Inside the NTSB’s General Aviation Investigative Process

Do We See and Avoid or Avoid Seeing?

 By John O’Callaghan

This is the fourth blog in a new series of posts about the NTSB’s general aviation investigative process. This series, written by NTSB staff, explores how medical, mechanical, and general safety issues are examined in our investigations.

John O’Callaghan at a runway friction test in Conroe, TX.

As a National Resource Specialist for Aircraft Performance, which is government-speak for a technical expert in the aerodynamics and flight mechanics of aircraft, I work to determine and analyze the motion of aircraft and the physical forces that produce that motion. In particular, following an accident or incident, I attempt to define an aircraft’s position and orientation during the relevant portion of the flight, and determine its response to control inputs, external disturbances, ground forces, and other factors that could affect its trajectory.

I recently reviewed a 2009 cockpit video taken while I was testing a video recording device in a Bellanca Citabria. The footage called to mind recent NTSB cases that highlight the fallacies inherent in one of aviation’s oldest mantras—“see and avoid.”

The video from the camera mounted over my left shoulder reveals a hazy blue sky above and the Potomac River winding lazily below the Citabria’s plexiglass windows. It shows my head dutifully swiveling as I scan the practice area for traffic in preparation for a series of aerobatic maneuvers intended to test a prototype “portable flight data recorder” developed by a friend of mine. I’m flying in the Washington, DC, Special Flight Rules Area so I’m in contact with Potomac Approach, which helpfully keeps a radar’s eye on me and nearby traffic and conveys what I fail to see.

“Citabria 758, traffic about a mile southwest of your position. A Cherokee is in the practice area, altitude indicates . . . I’m not showing an altitude right now.”

On the video, my head moves around a little more as I respond, “758 looking, thank you.”

The controller then alerts the Cherokee. “Cherokee [call sign], traffic seems to be about 1-mile orbiting, altitude indicates 3,600, a Citabria.”

I’m still looking with no success when Potomac advises that the Cherokee is at 2,200 ft. The controller lets me and the Cherokee pilot know that we are getting close to each other.

 “Cherokee [callsign], traffic just southeast of you, about less than 1 mile, Citabria in the practice area, altitude indicates 3,700.”

 “Roger, we’ll keep our eyes open for that Citabria in the practice area.”

“Citabria 758, that traffic is just northwest of you, less than a mile now, and his altitude still indicates 2,300, appears to be eastbound.”

“758 still looking, thank you.”

The video now shows me craning my neck left and right, leaning forward, scanning the entire symmetrical view offered by an airplane with its seats on the centerline. The airplane banks left and right in gentle turns as I maneuver, trying in vain to spot the Cherokee. A little over 3 minutes after Potomac’s initial advisory, I give up.

“Potomac, Citabria 758 still looking for that traffic . . . is he still a factor?”

“758, now he’s about 5 miles north of you, no factor.”

I don’t know if the Cherokee pilot ever saw me, but if he did, he didn’t announce it. I imagine that most general aviation pilots don’t need to accumulate too many hours before they have an experience much like mine, or its more unnerving inverse: suddenly seeing an airplane that you had no clue about whiz by close enough to read the N-number. Both situations point to the inherent limitations of the “see-and-avoid” concept: the foundation of collision avoidance in visual meteorological conditions (VMC) under visual flight rules (VFR).

My flight was a personal one, unrelated to my duties as an aircraft performance engineer at the NTSB. However, my fruitless search for the Cherokee was consistent with conclusions the NTSB has drawn from investigating a number of midair collisions, and which call to mind what can happen when traffic remains unnoticed.

As detailed in the NTSB reports concerning two midair collisions that occurred in 2015, described further below, the see-and-avoid concept relies on a pilot to look through the cockpit windows, identify other aircraft, decide if any aircraft are collision threats, and, if necessary, take the appropriate action to avert a collision. There are inherent limitations of this concept, including limitations of the human visual and information processing systems, pilot tasks that compete with the requirement to scan for traffic, the limited field of view from the cockpit, and environmental factors that could diminish the visibility of other aircraft.

In a collision between an F-16 and a Cessna 150 near Moncks Corner, South Carolina, in July 2015, the F-16 pilot was unable to spot the C150, even though the Charleston Approach controller had alerted him to the presence of the airplane. The F-16, call sign “Death41,” was flying under instrument flight rules and communicating with air traffic control (ATC); the C150 was flying under VFR and not communicating with ATC.

“Death41, traffic 12 o’clock 2 miles opposite direction 1200 indicated type unknown.”

“Death41 looking.”

“41 turn left heading 180 if you don’t have that traffic in sight.”

“Confirm 2 miles?”

“Death41, if you don’t have that traffic in sight turn left heading 180 immediately.”

[unintelligible reply]

Even before the controller finished her last instruction, the F-16 had begun a standard-rate turn to the left. The F-16 was heavy and, at 240 knots, moving relatively slowly—for a fighter jet. Contrary to what one might think, it could not turn much faster in those conditions. Twenty-three seconds after the controller’s last instruction, the F-16 and the C150 collided at about 1,470 ft above the Cooper River. The crippled F-16 flew for another 2.5 minutes before the pilot ejected safely, and the jet subsequently crashed. The C150 crashed almost directly beneath the collision site, and both the pilot and his passenger died.

We determined the probable cause of this accident was the approach controller’s failure to provide an appropriate resolution to the conflict between the F-16 and the Cessna. Contributing to the accident were the inherent limitations of the see-and-avoid concept, resulting in both pilots’ inability to take evasive action in time to avert the collision.

Midair collisions can happen even when both aircraft are in communication with ATC. A month after the Moncks Corner midair collision, a North American Rockwell Sabreliner collided with a Cessna 172 in the busy traffic pattern at Brown Field in San Diego. Both aircraft were under Brown Tower’s control and on a right downwind for runway 26R, with the Sabreliner outside of and overtaking the C172. The tower controller intended to instruct the C172 to perform a right, 360-degree turn to position him behind the Sabreliner; however, he mistakenly instructed a different C172 to perform the maneuver, and immediately after instructed the Sabreliner to turn right base.

The Sabreliner and C172 subsequently collided, and all five people on the two aircraft died. The cockpit voice recorder on the Sabreliner indicated that both Sabreliner pilots were aware of and concerned about the busy traffic pattern, pointing out other aircraft to each other. One of the nonflying crew in the back of the plane is even heard asking, “see him right there?” presumably referring to traffic. Yet the collision still occurred.

We determined the probable cause of the accident was the local controller’s failure to properly identify the aircraft in the pattern and to ensure control instructions provided to the intended Cessna on downwind were being performed before turning [the Sabreliner] into its path for landing. Contributing to the accident were the inherent limitations of the see-and-avoid concept, resulting in the inability of the pilots involved to take evasive action in time to avert the collision.

My role in the investigations of the Moncks Corner and San Diego collisions was to reconstruct the motion of the airplanes based on radar data and other information, and to evaluate the resulting visibility of each aircraft from the cockpit of the other. In addition, it was my job to evaluate how new collision avoidance technology—such as cockpit displays that provide a radar‑like view of surrounding traffic based on automatic dependent surveillance-broadcast (ADS-B) information—could have averted each accident.

One objective of these visibility studies was to determine whether either of the airplanes involved in the collision were obstructed from the other pilot’s field of view by cockpit structures, or whether the pilots had an unobstructed view of each other but simply failed to see one another (because seeing other traffic from the cockpit is hard!). To find out, we measured the geometries of the window and other structures of exemplar airplanes with laser-scanning equipment, and the resulting measurements were used to determine where the windows were in each pilot’s field of view and whether the other airplane appeared within the windows or not. The results were most intuitively presented by creating computer animations of the collision from the point of view of each pilot using flight simulation software (Microsoft Flight Simulator X) to create the outside scenery and airplane models.

Readers can watch the animations we created for the Moncks Corner and San Diego collisions on our YouTube channel and judge the visibility results for themselves. The performance studies for these accidents provide technical details about the reconstructions, and they note that periods when airplanes are obscured from a pilot’s nominal field of view “underscore the importance of moving one’s head (and occasionally lifting and dipping the wings) so as to see around structural obstacles when searching for traffic.”

Readers can also watch animations of cockpit display of traffic information (CDTI) displays for each of the airplanes involved in these midair collisions. The animations depict the information that these radar-like displays, fed by ADS-B, could have presented to the pilots involved. Had the airplanes been equipped with CDTI, the pilots could have been made aware of the presence and relative locations of the conflicting traffic minutes before the collisions.

In general, the timely and information-rich traffic picture offered by a CDTI can greatly improve a pilot’s ability to detect traffic threats and avoid a collision without aggressive maneuvering. We issued a safety alert, titled, “Prevent Midair Collisions: Don’t Depend on Vision Alone,” to encourage pilots to learn about the benefits of flying an aircraft equipped with technologies that aid in collision avoidance.

Much of flying is an exercise in mitigating or engineering out risk. Pilots are trained, examined, and reviewed; aircraft are certified and maintained; checklists are used; flights are planned; weather is studied. Great effort is made to leave little to chance. However, when it comes to collision avoidance in VMC, we wink at risk management (“see-and-avoid!” “Keep your head on a swivel!”), when the reality is that we rely in great measure on luck. It’s a big sky, and it would be hard to hit somebody if you tried. The odds are against a collision, but on occasion, disaster strikes.

Technologies such as CDTI provide rational risk reduction for the VMC collision avoidance problem. Guardian angels will never lack for work, but tools such as CDTI can help us to make their jobs a little easier.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s