Defining and Measuring Aircraft Delay and Airport Capacity Thresholds (2014)

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.

4The term delay is quite simple and generally applied when an event occurs later than it was planned, scheduled, or expected to happen. It is a common term used in everyday conversation, not a term that is unique to aviation. However, delay is used or interpreted differently by various stakeholders involved in airport planning studies, airline operational on- time performance analyses, and the public. In general, delays in aviation may describe one of two following situations: • Actual operational or real-time delay events, often compared to flight schedule. For actual flights—current or historical— delays are often measured as the actual times compared to the planned or scheduled times. Schedule may refer to times filed in a flight plan or a published airline schedule. • Mathematical or calculated estimates (using analytical or simulation models) for planning, often compared to un- impeded, nominal, or optimal travel time. Analysts often use computer simulation tools or other analytical procedures to evaluate delays and delay savings. These tools and methods typically calculate a nominal or unimpeded time, then mea- sure any additional time as delay. Although this report briefly discusses the actual operational events, the primary focus is on the analytical delay values used in airport planning. This comprehensive chapter on delay focuses on delay metrics, delay data sources, and approaches and methods to calculating delay. It includes the following items: • Delay used by various stakeholders – FAA Air Traffic – FAA Airports Office – NextGen – Airports – Airlines – Consumers/passengers – General public (including airport neighbors) • Historical/operational delay data – Descriptions of available delay databases – Examples of data analyses – Strengths/weaknesses • Calculations of analytical delay metrics – Spreadsheet/basic models – Simulation models – Average annualized delays – Other delay statistics – Comparison of mathematical delays to operational/ historical data 2.1 How Delay is Used and Defined by Various Stakeholders This section discusses definitions of delay used by various stakeholders involved in airport planning studies, as well as common understandings of delay by the public. 2.1.1 FAA Air Traffic In measuring flight delays, the FAA’s goal is meeting both airline and airport (user) expectations and passenger (end- user) expectations. Usually, FAA’s Air Traffic Organization (ATO) focuses on flights delayed 15 minutes or more from the flight plan that result from the ATC system detaining an aircraft at the gate, short of the runway, on the runway, on a taxiway, and/or in a holding configuration anywhere en route. This includes ground stop delays and delays in expected departure clearance time (EDCT). The cause of delay is also recorded (e.g., weather, volume, equipment, and runway). Data is reported for all U.S. ATC facilities and for all instru- ment flight rules (IFR)-filed aircraft in the United States. More detailed taxi and airspace travel times and calculated taxi delay based on typical unimpeded times for each airport also C H A P T E R 2 Delay

5 are recorded for 77 U.S. airports. This data is now maintained in several databases, which are more thoroughly described in Section 2.2. The current target for the FAA’s performance metric of National Airspace System (NAS) on-time arrival is that 88% of flights at the 30 core airports (determined by FAA to be 29 large hubs and Memphis International Airport) arrive no more than 15 minutes late, based on the flight plan filed with the FAA, and excluding minutes of delay attributed to weather, carrier action, security delay, and prorated minutes for late arriving flights at the departure airport. However, on-time arrivals are combined for all airports such that problems achieving the target at specific airports are not able to be identified and perhaps corrected, an issue noted by the U.S. Government Accountability Office (GAO) as early as 2010 when they rec- ommended airport-specific metrics be adopted and reported. In addition, FAA has estimated that 70% of all aviation delays are caused by weather events, and weather delays are excluded from the on-time metric altogether. However, there may be technology and/or new procedures that could be implemented such that the weather delays would not have such detrimental effects on flight delays. Attributing the lack of technology to a weather delay masks the issue that perhaps different procedures or technology are needed to allow flights to operate in low visibility. Airports Office For airport planning, delay is generally considered as excess travel time—the difference between actual operating time minus a nominal or optimal operating time. This is generally evaluated using analytical tools, spreadsheets, or simulation analysis. The nominal or unimpeded time is not readily avail- able from scheduled flight times. As described in the FAA’s Air- port Benefit-Cost Analysis Guidance (December 1999), “delay is the added trip time attributable to congestion at the study airport, where congestion constitutes any impediment to the free flow of aircraft and/or people through the system.” Several current FAA documents related to airport planning include discussion of delays, including the following: • National Plan of Integrated Airport Systems (NPIAS) 2013– 2017—Throughout the FAA’s current NPIAS, there are discussions of airport delay, congestion, and capacity. How- ever, there is no statement about what exact delay threshold defines capacity. – “Delay is an indicator that activity levels are approach- ing or exceeding throughput capacity levels.” – “The majority of airports in the national airport system have adequate airport capacity and few delays. However, there are airports that continue to experience delays. In 2011, there were five airports with average depar- ture delays of more than 12 minutes per operation and two airports with average arrival delays of more than 14 minutes.” – “The Nation’s air traffic delay problems tend to be con- centrated at certain large hub airports. Delays occur primarily during instrument weather conditions (i.e., reduced ceiling and visibility) when runway capacity is reduced below that needed to accommodate traffic levels. Because of the number of connecting flights sup- ported by these airports, delays among these busy large hub airports can quickly ripple throughout the system, causing delays at smaller airports nationwide.” • FAA’s Airport Master Plan AC 150/5070-6B with Change 1 (May 2007)—“Delay is typically expressed in minutes per aircraft operation, which can be translated into hours of annual delay and easily converted into dollar estimates to be used as a basis for comparison. Traditionally, 4 to 6 minutes of average delay per aircraft operation is used in annual ser- vice volume (ASV) calculation. When the average annual delays per aircraft operation reaches 4 to 6 minutes, the air- port is approaching its practical capacity and is generally considered congested.” • Airport Benefit-Cost Analysis Guidance (December 1999) – “Should simulation modeling reveal that the baseline traffic forecast would lead to average airside, terminal, or landside delays of more than 20 minutes per operation or passenger, the rate of growth in the baseline forecast would need to be adjusted downward. This revision is necessary because approximately 20 minutes represents the highest level of average delay realized in actual prac- tice, even at highly congested airports.” – “Airports experiencing severe delay due to congestion will not be able to accommodate rising demand for air service. Average delay per operation of 10 minutes or more may be considered severe. At 20 minutes of average delay (approximately the highest recorded average delay per operation known to FAA at an airport in the United States), growth in operations at the airport largely will cease. Prior to reaching these levels, airlines would begin to use larger aircraft, adjust schedules, and cancel or consolidate flights during peak delay periods. Passengers would make use of alternative airports, seek other means of transportation (e.g., automobile or train), or simply avoid making some trips.” • FAA Airport Benefit-Cost Analysis Guidance Addendum (June 2010)—Related to determining systemwide impacts in a BCA, this document acknowledges that some delay at a specific airport propagates delay downline as the aircraft continues through the day’s routing. Delay analysis related to airport planning typically is only focused on one specific airport, or “original” delay. This guidance contains multi- pliers for estimating “propagated” delay in a BCA.

6FAA Orders 1050.1, Environmental Impacts: Policies and Procedures, and 5050.4, National Environmental Policy Act (NEPA) Implementing Instructions for Airport Projects do not specifically address delay values or performance goals. Tech- nical analyses supporting the purpose and need in environ- mental studies often use the BCA guidelines noted above for determining when delays are excessive and unreasonable for air service providers and customers/passengers. Also, analy- ses for an environmental impact statement (EIS) at a large congested airport may use the BCA guidance regarding delays when estimating whether the traffic demand would continue to operate when simulated average delays exceed 20 minutes or whether they should attribute a cost to cancelled/diverted flights. NextGen The Next Generation Air Transportation System (NextGen) is a comprehensive initiative across multiple federal agencies to make air travel in the United States NAS more convenient and dependable, while ensuring flights are as safe, secure, and hassle- free as possible. This transformative change, which is already providing benefits, integrates new and existing technologies, including satellite navigation and advanced digital commu- nications. Some of NextGen’s goals include enhancing safety, reducing delays, increasing capacity, saving fuel and reducing aviation’s adverse environmental impact. NextGen involves many areas and disciplines, including new air traffic technology, weather information, data communications, environmental concerns, aviation security, and global harmonization. More specifically related to airports, one of NextGen’s delay reduction benefits is to reduce the impact of weather, achieving similar delays or capacity in instrument meteoro- logical conditions (IMC) as in visual meteorological con- ditions (VMC). Technology and procedures will allow for reduced dependencies between aircraft operating on closely spaced parallel runways. Airports’ runways that have not had instrumentation for arrivals during IMC may be able to have precision-based navigation (PBN) approach procedures to use those runways during low visibility/ceiling conditions. PBN includes area navigation (RNAV) and required navi- gation performance (RNP). RNAV enables aircraft to fly any course using ground- or space-based navigation aids. RNP is RNAV with onboard monitoring and alerting capability. Also, localizer performance with vertical guidance (LPV) approaches are being added to many general aviation air- ports. LPV is operationally equivalent to Category (CAT) I Instrument Landing System (ILS) approaches, and FAA plans to have LPV approaches to all qualified runway ends by 2016. The optimization of airspace and procedures in the metro- plex (OAPM) is an important method by which airspace and procedure design efforts are being incorporated into the NAS. Several OAPM efforts around major airports are cur- rently ongoing across the United States. The FAA is tracking airport performance for NextGen using key performance indicators of capacity, efficiency, and predictability at major airports. The metrics are based on taxi times/delays and flight travel times. The FAA has developed and continues to update a NextGen Performance Snapshot website that reports post-implementation performance data for metroplexes and airports. Currently, these performance snapshots are found at www.faa.gov/nextgen/snapshots. Additional metrics were being reported in the areas of efficiency (average delays/times), predictability (standard deviations), and capacity (average daily operations and peak throughputs) for the top 77 airports. Currently, the metrics reported are reduced to those most useful for tracking NextGen progress and for only the top 30 airports, using the following metrics on an efficiency scorecard: • Average gate arrival delay (minutes per flight), • Average gate-to-gate time (minutes per flight), • Average number of level-offs per flight (count per flight), • Distance in level flight from top of descent to runway threshold (nmi per flight), • Taxi-in time (minutes per flight), and • Taxi-out time (minutes per flight). Although FAA has these dashboards/performance snap- shots for the major airports and notes the NextGen improve- ments that are in place at these airports, GAO and others have reported that the agency still does not make a link between improvements and changes in delays; much of the recent decline in delays is more attributable to the declines in the traffic than NextGen or other improvements. 2.1.2 Airports An airport has an infrastructure that provides a certain throughput, but many operational delays—late arrivals or late departures—at airports are the result of issues out of their control (e.g., weather, airline scheduling practices) or else- where in the aviation system (e.g., airspace constraints, ATC, storms in other areas). In some cases, airports collect data on delays and are able to show that few delays are attributable to the actual airport, but are due to upline or downline con- straints. Operational delays are typically measured through FAA databases (see Section 2.2) or by comparing actual times to airline schedules. Airports typically focus both on throughput and average delay. In general, an airport measures its impact on delay based on the overall ability to stay below maximum airport

7 For Master Planning and Environmental Studies Airports typically follow the FAA’s guidance when consider- ing a capacity enhancement project and use delay estimates as the major master plan tool to help identify runway and taxiway needs to meet forecast demand levels. The cost-benefit analy- ses should result in a positive return of economic savings, and delay savings is a large part of this calculation. Airports gener- ally define delay as the difference between optimal/unimpeded travel times and expected times (often calculated with simula- tion tools), whether the delay is on the ground or in the air. The current standard metric for measuring delay at an air- port is average delay per operation. There is agreement that this metric is not adequate and does not tell the whole story. However, at a large airport, there is also general agreement that • Average delays below 5 minutes per operation are tolerable, • Average delays greater than 10 minutes are a problem, and • Average delays over 20 minutes indicate the airport is expe- riencing very significant congestion issues to the point of not being able to operate due to gridlock. Note that some of these metrics are based on FAA plan- ning documents (e.g., Airport Benefit-Cost Analysis Guidance, Airport Master Plan AC 150/5070-6B) as well as general expe- rience in aircraft operations/schedules regarding the level of delays that a particular airport or airline schedule can tolerate while still maintaining some reliability. Other times, airports simply consider whether the delay savings have a positive return, regardless of the absolute value of the delays. For Project Justification and Cost-Benefit Airports typically cannot justify projects that are needed just to accommodate the peak, especially knowing that airline scheduling practices could change. Airports do not build to meet peak demand, but try to provide an acceptable level of service for typical demand. Each airport has its own unique issues that may complicate the analysis. For instance, airports with gates proximate to runway ends cannot tolerate as much delay as airports with plenty of taxiway queuing space. Some have just one run- way; some have six or more, with most or all runways inter- secting. This makes comparisons difficult. Ten minutes of average delay at one airport is not comparable to 10 or even 6 minutes at another airport. For Comparison to Other Airports Significant delays can be considered reasonable given an airport’s history of delay. Some airports have run for years with average delays 10 to 13 minutes per operation. Figure 2-1 capacity. If the airport’s operations were under their maxi- mum capacity all day, from the airport’s point of view, there were no airport-caused delays, regardless of when a particu- lar aircraft was scheduled to depart versus when it actually departed. Those airports that have significant delays recognize that this affects their ability to compete for air service with other airports. International operations can have a big impact on, and be greatly affected by, delays. In some cases, if a flight misses its assigned slot time, the flight has to be cancelled and therefore process exceptions are made to make sure inter- national departures are not delayed. At many airports today, traffic demand is well below capac- ity. Where the demand is below VMC capacity, delays are small such that delay analyses, much less airline scheduling practices, are not of great concern. Other airports believe that most of these delays could be avoided by the airline schedul- ing additional time between flights. Most airlines take their own gate capacity at each airport into account when develop- ing their schedules, but do not necessarily consider runway capacity. This results in schedules being developed that can be maintained during VMC capacity, but these same schedules far exceed IMC capacity. Airports with similar VMC and IMC capacities tend to have somewhat reasonable delays during IMC. Also, airports that only encounter IMC a small amount of time can tolerate much higher IMC delays. However, airports with huge dif- ferences between VMC and IMC capacities that experience IFR conditions somewhat regularly experience significant delays. There is further information in Section 2.3 on ana- lyzing delays in various weather conditions and combining them into one overall delay value. Delays can be quite high during severe weather and it may take many hours for flights to get back on schedule. It is not unusual for extreme weather conditions, such as thunder- storms at the airport location or elsewhere, to result in cascad- ing delays at airports across the nation. A delay anywhere along the aircraft routing for that day—no matter how small—can have a domino effect and by the time the aircraft has reached the end of the day’s schedule, it could translate into hours of delays. Although many international airports coordinate airline schedules and have some control to prevent airlines schedul- ing more flights than can be accommodated, that does not occur in the United States. Only at the few slot-controlled airports (High Density Rule) is there oversight to the air- line flight schedules at airports. As an example, at LGA in 2000 when slots were removed, airlines scheduled 50% more flights than could be accommodated in an hour. That indi- cates that if the only control mechanism is market forces, then there will likely be significant delay at certain desirable airports.

8as that is the measure that the U.S.DOT reports. (For example, see Figure 2-2, which compares the on-time performance of the major carriers in North America.) Flights that are early are counted as on-time, even if they had to wait for a gate upon arrival, as long as they eventually pulled into the gate within 15 minutes of their scheduled time of arrival. If a flight arrives to the gate early, it is still considered on-time, and no “negative delay” offsets late arrivals in the calculations. In the competitive environment, airlines spend a great deal of effort to accurately schedule their flights and improve their on-time performance standings. For airlines, average delay information is not as meaningful as individual flight delay information. Some airlines specifi- cally focus on the first flight of the day for an aircraft (kickoff flights) because they recognize that this can lead to delayed flights all day. Airlines also record and analyze very detailed data on what caused the delays and trends in the types of delays, because they may be able to correct the items causing the delays. However, it can be very difficult to determine how to allocate delays if multiple items were impacting a plane’s departure. For example, if the bags were still being loaded late but main- tenance was fixing something as well, the delay will typically be coded to the item that took the longest, thus masking the other delay cause completely since it did not have a material effect. Airlines also specifically measure taxi delays and en route delays if they are recorded by the pilots in the onboard sys- tems. Long delays are monitored closely, especially since the new U.S.DOT tarmac rule can significantly fine airlines for long-delayed flights with passengers on board. However, delay metrics usually do not include cancelled flights. shows the arrival delay statistics for the core 30 airports (designated by FAA) for 2012. Note, however, that this data is compiled from data provided by the air carriers, not from the flight-plan data or from the airports. The on-time percent- age (blue bar) is evaluated as the flights that arrived within 15 minutes of their scheduled time. Note that for 2012, EWR and SFO had the lowest percentages of on-time arrivals and also the highest average delays per arrival flights (red dot). However, when looking at the amount of delay per delayed flight (green bar), ORD and EWR had similar values. 2.1.3 Airlines Airlines look at delays from a number of perspectives. Pri- marily, they look at delays as compared to their scheduled times. In general, “flight delay” is variance from schedule— that is, the actual gate arrival or departure versus the sched- uled time of arrival or departure. This simple calculation defines the airlines’ departure or arrival on-time performance. Airlines measure departure delay (e.g., leaving the gate within 5, 10, and/or 15 minutes of STD) because they know that it directly correlates to arrival performance. Airlines have some control in getting aircraft to leave the gate on time, for example, by setting policies to cut-off passenger boarding a specified number of minutes prior to scheduled departure. On-Time Performance Rankings Airlines also focus on arrival delays, specifically, A+15 (arrival at the gate within 15 minutes after scheduled arrival), Source: TransSolutions analysis of ASPM data 0 5 10 15 20 25 40 45 50 55 60 65 70 75 80 85 90 AT L BO S BW I CL T DC A DE N DF W DT W EW R FL L HN L IA D IA H JF K LA S LA X LG A M CO M DW M EM M IA M SP O RD PH L PH X SA N SE A SF O SL C TP A % On-Time Arrivals Avg Arr Delay per Delayed Arrival Av er ag e Ga te A rr iv al D el ay (i n m in ut es ) Figure 2-1. Arrival delays for core 30 airports, 2012.

9 “Padding” the Schedules/Block Times The published STD and STA are the only parts of the air- line schedule development that are publicly available. The duration from the STD to STA is the scheduled block time. Airlines include in the block time the expected taxi-out time, expected en route time, and expected taxi-in time at the des- tination airport. It is important for the scheduled block times to be accurate, but there are competing forces for setting the block times, as follows: • Passengers purchase tickets or flights based on the STA and/or STD. These times need to be both realistic and con- venient in order to meet passenger expectations. • An airline’s planning and staffing are based on scheduled times. For every minute in the block time, the crew and aircraft are not available to be scheduled for another flight/ trip. Each minute in the block time is equivalent to millions of dollars. • U.S.DOT rankings are based on percent on-time compared to the STA and STD. • Turnaround times at gates and minimum connection times for passengers are based on the block times. Airline block time for the same city-pair flight using the same aircraft type will vary by season of year and time of day. Out-Off-On-In (OOOI) Data commonly used for evaluating aircraft travel times and delays at an airport is the out-off-on-in (OOOI) data. Many airlines use onboard systems, such as the Aircraft Com- munications Addressing and Reporting System (ACARS) to automatically record these times, which are defined as follows: • Wheels “out” of the gate/parking position is the time an aircraft departed from the gate, typically measured when the parking brake is released. Also called the actual time of departure (ATD), which can be compared to the STD. • Wheels “off” the runway is the time an aircraft departed from the runway. • Wheels “on” the runway is the actual time an aircraft landed on the runway. • Wheels “in” the gate or parking position is the time an air- craft arrived at the gate, typically measured when the park- ing brake is set. Also called the actual time of arrival (ATA), which can be compared to the scheduled time of arrival (STA). Analysis of taxi times at an airport use “out-to-off” times for taxi-out or departure taxi time, and “on-to-in” times for taxi-in or arrival taxi time. Similarly, “out-to-in” times would require the entire time from one airport gate to another, which can be compared to the scheduled block time. Source: FlightStats Figure 2-2. Airline on-time performance (June 2012).

10 In the example above, while 70% of the flights had a total travel time of 150 minutes (the scheduled block time) or less, only 53.8% of the flights arrived at or before their scheduled arrival time, due to departing late from the origin airport. Still, 64.2% of flights arrived no more than 5 minutes after sched- uled arrival, 72.5% of flights arrived no more than 10 min- utes after scheduled arrival, and 77.5% of flights arrived no more than 15 minutes after scheduled arrival, which is con- sidered on-time by the common criteria of A+15 as reported by U.S.DOT. It is interesting to note that the median of the actual block times was 146 minutes. When airspace changes occur or a new runway is con- structed, airlines react by modifying their block time calcula- tions, although it may take a few schedule changes to fine-tune total travel times. Airlines also are very sophisticated at analyzing taxi-in and taxi-out times at their major hub airports and adequately accounting for the expected taxi times in the block timings throughout the day. At busy airports, the taxi-out component of the block time may vary by over 30 minutes throughout the day to account for typically higher taxi-out times during peak hours at the airport. However, when actual flight times are compared to block times, one does not obtain an accurate estimate of true delays, since the block times already include some typical or histori- cal delays. Flight Schedule Influence on Delays Airline scheduling practices contribute significantly to gate and airfield delays. Specifically, when airlines schedule a high Airlines add time to their block time schedules to accom- modate for historical actual times, which include some delay resulting from flight restrictions, congestion, and a variety of other factors. Although there is concern that airlines “pad” the block times to artificially improve their on-time performance, it is a costly endeavor to add minutes to the block time. Each carrier must make a decision as to what is the most realistic block time to apply to a given flight. Figure 2-3 depicts the vari- ation in the actual block time as recorded in the FAA’s Aviation System Performance Metrics (ASPM) database for all flights in a calendar year (2010) with a scheduled block time of 150 min- utes for a given airline (American Airlines) between the same city pair (DFW-MCO). If airlines use the average block time, then they are guar- anteed to be underestimating actual times for a large num- ber of flights. If they use a larger number for the block time, their on-time performance will improve, but it will be quite costly because that aircraft, crew, etc., cannot be available for another scheduled flight. In this example, the airline could schedule to a block time of 135 minutes and 8.5% of the flights would have experienced a block time of 135 minutes or less, showing that that smaller amount of block time is achievable, but not consistently since over 90% of the flights took more time than that. Even if airlines publish a sched- ule with the average block time of 147 minutes, 44% of the flights would experience a block time longer than that. Since the published time is used by passengers in anticipating not only their airport arrival, but their subsequent arrival at their ultimate destination, passengers prefer to have a scheduled arrival time that is reliable. Source: ASPM Figure 2-3. Distribution of actual block time (2010).

11 drive large delays. De-peaking tends to lengthen passenger layovers or connection times. If airline banking schedules include lulls or valleys in the daily schedule to allow catch up (recovery periods), then this practice does not create intolerable delays. However, without valleys or cancellations, delays will propagate throughout the day. Delays can be influenced by airline culture or strategy: some airlines have a “must complete” attitude about their schedule and will take very long delays to avoid cancellations. Others will cancel some flights to keep the network on time. Also, some have employee cultures that tend to make up time during turns to avoid delays. When the operational delays often motivate an airline to consider capacity improvement alternatives at an airport, they will typically rely on similar planning analyses (delay relative to unimpeded time) as those used by the FAA and airports. How- ever, for airlines, average delay information is not as meaning- ful as individual flight delays. Airlines are concerned about maximum delays, maintaining schedule integrity, and being able to turn the aircraft on the ground within the scheduled time (i.e., even if an arrival flight is a few minutes late, they still want to make an on-time departure). Figure 2-4 plots simu- lated departure delays for each individual aircraft throughout the day, with the different symbols representing the runways used by the flights. It is quite easy to see how delays grow sev- eral times throughout the day during the departure peaks, when there are more departures scheduled in a few minutes number of flights in a short period of time, airports experi- ence congestion and delays. Airline business model practices such as hub-and-spoke scheduling also play a part in capacity constraint delays. For example, with tightly scheduled flights, one thunderstorm with lightning lasting just 15 minutes can cause delays for the 25+flights on the ground that cannot be loaded for departure. At the same time an additional 25+ flights continue to arrive, waiting on the tarmac for these same occupied gates. Even if gates are available to accommo- date extra flights, the airlines typically do not have the extra ground crews available to handle the extra gates, thus driving significant delays. Notably, with non-hub carriers, this is not as much of an issue as the flights are spaced out with enough time between them that few delays are attributable to the domino effect. However, airlines overschedule capacity because it makes economic sense. Consumers want flights that leave at spe- cific, desirable times during the day and want connections scheduled that allow them to most quickly reach their desti- nation. Studies have shown the tradeoff between delay costs and revenue benefit of overscheduling. In an unrestricted competitive environment, airlines overschedule capacity for peak times at airports. Hub-and-spoke scheduling tends to overschedule for peak arrival and departure times—causing some delay—while this tends to be less of an issue for point- to-point carriers. Banks of flights, or complexes, create math- ematically optimal setups for passenger connections but can Source: TransSolutions 0 1 0 2 0 3 0 4 0 5 0 6 0 6 8 1 0 1 2 14 1 6 1 8 2 0 2 2 2 4 T im e o f D a y D ep t G ro un d De la ys (in m inu tes ) Figure 2-4. Sample chart of individual aircraft delays.

12 (even though this long taxi-out time may have been fully accounted for in the scheduled block time). • The flight is delayed en route by ATC through speed con- trol or vectoring; this is often unknown to the passengers, and is only recognized if communicated by the flight crew or when the flight arrives to the destination airport late. • The flight is rerouted around weather, adding unexpected flight time, resulting in a delayed arrival at the destination airport. • The flight lands at the destination airport, but has to wait for an open gate; this is often perceived as a delay even if the passengers are able to deplane prior to, or at, the scheduled arrival time. That said, not all delays are equal. A 5-minute delay en route is typically acceptable (and may not even be recognized by the passenger), while the same 5 minutes experienced waiting for the aircraft door to open after the plane has arrived can be extremely frustrating to passengers. In general, passengers are more accepting of delay if the airline does a good job of com- municating the reason for the delay and expected departure and/or arrival times. Likewise, passengers do not necessarily perceive an early arrival in a positive way. Passengers do not tend to “credit” airlines with arriving early and, in fact, often mock the airline for over-estimating the travel time just to improve their on- time performance. Although consumers dislike any delays, long delays inside an aircraft on the ground are of particular concern. The recently enacted passenger protection rule is a punitive way to force airlines to manage and avoid long tarmac delays. This law mandates, among other things, that airlines provide food, water, and other amenities to passengers kept waiting on tar- macs and give them the opportunity to get off the plane after a wait longer than 3 hours for domestic flights (4 hours for international flights). Passenger tolerance also is not consistent at all airports. Historical delays are so commonplace that people expect to be delayed at some airports. There is some thought that, when given a choice of regional airports, passengers are willing to tolerate longer delays at closer, convenient airports rather than at airports farther out (and perhaps more inconsistent surface travel time to/from). If so, this implies that passengers somewhat consider their total travel time from origin (e.g., home, business, hotel) to destination (hotel, home, business) rather than merely the scheduled airline travel time. However, many passengers are extremely sensitive to price. Although delays are of concern to passengers, they tend to make their ticket purchase decision based on price and sched- uled arrival (or departure) time rather than delay rankings. The airline consumer group Flyers Rights (formerly the Coalition for an Airline Passengers’ Bill of Rights) has been than what the airport can efficiently accommodate. Generally, the airport (and the airline flight schedule) is able to recover between the departure complexes as the delays go back to zero in between the complexes. If the airline were able to schedule the flights throughout the hour rather than bunched together in a few minutes, the average and maximum delays would be much lower. In this example, while the maximum delay is 50 minutes, the average is only 10.3 minutes. In this scenario, the airport might promote that 10 minutes of departure delay is reason- able, but the airline could be concerned about the variance in the delays and the flights that are delayed longer than can be added into the block time. When determining a reasonable amount of delay, airlines may use a low average delay goal at some airports but tolerate large amounts of delay at other airports—either due to their operations (connecting hub vs. destination/spoke airport) and/or marketing strategies to limit additional flights, airlines, and/or gates. 2.1.4 Consumers/Passengers The most effective definition of delay may be “it’s a delay if perceived as one by the traveling public.” In many con- sumers’ minds, the travel experience has gotten worse and, given that an increase in travel demand is expected, it is likely to worsen further. They perceive real delays as anything beyond the scheduled times. Consumers, in general, would like predictable departure and arrival times. Flights are purchased based on the expected departure and arrival times. Passengers would like to avoid lost productivity from unexpected delays. A passenger considers it a delay if they depart or arrive late as compared to schedule and do not recognize any extra time that might have been built into the scheduled block time by the airlines. However, if a flight departs late, that is easily forgotten and forgiven if the flight arrives on time. But if the aircraft lands early but has to then wait for a gate, this again is often perceived as delay—even if the flight still reaches the gate prior to its STA. Passengers may use the word “delay” in reference to any of the following situations: • Inbound flight is late so that no aircraft is available for the scheduled flight departure, causing a delayed departure. • Mechanical or aircraft cleaning or missing crew (or some other problem) prevents the passengers from boarding the aircraft, delaying the departure. • After the passengers board the flight, something prevents the aircraft from departing the gate on time. • The aircraft leaves the gate, but experiences a long taxi- out time; when the aircraft finally takes off the runway, passengers perceive that the runway departure is delayed

13 When local communities are affected by increased flight operations (e.g., noise) or proposed airfield expansion, esti- mating expected delay savings of potential capital projects at an average of 1 or 2 minutes does not sound very large or worthwhile. But when the average dollar savings is applied to each and every flight throughout the year, it has a dramatic impact on fuel burn and overall passenger delay. 2.2 Historical/Actual Delay Data Actual operational flight travel times and delays can be accessed through several data sources. This section contains a description of several FAA databases and other data sources with samples of data analyses. In addition, the databases are summarized in Appendix A. The following data sources are discussed in this report: • TFMSC—Traffic Flow Management System Counts, • PDARS—Performance Data Analysis and Reporting System, • OPSNET—Air Traffic Operations Network, • ASQP—Airline Service Quality Performance, • ASPM—Aviation System Performance Metrics, • BTS—Bureau of Transportation Statistics, and • Local airport systems. focused on extremely long flight delays for several years. They recognize that there is a need to invest in infrastructure, espe- cially noting the importance of gate capacity to improve air- port capacity and delays. Perceptions associated with delays for consumers/passengers mostly center around unpredict- able arrival/departure times and unexpected delay, leading to lost productivity of the passengers. 2.1.5 General Public Each month, media reports compare the on-time perfor- mance and delays at different airports (Figure 2-5). Flight delays are regularly reported in the mainstream media, whether due to severe weather or just typical operating conditions. As previ- ously stated, these delays are based on the airlines’ flight sched- ules, which already have some delay included in them, but this is not apparent to the casual observer. However, public opinion tends to remember the long opera- tional delays or the extreme cases. Even on a day where several flights experienced an hour or more of delay, the average delay might have been only 1 or 2 minutes. And, as reported every year, the vast majority of flights arrive “on time,” meaning within 15 minutes of their scheduled arrival. A large number of flights even arrive at the gate prior to their scheduled time. Source: FlightStats Figure 2-5. North America’s top 10 on-time departure airports (June 2012).

14 are made: “TFMS Operations” are the number of flights that departed or arrived at that terminal SDP. The TFMS data provides the capability to calculate types of operations (arrival, departure, or overflight for en route cen- ters), terminal operations counts (arrivals, departures, and overflights) and instrument operations (primary, secondary, and over) on a flight-specific basis. In addition, for the en route and oceanic environments, it also is possible to derive the time within the center’s airspace, actual distance flown within the center’s airspace, and the great-circle route distance between the entry and exit point of the center’s airspace. Actual/Historical Data vs. Calculated/Estimated Data The TFMS data provides the capability to calculate types of operations (arrival, departure, or overflight for en route centers), terminal operations counts (arrivals, departures, and overflights) and instrument operations (primary, second- ary, and over) on a flight-specific basis. In addition, for the en route and oceanic environment it is also possible to derive the time within the center’s airspace, actual distance flown within the center’s airspace, and the great-circle route distance between the entry and exit point of the center’s airspace. TFMS raw databases, used by aviation analysts for airport delay calculation, are always filtered for duplicate and missing data. The most reliable times (i.e., field data) are 1. OOOI (gate out, wheels off, wheels-on, and gate in) times: provided by most airlines and relay information about the actual aircraft movement times. When OOOI times are not available, they are estimated according to the guidelines provided by FAA. 2. Actual gate departure time: if an airline does not provide OOOI times, estimated gate departure times are calcu- lated as (estimated wheels-off time) minus (median taxi- out time by carrier by day by hour). 3. Actual gate arrival time: if an airline does not provide OOOI times, estimated gate arrival times are calculated as (estimated wheels-on time) + (median taxi-in time by carrier by day by hour). 4. Scheduled gate departure time: given by Official Airline Guide (OAG). 5. Scheduled gate arrival time: given by OAG. Depending on the available information from the TFMS database, delay is computed as the difference between the actual time and scheduled time. One of the standard steps in computing actual departure or arrival delay is to adjust for taxi time. This is because airlines publish gate times, while an analyst frequently only has actual runway times; therefore, adjustments by taxi time are required as follows: 2.2.1 TFMSC The FAA’s Traffic Flow Management System (TFMS) is a data exchange system for supporting the management and moni- toring of national air traffic flow. (Note: for those familiar with the Enhanced Traffic Management System [ETMS], TFMSC replaced that system in 2010.) TFMSC processes all available data sources such as flight-plan messages, flight-plan amend- ment messages, and departure and arrival messages. The FAA’s airspace lab assembles TFMS flight messages into one record per flight. TFMSC is restricted to the subset of flights that fly under IFR and are captured by the FAA’s en route computers. Visual flight rules (VFR) traffic is not included, and even some non en route IFR traffic is excluded. TFMSC includes informa- tion about commercial traffic (air carriers and air taxis), gen- eral aviation, and military to and from every landing facility, as well as airspace fixes in the U.S. and in nearby countries that participate in the TFMSC system. After processing, the TFMS file provides detailed flight records, including time, distance, aircraft type and user type. The Air Traffic Laboratory (ATA-100) provides Boundary Crossing File (BCF) records for each flight in the NAS. A flight segment for this purpose is one aircraft traveling through one air route traffic control center (ARTCC), so a flight that travels through three ARTCCs would be divided up into three records. The flight segment records are then grouped into flight records (one record per flight) using a unique flight identity code. The aircraft’s maximum takeoff weight (MTOW) is also added to the flight record using an aircraft type reference file (this file also contains seating capacity, cargo capacity, load factors and fuel consumption). For international arrivals and departures, the model also estimates flight time and distance outside of U.S. airspace, which is later used to calculate user revenues and costs. The model then sums the number of operations, actual flight miles, great-circle flight miles and flight hours at each ARTCC, and these data are used for further processing as follows: • Count of flights (departures, arrivals, and both departing and arriving in the same ARTCC), • Actual miles flown, • Great-circle equivalent of miles flown, and • Hours flown. The en route activity used is in the form of counts, hours, miles and great-circle miles for flights departing, arriving, both departing and arriving, or overflying an en route service delivery point (SDP). Additional records are created to turn a flight that both departs and arrives within one center into two operations for the en route SDP, which is a more accurate depiction of how en route activity is counted in other data systems. For terminal SDPs, two counts of TFMS operations

15 IFR traffic and some VFR traffic, it has several limitations and challenges. First, due to limited radar coverage and incomplete messaging, TFMS may exclude certain flights that do not enter the en route airspace and other low-altitude flights. Also, of the 35,000 location identifiers reported over time, only the top few thousand, accounting for over 95% of traffic, are reliable. The others are waypoints or other references to locations not associated with an airport. Access to the TFMSC database is restricted by password and authorized by FAA. Documentation, Data Access, and Point of Contact • To obtain a TFMSC login and password: https://aspm.faa. gov/Default.asp • About TFMSC: http://aspmhelp.faa.gov/index.php/ TFMSC • Address: Office of Aviation Policy and Plans Federal Aviation Administration 800 Independence Ave., SW Washington, D.C. 20591 Phone: (202) 267-3336 Fax: (202) 267-5370 2.2.2 PDARS As a result of collaboration between the FAA and the National Aeronautics and Space Administration (NASA), PDARS col- lects data every 5 to 6 seconds from Air Route Traffic Control Centers (ARTCCs) and Terminal Radar Approach Control facilities (TRACONs). PDARS’ raw data produces informa- tion such as the type of operation, aircraft identification, and actual runway threshold time. On the airport level, a signifi- cant piece of information lies in the information on aircraft- runway assignments. Delay measurements can be indirectly calculated when compared against the OAG databases. The raw PDARS database contains more than 90 fields for each flight. PDARS software calculates a range of performance mea- sures, including traffic counts, travel times, travel distances, actual departure delay = [(actual gate departure time – taxi time)] – scheduled gate departure time Table 2-1 provides an analysis example for one flight from IAD to SFO with TFMSC data. In this example: actual departure delay for flight UAL225 = [dept time – (off time – out time)] – fs_dept_time = [4:42:00AM – (4:41:00AM – 4:30:00AM)] – 3:05:00AM = 1:24:00 hrs actual arrival delay for flight UAL225 = [arr time – (in time – on time)] – fs_arr_time = [10:23:00AM – (10:23:00AM – 10:19:00AM)] – 9:16:00AM = 1:03:00 hrs An airport analyst interested in airborne delay can quickly and simply compute airborne delay by following schedule times and OOOI times; for example, if a flight departed 15 minutes late, and arrived 25 minutes late, its airborne delay was 10 minutes. Strengths TFMSC includes data for flights that fly under IFR and are captured by the FAA’s en route computers. This includes infor- mation about commercial traffic (air carriers and air taxis), general aviation, and military to and from every landing facil- ity as well as fixes, both in the U.S. and in nearby countries that participate in the TFMS system. TFMSC can calculate airborne delay such as delay in each ARTCC or sector (because it has sector boundary crossing information) and speed. Data is collected electronically. This is an input into ASPM. Weaknesses TFMSC is not appropriate for micro analyses of delays per runway. Although TFMS reliably captures the vast majority of ACID dept_arpt arr_arpt dept_time arr_time fs_dept_time fs_arr_time out time off time on_time in_time UAL225 IAD SFO 4:42:00 AM 10:23:00 AM 3:05:00 AM 9:16:00 AM 4:30:00 AM 4:41:00 AM 10:19:00 AM 10:23:00 AM Key: dept_time: actual gate departure time arr_time: actual gate arrival time fs_dept_time: scheduled gate departure time fs_arr_time: scheduled gate arrival time out time: time aircraft leaves gate or parking position (gate out time) off time: time aircraft takes off (wheels-off time) on_time: time aircraft touches down (wheels-on time) in_time: time aircraft arrives at gate or parking position (gate in time) Table 2-1. TFMSC data for one flight.

16 probability distribution for gate delays is useful in forecast- ing gate delays and can be used in planning or revising airline schedules, or in implementing alternative strategies for delay reduction. The PDARS database provides detailed information on aircraft-runway assignments and therefore enables an airport analyst to conduct micro-level analyses of delays for each run- way. To date, this is the only aviation database that provides such information. Figure 2-8 presents empirical probability distributions of gate delays for individual runways at SFO. traffic flows, and in-trail separations. It turns these measure- ment data into information useful to FAA facilities through an architecture that features (1) automatic collection and analysis of radar tracks and flight plans, (2) automatic gen- eration and distribution of daily morning reports, (3) shar- ing of data and reports among facilities, and (4) support for exploratory and causal analysis. One of the main PDARS functions is to provide FAA facili- ties with the capability to both identify air traffic situations that can be changed or improved and quantify the consequences of operational adjustments from safety and efficiency perspec- tives. However, the PDARS database can be used to calculate delays when combined with additional databases. Figure 2-6 presents the steps for calculating gate delays. The following databases are combined for this particular delay calculation: • ASPM: to obtain taxi-in aircraft time; • PDARS: to obtain information on flight number, date, runway assignment, airline, type of aircraft, and actual threshold arrival time; and • ASQP and OAG: to match flight number information, date, airline, and origin/designation to provided scheduled time at a gate. Once gate delays are calculated for each flight, a better understanding of the delay process could be achieved by fit- ting the best theoretical probability distribution against the calculated empirical delay data (Figure 2-7). Estimation of a Figure 2-6. Calculation of gate delays using ASPM, PDARS, ASQP and OAG databases. Figure 2-7. Probability distributions for aggregate gate delays at SFO using PDARS. Source: J. Rakas

17 ties serving the 34 (of 35) domestic Operation Evolution Plan (OEP) airports, the Air Traffic Control System Command Center (ATCSCC), and the Mike Monroney Aeronautical Cen- ter in Oklahoma City, Oklahoma. Therefore, only 34 airports’ delay data is included in PDARS, far less than ASPM. Surface delay metrics are not as elaborate as in ASPM. Current efforts include better integration of ASDE-X data in order to create advanced surface metrics measuring airport efficiency and safety. Although it is an excellent tool for airspace redesign, it is less suitable for airport capacity and delay analyses. Documentation, Data Access and Point of Contact • PDARS is only available to authorized FAA users at ATC facilities, but raw data could be available to FAA contractors. • ATAC’s article about PDARS: http://www.atac.com/docs/ MTS%20Nov%20Dec%202010%20PDARS.pdf • Address: ATAC Corporation 2770 De La Cruz Boulevard Santa Clara, CA 95050-2624 Phone: (408) 736-2822 NASA Ames Research Center Moffett Field, CA 94035 PDARS Program Office FAA Air Traffic Organization (ATO) Performance Analysis The Portals Building 1250 Maryland Ave., SW Washington, DC 20024 Actual/Historical Data vs. Calculated/Estimated Data PDARS software calculates a range of performance mea- sures, including traffic counts, travel times, travel distances, traffic flows, and in-trail separations. It turns these measure- ment data into information useful to FAA facilities through an architecture that features (1) automatic collection and analysis of radar tracks and flight plans, (2) automatic genera- tion and distribution of daily morning reports, (3) sharing of data and reports among facilities, and (4) support for explor- atory and causal analysis. Delay measurements are indirectly calculated when compared against the OAG databases. Strengths Data is collected electronically; information is updated every 5 to 6 seconds. Analysts can calculate delays on a micro level for each runway because of underlying information on aircraft- runway assignments. The raw database has over 90 fields, con- taining more information about each flight than the TFMSC database. Analysts generally find that PDARS is a better data source to measure system capacity in comparison with ASPM in that PDARS provides a higher level of fidelity that provides more accurate analytical data in comparison with ASPM. Weaknesses PDARS is not publicly available—it is only available to FAA, NASA, and ATAC Corporation. It collects its own target report data from radars from 20 (of 22) domestic ARTCCs, 27 (of 185) terminal radar approach control (TRACON) facili- Source: J. Rakas 0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 -80 -60 -40 -20 0 20 40 60 80 100 120 140 160 180 200 Delays at the gate (in minutes) 19L 19R 28L 28R Figure 2-8. Empirical gate-delay data at SFO for each runway.

18 For example, to produce the Standard Report for Airport Operations for a selected airport (Table 2-2), the following steps are performed, select: 1. Facility (SFO airport), 2. Date (1/2012–1/2013), 3. State (CA); Region (AWP—Western Pacific code), 4. Service Area (WT—Western Terminal), and 5. Class (facility type or classification—Towers with Radar). The output produces the total number of operations, divided between Itinerant (air carrier, air taxi, GA, military, total) and Local (civil, military, total). To obtain a Standard Report for Airport Delays (Table 2-3), an aviation analyst may select the following items from the input menu: Facility (35 OEP airports) and Date (01/01/12– 12/31/13). This standard report produces results for the fol- lowing variables: Total Operations, System Impact Delays, Total Delays, Traffic Management Initiative (TMI) to Delays, Occurred at Delays (departure delay, airborne delay, TMI from delays, airborne destination to delay), delays by class (air carrier, air taxi, general aviation, military), Delay by Cause (weather, volume, equipment, runway, other), and Delay by Time (average) and Time (total). To obtain the OPSNET delay data as a standard report for an airport (Table 2-4), an analyst selects a specific Facil- ity (e.g., SFO airport only), and the time period (01/2012– 12/2013) from the menu. A simple Ground Delay Report (Table 2-5) may be obtained by selecting the following items from the menu: Date (01/2012– 01/2013) and Facility (SFO airport). The out put produces number of Delays, Minutes, and Average delay for Ground Stops, Expected Departure Clearance Times (EDCT) and Total. Actual/Historical Data vs. Calculated/Estimated Data OPSNET Delays provides information about reportable delays provided daily through FAA’s Air Traffic Operations Network (OPSNET). A reportable delay recorded in OPSNET is defined in FAA Order 7210.55F as, “Delays to IFR traffic of 15 minutes or more, which result from the ATC system detaining an aircraft at the gate, short of the runway, on the runway, on a taxiway, or in a holding configuration anywhere en route, must be reported. The IFR controlling facility must ensure delay reports are received and entered into OPSNET.” These OPSNET delays are caused by the application of initia- tives by the Traffic Flow Management (TFM) in response to weather conditions, increased traffic volume, runway condi- tions, equipment outages, and other causes. Strengths The FAA’s OPSNET database contains delay causality infor- mation. Delays are assigned to five major categories within 2.2.3 OPSNET OPSNET is the official FAA aircraft delay reporting system. Data comes from observations by FAA ATC personnel, who manually record the number of aircraft delayed 15 minutes or more relative to nominal or unimpeded taxi-out and taxi- in times estimated for each airport. OPSNET data measures the efficiency of the FAA ATC system; it does not measure delays based on the scheduled times, but on the flight-plan times submitted to air traffic. OPSNET reports delays for each airport by the following categories: • Category of delay (e.g., departure delay vs. arrival delay); • Class (e.g., air carrier vs. general aviation); and • Cause (e.g., weather vs. traffic volume). OPSNET records the following information and data: • Airport operations: IFR and VFR arrivals and departures, and local operations at the airport as reported by Air Traffic Control Towers (ATCTs). It does not include overflights. • Tower operations: IFR and VFR arrivals and departures, IFR and VFR overflights, and local operations worked by the tower. • TRACON operations: IFR and VFR operations and over- flights worked by the TRACON. • Total terminal operations: Total operations worked by any facility based on the functions at the facility. If a facility has a tower and a TRACON present, the total terminal opera- tions are a sum of the tower operations and the TRACON operations for that facility. • Center aircraft handled: Domestic and oceanic departures and overflights and total aircraft handled by ARTCCs and center radar approach controls (CERAPs). • Facility information: Provides information about each ATC facility, such as facility name and type, region, state, hours of operation, etc. • Delays: Provides information about the reportable delays provided daily through FAA’s OPSNET. Delay results using OPSNET data are available through the FAA’s ASPM website. If an airport analyst is interested in obtaining performance reporting for a selected airport, or a group of airports, the results could be displayed with the follow- ing report types: standard report, ranking, comparison, peak days report, ground delay, day of the week, and report facility. As an illustration, Tables 2-2 through 2-5 present out- put results using various criteria that were selected from the main OPSNET delay menu: My Reports (report type), Facil- ity (airport—can be selected from ASPM 77, OPSNET 45, OEP 35, or core database), Dates (days, months, year range, period), Filters (default, only FAA staffed facilities, only FAA contract staffed facilities, only data by ARTCC), Groupings (date, facility, state, region, class), Output, Run.

19 2.2.4 ASQP ASQP contains data provided by the airlines by flight for airlines that carry at least 1% of all domestic passengers. The data is available from June 2003 and is updated on a monthly basis. The number of airlines providing data has varied from 10 to 20, with the current list of 14 carriers at http:/aspmhelp. faa.gov/index.php/ASQP:_Carrier_Codes_and_Names Actual and scheduled time is available for gate departure and gate arrival, based on the airlines’ block times (which include some expected delays). The airlines also provide the actual wheels-off time (so that taxi-out time can be computed) and wheels-on time (so that taxi-in time can be computed). In addition, the airlines provide causal data for all delayed flights arriving 15 minutes past their sched- uled arrival time. The causes of delay categories are airline, extreme weather, National Aviation System, security, and late arriving flight. Through ASQP, an analyst may select the following: • Report output type (standard, causal, on-time NAS, BTS, BTS TranStats, schedule reliability report, or dispatch and schedule reliability); • Dates; • Airports; OPSNET: weather, volume, equipment, runway, and other. This is a major strength. Weaknesses Delays are reported manually at ATC facilities at airports and can be inaccurate, are excluded if initiated by pilot/ airlines, and are not reported if less than 15 minutes. Delay reporting methods are also subjective and differ by facility. Access to the OPSNET database is restricted by password and authorized by FAA. Documentation, Data Access and Point of Contact • To obtain a login and password: https://aspm.faa.gov/ Default.asp • About OPSNET: https://aspm.faa.gov/opsnet/sys/main.asp http://aspmhelp.faa.gov/index.php/ Operations_Network_%28OPSNET%29 • Address: Office of Aviation Policy and Plans Federal Aviation Administration 800 Independence Ave., SW Washington, D.C. 20591 Phone: (202) 267-3336 Fax: (202) 267-5370 Itinerant Local Date Facility State Region Service Area Class Air Carrier Air Taxi General Aviation Military Total Civil Military Total Total Operations 01/2012 SFO CA AWP WT Towers with Radar 24,660 7,237 1,207 276 33,380 0 0 0 33,380 02/2012 SFO CA AWP WT Towers with Radar 23,250 7,244 1,078 295 31,867 0 0 0 31,867 03/2012 SFO CA AWP WT Towers with Radar 25,245 7,533 986 312 34,076 0 0 0 34,076 04/2012 SFO CA AWP WT Towers with Radar 25,981 7,208 1,049 289 34,527 0 0 0 34,527 05/2012 SFO CA AWP WT Towers with Radar 27,391 7,519 1,182 266 36,358 0 0 0 36,358 06/2012 SFO CA AWP WT Towers with Radar 28,129 7,377 1,176 259 36,941 0 0 0 36,941 07/2012 SFO CA AWP WT Towers with Radar 29,333 7,903 954 241 38,431 0 0 0 38,431 08/2012 SFO CA AWP WT Towers with Radar 29,480 8,010 1,196 302 38,988 0 0 0 38,988 09/2012 SFO CA AWP WT Towers with Radar 26,979 7,174 1,132 250 35,535 0 0 0 35,535 10/2012 SFO CA AWP WT Towers with Radar 27,090 7,813 1,447 400 36,750 0 0 0 36,750 11/2012 SFO CA AWP WT Towers with Radar 24,900 6,844 1,010 270 33,024 0 0 0 33,024 12/2012 SFO CA AWP WT Towers with Radar 25,192 7,158 864 231 33,445 0 0 0 33,445 01/2013 SFO CA AWP WT Towers with Radar 24,199 7,168 1,170 283 32,820 0 0 0 32,820 Sub-Total for Unknown 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Sub-Total for WT 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Sub-Total for AWP 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Sub-Total for CA 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Sub-Total for SFO 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Total: 341,829 96,188 14,451 3,674 456,142 0 0 0 456,142 Report created on Mon Apr 8 11:32:03 EDT 2013 Source: The Operations Network (OPSNET) Table 2-2. OPSNET Standard Report for Airport Operations, 01/00/12–1/31/13, SFO.

20 Facility Total Ops System Impact Delays Abrn Dest To Delays Occurred At Delays Total Delays TMI To Dep Abrn TMI From Total Occ At ATL 1003727 10192 6027 4165 0 9129 13294 2387 BOS 385816 3693 3231 462 0 5390 5852 478 BWI 287896 1831 1652 179 0 3854 4033 397 CLE 194949 96 50 0 46 2102 2148 72 CLT 597713 3810 2705 1105 0 3886 4991 1059 CVG 154483 73 13 38 22 2023 2083 93 DCA 314079 3574 1528 2046 0 6660 8706 790 DEN 666113 2123 1641 482 0 3060 3542 1125 DFW 705428 1745 1629 116 0 5892 6008 1456 DTW 460127 2010 1214 796 0 4359 5155 1046 EWR 454870 38010 30812 7198 0 6451 13649 2173 FLL 287224 2319 863 1456 0 4077 5533 470 HNL 301375 32 4 28 0 40 68 19 IAD 364297 851 732 119 0 4075 4194 382 IAH 552228 5180 1682 3498 0 4186 7684 1124 JFK 442688 9766 5282 4484 0 9119 13603 2414 LAS 568333 1951 824 1127 0 2840 3967 342 LAX 653793 2640 1671 969 0 5130 6099 227 LGA 404909 28391 20958 7433 0 9091 16524 3638 MCO 333749 56 56 0 0 3390 3390 355 MDW 267814 583 556 27 0 1476 1503 370 MEM 291232 681 565 116 0 1586 1702 142 MIA 427208 1567 481 958 128 2176 3262 879 MSP 457874 776 555 221 0 3052 3273 687 ORD 945902 22441 16524 5917 0 11373 17290 1841 PDX 232271 9 0 9 0 1052 1061 31 PHL 479038 15768 14013 1687 68 6029 7784 548 PHX 487192 5034 2498 2536 0 2285 4821 296 PIT 150284 30 0 0 30 1749 1779 52 SAN 202336 600 181 419 0 2212 2631 149 SEA 332781 118 13 105 0 1765 1870 64 SFO 456142 26770 23067 3703 0 2003 5706 1400 SLC 354808 348 167 181 0 1135 1316 271 STL 207954 4 4 0 0 1761 1761 92 TPA 204021 129 27 80 22 1478 1580 255 Total : 14630654 193201 141225 51660 316 135886 187862 27124 Report created on Mon Apr 8 11:42:36 EDT 2013. Source: The Operations Network (OPSNET) Table 2-3. OPSNET Standard Report for Airport Delays, 01/01/12–12/31/13, 35 OEP airports.

Facility Date Total Ops System Impact Delays Abrn Dest To Delays System Impact Delays Occurred At Delays By Class By Cause Time Total Delays TMI To Dep Abrn TMI From Total Occ At AC AT GA Mil Wx Vol Equip Rwy Other Avg (Min) Total (Min) SFO 01/2012 33380 1820 1794 26 0 157 183 167 1378 394 46 2 1792 6 0 3 19 88.00 161843 SFO 02/2012 31867 1059 915 144 0 88 232 52 743 287 29 0 990 13 8 7 41 57.00 60594 SFO 03/2012 34076 2813 2576 237 0 130 367 114 2205 564 41 3 2790 5 0 17 1 81.00 229294 SFO 04/2012 34527 1509 1299 210 0 87 297 96 1174 299 36 0 1302 39 0 168 0 68.00 102672 SFO 05/2012 36358 1672 1220 452 0 154 606 132 1250 379 43 0 1528 109 0 33 2 57.00 96100 SFO 06/2012 36941 2570 1918 652 0 164 816 100 2041 478 51 0 1467 149 0 732 222 76.00 195744 SFO 07/2012 38431 2502 2072 430 0 345 775 104 1904 557 41 0 2261 179 0 10 52 70.00 177080 SFO 08/2012 38988 2334 2165 169 0 223 392 36 1757 523 52 2 1526 158 0 650 0 67.00 157711 SFO 09/2012 35535 2855 2731 124 0 166 290 59 2135 663 55 2 1363 107 0 1374 11 60.00 172788 SFO 10/2012 36750 2763 2124 639 0 190 829 151 2080 602 81 0 2099 118 0 35 511 61.00 170050 SFO 11/2012 33024 2114 1882 232 0 70 302 124 1675 408 29 2 2072 12 0 24 6 81.00 171822 SFO 12/2012 33445 2170 1856 314 0 160 474 181 1681 457 32 0 2119 33 0 18 0 80.00 174695 SFO 01/2013 32820 589 515 74 0 69 143 84 402 171 16 0 334 20 0 235 0 42.00 24739 Sub-Total for SFO 456142 26770 23067 3703 0 2003 5706 1400 20425 5782 552 11 21643 948 8 3306 865 70.79 1895132 Total : 456142 26770 23067 3703 0 2003 5706 1400 20425 5782 552 11 21643 948 8 3306 865 70.79 1895132 Key : Abrn = Airborne; AC = Air Carrier; AT = Air Taxi; Avg = Average; Dep = Departure; Dest = Destination; Equip = Equipment; GA = General Aviation; Mil = Military; Min = Minute; Occ= Occurred; Ops = Operations; Rwy = Runway; TMI = Traffic Management Initiative; Vol = Volume; Wx = Weather. Report created on Mon Apr 8 11:53:26 EDT 2013. Source: The Operations Network (OPSNET) OPSNET: Delays: Standard Report From 01/2012 To 01/2013 | Facility=SFO Table 2-4. OPSNET delay data for SFO. Ground Stops EDCT Total Date Facility Delays Minutes Average Delays Minutes Average Delays Minutes Average 01/2012 SFO 32 1078 33.69 1748 160027 91.55 1780 161105 90.51 02/2012 SFO 6 146 24.33 893 56877 63.69 899 57023 63.43 03/2012 SFO 9 286 31.78 2533 223195 88.11 2542 223481 87.92 04/2012 SFO 20 634 31.70 1248 96911 77.65 1268 97545 76.93 05/2012 SFO 8 192 24.00 1174 83996 71.55 1182 84188 71.23 06/2012 SFO 25 1752 70.08 1871 176460 94.31 1896 178212 93.99 07/2012 SFO 11 343 31.18 2010 164800 81.99 2021 165143 81.71 08/2012 SFO 15 498 33.20 2091 151822 72.61 2106 152320 72.33 09/2012 SFO 14 521 37.21 2689 169461 63.02 2703 169982 62.89 10/2012 SFO 41 1877 45.78 1995 147912 74.14 2036 149789 73.57 11/2012 SFO 1 24 24.00 1833 165598 90.34 1834 165622 90.31 12/2012 SFO 1 41 41.00 1802 166344 92.31 1803 166385 92.28 01/2013 SFO 28 1184 42.29 406 19814 48.80 434 20998 48.38 Sub-Total for SFO 211 8576 40.64 22293 1783217 79.99 22504 1791793 79.62 Total : 211 8576 40.64 22293 1783217 79.99 22504 1791793 79.62 Key: EDCT = Estimated Departure Clearance Time. More information about this report. Report created on Mon Apr 8 12:05:27 EDT 2013. Sources: The Operations Network (OPSNET) Table 2-5. OPSNET data on ground delays at SFO, 2012–2013.

22 ASQP are international, cargo, and general aviation flights, as well as small commercial carriers. Access to the ASQP database is restricted by password and authorized by FAA. Documentation, Data Access, and Point of Contact • To obtain a login and password: https://aspm.faa.gov/ asqp/sys/ • About ASQP: http://aspmhelp.faa.gov/index.php/ Airline_Service_Quality_Performance_ %28ASQP%29 • Address: Office of Aviation Policy and Plans Federal Aviation Administration 800 Independence Ave., SW Washington, D.C. 20591 Phone: (202) 267-3336 Fax: (202) 267-5370 2.2.5 ASPM The FAA Office of Aviation Policy and Plans (APO) devel- oped ASPM to provide data on flights among the 77 ASPM airports, and all flights by the 22 ASPM carriers, including flights by those carriers to international and domestic non- ASPM airports. (Note that http://aspmhelp.faa.gov/index. php/ASPM_Carriers provides a list of 30 carriers, some of which are no longer operating or have merged with other air- lines also on the list.) Only specific air carriers’ data is recorded in ASPM for 77 U.S. airports—in other words, it is not a com- prehensive database for all filed flights in the United States. ASPM captures actual times for out-off-on-in (OOOI), then it calculates taxi-out delay and taxi-in delay based on typical unimpeded times established for each airport. The database has estimates of typical unimpeded times for each typical runway configuration at each of the airports. There- fore, the delay figures are not comparing actual times to flight schedule or block time estimates, but to FAA’s expected val- ues at that airport. In addition, ASPM reports gate departure delay (actual “out” time compared to filed flight-plan sched- uled “out” time), airport departure delay (actual “wheels off” • Grouping (by date, airport, etc.); and • Filters (by ASQP or ASPM). As an illustration, the Bay Area airports (OAK, SFO, and SJO) are analyzed using ASQP data for 2012–2013 (Table 2-6). Using the Standard Report format, an aviation analyst might be interested in the following delay metrics: On-Time Arrivals, Average Gate Departure Delay, Aver- age Block Delay, Average Taxi-Out Time, Average Taxi- In Time, Delayed Arrivals, and Average Delay per Delayed Arrival. Actual/Historical Data vs. Calculated/Estimated Data ASQP provides data such as departure, arrival, and elapsed flight times as shown by the OAG, the carrier’s computer res- ervations system (CRS), and the carrier’s actual performance; selected differences among the three sources, such as delay and elapsed time difference; and the causes of delays. Strengths Actual and scheduled time is available for gate departure and gate arrival. The airlines also provide the actual wheels- off time (so that taxi-out time can be computed) and wheels- on time (so that taxi-in time can be computed). In addition, the airlines provide causal data for all delayed flights arriving 15 minutes past scheduled arrival time. The causes of delay categories are airline, extreme weather, National Aviation System, security, and late arriving flight. This is an input into ASPM. Weaknesses ASQP covers flights within the Continental United States on airlines having at least 1% of the total scheduled domestic passenger revenues. Hence, at a large international airport such as EWR, the ASQP data covers only about 65% of the flights. For those flights that are not covered by ASQP, an analyst is not able to realistically link them into multi-leg itin- eraries. Significant categories of traffic that are missing from Facility Actual Departures Actual Arrivals Departure Cancellations Arrival Cancellations Departure Diversions Arrival Diversions On-Time Arrivals % On-Time Gate Departures % On-Time Gate Arrivals Average Gate Departure Delay Average Gate Arrival Delay Average Block Delay Average Taxi-Out Time Average Taxi-In Time Delayed Arrivals Average Delay Per Delayed Arrival OAK 52075 52117 521 531 74 22 44820 84.35 86.00 8.34 7.59 1.42 10.72 5.99 7297 45.23 SFO 191748 191602 4358 4597 436 380 140514 76.44 73.34 15.93 19.19 3.61 17.26 7.34 51088 67.95 SJC 44864 44928 455 410 56 37 39161 88.36 87.16 6.64 6.85 1.63 10.96 3.89 5767 44.45 Total : 288687 288647 5334 5538 566 439 224495 79.72 77.77 13.12 15.17 2.90 15.10 6.56 64152 63.26 Report created on Mon Apr 8 12:11:02 EDT 2013 Source: http://aspmhelp.faa.gov/index.php/ASQP Airline Service Quality Performance System : Airport View : Standard Report Calendar Year from 2012 to 2013 : Airport=OAK, SFO, SJC : ASQP Flights Table 2-6. Standard Report for OAK, SFO, and SJC with ASQP.

23 tion: efficiency counts (flights handled by air traffic control- lers) and metrics counts (a basis for delay calculations that only include complete flight records). An aviation analyst may select a wide variety of perfor- mance metrics from the FAA Operations and Performance Data website. One of the most common analyses is the air- port delay analysis by hour, where the following steps are followed: 1. From the ASPM menu, select Analysis function; 2. Select target area: airport, city pair; 3. Select airlines: all, exclude; 4. Select time period: quarterly, hourly, monthly, yearly; and 5. Select type of analysis: for all flights, delayed flights, ETMS, weather, schedule, delay counts, etc. After running the results, a large number of delay met- rics are output (Table 2-7), including the following: % On- Time Gate Departures and Arrivals, Gate Departure Delay, Average Taxi-Out Time, Taxi-Out Delay, Airport Depar- ture Delay, Airborne Delay, Taxi-In Delay, Block Delay, and Gate Arrival Delay. compared to flight-plan scheduled “out” time plus unim- peded taxi-out), airborne delay (actual airborne time com- pared to flight plan filed estimated time en route or ETE), block delay (actual out-to-in compared to scheduled gate-to- gate block time) and gate arrival delay (actual in time com- pared to the filed flight-plan arrival time). Using actual times directly from ARINC, ASPM calculates and reports the following: • Excess travel-time delays and • Arrival delay calculated as the difference between actual gate arrival times and either OAG scheduled arrival times or flight-planned arrival times ASPM provides a block delay metric: the difference in the actual gate-to-gate time computed from ARINC OOOI data and the scheduled gate-to-gate block time from the OAG. It also provides data on flight cancellations. The ASPM data- base includes IFR traffic, some VFR traffic, airport weather, runway configuration, and arrival and departure rates. ASPM provides two types of flight-related performance informa- From 2012 To 2013 : 'SFO': (Calendar Year) Local Hour Scheduled Departures Scheduled Arrivals Departures For Metric Computation Arrivals For Metric Computation % On-Time Gate Departures % On-Time Airport Departures % On-Time Gate Arrivals Gate Departure Delay Taxi-Out Delay Airport Departure Delay Airborne Delay Taxi-In Delay Block Delay Gate Arrival Delay 0 2263 1981 2008 2465 77.59 73.51 77.08 9.78 2.54 11.73 2.16 2.18 2.89 11.41 1 1770 281 1341 585 88.59 85.31 79.66 5.90 2.06 8.18 1.66 1.47 2.11 10.19 2 450 0 332 72 82.53 78.61 75 8.20 1.67 10.04 2.51 1.37 3.17 12.97 3 184 303 175 347 84 80 83.00 8.37 1.12 9.73 0.30 1.31 0.80 6.81 4 281 681 181 790 51.93 49.17 82.15 30.81 2.56 33.68 0.80 0.72 1.57 7.32 5 1125 790 1371 1180 84.76 76.08 78.47 8.39 4.17 11.95 0.94 1.60 2.17 8.57 6 13475 4356 13411 4532 92.91 85.22 91.62 4.01 5.44 7.78 2.84 2.25 2.39 4.52 7 14690 10391 15481 10128 87.93 81.53 90.43 6.29 4.84 9.66 3.66 3.79 2.93 5.12 8 18289 12309 18404 11937 87.89 80.36 83.98 6.13 4.98 9.72 2.71 4.09 3.36 8.87 9 15529 19688 15706 19464 85.08 77.59 76.90 7.44 4.81 10.90 3.74 3.45 4.12 13.84 10 18660 13607 18850 13573 78.22 66.76 71.33 10.87 5.73 15.58 3.91 3.81 4.26 18.59 11 14706 16795 14697 17260 75.35 64.61 73.62 11.71 5.59 16.33 3.33 3.35 3.57 15.72 12 18000 17176 18117 17211 73.94 62.26 70.95 12.74 6.13 17.92 3.47 2.92 3.55 18.27 13 17342 13461 17268 13809 73.23 59.92 72.42 12.75 6.47 18.18 2.93 3.05 3.65 16.80 14 13082 12691 13593 12870 71.99 61.66 75.13 13.11 5.30 17.62 2.77 2.53 3.17 14.93 15 13240 14950 13552 15295 74.53 65.86 74.72 11.69 4.77 15.59 2.47 2.72 3.03 15.87 16 14984 10778 15211 11287 77.48 69.17 77.14 11.04 4.54 14.56 2.86 2.02 2.88 13.02 17 9921 13590 9901 14002 80.35 75.39 77.28 9.47 3.32 11.78 2.61 1.68 2.77 13.74 18 10258 14527 10404 15197 80.72 74.69 75.30 9.10 3.51 11.77 2.91 1.77 3.18 14.23 19 12360 15729 12075 15437 79.40 71.44 77.35 9.83 4.13 13.05 3.55 2.10 3.37 14.18 20 8296 16768 8111 16528 80.38 74.22 76.29 9.61 3.61 12.15 2.96 2.41 3.90 15.02 21 8136 17921 8100 17422 79.88 75.02 75.15 8.85 3.35 11.28 3.39 2.83 4.00 16.00 22 13225 11268 13071 11051 80.02 67.44 73.14 9.29 6.17 14.20 3.21 3.75 4.40 16.02 23 7344 7037 7438 7151 81.12 73.69 74.90 8.83 4.31 11.97 2.48 3.08 3.38 13.98 247610 247078 248798 249593 79.98 71.13 76.26 9.67 4.97 13.55 3.11 2.84 3.49 14.36 Table 2-7. ASPM delay data by hour.

24 From 1/1/2013 To 1/1/2013: 'SFO' Airport Scheduled Departure Date Departures For Metric Computation Arrivals For Metric Computation Gate Departure Delay Taxi-Out Time Taxi-Out Delay Airport Departure Delay Gate Arrival Delay Total 60-119 120- 179 180+ 60- 119 90+ 120- 179 180+ 60- 119 120- 179 180+ Total 60- 119 120- 179 180+ Total 60- 119 120- 179 180+ SFO 01/1/2013 507 503 51 4 2 0 0 0 0 0 0 0 0 75 3 3 0 59 7 1 1 Total 507 503 51 4 2 0 0 0 0 0 0 0 0 75 3 3 0 59 7 1 1 Table 2-8. SFO Delay Counts by Airport Report for 1/1/2013. Airport : SFO Carrier : ALL Dates : From 4/7/2013 To 4/7/2013 % On-Time Operations Departures 85% Arrivals 74% Weather IA - Instrument Approach Conditions VA - Visual Approach Conditions Arrivals IA 88.36% VA 11.64% Efficiency Airport 96 Departure 98 Arrival 95 Capacity Arrival 797 Departure 1018 Total 1815 Traffic Counts Scheduled Operations 1035 Times (Average Minutes) Departures from SFO: Gate Delay 6.80 Taxi-Out Delay 3.02 Arrivals to SFO: Airborne Delay 5.51 Taxi-In Delay 1.23 Block Delay 5.59 Arrival Delay 12.11 Facility Reported Operations * Air Carrier 832 Air Taxi 239 General Aviation 22 Military 3 Total 1096 * - Data for all operations (AC, AT, GA, MIL) Table 2-9. ASPM standard reporting under the management reports function. If an aviation analyst is interested in obtaining delay counts only (Table 2-8), a delay count option is selected under the type of analysis function. Results, in the form of counts, broken down by minutes, are then displayed for the following metrics: Gate Departure Delay, Taxi-Out Time, Taxi-Out Delay, Air- port Departure Delay, and Gate Arrival Delay. The most common performance analyses are displayed under the Management Report Function in the ASPM stan- dard reporting. After an analyst selects a target airport from the U.S. map and a time period, the following information is displayed in a tabulated form (Table 2-9): % On-Time Opera- tions, Weather, Efficiency, Capacity, Traffic Counts, Times, and Facility Reported Operations. The ASPM Standard Reporting for % On-Time Gate Operations, Taxi-Out Delay and Airborne Delay can also be displayed using graphs (Figure 2-9). ASPM records are created using data from a variety of sources with varying update cycles. TFMS and ARINC sup- ply next-day operational data, and Innovata provides flight schedule data, while ASQP provides finalized schedule data, OOOI data, and delay causes as reported by the carriers after the close of each month. ASPM is further enhanced with weather data and airport-specific information. ASPM data sources and updated cycles are displayed in Table 2-10. Actual/Historical Data vs. Calculated/Estimated Data Calculated data for: 1. Actual gate-out time, 2. Actual gate-in time, 3. Actual wheels-off time, 4. Actual wheels-on time, 5. Average taxi-out time and average taxi-in time, 6. Unimpeded taxi-in time, 7. Unimpeded taxi-out time, and 8. Matching flight schedule data to flights in ETMS.

25 Metrics computed in ASPM are developed comparing actual time to scheduled time or flight-plan time. Taxi delays are determined based on unimpeded times. Delays are calculated for • % On-Time Gate Dept, • % On-Time Gate Arr, • Taxi-Out Delay, • Taxi-In Delay, • Gate Delay, and • Block Delay. Strengths ASPM efficiency rates were specifically created to measure an ATC facility’s ability—and by extension, that of the air traffic system—to do what it says it can do. The ASPM data- base also includes airport weather, runway configuration, and arrival and departure rates. Weaknesses ASPM covers only flights for ASPM airports (currently 77), and ASPM airlines (currently 22). Only some VFR traffic is included. Access to the ASPM database is restricted by pass- word and authorized by FAA. Documentation, Data Access, and Point of Contact • To obtain a login and password: https://aspm.faa.gov/ Default.asp • About ASPM: http://aspmhelp.faa.gov/index.php/ ASPM_System_Overview • Address: Office of Aviation Policy and Plans Federal Aviation Administration 800 Independence Ave., SW Washington, D.C. 20591 Phone: (202) 267-3336 Fax: (202) 267-5370 2.2.6 BTS The BTS provides several aviation databases, including Airline Traffic Data, Air Fares Data, and Airline On-Time Statistics. The Airline On-Time Statistics database provides useful information on delay causes and answers the follow- ing questions: • How do we know the reason for a flight being late or cancelled? • Which airlines report on-time data? • Do the airlines report the exact cause of the delay? • How are these categories defined? • What have the airline reports on the causes of delay shown about flight delays? • Is it true that weather causes only 6% of flight delays? • How many flights were really delayed by weather? • Why aren’t all weather-related delays reported as a single number? • Is more information available on the Air Carrier On-Time Reporting Advisory Committee? Figure 2-9. ASPM standard reporting samples.

Data Source Content Update Cycle Purpose Traffic Flow Management System (TFMS) Flight-level records assembled by combining electronic messages transmitted to the host (en route) computer. Data include aircraft ID (flight number or tail number), flight-plan times, AZ and DZ times, arrival and departure airport, and aircraft and flight type. Daily update with preliminary next-day TFMS data and enhanced 5-day data. Approximately 10 days after the end of each month, TFMS data for the previous month are finalized. Source of flight-level data for passenger, freight, general aviation, and military flights that have filed a flight plan or otherwise transmitted data to the host computer. Includes scheduled and non- scheduled flights. ARINC OOOI data. Updated daily. Source of actual flight times for ACARS-equipped aircraft for eight airlines: AAL, ACA, DAL, FDX, SWA, UAL, UPS, and USA. CountOps Arrivals and departure information for individual flights. Also identifies arrival and departure runway end. CountOps includes all flights captured by OPSNET, but CountOps field only used to supplement ASPM flights (flights to or from the ASPM 77 airports or operated by one of the ASPM carriers.) Updated daily. Additional source of next-day on and off data for flights not captured by ARINC. Innovata A private aviation data provider of carrier flight schedules. Data were previously obtained from OAG. See FSDS. Updated every 2 weeks for the current month and next 5 months (6 months total). Source of schedule data for carriers flying domestic flights and international flights to or from the United States. ASQP Data provided by BTS. Schedule, flight plan, OOOI data, and delay causes as reported by carriers that handle at least 1% of total domestic scheduled passenger service. Monthly file loaded approximately 25 days after the end of the month. ASQP correction files are occasionally submitted several months later, resulting in a reload of ASPM for the affected months. ASPM records are updated with new or revised flight information from ASQP, including OOOI data, final schedule data from the CRS, and carrier-reported delay causes. Unimpeded Taxi Times Unimpeded taxi data by airport and carrier. Every year (typically in March), new unimpeded taxi times are calculated based on observed data from the previous year. ASPM is reloaded from December forward to apply the updated taxi times. Used for calculating taxi delays for individual flights. Operational Information System (OIS) Runway configuration data and arrival and departure rate data. Updated daily for the last 30 days. Used for summary reports of airport efficiency by hour and quarter hour. National Weather Service Airport weather data. Weather data are retrieved from three sources: METAR – published hourly without quality controls; ASOS – hourly data with some quality controls, available next day but not current through the end of the prior day; and QCLCD – quality controlled month-to-date file also not current through the end of the prior day. The month is finalized on the 6th of the following month regardless of it being complete. Therefore, occasionally ASPM does not receive QCLCD data for the last day or two of the month. ASPM uses QCLCD when it is available, followed by ASOS, and then METAR to fill in where the other two sources are not current. For more information see ASPM Weather Processing. Used for Airport Efficiency Daily Weather by Hour Report and calculating weather impact in the weather factors module. Source: http://aspmhelp.faa.gov/index.php/ASPM:_Data_Sources_and_Update_Cycle Table 2-10. ASPM data sources and update cycle.

27 time, departure delay, wheels-off time, and taxi-out time) by airport and airline; airborne time, cancellation, and diver- sion by airport and airline. – Departures, – Arrivals, – Airborne time, – Cancellation, and – Diversion. Since mid 2003, BTS has also collected the causes of flight delays, as reported by the carriers. Depending on a particular study scope, an aviation analyst may perform a series of analyses, by following these general steps: 1. Access the website at http://apps.bts.gov/, 2. Select Data and Statistics tab, 3. Select Airline On-Time Data option where you can choose summary statistics or detailed statistics, 4. If interested in summary statistics for an airport (as depicted in Figure 2-10), select the following: – Origin airport, destination airport, or origin and desti- nation airport, • Is more information available on the causes of delays and cancellations? Airline on-time data are reported each month to BTS by the 16 U.S. air carriers that have at least 1% of total domestic scheduled-service passenger revenues, plus two other carriers that report voluntarily. The data cover nonstop scheduled- service flights between points within the United States (includ- ing territories) as described in 14 CFR Part 234 of DOT’s regulations. Data are available since January 1995. The follow- ing statistics are available: • Summary statistics—All (total number, average departure delay, average taxi-out, and average scheduled departure) and late flights (total and percent of diverted and cancelled flights). – Origin airport, – Destination airport, – Origin and destination airport, – Airline, and – flight number. • Detailed statistics—Departure and arrival statistics (sched- uled departure time, actual departure time, scheduled elapse Source: RITA BTS Figure 2-10. BTS summary statistics request interface.

28 Scheduled Departure Time, Actual Departure Time, Depar- ture Delay (Minutes), Delay Carrier (Minutes), Delay Weather (Minutes), Delay National Aviation System (Minutes), Delay Security (Minutes), and Delay Late Aircraft Arrival (Minutes). Actual/Historical Data vs. Calculated/Estimated Data Data was derived for some items as follows: all and late flights (total number, average departure delay, average taxi- out and average scheduled departure) and late flights (total and percent of diverted and cancelled flights). Strengths This is the only publicly accessible database that con- tains flight cancellation information and percent of diverted flights. Better for aggregate analysis than ASPM. – Airline, and – Time period. The output results display information as shown in Fig- ure 2-11, with an average of all flights operating at the selected airport, and statistics for a selected airline, where flights are divided into all flights, and late flights. Under the categories of All Flights and Late Flights, the following results are displayed: Total Number, Average Departure Delay (Minutes), Average Taxi-Out (Minutes), Average Scheduled Departure to Takeoff (Minutes). Other information is related to the following metrics: Total Number Cancelled, Percent Flight Cancelled, Total Number Diverted, Percent Flights Diverted, and Percent Flights Late. If interested in detailed statistics, an analyst may select infor- mation for departing aircraft (departures) for one airline at a selected airport, for a specific time period. A wide range of sta- tistics can be extracted from the BTS database (Figure 2-12). The detailed statistics for departures for a selected airport (Figure 2-13) display information on Destination Airport, Figure 2-11. BTS summary statistics. Source: RITA BTS

29 U.S. Department of Transportation 1200 New Jersey Avenue, SE Washington, D.C. 20590 2.2.7 Local Airport Systems The local airport may have computer systems installed for other functions (e.g., noise monitoring) that contain a record of aircraft operations and times. These data sources should not be overlooked when an analyst is evaluating delay times because the local data source may have a more complete set of the aircraft operations than some of the FAA or BTS sources that only have data for certain airports and/or airlines. Weaknesses This contains data for only 16 U.S. air carriers that have at least 1% of total domestic scheduled service passenger reve- nues, plus two other carriers that report voluntarily. Contains less data than ASPM. Documentation, Data Access, and Point of Contact • BTS is public data; user access is not restricted. • About BTS: www.bts.gov • Address: Bureau of Transportation Statistics Research and Innovative Technology Administration Source: RITA BTS Figure 2-12. BTS statistics selection.

30 Source: RITA BTS Figure 2-13. BTS analysis reports.

31 Hughes Technical Center also uses RDSIM-Runway Delay Simulation Model and ADSIM-Airfield Delay Simulation Model for some of their airport delay analyses; those tools are generally not used outside the FAA.) 1. SIMMOD calculates the nominal travel time for each flight in the simulation, assuming that the aircraft is able to tra- verse along its path at its nominal speed on each link. Delay is then accumulated whenever the aircraft must experience a wait to maintain the required separations. Thus, delay equals the actual simulated travel time less the nominal travel time. If only one aircraft is simulated in the model, then it will traverse each link at the nominal speed, and its delay will be zero. Although not common, SIMMOD may report a negative delay when the simulation increases the aircraft speed to resolve a potential separation conflict, thus the travel time is then less than the nominal time. Delays are reported for arrival airspace delay, arrival taxi-in delay, gate delay (if no gate is available), departure taxi-out delay, departure queue delay, and departure airspace delay. Users can analyze delays by cause of the delay and the location/ link where each delay occurred. For airport planning, ana- lysts typically report the following: • Arrival airspace delay, • Arrival taxi-in delay, and • Departure taxi-out delay + queue delay. 2. Total Airspace and Airport Modeler (TAAM) calculates delays for virtually every segment of a flight starting with gate delay and finishing with taxi-in delay, which also includes delay incurred due to an unavailable arrival gate. TAAM bases delay on a comparison of incurred time versus unimpeded times calculated by the model, based on the conditions along the route of flight and the aircraft operat- ing characteristics. Thus, TAAM calculates the shortest time a flight should incur during any phase (e.g., at gate, during taxi-out, sequencing, etc.) and then subtracts that from the time actually incurred by that flight and reports the differ- ence as a delay for that segment of the flight. Delay calcu- lations available through the TAAM reporting function include gate delay, taxi-out delay, runway delay, departure queue delay, in-trail delay, sequencing delay, positioning delay, flow management delay, gate-turn (link) delay, taxi-in delay, delay per taxiway segment, ground delay, and airborne delay. TAAM also offers a number of cumulative delay indices such as follows: • Departure delay per aircraft—the sum of gate delay, taxi- way delay and departure queue delay. There is no airspace delay for departures. • Arrival delay per aircraft—the sum of taxi delay, arrival gate delay, and arrival sequencing delay. • Airport delay—the sum of arrival and departure delays. Does not include en route delay but does reflect delay due to sequencing action for arrivals. Specifically, noise monitoring systems often contain the time the aircraft used the runway (wheels on for arrivals and wheels off for departures), aircraft type, flight number identi- fier, runway used, etc., for all flights; this can be correlated to other sources to obtain taxi-in and taxi-out times for specific runways and configuration. Noise monitoring systems and/ or gate assignment programs can be used to analyze vari- ances between actual gate in and out times to scheduled gate arrival/departure times. 2.3 How Delays Are Calculated For airport planning purposes, analysts typically estimate delays of current and proposed airport layouts using methods ranging from basic analytical tools to sophisticated, detailed computer simulation models. This section provides informa- tion on which methodologies and models are most appro- priate for different planning efforts. In addition, this section addresses which types of delays are most appropriate for spe- cific purposes (i.e., when to use average delay vs. maximum delay, average annualized delays, etc.). 2.3.1 Spreadsheet and Basic Models The FAA’s AC 150/5060-5 Airport Capacity and Delay and the subsequent Airport Capacity Model (ACM) provide straightforward calculations to estimate average delays for particular runway layouts, based on fleet mix and several other items. Using this approach, average hourly, daily, and annual delays can be estimated. The calculations were based on ATC rules and procedures that were in place when the model and advisory circular (AC) were developed. Analysts are not able to adjust the delay estimates as new procedures, technology, and/or separation rules are implemented. Some consultants and academic institutions have devel- oped and applied basic queuing models. Typically, the models require the analyst to input the hourly throughput under differ- ent arrival-departure ratios, then the queuing model estimates average delays based on the traffic demand profile provided. 2.3.2 Simulation Models When calculating aircraft delays for airport infrastructure projects, analysts often use computer simulation tools to evaluate delays and delay savings. Capacity driven delays can be predicted very accurately using these models. Regardless of the simulation tool used, a nominal or unimpeded time is calculated, then any additional time is measured as “delay.” A discussion of how delay is calculated by commonly used air- port simulation tools follows. Note that the four simulation models listed here all have some type of animation capability to visualize the aircraft moving through the airspace and on the airfield. (In addition to these four models, the FAA William J.

32 particular wind/weather configuration is multiplied by the annual percentage of time that wind/weather configuration is in use at that airport. This results in a weighted average annualized delay, which is the usual measure for comparing airport development alternatives. Whether using simulation or spreadsheet or other analytical methods, delay analyses will typically be run for several wind/ weather configurations that are used at an airport to evaluate delays in the various configurations. Even for an airport with a single runway or pair of parallel runways, the runway is likely used in both the primary direction and the opposite or sec- ondary direction, depending on the winds. Often, at least four operating configurations are modeled as follows: • Primary runway direction in VMC, • Secondary runway direction in VMC, • Primary runway direction in IMC, and • Secondary runway direction in IMC. Of course, some airports or airfield layouts will have more than four operating configurations. The analyst then has arrival and departure delay values for each configuration. An exam- ple of calculating an average annualized delay is depicted in Table 2-11. With 60% of the flights experiencing 2.0 minutes of delay during the year (primary VMC), 20% of flights experienc- ing 2.5 minutes of delay (secondary VMC), 10% of flights experiencing 13 minutes of delay and 10% of flights experi- encing 19 minutes of delay, the resulting weighted average is 4.9 minutes of average annualized delay. Average annualized delays (Table 2-12) are extremely helpful for initial comparisons of development alternatives. However, such high-level delay measures can mask large delays that occur in a particular wind/weather configuration when that configuration does not occur very frequently dur- ing a typical year. In the above example, 80% of the flights experience, on average, 2.5 minutes of delay or less; but the high delays in IMC result in an average annualized delay almost twice that level. Due to operational issues that occur in that configuration (e.g., congestion, irregular operations recovery, etc.), the airport may choose to develop improve- ments to that configuration even if it has little influence on or reduction of the annualized delays. Using the example again, when this calculation is con- ducted for each demand level (or number of flights) and for 3. ARCPort models a link-and-node representation of the airfield and airspace. Main sources of airfield delay are reported, including the following: • Air delay—on approach to the airport, • Taxi delay—on the airfield for arriving and departing aircraft, • Stand delay—on arrival waiting for a vacant stand, and • Takeoff delay—on departure for the runway clearance. 4. Comprehensive Airport Simulation Technology (CAST) aircraft simulation generates basic data on taxi times, average waiting times, queue length, flow rates, through- put times, process times, punctuality, delays, number of runway crossings, and runway occupancy times. However, airline scheduling practice changes may have a major impact on the model’s predictions of delays. For example, if 20 flights are scheduled within an hour, this may show very little delay; but if the 20 flights are scheduled within the first quarter of the hour, the model may show delays to be larger. Also, the models typically do not show how delays that occur at one airport propagate and impact other airports. The New York region—including LGA, JFK, EWR, TEB, PHL— handle roughly 12% of all domestic flights but, according to the FAA, a third to nearly half of all delays around the nation each year are caused, in some way, by the New York airports. This can be very relevant on overall planning efforts to improve airline predictability throughout the network. Also, these simulation models do not consider that delays also can have an impact on consumers choosing other modes of transportation. For cost-benefit analyses, the amount of delay on the ground and in the air is multiplied by typical operating costs to estimate economic savings that will be realized from par- ticular capital improvements. 2.3.3 Average Annualized Delay To easily compare airport development alternatives, hav- ing a single value for each option is useful. Analysts often will calculate a weighted average delay, based on the percent of time each wind/weather configuration is used throughout the year. The average annualized delay is a weighted average of the delays in the various wind/weather operating configurations used at an airport. Commonly, analysts will run entire days of flight demand for each of the typical wind/weather scenarios that occur at an airport. Then the average daily delay for each Arrival Departure Per Flight* Primary wind – VMC 60% 1.0 3.0 2.0 Secondary wind – VMC 20% 1.5 3.5 2.5 Primary wind – IMC 10% 6.0 20.0 13.0 Secondary wind - IMC 10% 8.0 30.0 19.0 Average Annualized Delay 4.9 *Assumes equal number of arrival and departure operations Table 2-11. Sample average delays (in minutes) for Demand Level 1.

33 turnaround times within scheduled times. Some delay mea- sures that may be useful to carriers include the following: • Average hourly delays for each wind/weather configura- tion (IMC and VMC), • Peak hour delays, • 95th percentile (i.e., 95% of the flights have a delay value less than the reported value; conversely, 5% of the flights are estimated/simulated to experience a delay above that value), and • Maximum delay, meaning the maximum time that any air- craft is delayed. For any of the simulated/calculated delay items, the delays can be reported for different time periods, as follows: • Daily average—may mask the delay problem during peak times when there is little or no delay during most hours of the day. • Annual average—after analyzing delays in each of the major wind/weather configurations, then a weighted average of the daily average delay is calculated. Producing a single value, this is a convenient measure for comparing develop- ment alternatives. each scenario, the resulting average annualized delays could be reported in a table, but the resulting delays are generally plotted in a graph similar to Figure 2-14 so that the delay curves can be visualized. These types of capacity curves are very typically used in analysis of capacity constraint delays. Analysts may then fit the calculated data to representative curves for presentation purposes, as shown in Figure 2-15. 2.4 Additional Delay Statistics Although average delay may work well for airport plan- ning, it is generally not detailed enough for airline decisions. Also, in the public arena, publicizing that an airport has an average delay of 10 minutes may seem to be “no big deal,” when in fact it is very costly to airline operations. Expressing delays in millions of dollars in lost time and operating costs is typically more effective. Also, average annualized delay may not be effective measures for some airports with high levels of special event traffic or great variance of flights in different seasons (e.g., beach, ski, resort locations). At a minimum, the average delay needs to have a cor- responding maximum delay or average peak hour delays reported. For scheduled carriers, this is a better indicator of whether they can maintain schedule integrity and keep their Current Traffic Demand Level 1 Demand Level 2 Demand Level 3 No Build/Do Nothing 4.0 4.9 7.2 15.6 Runway Option 1 2.5 2.7 3.7 5.9 Runway Option 2 2.4 2.6 3.0 4.0 Table 2-12. Average annualized delays per flight. Source: TransSolutions Figure 2-14. Average annualized delays for airport scenarios and future demand levels.

34 for an airport with a single runway, using a forecast traffic demand level approximately 60% higher than the current traffic level.) Typical delay statistics from a simulation model report delays for all flights. Table 2-13 shows the simulated delay results for all flights in the simulation model. It also has been suggested that only those aircraft that expe- rience some delay could be included in the calculations. When this approach is taken, one must be very clear that they are reporting an “average of the delayed aircraft” and not an “aver- age delay of all aircraft.” If only 50% of flights are delayed, then the average delay of all aircraft will be half of the average delay of the delayed aircraft. The same delay statistics are shown in Table 2-14 for only the flights that experienced a delay. For this particular simu- lation run, over 80% of the departure operations experienced a delay, so the average delay for all flights (6.07 minutes) and • Seasonal average or peak month average—similar to a weighted annual average, but would show average of peak months or peak seasons when it is important to accom- modate demand for all times of the year. • Hourly average—throughout the day, this provides an indication of delays during peaks and valleys of the daily flight schedule/demand. Since this produces 24 values for a single day, this is most useful as depicted in graphical form to see how the delay varies throughout the day. • Average of peak hour—reports the maximum value from the hourly average just described. It is important that the particular hour has a significant number of operations. • Individual aircraft delays—show the extreme maximum values and how the delays vary throughout the day with the flight demand; most easily displayed graphically. From the individual aircraft delays, one can obtain maximum delay, 95th percentile (only 5% of flights would have a higher delay than this), etc. • Total delays—rather than express the delays in averages, one would report the total delay of all the flights for a day or for a year. This may translate into a very large number (e.g., thousands of hours) when calculated for the year. 2.4.1 Example: Delay Outputs from a Simulation Scenario To demonstrate how delay measurements vary according to the way in which delays are reported, this section uses an example of a particular simulation run at an airport dur- ing VMC. (For informational purposes, this scenario was Taxi-Out Delay Arrival Taxi Delay Arrival Air Delay Average 6.07 0.08 4.23 95th Percentile 20.97 0.57 22.72 Maximum 39.60 5.00 44.30 Total Flights 282 287 287 Total Delay 1714.2 22.5 1254.5 Table 2-13. Simulated delay statistics for all flights. Source: TransSolutions Figure 2-15. Average annualized delays fitted to curves.

35 Similarly, Figure 2-17 plots the arrival airspace/runway delay for all arrival flights throughout the day. Many flights are shown to experience almost no delay. However, again in the 12:00 to 14:00 timeframe, all arrivals experience some delay, with the delay steadily increasing as both arrivals and departures attempt to access the same runway during the midday peak, growing to nearly 45 minutes of delay. These types of plots can be quite useful for airlines since the 6-minute average taxi-out delay (from Table 2-13) does not demonstrate the extremes (approximately 30 minutes of delay) that several flights experience during the midday peaks. Also, the average arrival air delay of 4.23 minutes masks the high delay during the peak day where delays steadily climb from 5 minutes to nearly 45 minutes. Although the average may be useful for overall planning purposes, the additional informa- tion can be critical for gaining airline and/or public support. Other charts that can be produced from the same simu- lation output include plotting hourly departure operations against hourly average taxi-out delays (Figure 2-18) and similarly, hourly arrival operations against hourly average airspace delays (Figure 2-19). 2.4.2 Comparing Simulation Delays to FAA Delay Databases This section presents actual results from a simulation analy- sis at a large airport and compares this to the ASPM data for the same airport and time period. This particular airport experiences its peak month of traffic in March of each year. The SIMMOD simulation study baseline was conducted with a peak month schedule for 2007 and calibrated to actual hourly throughputs and observed taxi times. In March 2007, the ATCT counts reported an average of 370 daily operations, ranging from a low of 336 daily opera- tions to a high of 405 daily operations. The airport provided operations counts that average 358 daily flights. ASPM for the same month has data for an average of 308 daily operations. the average delay for only delayed flights (7.27 minutes) are quite similar. Likewise, nearly 75% of the arrival flights experience an air delay, so again the average for all flights (4.23 minutes) and the average for only delayed flights (5.92 minutes) are similar. However, just over 10% of the arrival flights experience a taxi delay, so the difference is quite large: 0.08 minutes delay for all arrival flights compared to 0.73 minutes delay for just those 31 flights that experienced a delay. Figures 2-16 and 2-17 were produced for the same simula- tion run, but the tables above are for 10 iterations while the figures are for a single iteration. Thus, the maximum delays shown in Tables 2-13 and 2-14 may not be displayed in Fig- ures 2-16 and 2-17 if the maximum delay occurred in a dif- ferent iteration than is graphed. The figures display the delay experienced by each aircraft in the simulation, plotted according to the time of day the aircraft used the runway. Figure 2-16 plots the taxi-out delay for each departure operation. Most flights before 10:00 a.m. experience taxi-out delay of 5 minutes or less. Later in the day, from 12:00 to 14:00, all flights experience some delay due to the peaked demand during that time of day. Taxi-Out Delay Arrival Taxi Delay Arrival Air Delay Average 7.27 0.73 5.92 95th Percentile 21.87 2.23 26.40 Maximum 39.60 5.00 44.30 Total Delayed Flights 236 31 212 Total Delay 1714.2 22.5 1254.5 Table 2-14. Simulated delay statistics for delayed flights only. Source: TransSolutions 0 5 10 15 20 25 30 35 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Taxi-Out Delay Time of Day M in ut es o f D el ay Source: TransSolutions 0 5 10 15 20 25 30 35 40 45 50 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Arrival Air Delay Time of Day M in ut es o f D el ay Figure 2-16. Individual flight delays for all departures. Figure 2-17. Individual flight delays for all departures.

36 • Arrival air delay in a simulation model is a function of the amount of airspace modeled, whereas any similar ASPM statistic would encompass the entire en route travel delay from “off” time at the departure airport until “on” time at the arrival airport. • Arrival ground delay in a simulation model is typically very small as the aircraft is directed immediately to an open gate or parking position. In actual operations, a flight may be stopped on a taxiway to await entry into a congested ramp or to wait for a specific gate to be vacated. • Similarly, departure ground delay in a simulation model will not start accumulating until the aircraft can depart the gate without adjacent ramp congestion. In actual Recall that ASPM includes data for 22 domestic air carriers. For the year, this particular airport recorded 13.3% general aviation, military, and local flight operations that would not be included in ASPM. So with ASPM including data for about 85% of the total airport traffic, it is a good representation of the scheduled flights. For a comparable month, the simulated traffic demand had 358 flights, of which 14.5% were general aviation/military. Key delay statistics from the simulation model are sum- marized in Table 2-15. Note that while specific weather data is not analyzed for a single month, a weighted average for VMC/ IMC is shown for comparison purposes. Taking each of these individually: Source: TransSolutions Figure 2-19. Hourly arrival operations and average hourly airspace delays. Source: TransSolutions Figure 2-18. Hourly departure operations and average hourly taxi-out delay.

37 combination of planned times with actual operational times. For example, the on-time percentages could be used as inputs to the simulation model to offset flight earliness/lateness from the scheduled times, but they would not be simulation outputs for comparison with ASPM. Some of the additional ASPM data for the month of March 2007 includes the following statistics: • 77% on-time gate departures (compared to flight plan): Flights that departed within 15 minutes past the flight- plan gate out time. • 71% on-time airport departures (compared to flight plan): Flights that depart within 15 minutes of the flight-plan wheels-off time. • 70% on-time gate arrivals (compared to flight plan): Flights that arrive at the gate less than 15 minutes late compared to the flight-plan gate out time plus the scheduled block time. The last schedule before wheels-off is used in the cal- culation, thus an EDCT hold may cause a late arrival. • Average gate departure delay of 10.5 minutes (compared to flight plan): The difference between actual gate out time and the flight-plan gate out time, in minutes. • Average airport departure delay of 13.4 minutes: The actual wheels-off minus the flight-plan gate out plus the unim- peded taxi-out time, in minutes. Negative values are allowed if the report includes early flights. • Average airborne delay of 3.2 minutes: The difference between actual airborne time and the flight-plan estimated time en route, in minutes. • Average block delay of 5.6 minutes: The difference between actual gate-to-gate time and scheduled gate-to-gate time, in minutes. • Average gate arrival delay of 14.5 minutes (compared to flight plan): The sum of minutes of gate arrival delay of 1 minute or more, divided by all arrivals. Gate arrival delay equals the actual gate in time minus the flight-plan gate in time. operations, the pilot would typically release the brake to record an earlier “out” time (to improve on-time depar- ture performance). The one simulation output that can be compared most directly to ASPM data is average taxi-out time. For March 2007, ASPM shows average taxi-out time to be 14.1 minutes. However, even with this data, the brake may be released before the flight is ready to depart in actual operations, yet the simulation does not accumulate delay until the aircraft vacates the gate. There also may be ground stops to other airports such that the departures accumulate additional departure taxi time when they have vacated the gate but are awaiting clearance. These phenomena also make it dif- ficult to compare ASPM’s taxi-out delay and taxi-in delay times to simulated values. ASPM estimates unimpeded taxi times by calendar year for each carrier and airport based on observed values in the previous year during optimal oper- ating conditions, whereas the simulation model calculates unimpeded times from the planned taxi routes. So the delay figures are not comparing actual times to flight schedule or block time estimates, but compared to FAA’s expected val- ues at that airport. For March 2007, ASPM estimates aver- age taxi-out delay at 3.8 minutes and average taxi-in delay at 2.5 minutes. Some additional items in ASPM cannot be compared directly to simulation output because they each represent a Average daily statistic VMC IMC Weighted Average 96% of year 4% of year Arrival air delay 0.9 2.0 1.0 Arrival ground delay 0.1 0.1 0.1 Departure ground delay 2.3 4.0 2.4 Taxi-in time 3.7 3.6 3.7 Taxi-out time 8.1 9.9 8.2 Table 2-15. Simulation output for March 2007 (in minutes).