Declaration of Independence (Expanded Version)
The NTSB expects its investigations of TCAS RAs to complement separate government-industry analyses of shared data.
By Wayne Rosenkrans
AeroSafety World, May 2011
As government-industry exchanges of vast banks of operational data flourish, incident-level investigations by the U.S. National Transportation Safety Board (NTSB) can appear to be out of step with the times. Some aviation safety professionals have seen the board’s approach to near-midair collisions (NMACs) as a case in point, specifically the latest requirement for operators to report certain resolution advisories (RAs) issued by traffic-alert and collision avoidance systems (TCAS II).1
Yet early indications are that NTSB investigations help to rapidly mitigate underlying risk factors of mid-air collisions, even if limited sometimes to a local application, while large-scale data analysis may take years to deliver system-level risk mitigations. Finding solutions either way has been extremely difficult, NTSB and U.S. Federal Aviation Administration (FAA) officials admit (ASW, 8/09, p. 32).
Before its effective date of March 8, 2010, the requirement to report certain RAs had been widely opposed as an unwarranted duplication of effort, but the first 12 months of RA investigations reveal more about the board’s complementary, check-and-balance purposes.
Investigating RAs has been a long-established process falling “well within our mandate,” says Tom Haueter, director, NTSB Office of Aviation Safety. “Our decision to go after formalized reporting was basically because of the problem that we didn’t know how many RAs were out there,” he said. “We previously got this information second-hand many times, and we needed to have reliable reporting of the TCAS RA events in which aircraft are in the positive control area [i.e., Class A airspace, from 18,000 ft through Flight Level (FL) 600 (approximately 60,000 ft)] or at lower altitudes” under instrument flight rules (IFR) if compliance with the RA is necessary to avert a substantial risk of collision between two or more aircraft.
The final rule creating the requirement explained: “The NTSB fully recognizes that when [an RA] occurs, it does not necessarily follow that an unsafe encounter is about to occur. The NTSB intends to require reports only when failure to comply with [an RA] would lead to an unsafe encounter with another aircraft, that is, an encounter presenting a substantial risk of collision.”
In December 2004, the NTSB had proposed to add RAs to its list of events required to be reported immediately to the board under Title 49 of the Code of Federal Regulations, Part 830, “Notification and Reporting of Aircraft Accidents or Incidents and Overdue Aircraft, and Preservation of Aircraft Wreckage, Mail, Cargo and Records.”
Two influential factors were NTSB investigators’ desire to have consistent awareness of “events in which aircraft crews perceived that they had been exposed to a collision hazard” and the NTSB’s obligation to comply with the International Civil Aviation Organization standard that national accident investigation bodies receive immediate notification of “near collisions requiring an avoidance [maneuver].”
After reviewing public comments in 2005, the board decided to make refinements. The final regulation requires reporting RAs either “when an aircraft is being operated on an [IFR] flight plan and compliance with the advisory is necessary to avert a substantial risk of collision between two or more aircraft, or [when an RA occurs on] an aircraft operating in Class A airspace.”
Visitors to the NTSB Web site www.ntsb.gov now find on the home page a “TCAS RA” reporting link separate from the link for the nine-page, PDF-format NTSB Form 6120.1, “Pilot/Operator Aircraft Accident/Incident Report.” The TCAS RA link simply launches an empty email message from the sender to email@example.com but any email program can be used to send a message to this address. “The key for us is getting accurate reports quickly — as fast as we can get them — so we can pull the air traffic control [ATC] radar tapes and interview people if necessary, and make an evaluation,” Haueter said. “If we need more data, NTSB staff will contact any person or organization as needed to complete the investigation.”
A contractor, a retired air traffic controller, performed during this time the first-level sorting of each incoming report, tagging and forwarding — to an assigned NTSB investigator from the Aviation Operational Factors Division — those that fit the brief Part 830 criteria. The investigator then performed follow-up work that began the process for deciding which reports deserved a full investigation.
From March 8, 2010, through March 8, 2011, the NTSB received about 950 RA reports. “Of the 950, there were only 260 that we thought merited additional examination to see if something serious was going on,” Haueter said. Nine RAs investigated recently include seven that occurred in the 12 months after the effective date of the final rule, one RA from October 2009 and one RA from February 2010.
As to RAs screened so far, “there have been no real surprises … nothing that jumps out in terms of a trend or something unusual,” he said. Investigators’ reviews of the 260 reports did not support categorization or identification of “pockets” of airspace (hot spots) where more RAs occurred than normal. “The events were about what we have seen before, but we will keep collecting data … and each year we will know better which to investigate, and we will refine the process if necessary,” he said. “This is going to take a long time.”
From these investigations, the probable cause has been determined for one RA. For the remaining events, preliminary and/or factual reports have been issued. For the probable cause investigation and one factual report, more publicly accessible documents have been issued through the NTSB Docket Management System dms.ntsb.gov/pubdms/search. For the 12-month period before March 2010, the NTSB Aviation Accident Database contains two preliminary reports involving RA investigations identifiable by the term TCAS in the narratives.
From Haueter’s perspective, the most prominent of the nine RA investigations was an NMAC on Sept. 16, 2010. This collision was averted by an immediate climb maneuver performed by the flight crew of a US Airways Airbus A320. The A320 crew and the pilot of a Beech 99, operated by Bemidji Aviation Services, had been cleared to conduct takeoffs and departure turns in instrument meteorological conditions from parallel Runways 30R and 30L, respectively, at Minneapolis-St. Paul (Minnesota) International Airport. Airport weather conditions included a reported ceiling at 900 ft and visibility of 10 mi (16 km).
After takeoff, the A320 crew had received and complied with an ATC instruction to turn left to heading 260. The air traffic controller responsible for the Beech 99 cargo flight’s departure instructed the pilot to take off and turn left to heading 180.
However, the pilot delayed his compliance with the turn instruction for about 2.0 nm (3.7 km) until reminded, and the controller did not look at the radar display or otherwise realize that this delay was causing the path of the Beech 99 to intersect the path of the A320. The NTSB investigation found that about one minute after the TCAS RA, the same controller issued a vector to the Beech 99 pilot that caused a second, unreported loss of separation — a radar proximity of 500 ft and 1.23 nm (2.28 km).
The Beech 99 conflict with the A320 occurred because of the controller’s assumption that the Beech 99 pilot had turned immediately after takeoff, Haueter said. “That kind of assumption in the ATC system is one we have seen before,” he said. “By being able to see radar tracks and make safety recommendations, hopefully we can prevent this issue from leading to an accident.” The factual report noted that the incident controller was distracted by a taxiing aircraft pilot’s questions about an ATC instruction.
In another example from the set, the database narrative for an accident on Oct. 26, 2009, said that one flight attendant sustained a serious injury and one passenger sustained a minor injury when the flight crew of an AirTran Airways Boeing 717 responded to an RA during cruise-descent near Norfolk, Virginia, and maneuvered to avoid another airplane by suddenly stopping their descent.
The database narrative for an accident on Feb. 13, 2010, said that one flight attendant sustained a serious injury and one flight attendant sustained a minor injury when the flight crew of a Southwest Airlines 737 abruptly maneuvered for a descent and then a climb — both at rates estimated at 1,000 to 2,000 fpm — in response to an RA at about 6,000 ft over Santa Clarita, California, while complying with ATC radar vectors for initial approach to Bob Hope Airport, Burbank.
First Probable Cause
The probable cause has been determined for a serious incident that occurred March 25, 2010, when the flight paths of a Continental Airlines 737 and a Gulfstream II crossed within 1.04 nm (1.93 km) and 300 ft in Class A airspace over Worton, Maryland. Just before the incident, the GII was at FL 290 and the 737 was at FL 360. An operational error by the controller responsible for the GII occurred during her attempt to simultaneously vector this flight crew to pass clear of Aberdeen Restricted Area and to position the GII more than 5.0 nm (9.3 km) behind the 737, the report said.
The probable cause was, “The [radar controller for sector 10/12 of the Washington Air Route Traffic Control Center] issued an improper vector and descent clearance to the GII that put the airplane on a converging flight path with the B737.
Contributing to the incident was the failure of the FAA’s training program to correct ongoing controller performance deficiencies before certifying the [manual controller for sector 10/12] to work without immediate supervision.” The documents in the public docket are similar in scope to those for the Minneapolis incident.
The documents from the docket provide details from radar track replay analysis and the transcribed audio recordings of pilot-controller communication; interviews with pilots, local controllers and ATC supervisors; analysis of applicable ATC rules, procedures, typical route coordination, radar/visual separation practices, radar range setting, automatic acquisition of radar target data tags, position relief briefings, and duty assignments; and analysis of the incident controller’s training, fatigue, duties and past performance issues.
These associated reports also describe the FAA’s quality assurance investigation, include three local directives to controllers issued before completion of the NTSB investigation, and cite the planned follow-up actions by quality assurance staff from FAA headquarters. The docket also contains an NTSB comparison of similarities between this incident and an ATC operational error that resulted in loss of separation on Nov. 11, 2010, between two airliners departing from these same runways.
Unfinished investigations include an RA that the flight crew of a 777 responded to on March 27, 2010, at about 1,100 ft on departure climb from San Francisco International Airport. The crew reported seeing the underside of a light high-wing airplane at the time, and the slant range proximity was estimated by the pilots as 100 to 200 ft (30 to 61 m).
Another is an ATC operational error on April 21, 2010, at Chicago O’Hare International Airport, involving an Embraer 135 in a departure climb and a Raytheon King Air being flown to check the localizer of the instrument landing system; with closest proximity on radar 300 ft and 0.41 nm (0.76 km).
On Aug. 9, 2010, the flight crew of an Embraer 170 cruising at FL 290 abruptly climbed 600 ft in response to an RA after the flight crew of a converging U.S. Air Force Northrop T-38 deviated for 30 seconds from their assigned altitude, FL 280, by climbing 600 ft.
On Jan. 20, 2011, the flight crew of a 777 climbing in visual meteorological conditions after takeoff from John F. Kennedy International Airport and the flight crews of a flight of two U.S. Air Force Boeing C-17s received RAs, traffic advisories and conflict-resolving vectors around the time that their closest proximity on radar reached 200 ft and 0.84 mi (1.4 km).
Strict NTSB Independence
In response to the 2004 and 2008 notices of proposed rulemaking for Part 830, the airline industry and the FAA urged the NTSB to endorse, rely upon or — ideally — participate in the existing voluntary non-punitive FAA-industry processes for reporting and analyzing RAs. Often mentioned was joining in the FAA Aviation Safety Information Analysis and Sharing (ASIAS) program, which currently has 35 participating airlines. The NTSB declines to do so, although some have seen the resulting limited access to data as a disadvantage.
“Certainly the FAA and airlines can take their data and look at it through the ASIAS viewpoint; we can’t,” Haueter said. “We are not linked into ASIAS.”
Some observers may have misconstrued the statutory safety-oversight role of the NTSB, and how this limits relationships with the FAA and the industry. “We have a ‘watchdog’ function over the FAA, and one of our functions is to oversee ATC safety,” Haueter said. “As the regulatory agency running the ATC system, they can make changes. So they do their own investigations of RAs, and we do ours. This works quite well as a system. Certainly, we will share with the FAA any of our information.”
The Part 830 final rule also explained practical aspects of the NTSB’s stance. “The NTSB does not believe that the FAA’s processes for assessing and reporting incidents, particularly those involving losses of separation, are sufficiently reliable,” the NTSB said. “The NTSB has concluded that a source of safety reports not solely dependent on ATC will provide a useful means of ensuring that serious incidents receive adequate attention and will enable improvements to the ATC reporting process, where needed. … While the NTSB does support [ASIAS, flight operational quality assurance (FOQA) and aviation safety action programs (ASAPs)] in principle, the de-identified and otherwise filtered information available through them is not useful for investigative purposes. … The NTSB cannot delegate [investigative] responsibilities to external organizations, become wholly dependent on information such organizations may or may not see fit to share, or limit the investigative use of that information to comply with accompanying restrictions.”
Meanwhile, many advantages accrue from the increased RA reports reaching the NTSB. “We now have a better handle on what’s going on … numbers to back up what we have been looking at,” Haueter explains. “Yet each of these events is unique, so it has been hard to pin down exactly where we definitely see improvement necessary.”
The most important driver of these NTSB investigations, Haueter said, is ensuring a detailed awareness of how the few unsafe situations developed and resulted in the RAs. His basic message to pilots and airlines willing to read investigation reports is: “Be vigilant; watch out for situations where you might lead yourself or ATC may inadvertently lead you into another airplane’s airspace.”
Broader RA Interests
In March 2011, one briefing of airline representatives attending the FAA Infoshare meeting — a twice-a-year meeting of participants and prospective participants in ASAP, FOQA and advanced qualification programs — highlighted the evolution of RA investigations at the FAA.
“We know the days of a ‘find and fix’ approach to safety just won’t cut it anymore,” said Margaret Gilligan, FAA associate administrator for aviation safety. “We’ve looked at all major airports where airline concerns [about non-safety-critical RAs] were expressed. … We literally mapped all of the TCAS RAs over a certain period of time for all of the arrivals and departures at every runway in the United States [using] FOQA, ASAPs and radar data. After analyzing the data, we found out that unwarranted TCAS RAs can desensitize flight crews because [of RAs] in situations where aircraft are adequately separated in accordance with air traffic control rules and procedures. And aircraft arriving on closely spaced parallel runways under visual [meteorological] conditions may get a TCAS RA even if both aircraft are adequately separated for the arrival.”
Earlier this year, the FAA began conducting the Collision Avoidance Outreach Program for the pilot community, sharing results of TCAS and RA research and soliciting feedback about operational experiences and preferences to be factored into TCAS enhancements. The FAA also completed updates of several TCAS-related advisory circulars and other guidance materials, Gilligan said.2
Recent FAA briefings of RTCA Special Committee 147, which is devoted to TCAS issues, have covered solutions in progress, such as modifying airspace design and ATC procedures in light of RA hot spot knowledge, updating equipment for TCAS II Version 7.1 (ASW, 4/09, p. 34) and designing next-generation collision avoidance system or “NextCAS” equipment.
Government-industry discussions under way include the possibility of uplinking TCAS sensitivity-level control messages to aircraft via Mode S transponder near Denver International Airport, and perhaps other airports in mountainous areas; the feasibility of horizontal maneuvering in response to RAs; the pros and cons of downlinking RAs to ATC displays as in four European countries; how TCAS would perform if en route separation standards were reduced from 5.0 nm (9.3 km) to 3.0 nm (5.6 km); how technical standards would have to be updated to design combination TCAS–ADS-B (automatic dependent surveillance–broadcast) receivers; and certification of the Airbus TCAS Alert Prevention system (ASW, 2/11, p. 13).
The nation’s largest source of raw data about RAs is the FAA-funded TCAS Operational Assessment Program, managed by the Massachusetts Institute of Technology Lincoln Laboratory, which automatically monitors and records any TCAS-generated data embedded in Mode S transponder messages within the range of 20 long-range ground radar sites nationwide.3 The program’s database now stores about 100,000 RAs and its Web site solicits operational details from pilots and controllers through online questionnaires.
These researchers focus on “understanding and characterizing the performance of TCAS and its use by pilots” — such as calculating rates of compliance with climb RAs and descend RAs and assessing the relatively high rates of RAs on business jets — but avoid characterizing their work as a safety program.
The NTSB acknowledged in 2010 the promise of such research but decided that “the infrastructure to provide this capability is not sufficiently common to ensure that the NTSB would receive notification of an event” through Mode S datalink technology.
ASIAS Directed Study
From 2008 to 2010, the ASIAS issue analysis team studied RAs to determine areas within the national airspace system where the alerts have occurred with relatively high frequency, conducted focused investigations at problematic airports and characterized the causes of RAs, according to the FAA briefings. As the FAA’s Gilligan noted, the ASIAS directed study of RAs showed that they have occurred in many situations in which aircraft were adequately separated per ATC rules and procedures, prompting intense interest in current TCAS alerting parameters, high-altitude airports, airspace design processes and ATC procedures. One report cited said, “TCAS … may issue an RA to aircraft with as much as 600 ft of vertical separation.”
Examining fusions of data that represent aircraft operating under visual flight rules (VFR) and aircraft operating under IFR, ASIAS work has focused in part on general aviation traffic under airport tower control operating near structured IFR traffic at a nearby airport. Another context of relatively high-frequency RAs has been aerial work aircraft “loitering” adjacent to structured IFR traffic to/from a nearby airport, sometimes operating under VFR in or near Class B airspace and interacting with aircraft arrivals to a large airport under IFR. Another scenario identified was structured flows of VFR aircraft — such as VFR helicopter routes and VFR transition routes/flyways — interacting with nearby structured flows of IFR aircraft.
Educating the Industry
Uncertainty persists for now about how many RA reports typically will arrive per year at the NTSB, but outside predictions of many thousands have not materialized, and polite reminders have been effective in enforcing compliance by all operators involved in each reportable event. “One thing we do know from the first year is that there has been a lot of over-reporting,” the NTSB’s Haueter said. “Some people reported TCAS RAs that they did not have to report, so we are educating the industry, and I imagine in the following years, we will see the number decrease a bit.”
Flight crews, pilots and operators can use as a general guideline the FAA definition of an NMAC, given that “the infinite variety of encounter geometries does not lend itself to specific [RA-reporting] guidance that would apply to every possible scenario,” the NTSB said. An NMAC is ‘‘an incident associated with the operation of an aircraft in which a possibility of collision occurs as a result of proximity of less than 500 ft [152 m] to another aircraft, or a report is received from a pilot or a flight crewmember stating that a collision hazard existed between two or more aircraft.’’
An explanation in the final rule also clarified, “[RAs] that command maximum vertical speed, ‘reversal’ advisories that require a change in vertical direction after the initial advisory is issued, or encounters that result in zero vertical separation between the aircraft involved are all examples of the types of advisories that the NTSB believes may be indicative of substantial collision risk. Conversely, [RAs] issued to aircraft operating on closely spaced parallel approaches or in other circumstances where there is no substantial risk of collision need not be reported under this rule.”
- The NTSB uses the international term airborne collision avoidance system (ACAS).
- For example, the updated FAA AC 120-55C, “Air Carrier Operational Approval and Use of TCAS II,” was issued in February 2011.
- Data collected include the identity of the aircraft on which the TCAS alert was issued, the type of alert and radar track information for nearby aircraft. The program’s Web site tcasreport.com says that data are used only for TCAS event analysis, do not “support punitive or enforcement actions” and are de-identified within 45 days.
Top of page