interviews – Model users interview aircraft operators/pilots to determine “typical” flight trajectories, and speed/thrust if available. Air

13 Traffic Controllers are also interviewed to determine the “standard” flight procedures in use at an airport (headings, turns, holds, etc.).  Flight manual or manufacturer data – In the absence of the above data, or to supplement it, modelers review aircraft-specific performance data to verify that the modeled profile will conform to “standard operating procedures.” Trajectory-based profiles can be generated using aircraft position data (such as radar, ADS-B, ASDE-X, etc.). The key elements of the approach include:  Segregate position data by aircraft type, runway, operation type, and stage length (using trip distance).  Plot and review altitude and speed profiles as a function of ground track distance.  Compare altitude and speed with the corresponding standard model profile, using statistical methods to determine whether new profiles are necessary.  Translate position data into altitudes for INM modeling or altitude controls for input to AEDT. Trajectory-based profiles can also be created in the absence of radar data, using navigational charts (e.g., SIDs, STARs, or Approach Plates). Using these published flight procedures, the user can define the waypoints, and altitude controls for input to AEDT. FAA Profile Review Process This section describes the FAA’s review and approval process for custom profiles in INM and AEDT. The research team’s interviews and discussions about current practices were in reference to INM, whereas recommendations for future modeling consider the capabilities of AEDT. Therefore, this section discusses the approval processes for both models. INM Profile Review The guidance on the development and approval for INM customized profiles is included in the INM 7.0 User’s Guide as Appendix B, FAA Profile Review Checklist.2 This checklist includes the following elements: 1. Background – Description of the project. 2. Statement of Benefit – Description of the need for custom profiles (in terms of altitude, speed, and thrust). 3. Analysis Demonstrating Benefit – INM grid-point noise analysis in terms of Sound Exposure Level. 4. (SEL). 5. Concurrence on Aircraft Performance – Supported by either an aircraft flight manual or statement from an aircraft operator or manufacturer. 6. Certification of New Parameters – Definition of variables, units, and demonstration of compliance with SAE AIR 1845. 7. Graphical and Tabular Comparison – Custom altitude, speed, and thrust profiles compared to the INM standard profile.

14 In addition to the checklist, information on the procedure for submittal of this information is detailed in the 2009 memo AEE and Airports Coordination Policy for Non-Standard Modeling Procedures and Methodology. This memo details the procedures for the submittal of a Profile Review Checklist. It also clarifies the types of INM profile modeling which require FAA approval, as summarized in Table 3-2. Table 3-2 Profile Modeling Types Requiring FAA Approval Custom Profile Type Description All User-defined aircraft profiles (including modifications to standard profiles) developed by methods other than the FAA-accepted methodology built-in to AEDT (i.e., altitude controls or sensor paths). Departure Stage Length Profiles which are defined using methods other than: trip length, estimated takeoff weight, and documented procedures based on ICAO/SAE/ECAC standards. Touch-and-Go Operations Adjustments to touch-and-go and circuit profiles using procedure steps not found in the standard profiles. Military Operations Military profiles for aircraft not having any INM/AEDT standard profiles. Helicopter Operations Helicopter profiles that do not follow INM/AEDT-defined profiles and parameters. AEDT Profile Review The profile review process for AEDT is provided in an appendix to the FAA Order 1050.1F Desk Reference1. It contains the same six checklist elements discussed above, but includes several important differences from the INM guidance:  In the Statement of Benefit section, there is an additional requirement to discuss “why the default method or data are not sufficient.”  In the Analysis Demonstrating Benefit section, it is clarified that “fuel burn, emissions, and emissions dispersion results are not needed in the submittal package”.  In the Certification of New Parameters section, for profiles that are defined in terms of procedure steps with new aircraft performance coefficient data, “separate AEE review is required [and] direct manufacturer support or flight test information is required to facilitate that review.” This represents a new requirement.  In the Graphical and Tabular Comparison section, a “quantitative comparison, such as an estimate of the least mean square of differences, should be provided and explained.” This represents a new requirement. The Order 1050.1F Desk Reference1 also updates the process for AEE submittal and review. It provides additional clarifications for how the AEDT modeler, project sponsor, and various FAA offices should interact to complete the review process. Current Industry Practices This section presents the approach and the results of meetings and interviews conducted by the research team. The purpose of the meetings with FAA and SAE International was to coordinate directly with the regulatory and standards-developing groups involved with custom profile modeling. The purpose of the interviews was to survey industry practitioners about the methods used to model custom profiles.

15 Meetings with FAA and SAE International In late 2014 and early 2015, the research team conducted meetings with the following groups:  FAA Office of Environment and Energy, Noise Division (AEE-100)  FAA Office of Airport Planning and Programming, Planning and Environmental Division (APP-400)  SAE International, Aircraft Noise Measurement and Aircraft Noise/Aviation Emission Modeling Committee (A-21) The meeting with AEE-100 was held on December 8, 2014. The research team met with staff in the Noise Division who review requests for custom profiles. AEE-100 stated that the most common types of requests received are, in the following order: 1. Military aircraft arrival, departure, and touch-and-go profiles for joint-use facilities. 2. Civilian aircraft departure stage lengths determined from radar data. 3. Reduced-thrust takeoff profiles (i.e., using less than the maximum available thrust during takeoff ground roll and the early portion of initial climb). 4. Extended profiles beyond the standard altitude cutoffs (i.e., 10,000 ft. AFE for departures and 6,000 ft. AFE for arrivals). 5. Level segments on arrival profiles. 6. Taxiway profiles for civilian fixed-wing aircraft. The research team clarified that touch-and-go and taxiway profiles are beyond the intended scope of this project. AEE-100 explained that a request should provide background information about the project, give a clear explanation of the need for custom profiles, include concurrence with the profiles (from an operator/pilot or using radar data), and clearly show the noise benefit of the new profile. AEE-100 also suggested that the research team determine the typical motivators for model users to customize profiles. Next, the research team held a conference call with APP-400 on February 6, 2015. APP-400 is the office which issues the approval letters to entities which request custom profiles – once AEE technical approval is granted. The research team found that custom profile requests – for NEPA and 14 CFR Part 150 studies – are tracked on a per-project basis as opposed to being organized in a central database. Therefore, there was no straightforward method to collect and review past approvals. In addition, the research team briefed the SAE International A-21 committee on November 5, 2014. The team provided an overview of the project and the intended outcomes, and requested that the committee form a Project Working Team (PWT) to review and provide input on future project deliverables. Practitioner Interviews - Overview In order to document the current approaches to INM and AEDT profile modeling, the research team sought to conduct interviews of practitioners experienced with airport noise modeling. The team sent an invitation via email to 25 individuals affiliated with consulting firms and research organizations in the U.S. During the first quarter of 2015, the research team conducted one-on- one telephone interviews with 11 individuals representing the following organizations:  HMMH  HNTB

16  Hoyle Tanner  Landrum & Brown  URS/AECOM  VHB  Wyle Laboratories A survey was developed by the research team to guide the discussions held in the interviews. There were 14 questions in the survey, each designed to engage the interviewee in a discussion about their experience customizing profiles for various types of noise modeling projects. The names of individuals interviewed, and the specific airport noise studies discussed, are not included in this report at the request of those who were interviewed. Table 3-3 presents a compilation of the survey and responses. In this table, each survey question is listed along with the possible responses. The count and percentage of each response is shown. Note that the count of responses to each question varies because, for some questions, an individual respondent provided answers for more than one study. Also, some questions were phrased such that a respondent could provide multiple answers. The final column of the table provides a brief summary of the responses to the question.

17 5. What was the main reason/motivation for customization (e.g., sponsor, public, legal, technical, etc.)? There are a number of motivators for expending time and resources on custom profile modeling. Most often, the decision to Technical - impact on noise contours/radar indicates non-standard 5 31% customize profiles originates from a technical need, which may be Public 3 19% identified by the sponsor, the FAA, or the public. Sponsor 2 13% National Park nearby 2 13% No standard INM profile/using NOISEMAP profile 2 13% FAA air traffic 1 6% Other 1 6% 6. Have you customized profiles for: military (jet/prop), civilian (jet/prop), helicopters? Widebody, narrowbody, regional, and business jets comprised Air carrier jet 9 45% 70% of the responses. Military jets are important noise Business jet 5 25% contributors at joint use airports. Propeller aircraft are important at Military jet (Joint use bases) 3 15% Prop (GA airports) 3 15% Military prop 0 0% Helicopter 0 0% Table 3-3 Summary of Practitioner Interview Responses Question Count of Responses+ Percentage of Responses Notes 1. Have you submitted custom profile requests to FAA AEE in the past? All have submitted profiles through the FAA AEE profile review Yes 7 100% No 0 0% process. 2. If so, were they approved or disapproved? Disapproval is rare. Approved 6 86% Disapproved 1 14% 3. How long did FAA’s review take? Did the approval process result in a significant delay in the study? <1 month 0 0% 1-2 months 2 29% 3+ months 1 14% Unknown/Do not remember 4 57% In most cases, the study was not significantly delayed. 4. Have you customized profiles but decided not to submit for FAA review? Nearly half of respondents customized profiles but did not submit them to the FAA AEE profile review process, because it was not Yes 4 57% No 3 43% required (e.g., determined there was no noise benefit, study was not FAA-funded, study was for research, etc.).

18 Question Count of Responses+ 7. Of the custom profiles that were built, what percent would you estimate were departures and what percent were arrivals and what percent were touch-and go? Percentage of Responses Notes Departure customization includes takeoff weight, climb rate, or selection of an alternate stage length. Arrivals include level Departure 7 50% Arrival 6 43% Touch and Go (Military, including overhead break arrivals) 1 7% segments, and in a few cases, an approach angle other than the 3-degree standard. 8. What type of study were the custom profiles created for – EA, EIS, Part 150, Part 161? The predominant study types are EIS's and non-NEPA studies EIS 7 39% Other (research, state required, airport outside of US, etc.) 6 33% EA 2 11% Part 150 2 11% Part 161 1 6% including research, state-mandated noise contour mapping, and studies conducted outside of the U.S. Of note, for the three latter types, FAA AEE approval is not required. 9. What are your method(s) for radar altitude matching (mean, median, RMS, other?) Most often, a visual comparison to radar data is conducted to Statistical 3 43% Overlay/visual 4 57% validate profiles developed from operator or air traffic procedures. 10. Were the profiles verified by the operator or manufacturer? Operator concurrence is a requirement for FAA AEE approval. Yes 6 86% No 1 14% The issue of de-rated takeoff thrust is often cited by pilots. 11. Have you extended standard profiles higher than the max altitudes? This is most common when calculating noise levels beyond DNL Yes 4 57% No 3 43% 65 dB or in areas near National Parks. 12. Have you submitted level-segment profiles? Common during arrivals to small and large airports. Yes 5 71% No 2 29% 13. Have you submitted customized taxi profiles? Infrequent, but an emerging issue for some airports. Out of scope Yes 2 29% No 5 71% for ACRP 02-55. 14. Have you submitted helicopter profiles? When modeling helicopters, this is almost always required, Yes 4 57% No 3 43% + The count of responses to each question varies because, for some questions, an individual respondent provided answers for more than one study. Also, some questions were phrased such that a respondent could provide multiple answers. because operations at every airport/heliport are unique

19 Analysis of Practitioner Interviews There are a number of motivators for expending time and resources on custom profile modeling. Most often, the decision to customize profiles originates from a technical need (e.g., radar data show a clear difference between standard profiles and actual profiles). These may be identified by the study sponsor, the FAA, or the public. Departure profile customization most often addresses takeoff weights, climb rates, or selection of an alternate stage length. The motivations to customize arrival profiles include level segments, and in a few cases, an approach angle other than the 3-degree standard. In order to effectively utilize the time and resources available in a given project, modelers often: • Weigh the need for modeling custom profiles against the project sponsor’s needs and the benefit to noise modeling results. • Limit the analysis to the most frequently operated aircraft in the fleet mix. • Limit the analysis to the loudest aircraft – for example, the aircraft with noise certification levels within approximately 10 dB of the loudest aircraft. Every organization had submitted custom profile requests to the FAA – most commonly for Environmental Impact Statement (EIS) studies – and disapproval was rare. However, nearly half of respondents customized profiles but did not submit them to the FAA AEE profile review process, because it was not required (e.g., research studies, state-mandated noise contour projects, and studies conducted outside of the U.S.). Wide body, narrow-body, regional, and business jets comprised 70% of the responses. It was noted that older aircraft in the INM/AEDT database have standard profiles that do not necessarily reflect current operational characteristics. Also, military jets are important noise contributors at joint-use airports, and profiles must always be customized to match the unique procedures at the facility. One respondent suggested improving the process of converting and importing profiles from the NOISEMAP model to INM/AEDT. In order to model custom profiles, a visual comparison to radar data is typically conducted to validate profiles developed from operator or air traffic procedures. Operator concurrence is a requirement for FAA AEE approval. Presenting noise-model profiles to operators can be challenging; one respondent suggested graphing profiles with time on the horizontal axis, instead of track distance. Another obstacle to operator concurrence can be the issue of de-rated takeoff thrust, a practice in which pilots choose an engine power setting that reduces the upper limit on thrust to a value that is less than the maximum available value. Since it is difficult to model de-rated takeoff thrust in INM/AEDT (it requires modification of aircraft performance coefficients), custom profiles typically take off at maximum available thrust. However, pilots often find this unrealistic, as the use of de-rated takeoff thrust is so common. Note that this topic is being investigated by other ACRP research projects. More than half of the respondents had extended standard profiles higher than the maximum altitudes. This is most common when calculating noise levels lower than DNL 65 dB, or in areas near National Parks. Taxi and helicopter profiles were modeled by some practitioners; however, they are not a main focus for this research project. Finally, one practitioner noted that the models do not check for intersections of terrain with flight paths, and suggested such a check be added to AEDT.

20 Industry Review Conclusions The Industry Review documents the current approaches to INM and AEDT arrival and departure profile modeling. The research team interviewed practitioners experienced with airport noise modeling and also met with key industry stakeholders. Based on our findings, the conclusions below were used to refine and advance the project. The research team also considered the motivations for customizing profiles and provides situation-specific guidance on AEDT profile modeling. (1) There are several common motivations for customizing profiles, each of which the research will address. In some cases, the formal FAA AEE profile review process is not used, depending on the noise study purpose and sponsor requirements. The research will provide guidance and tools to ensure that modelers are properly creating custom profiles in the absence of FAA AEE review. (2) The research should focus on air carrier and business jets. As a result of the practitioner interviews, we found that custom profiles are often needed for civilian jet aircraft (e.g., wide body, narrow-body, regional and business jets) and also for military jet aircraft operating at civilian and joint-use facilities. However, INM and AEDT do not contain the necessary underlying aircraft performance data needed to customize profiles for military aircraft. Propeller aircraft and helicopters are a lower priority for this research, although they sometimes contribute to noise impacts at smaller airports. (3) The research will address arrivals and departures equally because both are typically customized. There are several methods for customizing profiles, and AEDT provides new capabilities not included in INM (i.e., altitude controls and sensor paths). However, for some aircraft types, the AEDT altitude control capability is not available for arrivals because the standard profiles are not defined in terms of procedure steps. (4) Updated guidance in FAA Order 1050.1F, which includes new FAA AEE profile review requirements, would need to be addressed. There are two significant requirements for AEDT custom profiles that were not previously included in the INM guidance. The research should focus on providing methods that avoid the need to modify performance coefficients, when possible. In addition, the research should discuss methods for graphical comparison of profiles (such as an estimate of the least mean square of differences, or other statistical methods). New Profile Identification and Prioritization This effort focused on determining the extent of the need in addition to the best way to improve the available set of standard arrival and departure profiles in AEDT. As noted above, the current limited set of standard arrival profiles cannot fully represent the variety of real-world profiles flown today. They do not account for many factors including the local airspace, ATC interventions, and differences in Flight Management System (FMS) logic across the fleet. Historically they did a good enough job for the purpose they were developed for, but many environmental stakeholders have recently noted their shortcomings. As we learned while conducting our Industry Review, industry stakeholders are also interested in an expanded set of departure profiles as departure profiles are often customized as well. While that effort also discovered a stakeholder focus on military aircraft, subsequent planning determined that it would not be practical to create new profiles for military aircraft due to the lack of underlying performance data for those aircraft types in AEDT. That data is needed to both define new flight procedures within AEDT and allow AEDT to calculate trajectories based on those flight procedure definitions.

21 Process This effort consisted of four basic phases: acquisition and manipulation of trajectory data, profile grouping and creation, candidate profile evaluation, and prioritization of candidate profiles. We followed an iterative process that allowed the grouping and profile creation methodology to be revised as shortcomings became apparent during the process. This was important because information learned and insights gained during later phases of the study necessitated revisiting earlier phases before proceeding. Exhibit 3-1 is a flow chart of our process with a box for each phase. Once the prioritization of the candidate profiles was completed, the process was reviewed for any shortcomings. If any were found, the process was modified to correct these shortcomings and then iterated through again. Exhibit 3-1 Process Flow Chart Baseline and Candidate Profile Creation and Evaluation This section begins with an explanation of the data selected. Following that are details about creating baseline and candidate profiles. The section closes with a walkthrough of the metric used to evaluate candidate profiles.

22 Data Selection The main data source used to identify common profiles that are not currently represented in AEDT was historical trajectory data, specifically radar data from the Performance Data Analysis and Reporting System (PDARS). PDARS is an FAA product that collects detailed air traffic management system data across most of the United States from 20 domestic Air Route Traffic Control Centers (ARTCC’s), 28 Terminal Radar Approach Control (TRACON) facilities, and 27 Airport Surface Detection Equipment, Model X (ASDE-X) equipped airports. Among the various PDARS data sets that exist, the “USAMERGE” data set was chosen because it contains full flight trajectory data from ARTCCs and TRACONs that cover the entire country. The trajectory data is accompanied with detailed flight characteristics such as origin/destination airport, departure/arrival runway, aircraft type, aircraft id, operational times, flight plan information, beacon code, and more. This data set is collected and archived for each day. The FAA has identified 30 airports throughout the country that exist in major metropolitan areas with the highest volume of air traffic that have been labeled the “Core 30” airports. Our analysis focused on these airports. The Core 30 airports are:  Hartsfield-Jackson Atlanta International (ATL)  Boston Logan International (BOS)  Baltimore-Washington International (BWI)  Charlotte Douglas International (CLT)  Ronald Reagan Washington National (DCA)  Denver International (DEN)  Dallas Fort-Worth International (DFW)  Detroit Metropolitan Wayne County (DTW)  Newark Liberty International (EWR)  Fort Lauderdale Hollywood International (FLL)  Honolulu International (HNL)  Washington Dulles International (IAD)  George Bush Houston Intercontinental (IAH)  New York John F. Kennedy International (JFK)  Las Vegas McCarran International (LAS)  Los Angeles International (LAX)  New York LaGuardia (LGA)  Orlando International (MCO)  Chicago Midway (MDW)  Memphis International (MEM)  Miami International (MIA)  Minneapolis St. Paul International (MSP)  Chicago O’Hare International (ORD)  Philadelphia International (PHL)  Phoenix Sky Harbor International (PHX)  San Diego International (SAN)  Seattle-Tacoma International (SEA)  San Francisco International (SFO)  Salt Lake City International (SLC)  Tampa International (TPA)

23 The PDARS USAMERGE data set has no coverage for Honolulu International (HNL); therefore, the airport is not represented in this project. The remaining 29 airports are referred to as “Study Airports”. Thirty days of USAMERGE data beginning on April 1, 2014 and ending on April 30, 2014 were selected as the date range to be analyzed. The data were consolidated into three categories on a per airport basis based on operation type: arrival, departure, and intraflight. An intraflight is defined as an operation that occurs between 2 of the 29 Study Airports, with the operation being assigned to the departure airport. For example, an operation that departs ATL and arrives at SFO is assigned to the ATL intraflight category. Due to the large number of operations in this dataset only the departure halves of intraflights were analyzed. Table 3-4 reports the arrival and departure operation counts for each of the Study Airports. Table 3-4 PDARS Operation Counts By Airport & Operation Type Airport Count of Arrival Operations Count of Departure Operations Total Count of Operations ATL 22,615 35,315 57,930 BOS 6,495 15,917 22,412 BWI 5,166 9,470 14,636 CLT 13,853 20,995 34,848 DCA 5,762 11,754 17,516 DEN 12,742 21,943 34,685 DFW 16,809 26,906 43,715 DTW 9,266 16,290 25,556 EWR 8,599 15,257 23,856 FLL 5,469 10,541 16,010 IAD 8,343 12,355 20,698 IAH 12,782 19,735 32,517 JFK 12,242 18,759 31,001 LAS 8,276 16,248 24,524 LAX 12,786 25,007 37,793 LGA 7,432 15,157 22,589 MCO 5,735 12,282 18,017 MDW 5,663 9,801 15,464 MEM 5,228 8,772 14,000 MIA 11,271 16,698 27,969 MSP 9,537 16,546 26,083 ORD 22,737 35,963 58,700 PHL 9,337 15,916 25,253 PHX 9,378 18,197 27,575 SAN 2,579 7,391 9,970 SEA 6,858 12,595 19,453 SFO 7,791 17,593 25,384 SLC 6,389 11,250 17,639 TPA 3,307 7,627 10,934 Grand Total 274,447 482,280 756,727 Baseline Trajectory Creation For this effort baseline trajectories were calculated using AEDT default profiles by running AEDT 2a Service Pack 2 (SP2), which was released to the public on February 15, 2014. There

24 are 3 system databases that get installed as a part of the standard AEDT 2a SP2 installation: (1) AIRPORT, containing a global set of airport data, (2) FLEET, containing aircraft models and performance data, and (3) STUDY, containing the baseline schema for creating and importing new studies. The FLEET database contains all of the approach and departure profile information for the model. These AEDT databases are updated with each new version of AEDT, however the key data for this task, including the default flight profiles and basic airport data, are not often updated between versions. There are two main aircraft flight performance specifications in the AEDT FLEET database, the European Civil Aviation Conference (ECAC) Doc 295 (based on Society of Automotive Engineers (SAE) - Aerospace Information Report No. 1845) and EUROCONTROL’s User Manual for the Base of Aircraft Data (BADA). AEDT applies the flight procedures, atmospheric model, performance-related equations, and proper coefficients of each model depending on the aircraft’s phase of flight. In the terminal area, defined as altitudes below 10,000 feet Above Field Elevation), Doc 29 is primarily used. In the enroute area, defined as altitudes above 10,000 ft. Above Field Elevation), BADA is primarily used. Under Doc 29, an approach profile is defined from 6,000 feet Above Field Elevation (AFE) to the end of runway roll and a departure profile is defined from the beginning of runway roll up to 10,000 feet AFE. This task is limited to expanding the list of available default Doc 29-based procedural flight profiles. Terminal area flight profiles in AEDT are defined by a combination of four parameters: operation type, aircraft type, profile identifier, and stage length. operation type - AEDT contains profile information for approaches, departures, touch & go’s, circuit flights, and overflights. For this project, only approach and departure profiles were investigated. These operation types are commonly abbreviated as “A” and “D”, respectively. aircraft type - The aircraft type is normally identified by a four-letter code. profile identifier - Each profile has an identifier, generally limited to either “STANDARD”, “ICAO_A”, or “ICAO_B” in the AEDT FLEET database. For this project, profiles that are identified as “STANDARD” were used as the baselines, as these profiles are the ones typically used in FAA regulatory environmental studies. stage length - Stage length is a parameter AEDT uses to set aircraft weight, using trip distance as a proxy. Trip distance data is generally much more available to AEDT users than aircraft weight data, and AEDT can calculate the distance itself if a user defines the origin and destination airports. The AEDT FLEET database contains flight profiles for aircraft weights that span the range of usage for most aircraft types. The trip distance associated with each stage length value is shown in Table 3-5. Table 3-5 Stage Length Definition Stage Length RANGE_MIN (NM) RANGE_MAX (NM) 1 0 499 2 500 999 3 1,000 1,499 4 1,500 2,499 5 2,500 3,499 6 3,500 4,499 7 4,500 5,499 8 5,500 6,499 9 6,500 11,000

25 Each aircraft type in AEDT has a different number of stage lengths available according to its range. In most cases, approach profiles are only defined for stage length 1 as it is assumed that most airplanes only carry enough fuel to get them to their destination, in addition to emergency reserves. In general for departures, larger, more powerful aircraft have longer maximum ranges which enable them to have more available stage lengths. The set of available stage lengths for each aircraft is defined by AEDT in the FLEET database. A unique key was assigned to each baseline profile to distinguish one from another based on the aforementioned four parameters. The generic format for the baseline profile and example are provided here: Generic: Airport-OperationType-AircraftType-Stagelength Example: JFK-A-B737-1 AEDT profiles are one of two types, procedural or fixed-point. Procedural profiles are a series of ordered aerodynamic and thrust instructions used to calculate aircraft trajectories. They are dynamic and adapt to study and track data (such as airport elevation, temperature, pressure, bank angle, etc.). They work in conjunction with other aircraft-type specific flight performance data to allow AEDT to calculate trajectories. Fixed-point profiles consist of a static set of points each with its own defined distance, altitude, speed, and thrust values irrespective of study and/or track data. Their trajectories are explicitly defined. This research only works with aircraft types with procedural profiles, as they are the only aircraft types for which the underlying flight performance data needed to calculate new trajectories are available. The set of “STANDARD” profiles in AEDT 2a SP2 is a mix of both procedural and fixed-point types totaling to 732 profiles, 155 approach profiles and 577 departure profiles. An inventory of all 732 “STANDARD” profiles at each of the 29 Study Airports was created to represent all possible baseline profiles. Calculating trajectories at each airport for the 732 profiles was necessary for procedural profiles due to their dynamic nature. Technically, each runway would also produce a slightly different profile as well due to elevation and slope differences, but it is assumed that these differences are negligible for this study. Baseline trajectories were calculated by first creating an AEDT Standard Input File (ASIF) of all profiles in the inventory. The ASIF is in Extensible Markup Language (.xml) format and is the standard format for inputting data into AEDT. The file included a straight vector ground track either arriving or departing from a selected runway at the given airport for each profile in the inventory. The input vector tracks were created with sufficient length to allow the profiles to be modeled in their entirety. If the length of the input vector track exceeds the distance necessary for that profile to reach its “end” as dictated by the profile definition in AEDT (i.e., 6,000 ft. AFE for approaches or 10,000 ft. AFE for departures), then the output trajectory is truncated by AEDT during runtime at the final altitude and the output trajectory ends. Each profile in the inventory was modeled in AEDT 2a SP2 using the appropriate input vector track for the given airport and operation type with outputs archived for later use. The AEDT runs were all done using “Airport Weather”, or airport-specific average annual meteorological data that is stored in the AEDT AIRPORT database. This is the most commonly used weather setting used for regulatory analyses. Candidate Trajectory Creation Creating a prioritized list of candidate profiles began with analyzing the PDARS data to create of a set of all possible candidate trajectories.

26 PDARS Trajectory Manipulation The first task was to filter the PDARS data based on two criteria. The first criterion was a check that both the origin and destination airports are known. This data must be known in order for AEDT to accurately calculate stage length. The second filtering criterion is a test of the PDARS operation’s aircraft type. Within the 732 “STANDARD” profiles, there are 155 different aircraft types. To make an equitable comparison, the aircraft type of the baseline trajectory would only be compared to candidate trajectories with a matching aircraft type. For example, if a PDARS operation has an aircraft type of “B737”, representing a Boeing 737-700, then this operation was kept because AEDT has a “737700” aircraft type representing the same aircraft. If a PDARS operation did not have matching AEDT aircraft type, the operation was removed from the candidate trajectory process. The subset of aircraft types found in the PDARS data after the completion of the filtering process mapped to 69 aircraft types which are shown in Table 3-6. This table represents the set of AEDT aircraft that were available in the PDARS data to use for the creation of candidate profiles. Table 3-6 Set of AEDT Aircraft Types Used for Candidate Profiles AEDT Aircraft Type Aircraft Description 727100 Boeing 727-100 737300 Boeing 737-300 737400 Boeing 737-400 737500 Boeing 737-500 737700 Boeing 737-700 737800 Boeing 737-800 747200 Boeing 747-200 747400 Boeing 747-400 757300 Boeing 757-300 767300 Boeing 767-300 767400 Boeing 767-400 Extended Range 777200 Boeing 777-200 Extended Range 1900D Raytheon Beechcraft 1900-D 727D17 Boeing 727-200 727EM1 Boeing 727-100 (Cargo) 727EM2 Boeing 727-200 (Cargo) 737N17 Boeing 737-200 74710Q Boeing 747-100 747SP Boeing 747SP 757PW Boeing 757-200 757RR Boeing 757-200 A300-622R Airbus A300-622R A300B4-203 Airbus A300B4-200 A310-304 Airbus A310-304 A319-131 Airbus A319-131 A320-211 Airbus A320-211 A321-232 Airbus A321-232 A330-301 Airbus A330-301 A330-343 Airbus A330-343 A340-211 Airbus A340-211 A340-642 Airbus A340-642 A380-841 Airbus A380-841 BAC111 British Aircraft Corporation One-Eleven

27 AEDT Aircraft Type Aircraft Description BEC58P Beechcraft 58P Baron C130 Lockheed C-130 Hercules CL601 Bombardier Challenger 601 CNA172 Cessna 172 Skyhawk CNA182 Cessna 182 Skylane CNA441 Cessna 441 Conquest II CNA500 Cessna 500 Citation I CNA510 Cessna 510 Citation Mustang CNA525C Cessna 525 Citation Jet CNA55B Cessna 550 Citation Bravo CNA680 Cessna 680 Sovereign CNA750 Cessna 750 Citation X DC1030 McDonnell Douglas DC-10-30 DC3 Douglas DC3 DC870 McDonnell Douglas DC-8-70 DC910 McDonnell Douglas Tanker 910 DC93LW McDonnell Douglas DC-9-30 DHC6 de Havilland Canada DHC-6 Twin Otter DHC8 De Havilland Canada Dash 8 - Q400 DHC830 De Havilland Canada Dash 8 - DHC-8-300 DO328 Dornier 328-100 ECLIPSE500 Eclipse 500 EMB120 Embraer EMB120 Brasilia EMB145 Embraer ERJ 145 FAL20 Dassault Falcon 20 GII Grumman Gulfstream II GIIB Grumman Gulfstream IIB GIV Grumman Gulfstream IV GV Grumman Gulfstream V IA1125 IAI Astra 1125 KC135 Boeing KC-135A Stratotanker KC135R Boeing KC-135R Stratotanker LEAR25 Bombardier Learjet 25 LEAR35 Bombardier Learjet 35 MD11GE McDonnell Douglas MD-11 MD9025 McDonnell Douglas MD-90 Once the PDARS operations with insufficient or non-matching data were removed, there were a total of 319,945 operations to be further processed. This remaining set of operations had their trajectory converted into a two-dimensional representation of the trajectory based on cumulative ground track distance and altitude. The conversion of each trajectory eliminated lateral differences among trajectories and allowed for equitable comparisons, not only to each other, but also to the straight-line ground track-based trajectories used in the baseline trajectory creation. The transformation was done on a per airport basis so that all tracks arriving to and departing from a given airport were referenced to/from the same runway. This method resulted in a straight trajectory with no turns that extended from the runway end chosen to represent that airport. Some PDARS trajectories contain ground roll and/or taxi track points. This generates some bias because the conversion of trajectories is based on cumulative track distance and not all trajectories contain analogous ground roll and/or taxi segments. To remove variability of

28 trajectories close to the runway, any and all ground roll or taxi track points were identified and removed. TRACON radar data, including that from PDARS, is often erratic close to the ground for various reasons. To remove any influence from that issue, all of the PDARS trajectories were normalized in reference to their runway end by manipulating the trajectories at altitudes below 1,000 ft. AFE. This manipulation is not expected to adversely impact the details of the generated candidate profiles as flown trajectories are generally very consistent this close to the ground. For departure operations, the first track point was placed at the runway end and the second track point was placed at a point equal to 2/3rds of the length of the selected runway, measured from the runway end. The second point’s altitude was determined by linearly interpolating the altitudes of the two runway ends, thus falling on the runway and accounting for the slope of the runway. The first step in determining the third track point’s location was to calculate an average slope based on the section of radar track that was 1,000 feet above the runway end altitude. Next, the third point’s location was forced to fall along the calculated slope when projected from the second track point. The location of all track points were adjusted as necessary to ensure that each track point’s relative location to each other remained constant. Exhibit 3-2 displays how a departure trajectory was manipulated to ensure that trajectories around the runway were treated similarly in order for fair comparisons to be made among trajectories. From this diagram, it can be seen that the slopes of the section of the original PDARS trajectory (yellow line) under 1,000 ft. AFE are averaged to calculate a new slope that is projected from the second point on the adjusted PDARS trajectory (dashed red line). Exhibit 3-2 PDARS Trajectory Manipulation - Departures For arrival operations, the initial position of the final track point was placed at a point equal to 1/3rd the length of the selected runway beginning at the runway end. The initial position of the penultimate track point was computed by analyzing the section of radar track that was 1,000 feet above the runway end altitude and calculating an average slope. This calculated slope was then projected from the final track point. The initial position of the penultimate track point was moved to fall along the calculated slope over the runway end. Due to the AEDT requirement that the final track point of an arrival must be located at the runway end, all track point locations were translated as necessary to achieve this and guarantee all spacing between track points remained relatively correct. Exhibit 3-3 displays the trajectory manipulation for arrival trajectories

29 before translating the track so that the final track point is at the runway end. The figure shows that the slopes of the section of the original PDARS trajectory (yellow line) under 1,000 ft. AFE are averaged to create a new slope that is projected from the last point on the adjusted PDARS trajectory (dashed red line). The penultimate track point is moved to be along the projected slope at a location directly above the runway end (grey dashed line). Exhibit 3-3 PDARS Trajectory Manipulation - Arrivals After PDARS trajectories were manipulated to account for variability around the runway, the trajectories were visually inspected. These trajectories were evaluated to identify trajectories that terminated/began a significant distance from the runway end. If trajectories with this quality were identified, they were removed from the data set. In parallel with that effort, an inspection to recognize abnormal trajectories was performed and these tracks were also removed from the data set. After the visual inspection, there were 311,254 operations remaining in the data set. Level-Off Identification & Analysis From observing PDARS radar trajectories, it was apparent that the most prominent characteristic that defines one profile as distinct from another profile is the nature of its level segments. Level-off information such as altitude of level-off, distance of level-off, duration of level-off, and location of level-off was generated for each of the 311,254 PDARS radar trajectories. Since the “STANDARD” profiles generated from the baseline trajectory creation end at 10,000 ft. AFE, the level-off analysis did not extend past this altitude. AEDT’s most commonly used profile customization mechanism, altitude controls, does not allow users to modify an aircraft’s profile below 500 ft. AFE; therefore, this was set as the lower bound of the level-off analysis. For the level-off analysis, each radar trajectory was analyzed by computing and testing the slope and altitude difference of each segment along the track. The first step in identifying a level-off was to identify trajectory segments that had a slope between -0.4° and 0.4° or an altitude difference less than 100 ft. between consecutive trajectory points. The next step was to sum the distances of consecutive track segments that were identified to have a level segment. If the sum was greater than 3 nautical miles (NM), the level segment was considered to be a level-off. If a radar trajectory’s longest level segment was less than 1 NM, the trajectory is

30 considered to not have a level-off. If a radar trajectory’s longest level segment is between 1 and 3 NM, the track was excluded from the analysis. These trajectories were excluded because they would dilute the trajectories that were characterized by significant level-off as their level-off was brief. It would also dilute the trajectories that were not considered to have a level-off because a minor level segment did exist on this trajectory. This filter caused 20,530 radar trajectories to be removed from the analysis. Creating level-off statistics at regular altitude levels simplifies the analysis and reduces the number of unique groups by pushing level-offs occurring at similar altitudes into the same group. To accomplish this, each identified level-off is forced or “snapped” to an altitude. The only altitudes available are altitude levels that are commonly flown at by aircraft, which are traditionally in 500 ft. increments in mean sea level (MSL) altitude (i.e., 6,500 ft., 7,000 ft., 7,500 ft., etc.). Thus, each identified level-off is “snapped” to the nearest 500 ft. altitude increment. After all trajectories were tested, each was analyzed in an attempt to determine the trajectory’s primary and secondary level-off, if applicable. If a trajectory did not have an identified level-off, it was given a “NoLVL” designation. If exactly one level-off was identified on a trajectory, then it was assigned to be the primary level-off and the secondary level-off was null. Any trajectory that had two or more identified level-offs required them to be ranked. This ranking was determined by ordering the level-offs by distance (NM). The longest level-off was recorded as the primary level-off and the second longest was recorded as the secondary level-off. Once the level-off data was generated, a visual inspection of the level-offs was performed to ensure that the accuracy of the results of this methodology was sufficient. After this validation was done, the radar trajectories were geometrically averaged together and then evaluated against the analogous baseline profile trajectories as described below. Candidate Profile Evaluation In order to evaluate all of the candidate profiles, a metric was created that could rank the candidates based on two factors: how different their trajectory was from the analogous baseline trajectory and the frequency of the candidate profile in the PDARS data. These two factors together balance how different a baseline and candidate trajectory are with how frequently that candidate trajectory is flown, according to PDARS radar data counts. Trajectory Score The first evaluation factor was the trajectory score. The objective of the trajectory score is to measure the difference between a candidate trajectory and the analogous baseline trajectory. This can be simplified to strictly differences in profile (i.e., altitude) only as the two-dimensional plan view (i.e., with respect to latitude and longitude) of both trajectories is identical and has been created to be a straight track. By plotting cumulative track distance versus altitude and taking samples every 1 NM along the track, the absolute difference between the baseline trajectory altitude and the candidate trajectory altitude was recorded at each sampled distance and then summed. In an effort to equally weight longer profiles and shorter profiles, the sum over the trajectory was then normalized by the number of samples taken. The resulting number was the trajectory score for that candidate profile. Trajectory scores are dimensionless and, in theory, can range from zero to positive infinity. The minimum trajectory score is zero, indicating that the baseline and candidate trajectories were identical, matching in altitude at every sample taken. The difference in altitudes is an absolute value, as only the magnitude of deviation from the baseline trajectory is considered. Whether

31 the candidate trajectory is above or below the baseline trajectory is not taken into consideration in the trajectory score. Frequency of Profile in PDARS Data The second factor that was used to evaluate candidate profiles was the frequency of which that particular profile appeared in the PDARS data. During the evaluation and prioritization of the candidate profiles, all frequency counts were kept and updated as necessary. Candidate Profile Iteration Process The analysis to prioritize the candidate profiles was done in an iterative fashion, with the results of each iteration being evaluated to inform the process used in the next. This section describes the three different interactions that were performed to achieve a prioritized list of candidate profiles. Iteration #1 Iteration #1 was the first effort to generate a prioritized list of candidate profiles; however this iteration was imperfect and shortcomings were revealed that would be corrected in the next iteration. 3.2.6.1.1 Profile Grouping Methodology The first step in the first iteration of creating candidate profiles was to group profiles together based on the four aforementioned parameters (airport, operation type, aircraft type, and stage length), as well as the primary and secondary level-offs calculated from the level-off analysis. Each group’s set of trajectories was then averaged to generate one trajectory to represent the candidate profile that would be modeled in AEDT. Each group of radar tracks was given a unique key to define its candidate profile. The generic format and example are provided here: Generic: Airport-OperationType-AircraftType-Stagelength-PrimaryLevelOff-SecondaryLevelOff Example: JFK-A-B737-1-9000-6000 The example can be understood as the average track to represent the candidate profile of all radar trajectories that arrived at JFK in a Boeing 737-700 with a stage length of 1 and had a primary level-off at 9,000 ft. MSL and a secondary level-off at 6,000 ft. MSL. The average track was constructed by sampling the composite tracks of the group every 1 NM in ground distance and computing an average location at that sample. This process resulted in the construction of 22,251 average tracks. A trajectory score was computed and the frequency was updated to correctly sum based on the count of radar trajectories belonging to that candidate profile group. The trajectory score was computed between the average trajectory and the analogous baseline trajectory. For example, a trajectory score was computed between the trajectory of the “JFK-A-B737-1-9000-6000” candidate profile and the AEDT output trajectory of the “JFK-A-B737-1” baseline profile. The next step was to consolidate aircraft types into one of seven aircraft classes. The seven aircraft classes were heavy jet (HJ), large jet (LJ), small jet (SJ), large turbo (LT), small turbo (ST), large piston (LP), and small piston (SP). A mapping of aircraft type to aircraft class was created and used to eliminate aircraft type from the profile definition, thus making the profile more generic. For example, a Boeing 737-700 and an Airbus A320-211 both map to the large jet aircraft class, therefore both of these candidate profile groups would consolidate into the large jet aircraft class, provided that all of the other four parameters of the profile group key matched.

32 It was during this step that it was realized that the LP aircraft class was composed of only one aircraft type (Douglas DC3, “DC3”) with few counts. It was decided that this aircraft class be dropped from any further analysis. Based on the aircraft type/aircraft class mapping, a new average track was generated to represent the aircraft class candidate profile. This new average track is an average of aircraft type specific average tracks that replaces aircraft type with the more generic category of aircraft class. Each group of aircraft class average tracks was assigned a new unique key to define its candidate profile based on the aircraft type/aircraft class consolidation. The generic format and example are provided here: Generic: Airport-OperationType-AircraftClass-Stagelength-PrimaryLevelOff-SecondaryLevelOff Example: JFK-A-LJ-1-9000-6000 The example can be understood as the aircraft class average track to represent the candidate profile of all averaged large jet tracks that arrived at JFK with a stage length of 1 and had a primary level-off at 9,000 ft. MSL and a secondary level-off at 6,000 ft. MSL. This consolidated the number of aircraft type specific average tracks down to 7,463 aircraft class averaged tracks. Exhibit 3-4 summarizes the process of creating an aircraft class average track beginning with PDARS radar trajectories.

33 Exhibit 3-4 Creating an Aircraft Class Averaged Track The third step was concerned with providing AEDT an aircraft type to model the aircraft class average track representing the candidate profile. This step was necessary because AEDT requires an aircraft type to be specified in order to model an operation. However, the construction of the aircraft class averaged tracks specifically removed the aircraft type in an effort to make the profiles more generic. The aircraft type selected was the aircraft type that had the worst characteristics based on a combination of trajectory score and frequency. For example, if the aircraft type chosen to serve as a proxy for the aircraft class average track of “JFK-A-LJ-1-9000-6000” was an Airbus A319-131 (“A319”), the aircraft class average track’s data was updated to reflect that (“JFK-A-A319-1-9000-6000”) for use in AEDT. After choosing an aircraft type, the aircraft class average track still matched the candidate profile, it just had a specific aircraft type assigned to it.

34 After the representative aircraft type was chosen, a trajectory score for each aircraft class average track was generated by computing an average of the trajectory scores of the average tracks that comprised the aircraft class average track. Although the trajectory score is not actually computed between the aircraft class average track and the baseline profile, it still represents a metric to describe the similarity of the two profiles. Extending the previous example, the trajectory scores of the average tracks that were consolidated into the “JFK-A-LJ- 1-9000-6000” candidate profile were averaged to compute a new trajectory score for that aircraft class averaged candidate profile. The frequency of each aircraft class averaged track was again updated based on the aircraft class consolidation. The next step was to categorize each of the aircraft class average tracks by operation type and aircraft class (i.e., heavy jet arrivals) to prioritize the candidates within their respective groupings. Within a given operation type/aircraft class combination, each aircraft class average track was given a rank based on that candidate’s trajectory score where a rank of “1” indicated the highest trajectory score (i.e., the most different from the analogous baseline profile). Each aircraft class average track was given another rank based on that candidate’s frequency where a rank of “1” indicated the highest frequency. In either case, aircraft class average tracks with the same value were given the same rank. To prioritize the candidate profiles, the two rankings were summed and then sorted in ascending order. At this time, the candidate profiles with the lowest combined rank represented the prioritized list of profiles that were most poorly represented by AEDT “STANDARD” profiles. The six worst ranked arrival profiles and four worst rank departure profiles in each aircraft class were identified as the common approach and departure profiles that not currently available in AEDT. 3.2.6.1.2 Shortcomings of Iteration #1 The results of iteration #1 were deemed to be too airport dependent. Using the developed metrics would result in prioritized candidate profiles being biased towards airports at high elevations or with more unusual meteorological data, with no clear way forward for generalizing them to create candidate profiles that are suitable for use at all airports as is the intent of default AEDT profiles. Also, there was a desire to simplify the grouping process and recording additional information, both for the sake of simplifying the process of developing the candidate profiles. Iteration #2 Iteration #2 removed the airport grouping from the candidate profile generation process and made several other adjustments relative to iteration #1 as described below. 3.2.6.2.1 Profile Grouping Methodology The first step in the second iteration of creating candidate profiles was to group profiles together based on operation type, aircraft type, stage length, and level-off information. To correct for bias of airport present in Iteration #1, airport was removed as a defining parameter for candidate profiles. To reconcile this, all PDARS trajectories were translated to arrive or depart from the selected runway at ATL. The altitudes for each trajectory were translated so that they maintained the same relative profile in AFE from their respective airport when translated to ATL. As before, all trajectories translated remained “flattened” into a straight line based on cumulative ground track distance and altitude. The methodology behind how level-offs were calculated differed from the Iteration #1 process in five ways. First, the lower bound of the analysis changed from 500 ft. AFE to 1,000 ft. AFE. Second, the available set of altitudes that level-offs were forced or “snapped” to were modified

35 from 500 ft. increments to 1,000 ft. increments (5,000 ft., 6,000 ft., 7,000 ft., etc.). Third, secondary level-offs were no longer considered; only primary level-offs were recorded. Fourth, all level-offs were recorded in AFE with respect ATL’s field level, rather than MSL. Fifth, the method in which the primary level-off was identified was altered. Two new pieces of information were recorded when analyzing radar trajectories for level-offs: the distance from the airport to the level-off and the length of the level-off. For an arrival, the distance from the airport to the level-off is measured to the end of the primary level-off. For a departure, the distance from the airport to the level-off is measured to the beginning of the primary level-off. Exhibit 3-5 depicts how this measurement was calculated for arrivals and departures. Exhibit 3-5 Distance From Airport to Level-off Any trajectory that had one or more identified level-offs required them to be ranked. This ranking was executed by weighting the distance from the airport to the level-off and distance of the level-off itself. Level-offs occurring closer to the airport and for longer distances were given more weight than those occurring further away and for shorter distances. After ranking all level- offs for a trajectory, the level-off atop the rankings was recorded as the primary level-off. These two new parameters expanded the current definition of a candidate profile. Due to these two new parameters containing a large spectrum of values, each of them was placed into ranges within their own respective operation type/aircraft class combination by analyzing the distribution of each data subset. Critical points of the distribution were determined and cutoffs for the ranges were made at these locations in the distribution. For example, the entire set of calculated distances of a level-off for all heavy jet arrivals was analyzed and it was determined that each candidate profile’s primary level-off would fall into exactly one of the following ranges (in NM): 3 to 9, 10 to 19, 20 to 29, 30 to 39, and 40 and up. Creating ranges for both values was repeated for each of the twelve aircraft class/operation type combinations (6 aircraft classes, 2 operation types). The ranges identified and the counts of each range per aircraft class/operation type combination are displayed in Table 3-7.

36 Table 3-7 Iteration #2 – Ranges & Counts of Distance of Level-off Operation Type Aircraft Class Distance of Level- Off (NM) Count Arrival Heavy Jet 3 to 9 384 10 to 19 382 20 to 29 255 30 to 39 141 40 and up 86 Large Jet 3 to 9 567 10 to 19 592 20 to 29 457 30 to 39 303 40 to 49 187 50 to 59 107 60 to 69 64 70 and up 66 Large Turbo 3 to 9 120 10 to 19 162 20 to 29 159 30 to 39 126 40 to 49 108 50 to 59 87 60 to 69 61 70 and up 56 Small Jet 3 to 9 279 10 to 19 272 20 to 29 151 30 to 39 77 40 and up 65 Small Prop 0 to 20 18 21 to 40 19 41 to 60 19 61 to 80 11 81 and up 44 Small Turbo 3 to 9 67 10 to 19 78 20 to 29 66 30 to 39 47 40 to 49 33 50 to 59 29 60 to 69 17 70 and up 25 Departure Heavy Jet 3 to 9 345 10 to 19 145 20 and up 24 Large Jet 3t o 9 521 10 to 19 287 20 to 29 73 30 and up 53 Large Turbo 3 to 9 39 10 to 19 33 20 to 29 29 30 to 39 20

37 Operation Type Aircraft Class Distance of Level- Off (NM) Count 40 and up 41 Small Jet 3 to 9 74 10 to 19 40 20 and up 36 Small Prop 0 to 20 14 21 to 40 25 41 to 60 20 61 to 80 18 81 and up 37 Small Turbo 3 to 9 32 10 to 19 24 20 to 29 17 30 to 39 12 40 to 49 12 50 to 59 10 60 to 69 9 70 and up 12 This same analysis was done for the length of the primary level-off. For example, the entire set of calculated distances from the airport to the level-off for all heavy jet arrivals was analyzed and it was determined that each candidate profile’s primary level-off would fall into exactly one of the following ranges (in NM): 0 to 4, 5 to 14, 25 to 34, 35 to 44, and 45 and up. The ranges identified and the counts of each range per aircraft class/operation type combination are displayed in Table 3-8. Table 3-8 Iteration #2 – Ranges & Counts of Distance from Airport to Level-off Operation Type Aircraft Class Distance From Airport To Level-off (NM) Count Arrival Heavy Jet 0 to 4 78 5 to 14 324 15 to 24 295 25 to 34 288 35 to 44 164 45 and up 99 Large Jet 0t o 4 196 5 to 14 530 15 to 24 498 25 to 34 462 35 to 44 324 45 to 54 193 55 to 64 87 65 and up 53 Large Turbo 0 to 5 56 6 to 15 178 16 to 25 169 26 to 35 163 36 to 45 136 46 to 55 90 56 to 65 49

38 Operation Type Aircraft Class Distance From Airport To Level-off (NM) Count 66 and up 38 Small Jet 0 to 4 47 5 to 14 224 15 to 24 249 25 to 34 195 35 to 44 90 45 and up 39 Small Prop 0 to 4 9 5 to 14 36 15 to 30 25 30 to 50 30 51 and up 11 Small Turbo 0 to 4 23 5 to 14 78 15 to 24 84 25 to 34 77 35 to 44 58 45 and up 42 Departure Heavy Jet 0 to 8 80 9 to 18 307 19 to 28 114 29 and up 13 Large Jet 0 to 8 204 9 to 18 511 19 to 28 198 29 and up 21 Large Turbo 0 to 6 18 7 to 16 57 17 to 26 53 27 to 36 21 37 and up 13 Small Jet 0 to 7 33 8 to 14 75 15 to 24 37 25 and up 5 Small Prop 0 to 9 15 10 to 24 60 25 to 40 28 41 and up 11 Small Turbo 0 to 7 11 8 to 14 41 15 to 24 46 25 and up 30 After these changes were implemented, the number of trajectories removed from the dataset due to their longest level segment being between 1 and 3 NM slightly decreased from 20,530 to 20,439 for Iteration #2. Subsequent to all level-off information being re-calculated, the second step in Iteration #2 was to group the profiles together into an average track using the identical methodology described

39 for Iteration #1 for creating an average track. Each average track was given a unique key based on the new set of parameters being used to define a candidate profile. The generic format and example are provided here: Generic: OperationType-AircraftType-StageLength-PrimaryLevelOffAltitude(AFE)- LengthofLevelOffRange-DistFromAirportRange Example: A-B737-1-3000-40to49-5to14 The example can be understood as the average track to represent the candidate profile of all radar trajectories arriving in a Boeing 737-700 with a stage length of 1 that has a primary level- off occurring at 3,000 ft. AFE for a distance between 40 and 49 NM where the primary level-off ended at a distance between 5 and 14 NM from the airport. This process resulted in 8,470 average tracks. As previously done in Iteration #1, a trajectory score was computed between each average trajectory and the analogous baseline trajectory. For example, a trajectory score was computed between the trajectory of the “A-B737-1-3000- 40to49-5to14” candidate profile and the AEDT output trajectory of the “A-B737-1” baseline. The frequency was updated as well to accurately sum based on the count of radar trajectories comprising the candidate profile. Akin to Iteration #1, the next step was to create a new aircraft class average track. Assuming the five parameters aside from aircraft type matched, aircraft type was again consolidated into its mapped aircraft class and a new aircraft class average track was created. The aircraft class average track was assigned a new unique key to define its candidate profile. The generic format and example are provided here: Generic: OperationType-AircraftClass-StageLength-PrimaryLevelOffAltitude(AFE)- LengthofLevelOffRange-DistFromAirportRange Example: A-LJ-1-3000-40to49-5to14 The example can be understood as the aircraft class average track to represent the candidate profile of all averaged large jet tracks arriving in a Boeing 737-700 with a stage length of 1 whose primary level-off occurred at 3,000 ft. AFE for a distance between 40 and 49 NM where the primary level-off ended at a distance between 5 and 14 NM from the airport. This consolidation process generated 2,199 aircraft class averaged tracks. The issue of providing AEDT an aircraft type to model the aircraft class averaged track was resolved in the same manner as Iteration #1. The aircraft type selected was the aircraft type that had the worst characteristics based on a combination of trajectory score and frequency. Repeating the process of Iteration #1, a trajectory score for each aircraft class average track was generated by computing an average of the trajectory scores of the average tracks that comprised the aircraft class average track. The frequency of each aircraft class averaged track was again updated based on the aircraft class consolidation. Similar to Iteration #1, the final step of prioritizing the candidate profiles was to categorize and rank each of the candidate profiles respective to its operation type/aircraft class combination. The Iteration #1 method of ranking the trajectory score and frequency values and then summing the rank to prioritize the candidate profiles was re-used for Iteration #2. The six worst ranked arrival profiles and four worst rank departure profiles in each aircraft class were identified as the highest priority approach and departure profiles that are not currently available in AEDT.

40 3.2.6.2.2 Shortcomings of Iteration #2 There were two shortcomings of the process used in Iteration #2: first, the level-off analysis extended past the altitude in which baseline approach profiles exist and second, using an average of trajectory scores to be the trajectory score of the aircraft class average track was a biased representation of that metric. When analyzing radar trajectories for level-offs, the analysis ended at 10,000 ft. AFE. This shortcoming reveals itself in arrivals because “STANDARD” approach profiles only extend to 6,000 ft. AFE; therefore, any level-off occurring above 6,000 ft. AFE for an arrival is occurring outside the range of the “STANDARD” profile and should be excluded from the analysis. This shortcoming created a bias by grouping candidate profiles at a level-off altitude that is beyond the extent of the “STANDARD” profile, thus grouping by a variable that only exists in one side of the comparison. This shortcoming did not exist for departures as the “STANDARD” profile for departures extends to 10,000 ft. AFE, matching the extent of the level-off analysis to confirm a fair comparison. The second shortcoming of Iteration #2 was the method used to average trajectory scores of the components of the aircraft class average track. The unintended consequences of this averaging become apparent when the aircraft class average track approximated the baseline trajectory when no individual average track was similar. This shortcoming can best be understood with an example as depicted in Exhibit 3-6. Beginning with the top left chart (yellow “1”) in Exhibit 3-6, imagine an aircraft class average track (red dotted line) with two component tracks, Component Track A (gold line) and Component Track B (orange line), that each has a trajectory score of 1,000. Moving to the top right chart (yellow “2”) in Exhibit 3-6, it can be noted that Component Track A is higher in altitude than the baseline profile (teal line) at every sample and Component Track B is lower in altitude than the baseline profile at every sample. The magnitude above/below the baseline profile of Component Track A/B is equivalent, approximately 400 ft. As depicted in the bottom middle chart (yellow “3”), when these two component tracks get averaged together, they form an aircraft class average track that approximates the baseline profile. This is indicated by the overlap of the red dotted line and the teal line in Exhibit 3-6. The shortcoming of averaging the trajectory scores of the component tracks to resolve the trajectory score of the aircraft class average track as opposed to calculating the resultant trajectory score between the aircraft class average track and the analogous baseline profile becomes apparent during this stage of the example. Averaging the trajectory scores of Component Track A and B produces a trajectory score of 1,000 for the aircraft class average track. As the trajectory score was created to indicate differences in trajectory, a trajectory score of 1,000 would indicate some level of difference between comparing trajectories. However because the aircraft class average track approximates the baseline profile a trajectory score significantly lower than 1,000 would be generated.

41 Exhibit 3-6 Averaging Trajectory Scores Using a trajectory score that is constructed based on averaging the trajectory scores of component parts is inaccurate. A trajectory score must be computed between the aircraft class average track and the analogous baseline profile in order to accurately measure the differences between profiles. Iteration #3 Iteration #3 corrected for the first shortcoming of Iteration #2 by re-performing the level-off analysis to stop after 6,000 ft. AFE for arrivals. Because the level-off analysis occurs at the beginning of the process, the entire set of processes was re-done for arrivals only. The second shortcoming of Iteration #2 was resolved by calculating a trajectory score between the aircraft class average track and the analogous baseline profile rather than computing an average of the trajectory scores of the component tracks. All updates made from Iteration #1 to Iteration #2 were carried over for Iteration #3. The level-off analysis was re-created and candidate profiles were grouped together based on the parameters of the unique key assigned to them. Due to the re-processing of level-offs, the determination of the critical points of the distribution for the distance from the airport to the level- off and distance of the level-off was also re-performed. The creation of ranges at critical points was only necessary for arrivals in Iteration #3 as they were the only subset that re-processing was necessary for. The ranges identified and the counts of each range per aircraft class for arrivals are displayed in Table 3-9.

42 Table 3-9 Iteration #3 – Ranges & Counts of Distance of Level-off (Arrivals Only) Operation Type Aircraft Class Distance of Level- off (NM) Count Arrival Heavy Jet 3 to 9 215 10 to 19 182 20 to 29 124 30 to 39 73 40 and up 38 Large Jet 3 to 9 290 10 to 19 262 20 to 29 205 30 to 39 142 40 to 49 100 50 and up 64 Large Turbo 3 to 9 72 10 to 19 78 20 to 29 59 30 to 39 48 40 to 49 37 50 and up 43 Small Jet 3 to 9 168 10 to 19 128 20 to 29 86 30 and up 60 Small Prop 0 to 20 22 21 to 40 13 41 to 60 14 61 and up 15 Small Turbo 3 to 9 38 10 to 19 33 20 to 29 23 30 to 39 15 40 and up 11 This same analysis was re-done for the length of the primary level-off for arrivals. The ranges identified and the counts of each range per aircraft class are displayed in Table 3-10. Table 3-10 Iteration #3 – Ranges & Counts of Distance From Airport To Level-off (Arrivals Only) Operation Type Aircraft Class Distance From Airport To Level-off (NM) Count Arrival Heavy Jet 0 to 4 74 5 to 13 233 14 to 23 191 24 to 33 112 34 and up 22 Large Jet 0 to 4 152 5 to 14 348 15 to 24 285 25 to 34 185 35 to 44 76 45 and up 17

43 Operation Type Aircraft Class Distance From Airport To Level-off (NM) Count Large Turbo 0 to 5 49 6 to 15 105 16 to 25 84 26 to 35 64 36 and up 35 Small Jet 0 to 4 46 5 to 14 177 15 to 24 143 25 to 34 67 35 and up 9 Small Prop 0 to 20 40 21 and up 24 Small Turbo 0 to 5 24 6 to 15 49 16 to 25 34 26 and up 13 After the level-off data was re-calculated, average tracks were generated for each candidate profile totaling to 2,731 average approach tracks. A trajectory score was computed between the average trajectory and the analogous baseline trajectory. The frequency of each candidate profile was also revised during this phase. At this stage, the departure aircraft class average tracks created in Iteration #2 were added to the 507 approach aircraft class averaged tracks created in Iteration #3 to yield 1,320 aircraft class average tracks for Iteration #3. It was during this step that frequency data was updated and the improved method of calculating a trajectory score between the aircraft class average track and the analogous baseline profile was implemented. Finally, each of the approach candidate profiles was ranked within their respective operation type/aircraft class combination. Prioritizing the six worst approach profiles in each aircraft class was carried out as before by summing the rank of the trajectory score and frequency rank of each and sorting in ascending order. Prioritized Candidate Profiles Table 3-11 displays the final list of prioritized candidate profiles from Iteration #3 along with an identification number, calculated trajectory score and rank, frequency count and rank, and the total number of candidate profiles in that particular operation type/aircraft class combination. The identification number is an index to be used as a reference in this report for other tables. The thick black lines in Table 3-11 delineate one operation type/aircraft class combination from the next. The aircraft type in the candidate profile key is the representative aircraft type used in AEDT modeling. For a definition of each aircraft type see Table 3-12. Table 3-11 Iteration #3 – Values & Ranks for Trajectory Metric and Frequency Trajectory Metric Frequency Total Count of Profiles In Optype/Aircraft Class ID Candidate Profile Key Score Rank Count Rank 1 A-B753-HJ-1-6000-40andup-34andup 2,289.0 2 8 51 95 2 A-B763-HJ-1-3000-20to29-5to13 1,684.4 33 161 13 95

44 Trajectory Metric Frequency Total Count of Profiles In Optype/Aircraft Class ID Candidate Profile Key Score Rank Count Rank 3 A-B763-HJ-1-4000-20to29-14to23 1,870.5 18 41 29 95 4 A-B763-HJ-1-4000-20to29-5to13 1,855.1 20 156 14 95 5 A-B763-HJ-1-5000-20to29-24to33 1,890.9 17 29 33 95 6 A-B764-HJ-1-5000-30to39-14to23 2,182.5 6 10 45 95 7 A-B737-LJ-1-2000-20to29-5to14 2,210.6 15 67 50 136 8 A-B737-LJ-1-4000-40to49-5to14 2,249.1 12 30 64 136 9 A-B738-LJ-1-5000-30to39-15to24 2,046.5 29 77 45 136 10 A-E145-LJ-1-3000-20to29-5to14 1,925.7 46 948 15 136 11 A-E145-LJ-1-3000-40to49-5to14 2,372.5 8 27 67 136 12 A-E145-LJ-1-3000-50andup-5to14 2,500.8 5 24 70 136 13 A-C130-LT-1-3000-50andup-6to15 2,751.3 2 12 54 106 14 A-DH8A-LT-1-4000-10to19-16to25 2,017.2 51 311 6 106 15 A-DH8A-LT-1-4000-40to49-16to25 2,405.3 16 26 39 106 16 A-DH8A-LT-1-5000-40to49-16to25 2,377.7 21 31 34 106 17 A-DH8A-LT-1-5000-40to49-6to15 2,429.2 14 16 44 106 18 A-DH8A-LT-1-5000-50andup-16to25 2,484.4 8 28 37 106 19 A-C550-SJ-1-2000-30andup-5to14 2,529.3 2 10 34 76 20 A-C680-SJ-1-3000-20to29-5to14 2,037.2 17 36 15 76 21 A-C680-SJ-1-3000-30andup-5to14 2,101.1 14 20 22 76 22 A-C750-SJ-1-3000-20to29-15to24 2,170.3 11 14 30 76 23 A-C750-SJ-1-3000-30andup-15to24 2,386.2 5 10 34 76 24 A-C750-SJ-1-4000-20to29-5to14 1,833.5 24 24 18 76 25 A-BE58-SP-1-5000-61andup-21andup 2,608.6 11 9 8 33 26 A-BE58-SP-1-6000-61andup-21andup 2,512.7 14 16 4 33 27 A-C172-SP-1-4000-61andup-0to20 2,563.8 12 6 13 33 28 A-C172-SP-1-5000-61andup-0to20 2,729.1 3 5 16 33 29 A-C182-SP-1-3000-61andup-0to20 2,640.1 7 7 9 33 30 A-C182-SP-1-4000-0to20-0to20 2,492.4 16 13 6 33 31 A-B190-ST-1-2000-3to9-6to15 1,792.4 22 94 9 61 32 A-B190-ST-1-5000-3to9-16to25 1,766.6 24 61 11 61 33 A-E120-ST-1-4000-10to19-6to15 1,704.0 26 185 5 61 34 A-E120-ST-1-6000-30to39-16to25 2,227.5 7 28 19 61 35 A-E120-ST-1-6000-40andup-16to25 2,063.1 11 99 8 61 36 A-E120-ST-1-6000-40andup-26andup 2,377.3 5 34 17 61 37 D-B763-HJ-5-4000-3to9-9to18 1,867.9 52 32 26 224 38 D-B763-HJ-5-5000-3to9-9to18 1,839.7 54 108 10 224 39 D-B763-HJ-5-7000-3to9-9to18 1,979.3 33 17 36 224

45 Trajectory Metric Frequency Total Count of Profiles In Optype/Aircraft Class ID Candidate Profile Key Score Rank Count Rank 40 D-B763-HJ-6-4000-3to9-9to18 1,925.1 44 58 12 224 41 D-B737-LJ-2-2000-3to9-0to8 2,617.6 9 38 57 254 42 D-B737-LJ-2-5000-3to9-9to18 1,959.9 65 458 8 254 43 D-B738-LJ-2-4000-3to9-9to18 2,012.1 58 165 19 254 44 D-B738-LJ-3-9000-3to9-19to28 2,091.8 46 122 25 254 45 D-C130-LT-2-NoLVL-NULL-NULL 3,220.8 1 14 9 101 46 D-DH8A-LT-1-5000-40andup-7to16 2,275.8 27 16 6 101 47 D-DH8C-LT-1-4000-10to19-7to16 2,228.4 32 18 5 101 48 D-DH8C-LT-1-8000-40andup-37andup 2,482.5 16 12 14 101 49 D-C25B-SJ-1-3000-20andup-8to14 3,065.9 5 5 11 54 50 D-C680-SJ-1-2000-20andup-0to7 3,433.3 2 3 20 54 51 D-C750-SJ-1-10000-3to9-8to14 2,523.1 9 5 11 54 52 D-LJ35-SJ-1-4000-3to9-0to7 1,981.9 16 10 5 54 53 D-BE58-SP-1-7000-81andup-25to40 2,114.7 19 7 6 74 54 D-C182-SP-1-4000-21to40-0to9 1,965.7 21 7 6 74 55 D-C182-SP-1-4000-81andup-10to24 2,357.7 10 3 21 74 56 D-C182-SP-1-5000-21to40-10to24 2,221.2 13 6 13 74 57 D-B190-ST-2-NoLVL-NULL-NULL 4,032.0 1 5 32 106 58 D-E120-ST-1-8000-30to39-15to24 2,574.6 17 29 10 106 59 D-E120-ST-1-8000-40to49-15to24 2,508.9 21 127 2 106 60 D-E120-ST-1-9000-50to59-15to24 2,378.1 27 108 3 106 Table 3-12 Candidate Profile Aircraft Types Aircraft Type AEDT Aircraft Type Aircraft Description B190 1900D Raytheon Beechcraft 1900-D B737 737700 Boeing 737-700 Series B738 737800 Boeing 737-800 Series B753 757RR Boeing 757-200 Series B763 767300 Boeing 767-300 Series B764 767300 Boeing 767-300 Series BE58 BEC58P Beechcraft 58P Baron C130 C130 Lockheed C-130 Hercules C172 CNA172 Cessna 172 Skyhawk C182 CNA182 Cessna 182 Skylane C25B CNA525C Cessna 525 CitationJet

46 Aircraft Type AEDT Aircraft Type Aircraft Description C550 CNA55B Cessna 550 Citation II C680 CNA680 Cessna 680 Citation Sovereign C750 CNA750 Cessna 750 Citation X DH8A DHC8 De Havilland Canada Dash 8 - Q400 DH8C DHC830 De Havilland Canada Dash 8 - DHC-8-300 E120 EMB120 Embraer EMB120 Brasilia E145 EMB145 Embraer ERJ 145 LJ35 LEAR35 Bombardier Learjet 35 After the candidate profiles were prioritized, the six worst ranked approach profiles and four worst ranked departure profiles for each aircraft class were modeled using AEDT’s altitude controls functionality at the completion of Iteration #1, #2, and #3. For all three iterations, the trajectory, fuel burn, and noise values were calculated for each candidate profile. At the completion of any given iteration, trajectory data, fuel burn data, and noise values were available for both the candidate profile and the corresponding baseline profile. Each of these was used to evaluate the deviation from currently available profiles in AEDT. Viewing Iteration #1 and #2 as intermediate and working steps to eventually populate the set final candidate profiles in Iteration #3, the results of Iteration #1 and #2 are not be presented. Only the results of Iteration #3 are included in this report. The percent change in fuel burn between the baseline and candidate profiles from Iteration #3 can be found in Table 3-13. Table 3-13 Baseline & Candidate Profile Fuel Burn Comparison ID Candidate Profile Key Percent Change in Fuel Burn from Baseline 1 A-B753-HJ-1-6000-40andup-34andup 16.6% 2 A-B763-HJ-1-3000-20to29-5to13 -1.5% 3 A-B763-HJ-1-4000-20to29-14to23 -7.4% 4 A-B763-HJ-1-4000-20to29-5to13 -7.4% 5 A-B763-HJ-1-5000-20to29-24to33 1.9% 6 A-B764-HJ-1-5000-30to39-14to23 -4.5% 7 A-B737-LJ-1-2000-20to29-5to14 13.2% 8 A-B737-LJ-1-4000-40to49-5to14 -1.1% 9 A-B738-LJ-1-5000-30to39-15to24 -3.8% 10 A-E145-LJ-1-3000-20to29-5to14 -13.3% 11 A-E145-LJ-1-3000-40to49-5to14 -13.3% 12 A-E145-LJ-1-3000-50andup-5to14 -14.1% 13 A-C130-LT-1-3000-50andup-6to15 23.1% 14 A-DH8A-LT-1-4000-10to19-16to25 9.3% 15 A-DH8A-LT-1-4000-40to49-16to25 10.2% 16 A-DH8A-LT-1-5000-40to49-16to25 8.8% 17 A-DH8A-LT-1-5000-40to49-6to15 10.2%

47 ID Candidate Profile Key Percent Change in Fuel Burn from Baseline 18 A-DH8A-LT-1-5000-50andup-16to25 8.1% 19 A-C550-SJ-1-2000-30andup-5to14 29.4% 20 A-C680-SJ-1-3000-20to29-5to14 20.9% 21 A-C680-SJ-1-3000-30andup-5to14 14.2% 22 A-C750-SJ-1-3000-20to29-15to24 20.8% 23 A-C750-SJ-1-3000-30andup-15to24 20.1% 24 A-C750-SJ-1-4000-20to29-5to14 -5.5% 25 A-BE58-SP-1-5000-61andup-21andup 21.6% 26 A-BE58-SP-1-6000-61andup-21andup 6.8% 27 A-C172-SP-1-4000-61andup-0to20 9.5% 28 A-C172-SP-1-5000-61andup-0to20 28.4% 29 A-C182-SP-1-3000-61andup-0to20 33.0% 30 A-C182-SP-1-4000-0to20-0to20 20.5% 31 A-B190-ST-1-2000-3to9-6to15 1.6% 32 A-B190-ST-1-5000-3to9-16to25 -1.3% 33 A-E120-ST-1-4000-10to19-6to15 -0.6% 34 A-E120-ST-1-6000-30to39-16to25 1.2% 35 A-E120-ST-1-6000-40andup-16to25 -1.6% 36 A-E120-ST-1-6000-40andup-26andup 0.1% 37 D-B763-HJ-5-4000-3to9-9to18 -8.2% 38 D-B763-HJ-5-5000-3to9-9to18 -0.8% 39 D-B763-HJ-5-7000-3to9-9to18 50.3% 40 D-B763-HJ-6-4000-3to9-9to18 -1.4% 41 D-B737-LJ-2-2000-3to9-0to8 1.6% 42 D-B737-LJ-2-5000-3to9-9to18 -0.4% 43 D-B738-LJ-2-4000-3to9-9to18 -2.9% 44 D-B738-LJ-3-9000-3to9-19to28 5.4% 45 D-C130-LT-2-NoLVL-NULL-NULL -9.2% 46 D-DH8A-LT-1-5000-40andup-7to16 27.9% 47 D-DH8C-LT-1-4000-10to19-7to16 49.3% 48 D-DH8C-LT-1-8000-40andup-37andup 0.9% 49 D-C25B-SJ-1-3000-20andup-8to14 -5.4% 50 D-C680-SJ-1-2000-20andup-0to7 4.9% 51 D-C750-SJ-1-10000-3to9-8to14 0.3% 52 D-LJ35-SJ-1-4000-3to9-0to7 -13.5% 53 D-BE58-SP-1-7000-81andup-25to40 3.3% 54 D-C182-SP-1-4000-21to40-0to9 1.4% 55 D-C182-SP-1-4000-81andup-10to24 3.7% 56 D-C182-SP-1-5000-21to40-10to24 3.6% 57 D-B190-ST-2-NoLVL-NULL-NULL -50.3% 58 D-E120-ST-1-8000-30to39-15to24 -2.7%

48 ID Candidate Profile Key Percent Change in Fuel Burn from Baseline 59 D-E120-ST-1-8000-40to49-15to24 -2.8% 60 D-E120-ST-1-9000-50to59-15to24 -3.0% The fuel burn changes between the baseline and candidate profiles vary widely in both the positive and negative direction as a function of the specific differences between the profiles. The majority of the differences are quite significant, indicating that there is a meaningful difference between the baseline and candidate profiles. Fuel burn can be thought of as a proxy for emissions, so using the candidate profiles (if and when appropriate) could have a meaningful impact on airport emissions calculations. A receptor set is required by AEDT to compute noise, thus a set of 10 receptors were created beginning at, but not including, the runway end of the input vector track and spaced every 2 nautical miles (NM) extending away from the airport in line with the runway. This created a receptor set directly beneath the input vector track. See Exhibit 3-7 for details, with the red dots identifying the receptor locations. Exhibit 3-7 AEDT Receptor Locations AEDT is able to adjust receptor elevations for terrain when the data is provided, however for this project, terrain data was not used. When terrain data is absent, AEDT sets the elevation of each receptor to the field elevation of the airport as defined in the AIRPORT system database. For each profile in the inventory, noise values were calculated using both LAMAX and Sound Exposure Level (SEL) noise metrics. LAMAX is the maximum A-weighted sound pressure level recorded for that receptor and SEL is the logarithmic measure of A-weighted sound pressure

49 level normalized to a 1 second period, relative to reference sound pressure level. In lay terms, LAMAX is the loudest noise registered at that receptor and SEL is the total amount of noise energy registered at that receptor averaged over time. Table 3-14 shows differences in LAMAX values (in decibels, dB) between the baseline and candidate profiles calculated by AEDT. Table 3-14 Baseline Minus Candidate LAMAX Values (dB) Distance of Receptor from Runway End (NM) ID 2 4 6 8 10 12 14 16 18 20 1 0.0 0.1 0.6 -2.1 -2.3 -4.2 -6.0 -5.9 -6.7 -15.0 2 0.0 0.2 1.1 0.4 0.7 -3.7 -6.0 -8.0 -9.8 -17.6 3 0.0 0.2 1.1 0.4 0.8 0.6 -2.8 -4.8 -6.6 -14.4 4 0.0 0.2 1.1 0.4 0.8 0.6 -2.8 -4.8 -6.6 -14.4 5 0.0 0.2 1.1 0.4 0.7 -2.7 -5.1 -6.9 -7.7 -14.8 6 0.0 0.2 1.1 0.4 0.8 -2.6 -1.8 -3.3 -4.8 -11.9 7 0.0 0.5 0.9 -4.6 -6.3 -8.5 -10.5 -13.2 -15.3 -24.0 8 0.0 0.5 0.9 0.3 0.8 0.6 -4.3 -5.0 -7.6 -15.9 9 0.0 0.5 0.9 0.3 0.8 0.6 -3.3 -3.8 -5.5 -13.9 10 0.0 0.0 2.9 1.2 3.3 -3.1 -6.4 -9.0 -10.5 -20.4 11 0.0 0.0 2.9 1.2 3.3 -3.1 -6.4 -9.0 -8.3 -20.1 12 0.0 0.0 2.9 1.2 3.3 -3.1 -6.4 -9.0 -10.6 -20.6 13 7.0 1.5 -0.5 0.2 -4.5 -5.9 -7.9 -9.6 -11.2 -19.8 14 0.0 0.0 0.0 -1.2 -2.4 -3.0 -3.5 -3.6 -4.1 -9.8 15 0.0 0.0 0.0 0.0 -2.0 -3.2 -3.7 -4.2 -4.7 -10.1 16 0.0 0.0 0.0 -1.2 -2.1 -2.9 -3.2 -3.4 -3.6 -9.0 17 0.0 1.5 1.4 1.0 0.5 0.1 -0.4 -1.5 -2.6 -8.5 18 0.0 0.0 0.0 0.0 -1.5 -1.9 -2.9 -3.1 -3.2 -8.5 19 0.0 1.2 0.2 -5.2 -7.9 -9.9 -11.7 -14.1 -15.7 -25.8 20 0.0 0.8 0.3 0.1 -2.0 -4.3 -5.5 -7.3 -9.0 -18.7 21 0.0 0.8 0.3 0.1 -3.4 -4.2 -5.9 -7.4 -9.3 -19.0 22 0.0 0.0 1.5 -3.1 -6.7 -8.1 -9.5 -10.1 -9.8 -18.0 23 0.0 0.0 1.5 -2.3 -5.4 -7.2 -8.2 -9.1 -9.6 -18.1 24 0.0 0.0 1.6 0.7 0.2 0.0 -2.1 -3.8 -5.1 -14.7 25 0.0 0.2 0.4 -5.3 -7.5 -6.4 -5.7 -7.8 -9.2 -16.5 26 7.6 1.9 -2.3 -2.5 -2.7 -4.3 -4.8 -6.7 -7.6 -14.6 27 0.0 5.0 2.7 0.4 -1.3 0.0 -6.2 -4.8 -8.4 -16.0 28 8.5 2.1 -0.3 -2.3 -6.5 -6.8 -8.3 -10.8 -9.8 -16.1 29 0.5 -0.1 0.0 -0.1 0.0 -2.5 -4.1 -5.5 -6.7 -13.7 30 0.5 -0.1 1.1 1.4 1.9 -0.5 -1.3 -3.0 -2.9 -9.8 31 0.0 0.4 0.0 -3.5 -6.0 -5.2 -4.8 -2.8 -3.6 -9.9 32 0.0 0.4 0.0 0.0 0.0 -0.1 -3.2 -2.0 -3.1 -12.5 33 0.0 0.1 -0.2 -0.1 0.3 0.2 -2.5 -4.2 -5.7 -13.3 34 0.0 0.1 -0.2 1.2 0.9 -0.2 -0.9 -1.6 -2.0 -9.3 35 0.0 0.1 -0.2 -0.1 0.3 0.2 0.0 -1.2 -1.4 -8.2 36 0.0 0.1 -0.2 -0.1 0.2 -2.7 -4.1 -5.7 -6.1 -12.7

50 Distance of Receptor from Runway End (NM) ID 2 4 6 8 10 12 14 16 18 20 37 -0.1 -0.2 -2.6 -4.4 -4.3 1.7 -9.5 -16.2 -22.6 -26.9 38 -1.6 -2.7 -3.3 -4.6 -5.8 -6.9 1.8 -15.1 -23.6 -30.5 39 -9.5 -11.2 -12.7 -11.3 -10.3 -10.1 -10.6 -18.2 -25.6 -32.7 40 0.6 0.6 -0.5 -3.1 -5.3 2.5 -8.7 -9.1 -16.0 -23.1 41 -3.0 -0.5 -9.2 -9.5 -9.0 -10.3 -17.8 -23.4 -27.3 -31.6 42 -3.8 1.3 -4.1 -5.7 -7.3 -9.2 -19.9 -25.6 -29.5 -32.5 43 -1.8 -2.6 -3.9 -5.7 -5.3 -7.2 -14.8 -22.0 -26.6 -30.0 44 -6.2 -5.9 -6.0 -5.6 -5.1 -5.2 -6.5 -14.7 -21.3 -26.2 45 -0.3 0.4 2.0 2.8 4.8 5.1 5.2 5.4 5.7 5.7 46 -10.7 -12.5 -13.0 -13.4 -14.1 -7.2 -6.6 -16.7 -17.1 -22.0 47 -8.8 -10.3 -10.4 -11.0 -11.9 -4.4 -5.5 -6.5 -7.4 -8.2 48 -3.0 -4.4 -4.1 -4.8 -5.5 -5.5 -5.8 -6.0 -5.8 -5.9 49 -7.3 -6.6 -10.1 -12.2 -2.6 -4.8 -12.2 -23.2 -31.2 -37.1 50 -3.6 -7.7 -5.5 0.3 -2.5 -4.8 -6.6 -16.7 -15.3 -26.2 51 7.6 6.7 7.9 7.6 5.7 4.1 3.5 3.3 4.5 3.4 52 -0.1 -0.2 10.9 -5.8 -4.9 -10.3 -10.0 -16.3 -26.1 -33.2 53 -7.3 -7.2 -7.5 -7.0 -7.3 -7.3 -7.3 -7.2 -7.1 -7.0 54 -0.1 -0.1 -0.1 -0.1 1.7 0.5 15.0 14.0 7.7 -3.1 55 -3.0 -2.3 -1.9 -1.6 -2.7 -0.4 9.2 7.9 -6.1 -6.6 56 -11.1 -10.4 -10.1 -9.4 -8.7 -8.6 -8.9 -9.5 -9.6 -9.9 57 -0.4 -0.2 -0.2 -0.2 -0.3 -0.2 -0.2 -0.1 -0.1 -0.1 58 -5.3 -5.1 -4.9 -4.4 -4.4 -11.4 -17.6 -22.6 -26.5 -24.4 59 -5.2 -5.1 -4.7 -4.1 -4.1 -11.1 -17.4 -22.4 -25.7 -22.2 60 -4.7 -4.6 -4.3 -3.8 -3.8 -10.8 -17.2 -22.1 -26.0 -24.5 As with fuel burn, there were significant differences between the noise calculated by AEDT for the baseline and candidate profiles. The noise differences generally increase with distance from the airport, which would be expected as the difference in the profiles is generally greater with greater distance from the airport. As with fuel burn the noise changes are positive for some profiles and negative for others generally as a function of increased or decreased level-offs in the candidate profiles versus the baseline profiles. Of course the most interesting comparison between the baseline and candidate profiles are the trajectories themselves. Exhibit 3-8 displays a comparison of trajectory data for candidate profile “A-B753-HJ-1-6000-40andup-34andup”. The gold line represents the trajectory of the baseline profile that this candidate profile was compared to when computing the trajectory score. The orange line represents the input trajectory for the candidate profile “A-B753-HJ-1- 6000-40andup-34andup” and the teal line represents the output trajectory for the candidate profile “A-B753-HJ-1-6000-40andup-34andup”. Altitude (AFE) is the y-axis of the chart and cumulative track distance (NM) is the x-axis.

51 Exhibit 3-8 A-B753-HJ-1-6000-40andup-34andup Trajectory Comparison It can be noted that the input and output trajectories are approximately the same, but not identical. The AEDT altitude control mechanism used here for expediency during the prioritization process attempts to produce an output trajectory that exactly matches the input trajectory within the constraints of the model. In most cases, AEDT builds the output trajectory very similarly to the input trajectory as seen in Exhibit 3-8. The trajectory scores needed for this step of the research were computed between the baseline and output trajectories. The initial candidate profiles generated using altitude controls for expediency were replaced with candidate profiles built using Aircraft Noise and Performance (ANP) procedure steps in subsequent steps of this research project (see Section 4). Deviations between the input candidate profiles and the AEDT output used to quantify fuel burn and noise differences still exist, but are different from what is shown here due to differences in the AEDT calculation methods. Nevertheless, these altitude control derived candidate profiles are suitable for calculating the fuel burn and noise differences needed for the candidate profile prioritization process. We have produced a comparison chart for each of the final 60 candidate profiles, generated using altitude controls, in Appendix A of this report for reference. (Appendix A of this document is available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx.)

52 New Profile Development and Validation This effort included developing new AEDT Aircraft Noise and Performance (ANP) departure and arrival procedures for each aircraft type included in the AEDT FLEET database based on the candidate profiles described in Section 3.3. Methodology Overview The effort described in Section 3.3 resulted in a set of 60 candidate profiles, of which 36 were approaches, and 24 were departures. Each of these candidate profiles characterizes flight for a certain class of aircraft (heavy jet, small prop, etc.) undergoing an operation within a certain distance range (classified by the "stage length" metric). A procedure to match each radar profile was developed for each AEDT aircraft with the same aircraft classification that has a standard procedure corresponding to the same stage length. The procedures were developed using software capable of modifying and iterating over ANP procedural step types to produce a set of procedural profile steps that best matches a target altitude profile within the limitations of the ANP performance model and the ANP performance data specified for each aircraft type. In all cases the new procedures were created using:  straight ground tracks  airports and runways at mean sea level  the International Standard Atmosphere (ISA)  a headwind of 8 knots Reduction of Targets The candidate altitude profiles created from radar data were treated as a sequence of "targets", and a set of corresponding procedure steps was designed to reach each of their respective targets. The candidate profiles are the result of measured and averaged data that includes more points than are desirable for procedural profile definitions. Therefore, a reduced set of targets was developed subjectively for each radar profile. The resulting "simplified altitude profiles" capture the overall shape of the candidate altitude profiles while avoiding conditions that AEDT is not capable of handling (such as procedure steps with climb angles between 0 and 0.5 degrees). The difference between the original candidate profile points and the reduced set of targets can be seen for each newly developed arrival procedure in Appendix B, and for each newly developed departure procedure in Appendix C. (Appendices B and C of this document are available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx.) Arrival Profile Development Arrival procedures were developed from the ground up using ground roll steps from the standard arrival procedure. For each target, working "backward" from touch-down, a new procedure step was added to the beginning of the procedure. Descent steps were created for non-zero targeted descent angles, and level steps were created where the targeted descent angle was zero. Calibrated airspeed (CAS) at any given point in the procedure was assigned according to the standard procedure's speed schedule (CAS as a function of altitude). Flaps for any step were chosen according to the flap schedule (flap setting as a function of calibrated airspeed) of the standard procedure, based on the following step's initial calibrated airspeed.

53 Departure Profile Development During the process of developing AEDT Aircraft Noise and Performance (ANP) departure procedures that match the averaged radar departure profiles, it quickly became evident that, while steps can be designed to ensure that computed profiles conform to the radar profiles, the overall result is not realistic. In particular, when modeling aircraft to follow the radar profiles, the modeled aircraft speeds reach well beyond the speeds at which they are meant to fly according to their standard ANP procedures and that violate the 250 knot indicated airspeed limit for altitudes below 10,000 ft. MSL. One factor that likely contributes to this phenomenon is that the radar profiles are based on measurements of aircraft operating in real-world conditions, where it is common to use reduced thrust takeoffs. While it is generally accepted that de-rated thrust is often used for commercial aviation departure procedures there are currently no provisions within AEDT for the use of de-rated thrust. It is also generally recognized that the load factor assumptions that went into the generation of AEDT procedure stage length weights are lower than modern load factors resulting in potentially under-predicted aircraft weights. For an altitude profile that can be achieved at lower thrust, the "extra" energy added to the aircraft by the additional thrust is converted to additional speed. Similarly, real-world aircraft do not necessarily fly at the weights specified in ANP profiles, and a deficit in ANP weight compared to in-service weights can also lead to excess energy that is converted to speed. These differences between the real-world and AEDT created an issue regarding the best way for us to represent the real world within the model which is the ultimate goal of this project. To maximize the potential future usefulness of the departure procedures generated in this project, while acknowledging our lack of control of future FAA policy and AEDT implementation decisions, we believe the best approach is to create a range of procedures using a variety of assumptions that bound the problem. This will provide multiple procedures that can be used later as appropriate depending on the results of those future FAA policy and AEDT implementation decisions. Our results could also provide information that would help guide those decisions Although the scope of the present research project is limited to developing profiles without a focus on thrust reduction or aircraft weight, the unrealistic departure procedures that are required in order to match the real-world altitude profiles represented by the candidate profiles prompted additional investigation into the possible role of these peripheral aspects in developing realistic procedures. Specifically, two scenarios were explored in addition to the originally- scoped scenario of procedures with standard thrust and standard weight. In the first expanded scenario, procedures were developed to match the radar profiles using reduced thrust and standard weight. In the second expanded scenario, matching procedures were calculated using both reduced thrust and increased weight. Departure Procedures at Standard Thrust, Standard Weight, Unconstrained Speed 3.3.3.1.1 Thrust Rating Schedule The generated departure procedures use maximum takeoff (MTO) and maximum climb (MCL) thrust. The standard maximum MTO altitude, above which the use of MTO is avoided, is determined from the standard AEDT profile. The actual maximum MTO altitude of the procedure depends on the relationship between the standard limit and the simplified profile. If the simplified profile does not cross the standard maximum MTO altitude near a target point, a new target point is created at that crossing point, and the standard and actual MTO limits are one and the same. If the simplified profile does cross the standard maximum MTO altitude near

54 a target point, the altitude of that target point marks the actual MTO-thrust limit for the procedure. 3.3.3.1.2 Procedure Generation All departure procedures begin with the standard profile takeoff ground roll step. From that point, step creation is driven by targets in the simplified profile; a procedure step is created to reach each target in the simplified profile. The following is done for each target:  Select the step's thrust rating: if below the actual maximum MTO altitude, use the previous step's rating; otherwise use MCL.  Adopt the previous step's final CAS as the present step's initial CAS, and select the step's flaps based on the standard profile flap schedule at that CAS.  Imposing the geometry of flight from the end of the previous step to the target, calculate the acceleration energy share percentage and final CAS (the latter two are fully resolved by specifying thrust rating, flaps, step geometry, and initial CAS).  If the computed final CAS is not at least 0.1 knots faster than the latest CAS, or the computed acceleration is not at least 0.02g, or the target is the last point in the simplified profile, the step is a constant-CAS climb (typically referred to by the shorthand "climb") step to the target altitude (the step will reach the target altitude, but not necessarily at the target distance, and its final CAS is the same as its initial CAS).  Otherwise, the step type is an increasing CAS climb step (typically referred to by the shorthand "accel-percent") based on the computed final CAS and acceleration percentage. Departure Procedures at Reduced Thrust, Standard Weight, Constrained Speed The process of creating a best-match procedure at reduced thrust under a speed constraint involves developing a set of procedures over a discrete space of thrust reduction parameters, identifying which of those procedures satisfy the speed constraint, and then choosing the "best" match from among those candidates. The space of thrust reduction parameters is limited to thrust reductions of 0 to 44 percent, in increments of 2 percent. A 44 percent thrust reduction could be considered very high but we extend our range to this value in order to account for the possibility of combined use of assumed temperature (flex thrust) and de-rate for thrust reduction. For each of these thrust reductions, a candidate procedure is developed through the same process as outlined for procedures using standard thrust, standard weight, and unconstrained speeds, except that the thrust reduction parameter is applied to all rated thrust calculations. The speed constraint is determined by the standard procedure. It is the maximum calibrated airspeed (CAS) called for within the procedure. It is procured by scanning all acceleration steps in the procedure and taking the largest final CAS imposed by those steps. Among the procedures developed at reduced thrust, those that exceed this constraint by more than 1% are eliminated from consideration as candidates for the best fit. Since the primary goal of Task 4 is to match altitude profiles, the metric that determines the "best" match among candidate procedures is a measure of the error in the computed procedure

55 altitude profile with respect to the targeted altitude profile. The error is computed as the magnitude of the difference between the target profile distance and the procedure distance, integrated from the bottom to the top. This is the magnitude of the area of the discrepancy between the altitude profiles. The error metric is the error divided by the integral of the target profile distance from bottom to top (essentially, the area of the “correct” answer). Although the error alone could be used to determine the best match, this metric makes it easier to conceptualize how the error compares to the correct answer independently of the target profile length. The candidate profile with the smallest error metric is selected as the best match at reduced thrust and standard weight. In cases where multiple procedures have the smallest error metric, the one whose maximum CAS is closest to the standard profile maximum CAS is taken as the best match. Departure Procedures at Reduced Thrust, Flexible Weight, Constrained Speed A best-match procedure with reduced thrust and flexible weight under constrained speed is generated similarly to that for its standard-weight counterpart, except that there is an expansion of the space of parameters for which candidate procedures are generated. For this best-match, the parameter space consists of all combinations of the following settings:  thrust reductions of 0 to 44 percent, in increments of 2  all weights from 0 to 100%, in increments of 2, that exceed the standard procedure weight Note that these weight percentages indicate how far the weight is between the BADA minimum and ANP maximum aircraft weights. That is, the actual weight is the BADA minimum weight plus the given percentage of the difference between the ANP maximum weight and the BADA minimum weight. The same selection process that was described for reduced-thrust, standard-weight best- matches is also used for reduced-thrust, flexible-weight procedures. The error metric is the same, as is the speed constraint. Omitted Procedures There are two situations that prevent the generation of a procedure for a given combination of radar profile and ANP airplane. The first occurs when creating departure procedures at standard thrust and weight. In this case, since the speed is unconstrained, it is possible for the airplane to reach speeds at which the available thrust is so low that the maximum supported climb angle is less than the minimum required climb angle of 0.5 degrees. This is an indication that the radar profile is too different from the aircraft performance capabilities represented within AEDT, and no standard-thrust, standard-weight procedure can be recommended. All aircraft for which this situation arises are small piston aircraft. The second situation occurs when creating departure procedures with flexible thrust and/or weight. In this case, sometimes even an aggressive thrust reduction of 44% and the use of the maximum aircraft weight are not enough to keep the airplane slower than the constrained CAS while following the altitudes dictated in the radar profile. Procedure Generation Metrics Summary Tables in Appendix E detail which departure procedures were generated and how well those procedures agree with the corresponding radar profile. Table 3-15 summarizes these outcomes

56 in aggregate, both for each class of airplanes modeled, and for all airplanes overall. (Appendix E of this document is available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx .) Table 3-15 Departure Procedure Generation Metrics Standard Thrust, Standard Weight Reduced Thrust, Standard Weight Reduced Thrust, Increased Weight Pass Fail Err Spd Pass Fail Err Thr Pass Fail Err Thr Wgt All 481 26 8.53 32.39 443 64 14.49 27.61 486 21 12.95 25.58 7.32 HJ 151 0 3.77 39.87 137 14 12.73 28.58 147 4 11.17 25.48 6.97 LJ 228 1 3.32 36.52 196 33 13.2 27.67 227 2 11.93 25.77 8.33 LT 40 8 21.4 24.59 42 6 16.94 27.76 42 6 13.92 25.33 12.8 SJ 26 2 14.8 -1.49 22 6 13.4 26.09 23 5 13.59 26.87 0.96 SP 21 15 66.95 -12.74 31 5 33.25 20.45 32 4 31.09 20.06 2.34 ST 15 0 8.83 36.94 15 0 3.47 34.4 15 0 3.37 34.27 0.37 In this table:  The first column indicates the scope of the row, which is either all aircraft and radar profiles, or just the ones for a certain class. Classes are a combination of weight class and engine type: o Weight classes:  H = Heavy  L = Large  S = Small o Engine types:  J = Jet  T = Turboprop  P = Piston  The "Pass" and "Fail" columns indicate how many procedures were successfully created, and how many were not created, for reasons described previously.  The "Spd" column indicates the average amount by which the generated procedures exceed the standard procedure's maximum CAS, as a percentage.  The "Err" columns indicate the average value of the altitude error metrics, expressed as a percentage.  The "Thr" columns indicate the average amount by which the generated procedures reduce thrust, as a percentage.

57  The "Wgt" column indicates the average amount by which the generated procedures increased weight from the standard procedure's weight, expressed as a percentage of the aircraft weight range. Some observations:  The errors and failure counts in the SS ("standard" thrust, "standard" weight) profiles tend to be lower than the other profile types. This is as expected. The SS profiles are less restricted since they are allowed to exceed the standard maximum CAS.  The SS profiles tend to exceed the maximum CAS. The average speed overshoot of 32% corresponds to about 80 knots for a 250 knot maximum CAS.  The errors and failure counts of the FF ("flexible" thrust, "flexible" weight) profiles tend to be lower than the FS ("flexible" thrust, "standard" weight) profiles. This is as expected. The option to reduce the weight allows more flexibility in keeping speed below the maximum CAS and conforming to the targets.  The thrust reductions in the FF profiles tend to be lower than the FS profiles. This is consistent with expectations: when there is an option to increase weight, there is less need to reduce thrust.  The average thrust reductions are on the order of 25%.  The average weight increases are 0-12% of the weight range. Approach profiles are not included in this table because they all were successfully created, with error metric values of zero, and without any thrust reductions, weight increases or speed violations. Validation Process Our validation process was similar to the one used when checking new procedural profiles supplied by aircraft manufactures before adding them to AEDT and included 1) Internal consistency check, 2) Reasonableness check, and 3) suitability across different atmospheric conditions. We had originally planned to include a consistency check with prior submittals, i.e., custom profiles previously submitted to the FAA for approval; however that was not possible due to the unavailability of details on prior profile approval submittals from the FAA. Summary of Procedure Results Plots of each of the newly generated procedures are presented in Appendix B (arrivals) and Appendix C (departures). (Appendices B and C of this document are available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx.) These plots show altitude versus ground track distance for the targeted original candidate profile and the reduced set of target points used to calculate the new procedures. They also show altitude, speed, and thrust values versus ground track distance for the relevant exiting standard profile and the newly created profile. Some observations from these plots:  The restricted-CAS departure profiles do not match targets as well as the unrestricted case. This is expected. Having full thrust allows the unrestricted-CAS procedures to meet the steeper targets, but this means that there is a lot more thrust than needed to meet the shallower targets, and that excess energy has to be diverted into speed. This is of no consequence for this set of profiles because the speeds are allowed to grow unchecked. In the restricted-CAS cases, the speed has

58 to be contained, so to avoid accumulating a lot of speed during shallow target segments, the thrust needs to be too low to reach the steeper targets exactly.  When a segment ends at a target altitude before the target's distance, this generally indicates that there was not enough thrust available to fly directly to the target while increasing speed enough to satisfy AEDT's minimum CAS change (0.1 knots) or acceleration requirements (0.02g), and so flight had to be modeled by a climb step instead of an acceleration step.  When a segment ends at a target altitude beyond the target's distance, this generally indicates that there was not enough thrust available to fly directly to the target while maintaining CAS, so a climb step was created to reach that altitude. This occurs very often in the CAS-restricted cases, where lower thrusts and higher weights help to prevent CAS increases. When looking at the profiles, it might seem intuitively like a closer match could have been achieved by giving up some acceleration in return for a steeper climb angle, but in reality, that is not possible because the ANP model doesn't support CAS reductions in procedure steps using rated thrusts. New Profiles Developed for AEDT For this project, 840 new approach and 1410 new departure procedural profiles were created. These would provide a much wider selection of choices for AEDT users should they be adopted by and incorporated into AEDT. Exhibit 3-9 overlays all of the existing standard departure profiles (covering all available stage lengths) for the 757PW AEDT ANP aircraft type along with the new 757PW departure procedures developed during this research on an altitude versus distance basis to provide a visual overview of the increase in options. In this Exhibit:  Existing standard profiles have grey lines.  New profiles are displayed as follows: o Line color indicates the candidate profile ID (i.e., which altitude profile was targeted) o Line solidness indicates the level of flexibility in customization:  Solid is standard thrust, standard weight (speed unlimited)  Dashed is flexible thrust, standard weight (speed limited)  Dotted is flexible thrust, flexible weight (speed limited) Detailed plots for each of the new departure procedures developed during this research, including altitude, speed, and thrust values, comparisons to the existing standard AEDT procedures, and the altitude targets used to create the new procedures are included in Appendix C. Separate Excel workbooks containing altitude, speed, and thrust data and plots for the new departure procedures and the relevant STANDARD procedure(s) for each aircraft type are also available. (Appendix C of this document is available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx.)

59 Exhibit 3-9 757PW Existing Standard and New Departure Procedures Exhibit 3-10 overlays the single existing standard arrival profile for the 757PW AEDT ANP aircraft type along with the new 757PW arrival procedures developed during this research on an altitude versus distance basis to provide a visual overview of the increase in options. In this Exhibit:  Existing standard profile is grey.  For the new profiles, line color indicates the candidate profile ID (i.e., which altitude profile was targeted). Detailed plots for each of the new arrival procedures developed during this research, including altitude, speed, and thrust values, comparisons to the existing standard AEDT procedures, and the altitude targets used to create the new procedures are included in Appendix B Separate Excel workbooks containing altitude, speed, and thrust data and plots for the new arrival procedures and the relevant STANDARD procedure(s) for each aircraft type are also available. (Appendix B of this document is available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx .)

60 Exhibit 3-10 757PW Existing Standard and New Arrival Procedures As in AEDT, the procedural profiles and their overall properties produced during this research are defined in one table (the "profiles" table), and the individual steps composing the procedural profiles are defined in a separate table (the "steps" table). Procedural profiles are defined in terms of six properties:  ACFT_ID: the ANP aircraft ID  OP_TYPE: the type of operation  PROF_ID1: the name of the operation  PROF_ID2: the stage length of the operation  WEIGHT: the aircraft weight associated with the profile  DERATE: the percentage by which rated thrusts are reduced This matches how profiles are defined in AEDT, except that DERATE is not present or supported in AEDT. Procedural Profile Steps are defined in terms of 11 properties:  ACFT_ID: the ANP aircraft ID  OP_TYPE: the type of operation  PROF_ID1: the name of the operation  PROF_ID2: the stage length of the operation  STEP_NUM: the order of the step within the profile 0 1000 2000 3000 4000 5000 6000 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 Al tit ud e  (ft  A bo ve  F ie ld  E le va tio n) Distance From Touchdown (AFE) StdA1 Custom; OpType=A; T3ID=07(StageLength=1); StdThrStdWgtFit; StdThr;StdWgt: Custom; OpType=A; T3ID=08(StageLength=1); StdThrStdWgtFit; StdThr;StdWgt: Custom; OpType=A; T3ID=09(StageLength=1); StdThrStdWgtFit; StdThr;StdWgt: Custom; OpType=A; T3ID=10(StageLength=1); StdThrStdWgtFit; StdThr;StdWgt: Custom; OpType=A; T3ID=11(StageLength=1); StdThrStdWgtFit; StdThr;StdWgt:

61  FLAP_ID: the name of the flap setting to use during the step  STEP_TYPE: the type of step  THR_TYPE: the type of thrust calculation  PARAM1: the first step-type-dependent parameter  PARAM2: the second step-type-dependent parameter  PARAM3: the third step-type-dependent parameter This is the same way that steps are defined in AEDT. In the case of these profiles, PROF_ID1 is constructed from the level of flexibility allowed in customizing the profile. The prefix SS ("standard" thrust, "standard" weight) indicates that the profile was determined without de-rating thrust or altering the weight, while the prefix FS ("flexible" thrust, "flexible" weight) indicates that the profile was determined with thrust de- rate allowed but without altering the weight, and the prefix FF ("flexible" thrust, "flexible" weight) indicates that the profile was determined with both thrust de-rate and weight alterations allowed. The 2-digit suffix indicates which of the candidate altitude profiles was targeted. Due to their size, these tables are contained in Appendix D. Tables 3-16 and 3-17 provide a sample representing one arrival profile. (Appendix D of this document is available at: http://www.trb.org/acrp/ACRPWOD36Materials.aspx.) Table 3-16 Sample New Procedural Profiles Table ACFT_ID  OP_TYPE  PROF_ID1  PROF_ID2  WEIGHT  DERATE  707320  A  SS01  1  222300  0  Table 3-17 Sample New Procedural Profile Steps Table ACFT_ID  OP_TYPE  PROF_ID1  PROF_ID2  STEP_NUM  FLAP_ID  STEP_TYPE  THR_TYPE  PARAM1  PARAM2  PARAM3  707320  A  SS01  1  1  ZERO  V  6000  250  314958  707320  A  SS01  1  2  ZERO  V  6000  250  1000  707320  A  SS01  1  3  ZERO  D  6000  250  0.785757  707320  A  SS01  1  4  ZERO  D  5013.715  220  0.785757  707320  A  SS01  1  5  ZERO  V  5000  220  68875.33  707320  A  SS01  1  6  ZERO  V  5000  220  1000  707320  A  SS01  1  7  14  D  5000  220  1.178543  707320  A  SS01  1  8  14  D  3020.572  160  1.178543  707320  A  SS01  1  9  D‐40  D  3000  160  2.356089  707320  A  SS01  1  10  D‐40  D  41.14471  131.5996  2.356089  707320  A  SS01  1  11  D‐40  L  410.6  0  0  707320  A  SS01  1  12  B  3695.4  124.9  40  707320  A  SS01  1  13  B  0  30  10 

62 New Methods to Customize Profiles The new profiles described in Section 3.3 will provide increased coverage of frequently used real-world flight profiles. However, any finite set of default profiles cannot hope to fully cover the entire range of flown profiles. In this section we describe additional methods for AEDT users to fill in the gaps where necessary. Considering the other profile customization methods either already implemented within AEDT or slated for future implementation, there is a need for a mid- level profile customization functionality in terms of complexity and input data requirements. This can be achieved by providing profile editors that can be leveraged and accurately used with less than full trajectory data, and for this step we have developed pseudo code for use by AEDT developers in implementing such editors. Pseudo Code Description The best way to provide profile customization is through profile editors that allow users to simply and quickly modify profiles. The present effort limited the scope of such capabilities to customization based on a small number of easily-obtained parameters that can automatically be used to make the corresponding changes that are anticipated to be most common. The purpose of this limitation is to focus available resources on supporting users with mid-level trajectory/profile information similar to screening data. Due to the heterogeneous nature of the set of available procedure steps, and the tightly-coupled dependencies between step parameters, designing rigorous support for arbitrarily complex approach procedures is resource- prohibitive for the current project. Instead, we restrict support to standard procedures that meet certain structural requirements. These requirements are based on standard profile structures presently encountered, procedure structures implied by the performance specification within AEDT, and standard profile structures implicitly or explicitly envisioned by the proposed customization. Procedures often feature groups of steps of roughly (or precisely) the same type in succession. For the following discussions, we refer to such groups as “sequences” (ex. “a sequence of descents” or a “descent sequence”). Note that the steps that comprise a sequence could have different characteristics from each other, such as specific step type (ex. a descent sequence can feature descend, descend-decel, and descend-idle steps), or drag-to-lift ratio. Furthermore, we refer to the progression of sequences comprising a procedure as the “profile structure”. Custom Approach Treatment The goal of this approach customization is to alter the chosen standard approach procedure such that it features a level step at a user-specified length and altitude. Honoring a user constraint on distance in addition to altitude and length would generally require modifications to the geometry of the descent that follows. That is, we would need to change the descent angle to touch-down. This could easily be inconsistent with the flap setting (R values are calibrated to flight at the corresponding angle). It’s also somewhat suspect in realism, since the approach altitude data described in Section 3.2.1.1 seem to best agree with the portion of standard descent just before touch-down. It is therefore preferable to leave descent angles as defined, and omit support for user control of the track distance at which the level-off occurs. Since 90% of all standard approach procedures in AEDT are comprised of a single descent sequence, followed by landing, we will limit our focus to this procedural structure. All of these AEDT profiles have monotonic requested calibrated airspeed (CAS), so our analysis is also based on this restriction.

63 The recommended customization treatment for approach profiles, in terms of the calibrated airspeed schedule, is illustrated in the figure below, in terms of the altitude profile. The standard altitude profile is shown in blue, and a sample customized profile is shown in red. Each line segment represents one procedure step. Exhibit 3-11 Recommended Approach Customization Treatment As illustrated, a level segment of target length is inserted where the profile reaches the target altitude (splitting the descent step there, if necessary). The CAS of the inserted step is equal to the initial CAS of the descent step it precedes (if a descent step was split to accommodate the custom level step, this CAS will be the result of interpolating the initial CAS values of the step that was split and its original successor). The flap setting of the inserted step matches that of the adjacent step with a lower drag-to-lift ratio. Pseudocode for approach customization is as follows: 1. Gather from user a. desired level altitude b. desired level length 2. Validate standard profile a. only descent, landing, and braking steps 3. if a descent step passes through tgt alt, split it there (interpolate CAS) 4. if tgt alt is above first descent, change first descent altitude to tgt alt 5. add a level step of desired length preceding descent from tgt alt 6. propagate flaps of adjacent descent with lower R to level step Custom Departure Treatment The goal of this departure customization is to alter the chosen standard departure procedure such that it features an acceleration step to fit a user-specified length, initial altitude, and rate of climb (ROC). Attempting to match initial distance would require customization of the steps preceding that acceleration, most likely by iterating a thrust reduction factor, which is beyond the intended scope of the proposed functionality. Distance A lti tu de Top of Standard Ground Target

64 All standard departure procedures in AEDT begin with a takeoff ground roll step. For 71% of them, this is followed by three climb sequences that are separated by two acceleration sequences (a profile structure we will call “TCACAC” for Takeoff, Climb, Accelerate, Climb, Accelerate, Climb). For 22% of them, ground roll is followed by two climb sequences that are separated by a single acceleration sequence (a profile structure we will call “TCAC” for Takeoff, Climb, Accelerate, Climb). We will limit our focus to these two profile structures. Departure profiles derived from averaged radar data generally follow the standard profile to 3k (which includes the first acceleration sequence). This is because the behavior of aircraft nearest the runway is the least flexible. For this reason, we will only ever consider modifying the last increasing CAS climb sequence in a profile. The recommended customization treatment for TCACAC profiles is illustrated in the figure below, in terms of the calibrated airspeed schedule. The standard profile schedule is shown in blue, and sample customized profile schedules are shown in red (one for a target altitude that is lower than the last standard acceleration, and one for a target altitude that is higher). Each line segment represents one procedure step. Vertical segments represent takeoff ground roll or acceleration segments. Relocation of acceleration steps (and adjustment of final CAS, where appropriate) is indicated by dashed arrows. Exhibit 3-12 Recommended TCACAC Customization Treatment As illustrated, when the target altitude is below, the last (second) standard acceleration sequence is moved to the target altitude (splitting the climb step there, if necessary). The final CAS of the sequence is changed to the value achievable over the target length, limited by the maximum CAS in the schedule; acceleration steps that cross that CAS are truncated there, and additional acceleration steps beyond that are discarded. All remaining steps are climbs, and their flaps are changed to match the final remaining acceleration step. If the target is above the second standard acceleration, the treatment is similar, but the flaps for the climbs between the standard acceleration altitude and the target altitude have to be changed to those of the last climb preceding the standard acceleration. No modifications are made to step thrust assignments. Note that conforming to the user’s length will often mean that maximum CAS is not reached. By capping the final CAS to the maximum CAS, this treatment essentially trusts user underestimations of length/CAS, but rejects overestimations. Altitude C A S Top of Standard Takeoff Low Target Standard Acceleration Altitudes High Target Max CAS Min CAS

65 For TCAC profiles, the first acceleration sequence constitutes acceleration to the maximum CAS. There are two options for a custom acceleration beginning at a target altitude. They are illustrated below: Exhibit 3-13 Recommended TCAC Customization Treatment As shown, the options are: 1. reduce the CAS of the first acceleration sequence, and delay the remaining portion to the target altitude, or 2. delay the first acceleration sequence to the target altitude, and change its final CAS to the value achievable over the target length Option 1 conforms to the original procedure for as long as possible, while still leaving room for the requested acceleration length. Option 2 conforms better to what we did for TCACAC (change the last acceleration sequence to match inputs) and would be the simplest algorithm. Option 2 is the recommended treatment. Pseudocode for departure customization is as follows: 1. Gather from user a. desired acceleration initial altitude (tgt alt) b. desired acceleration length (tgt length) c. desired acceleration ROC (tgt ROC) 2. verify a. profile is TCAC or TCACAC b. tgt alt is lower than top of procedure c. tgt alt is higher than i. TGR sequence for TCAC ii. First acceleration sequence for TCACAC 3. if a climb step passes through tgt alt, split it there 4. for first (TCAC) or second (TCACAC) acceleration sequence: a. move to tgt alt b. change ROC to tgt ROC Determined by target length Max CAS Determined by target length C A S Min CAS Altitude Top of Standard Takeoff Target Standard Acceleration Altitude