September 2006 Volume 49 / Number 9
Landmark Accidents: A Fix Too Far
Too much information confuses a flight crew
Knowing where you are when operating in instrument conditions close to the ground, such as during an approach, is of the highest priority. Out of tens of thousands of IFR approaches flown in real weather each year, only a few end in disaster. But statistics are scant comfort for those involved in such a mishap. The two questions every IFR pilot must be able to answer immediately are "Where am I?" and "What is the appropriate altitude for this segment of flight?"
Professional pilots don't make many mistakes, and the turbine equipment that is usually flown is equipped with excellent warning systems — usually. The avionics can help, or hinder if they are mis-programmed or misinterpreted. The potential for enhanced situational awareness provided by GPS also must be tempered with the knowledge that with more data there remains the opportunity for a catastrophic mistake.
On the approach
Take the case of a Beech King Air 200 flying into Blue Ridge Airport in Martinsville, Virginia, in October 2004.
A warm front blanketed the area with low overcast clouds, patchy fog, and low visibilities. The flight departed Concord, North Carolina, about 11:56 a.m. Eastern Daylight Time with a crew of two and eight passengers. As the aircraft approached Martinsville, air traffic control advised the crewmembers that they were second in line for the localizer Runway 30 approach and to hold as published on the localizer course at 4,000 feet and to expect a 28-minute delay. The crew requested five-mile legs in the holding pattern, and the controller approved that.
The King Air approached the BALES Locator Outer Marker (LOM) from the south at 4,000 feet, and made a direct entry, turning right for the outbound leg of the holding pattern. At about that time, the pilot of the airplane ahead of the King Air on the approach canceled the IFR clearance and landed at Martinsville.
At 12:24:19 p.m., while the King Air was turning right to the outbound leg, the crewmembers confirmed they were established in the holding pattern. At 12:24:26, the flight was cleared for the localizer Runway 30 approach and requested to advise when inbound on the approach. At this point, it should have been a simple approach: Continue outbound for one minute or go a bit farther — the five miles ATC had approved — descend to 2,600 feet, turn right to intercept the localizer, and descend as published.
The situation was about to unravel. Instead of extending the pattern to lose altitude, the King Air completed a continuous right turn toward the inbound course and crossed BALES LOM at 3,900 feet. At 12:26:53, the crew advised that the aircraft was "established inbound" on the approach. At 12:27:04, the controller apparently issued the approach clearance a second time and the flight was cleared to the common traffic advisory frequency. This second approach clearance may have been issued when the controller observed that the King Air was still 1,300 feet above the crossing altitude for the final approach fix, or perhaps the controller merely had forgotten about the earlier issuance.
At 4 DME, or two nautical miles after crossing the LOM, the flight started to descend and leveled at 2,600 feet as it passed the missed approach point. About one nm past the MAP and over the runway, the King Air descended and leveled off at 1,400 feet, about four nm beyond the airport and five nm beyond the MAP, and maintained that for the next 1 minute and 13 seconds. At 12:32:13, eight nm beyond the airport, the King Air began a straight-ahead climb.
At 12:33:08 p.m., the crew reported, "We're going missed at this time." The controller asked the crewmembers to repeat the radio transmission, which they did, and the controller acknowledged. Thirteen seconds later the controller advised the flight to climb and maintain 4,400 feet. It was too late. At 12:33:24, the radar target was lost as the King Air impacted Bull Mountain in Stuart, Virginia, at an elevation of about 2,400 feet and near the extended centerline of the runway.
IMC but flyable
Witnesses at the airport heard the flight pass overhead but did not see it. Two witnesses, who were three to four miles southeast of Bull Mountain and well past the airport, saw an airplane fly past them at a low altitude. A Virginia state trooper stated that at the time of the accident and throughout the search and recovery efforts, Bull Mountain was completely obscured by clouds and fog.
The official Martinsville weather at 12:20 p.m. showed calm winds, the ceiling at 600 feet, and a visibility of six miles. The captain of the airplane ahead of the King Air on the approach broke out about 150 feet above minimums and reported that the visibility below the clouds was about two miles.
The 51-year-old captain held an airline transport pilot certificate rated for airplane multiengine land with a commercial certificate for single-engine aircraft. He had been flying for the company for about three and a half years, and his training records showed total flight time of 10,733 hours, including 8,600 hours in the Beech 1900, 210 hours in the Beech King Air 200, with 67 hours in the previous 90 days, and 18 hours in the previous 30 days. He had attended regular professional simulator training annually and had completed Beech 1900 refresher training in May 2004.
The copilot, age 31, held a commercial pilot certificate with ratings for single-engine and multiengine aircraft and an instrument rating. She had been employed by the company for nearly three years. She had accumulated nearly 2,100 hours of flight time, with 860 in multiengine airplanes and 121 hours in the Beech King Air 200. She, too, had received annual simulator training in Beech turboprops and had completed her latest training in May 2004.
The 1981 Beech King Air 200 had accumulated 8,140 total flight hours and had been inspected in June 2004, at 8,079 flight hours. The aircraft was equipped with a Bendix/King KLN 90B GPS receiver. The GPS database was not current and therefore was not certified for IFR approaches. (Nor was it required for the localizer Runway 30 approach, assuming the aircraft had an operable automatic direction finder, as required by the approach.) The investigation uncovered no pre-impact malfunctions of the aircraft or installed equipment.
The airplane was not equipped with an enhanced ground proximity warning system (EGPWS). Terrain awareness warning systems are now required on turbine-powered aircraft with six or more passenger seats, but the installation was not mandatory until March 29, 2005. The King Air had been for sale and had been "parked." When it was not purchased, it was returned to service and scheduled for a GPS upgrade and an EGPWS installation in the very near future.
Air traffic control
At the time of the accident, the controller responsible for approach control services was managing Piedmont Triad International Airport in Greensboro and 11 satellite airports, including Martinsville. Radar contact was normally lost when an aircraft on the Martinsville localizer Runway 30 approach reached 2,000 feet, about two nm before the airport.
The controller observed the King Air's radar track descending from 3,600 feet when the airplane was about three nm southeast of the airport. He thought the aircraft was high but could not question the pilots because they had already changed to the common traffic advisory frequency. At 12:30:07 p.m., when the airplane was 2.5 nm northwest of the airport at an altitude of 1,800 feet, the controller's minimum safe altitude warning (MSAW) system generated an alert that lasted for 15 seconds.
MSAW provides visual and aural alerts to controllers if a Mode C-equipped aircraft is, or is projected to be, below an appropriate terrain clearance altitude. A review of the radar data indicated that an MSAW alert had sounded and was visually displayed on the controller's radar display as a flashing "LA" (low altitude) in the data block for the accident airplane. The controller stated that he did not see or hear the MSAW alert.
The last radar return from the airplane was received at 12:30:12. The MSAW predicted the projected path of the airplane for another 10 seconds, and the alert ended at 12:30:22. The airplane continued to fly beneath radar coverage for about 3 minutes before the flight crew declared the missed approach on the approach control frequency.
The NTSB conducted a demonstration flight in a company Beech 1900 that was equipped with similar avionics as the accident King Air to review the operation of the Bendix/King KLN 90B GPS, the GPS annunciation control unit, and other associated instrument displays. On the demonstration airplane, the GPS was installed in the front instrument panel; as a result, the flight crew could easily view the GPS screen. But on the accident airplane the GPS was installed between the cockpit seats so both pilots had to turn their heads 90 degrees toward the center of the airplane and then downward to view the screen. This was not an optimal location and certainly wasn't convenient for checking position. It required a disruption of the instrument scan. The GPS unit displayed the database expiration date and if it was expired, the pilot had to acknowledge the expiration during the unit start-up self-checks.
With the GPS mode selected, when a localizer frequency was tuned on the number-one nav radio, displayed on the captain's (left) side of the cockpit, the auto-select frequency feature automatically replaced the GPS information from the captain's horizontal situation indicator (HSI) with information pertaining to the nav mode.
On the demonstration flight, the Danville VOR, BALES LOM, and Blue Ridge Airport were programmed as waypoints. According to the NTSB report, when approaching the Danville VOR, a flashing waypoint light illuminated on the GPS annunciation control unit on the front instrument panel for about 1.5 nm until the airplane passed the Danville VOR. The GPS auto-sequenced to the BALES LOM, a message light then illuminated on the annunciation control unit, and the message on the GPS screen instructed the pilot to set the course deviation indicator on the course to the BALES LOM. Once the CDI was set, the message light extinguished.
When the airplane was about 1.5 nm from the BALES LOM, the flashing waypoint light illuminated on the GPS annunciator panel, and the waypoint identifier (BALES) flashed on the GPS screen. When the airplane was over the BALES LOM, the automatic direction finder (ADF), which was tuned to BALES, indicated a needle swing on both the captain's and the first officer's radio magnetic indicators. The needle swing confirmed that the airplane was passing directly over the BALES LOM. After the airplane crossed the BALES LOM, the flashing waypoint light went out. The GPS auto-sequenced to the Blue Ridge Airport waypoint, and the HSI showed that the distance to the waypoint was 5.2 nm.
The message light on the annunciation control unit illuminated, and the message reminded the pilot to set the CDI on a course of 305 degrees (the inbound course for the localizer Runway 30 approach). The pilot set the CDI on a course of 305 degrees. After the airplane crossed the BALES LOM, the pilot initiated a right turn and continued that turn to the inbound course.
Once the airplane was established on the inbound course, the captain's HSI showed a 5.1-nm distance readout to the Blue Ridge Airport waypoint, and the first officer's HSI showed a 5.6-nm DME distance readout to the localizer/DME station. After the airplane had overflown the airport, the distance readouts increased on the captain's and the first officer's HSIs. After the airplane had passed the departure end of the runway, the distance readouts between the GPS distance on the first officer's side and the DME distance on the captain's side showed a 0.5-nm difference (which was the same as the difference shown inbound). On the GPS screen, no course line was depicted beyond the airport; only the moving airplane symbol was displayed with a line drawn from the airplane symbol back to the airport.
Interviews with other company pilots who flew an identically equipped King Air 200 revealed that the KLN 90B was used for backup and for position awareness only. Furthermore, the pilots who had flown the accident airplane and other company King Airs to Blue Ridge Airport on the localizer Runway 30 approach had used the GPS and were familiar with the approach and the waypoints. The KLN 90B pilot's guide states, "A good pilot never relies on just one source of navigation for either VFR...or IFR flying. Cross-check your position using VOR/DME, ADF, or other navigational devices...."
According to the NTSB, two distinct differences emerge from the published approach procedure and how the accident pilots flew the approach. First, once the King Air was established on the inbound course, the descents to 2,600 and 1,400 feet appeared to be keyed to the MAP or Blue Ridge Airport rather than to the BALES LOM.
Second, the missed approach was initiated seven to eight nm past the MAP and was not properly executed. Even after the flight had overflown the airport, according to a performance re-creation, had the crew executed a climbing right turn, as required by the procedure instead of climbing straight ahead, the flight would have cleared the terrain.
The King Air was not equipped with a cockpit voice recorder or a flight data recorder, nor was it required to be, which prevented the safety board from determining for certain what occurred in the cockpit. However, the approach profile suggests that the pilots may have been using GPS to navigate to the BALES LOM and then failed to notice that the GPS had auto-sequenced to Blue Ridge Airport after the airplane passed BALES. Did the crew have the localizer tuned, which would have provided DME readout? We'll never know.
Having turned inbound toward BALES so quickly in the holding pattern, the crewmembers may easily have missed the auto sequence to the MAP. They might have mistakenly believed that the distance readout was to BALES LOM when, in fact, the distance readout was to Blue Ridge Airport.
That is consistent with the altitude profiles subsequently flown. Additionally, they might have needed to suspend auto-sequencing before crossing BALES outbound to prevent the unit from switching to Blue Ridge Airport prematurely.
On some older GPS units a course reversal, as required during this approach, requires suspension of waypoint sequencing. This may not have been required on the KLN 90B. Given the somewhat awkward location of the unit to read distance and waypoints, the crew may have relied on the panel-mount waypoint annunciator without verifying which waypoint the aircraft was passing.
The NTSB determined the probable cause to be the crew's failure to properly execute the published instrument approach procedure, including the published missed approach procedure, which resulted in controlled flight into terrain. Contributing to the cause of the accident was the flight crew's failure to use all available navigational aids to confirm and monitor the airplane's position during the approach.
If the approach is predicated on ground-based navaids, they are the primary guidance, and pilots must be sure that the distance readout is from the correct waypoint and that flight altitudes correspond. With so much data in the cockpit it's easy to become confused. In Martinsville, the crew may have been reading from a fix too far.
Bruce Landsberg is executive director of the AOPA Air Safety Foundation.
Links to additional information about this and other Landmark Accidents may be found on AOPA Online.