N4267RPiper PA-32-300 · 1969

On November 19, 2012, about 0618 local time (2018 November 18 Universal Coordinated Time), a Piper PA-32-300, N4267R, was destroyed when it impacted airport terrain during an attempted turnback immediately after takeoff from Francisco C. Ada/Saipan International Airport (PGSN), Obyan, Saipan, Northern Mariana Islands, a United States territory. One passenger sustained fatal injuries, the pilot and four passengers sustained serious injuries, and one passenger sustained minor injuries. The on-demand charter flight was operated by Star Marianas Air, Inc. (SMA), under the provisions of Title 14 Code of Federal Regulations Part 135. Dawn visual meteorological conditions prevailed, and no Federal Aviation Administration (FAA) flight plan was filed for the flight.

According to a representative of the airline, the airline typically shuttled tourists between Saipan and Tinian, an island about 10 minutes' flying time south of Saipan. The passengers on the accident flight included five Chinese nationals and one Filipino national, and the flight was destined for Tinian.

According to information from the FAA and Serco, the PGSN air traffic control tower (ATCT) service provider, the pilot first contacted the ATCT at 0604, requesting taxi clearance. The flight was instructed to taxi to the end of runway 7, and was cleared for takeoff about 4 minutes later. About 4 minutes after takeoff, the pilot radioed the ATCT that he wanted to "come back in for an immediate landing if possible." About 3 minutes later, the airplane landed uneventfully on runway 7, and exited the runway at taxiway B. The pilot briefly ran up the engine, and about 3 minutes after landing, informed the ATCT that he was ready for an intersection departure from runway 7. About 45 seconds after that, the airplane was observed turning back to the left, and some garbled radio transmissions were received from the airplane. Shortly thereafter, the airplane impacted airport property near the northeast end of runway 6, a smaller parallel runway situated northwest of runway 7.

The bulk of the airplane came to rest at the treeline northwest of runway 6, and a post-accident fire began. The pilot and most passengers exited or were helped from the airplane, but the female passenger seated in the front right seat remained in the airplane, and was fatally injured.

About 2 days after the accident, the wreckage was examined on scene by personnel from the FAA, and was then recovered to a secure location.

Pilot Experience and Medical Information

FAA and airline information indicated that the pilot held a commercial pilot certificate, with an instrument airplane rating. The pilot began flying for SMA in January 2012. The airline initially qualified the pilot for "PIC" (pilot-in-command) authority in the PA-32-300 on January 11, 2012, and for flight instructor/check airman authority in the airplane on February 3, 2012. His most recent flight review was completed in January 2012, and his most recent FAA first-class medical certificate was issued in August 2012.

The pilot had a total flight experience of about 1,238 hours, including about 674 hours in the accident airplane make and model. Airline records indicated that the pilot had flown about 54 hours in October 2012, and about 31 hours in November, the month of the accident. According to the airline's records, the pilot was on duty from 0200 to 0500 on November 15, and did not have another duty period until 1900 on November 17. He was then on duty for 11 hours, had a rest period of 12 hours and 40 minutes, and then went back on duty at 1840 the evening before the accident. At the time of the accident, the pilot had been on duty for almost 12 hours.

The FAA Civil Aeromedical Institute conducted forensic toxicology examinations on "blood, serum" specimens from the pilot, and reported that no carbon monoxide, ethanol, or any screened drugs were detected.

Pilot Flight Recollections

A brief telephone interview was conducted with the accident pilot about 2 weeks after the accident. The pilot remembered most of the accident flight, including events leading up to it, but he had no recollection of the takeoff and return just prior to the accident takeoff.

According to the pilot, the flight/load manifest was satisfactory, and he did not note any defects during his preflight inspection, taxi-out, or engine runup. The initial takeoff and climbout was normal, but when the airplane reached an altitude of about 400 to 500 feet above ground level (agl), he noticed a "significant change" in engine sound and power. He observed that the manifold pressure and rpm gauges did not indicate full takeoff power, but stated that the engine gauges in the airline's fleet are "never accurate." He did not observe any unusual instrument indications, but his sense was that the engine's power and sound was less than that normally experienced for the indicated power settings.

Upon sensing the power loss, the pilot immediately pushed the throttle, propeller, and mixture controls full forward, but the power continued to decrease. The pilot then turned on the electric fuel boost pump, and switched the fuel selector valve from the left to the right tank.

The pilot stated that because he had conducted an intersection takeoff, he did not believe that there was enough runway remaining to land on, and he was therefore faced with the choice of putting the airplane in the ocean ahead, or attempting a turnback for landing on the airport property. He decided to attempt a turnback, and was cognizant of the decreasing power, airspeed, and altitude, as well as the danger of stalling. Initially the pilot intended to land on runway 24, but he overflew that, and then decided to try to land on the ramp. He said that he maintained a bank angle of about 35 to 45 degrees in the turn, but when the airplane was about 15-20 feet above the ground, the airplane suddenly "dropped to the ground." That was the last item he recalled about the accident sequence.

When asked about the air turnback, landing, and runup that he had conducted just prior to the accident takeoff, the pilot recalled such an event, but he could not recall whether that event was the night before, or 2 weeks prior to, the accident flight. Regarding that event, the pilot recalled that on departing Saipan, he experienced a "slight power loss" of about 100 rpm, but was able to climb to 1,500 feet on the right downwind traffic pattern leg, and therefore, elected to return to the departure runway. He "landed long," but otherwise normally. He exited the runway, leaned the mixture to "clean the plugs," conducted an engine runup, and determined that the airplane was "fine." The pilot then departed uneventfully, and landed successfully at Tinian, where he informed maintenance or company personnel about the engine event.

When informed during a telephone interview with NTSB and FAA personnel that the ATCT records indicated that he had conducted a successful air turnback just prior to the accident, again the pilot had no specific recollection of that event. The pilot did not reconcile his lack of recollection of a previous turnback with his recollection that he had conducted an intersection takeoff during the accident flight.

General Information

The airplane was manufactured in 1969, and was equipped with a Lycoming IO-540 series engine. It was an all-metal low-wing monoplane design with fixed, tricycle-style landing gear. The airplane seated seven persons in a 2-3-2 arrangement from front to back. The cabin was equipped with two entrance doors, one on the front right side (for pilot and co-pilot/front seat passenger), and one at the left rear for the other five passengers. The airplane had two baggage compartments, one forward and one aft.

Flap positions include retracted (up/0 degrees), 10, 25, and 40 degrees, and were selected by moving the flap handle to one of the corresponding pre-select notches. The flaps were spring-loaded to the retracted position.

Fuel System and Usage Procedures

The airplane was equipped with four individually-selectable fuel tanks; left tip/aux, left main, right main, and right tip/aux. Total fuel capacity was 84 gallons. The four-position wing flaps were manually actuated via a cockpit handle and torque tube arrangement.

The manufacturer's Owner's Handbook (OH) specified that for takeoff, the fuel selector valve should be set to the "fullest main tank." The OH specified that once in cruise, "in order to keep the airplane in best lateral trim...the fuel should be used alternately from each tip tank" until the tip tank quantities were "nearly exhausted." At that point the OH specified switching to the main tanks.

Maintenance Information

According to the maintenance records, the airplane had accumulated a total time (TT) in service of about 6,805 hours at the time of the accident. An overhauled engine was installed in July 2012, when the airplane had a TT of about 6,290 hours.

The airline maintained the airplane on a progressive, cyclic inspection program which included four elements or segments, designated as "AAIP-1" through "AAIP-4." The most recent completed inspection cycle was the AAIP-1, which was completed on November 17, 2012, at an airframe TT of about 6,799 hours, and an engine time since overhaul of about 516 hours.

The PGSN 0554 automated weather observation included winds from 060 degrees at 10 knots, visibility 10 miles, scattered clouds at 2,000 feet, broken cloud layer at 5,000 feet, overcast cloud layer at 11,000 feet, temperature 27 degrees C, dew point 23 degrees C, altimeter setting of 29.91 inches of mercury.

PGSN was equipped with an ATCT that was operating at the time of the accident, and was operated and staffed under contract to the FAA by Serco Management Services, Inc. A transcript of the communications between the ATCT and the accident airplane was prepared and provided by Serco.

According to the transcript, the pilot first contacted the ATCT at 0604:32, and advised that he was at the "commuter ramp" with the terminal information, and was ready to taxi. The controller cleared the airplane to taxi to the "end" of runway 7 via taxiway B and runway 6. The pilot acknowledged "full length," and was advised to expect a 2-minute delay for wake turbulence separation. At 0608:27, the flight was cleared for takeoff and an "early right turnout," which the pilot read back correctly. At 0612:08, the pilot radioed "Saipan tower, uh" but did not continue. At 0612:35, the pilot radioed that he wanted to come back for an "immediate landing if possible," and after a brief discussion about his position, was cleared to land on runway 7. When asked if he required emergency assistance, the pilot replied "negative."

After the airplane landed, at 0614:28, the controller instructed the pilot to exit the runway at taxiway B, and remarked that this was the "second time it happened to you in less than two weeks, huh?" to which the pilot responded in the affirmative, and stated that he "just want[ed] to make sure everything's good to go." At 0614:54, the pilot told the controller that he just wanted to do a "quick run-up," and the controller told him to do so on taxiway B, and to advise when ready for departure.

At 0616:17, the pilot advised the controller that he was ready for an intersection departure, and was cleared a few second later. At 0617:17, the pilot radioed "six seven romeo would like" but he did not complete that request. At 0617:27, in response to observing the airplane turning and descending, the controller cleared the airplane to land, and a partial, unintelligible response was received from the airplane. Shortly after that, the controller initiated the accident response procedures.

PGSN was equipped with two parallel runways, designated as 6/24 and 7/25. Runway 7/25, which was the primary runway and the one used for the two departures, measured 8,700 by 200 feet. Taxiway B, which was perpendicular to the two runways, was located about 3,150 feet beyond the threshold of runway 7. A runway 7 intersection departure from taxiway B would have about 5,550 feet of available runway. The specified traffic pattern direction for runway 7 was left-hand.

Runway 6/24 was a designated portion of the pavement that paralleled the full length of runway 7/25. Runway 6/24 measured 7,000 by 100 feet, and the threshold of runway 24 was directly abeam that of runway 25. The centerline of runway 6/24 was offset about 750 feet northwest of that of runway 7/25.

The shoreline to the northeast of the airport was oriented approximately perpendicular to the runways, and was situated approximately 2,600 feet beyond the end of runway 7. The shoreline to the southwest of the airport was also oriented approximately perpendicular to the runways, and was situated approximately 4,300 feet prior to the runway 7 threshold. The airport elevation was listed as 211 feet above mean sea level.

Review of commercially-available aerial imagery revealed that beyond the northeast airport boundaries, most of the vicinity was either ocean or heavily-vegetated terrain, with very few roads or open fields.

According to information provided by FAA inspectors who responded to the accident site, the main wreckage came to rest at the junction of the grassy area adjoining runway 6 and the parallel treeline situated about 290 feet northwest of the centerline or runway 6/24. The main wreckage was located about 1,050 feet from the runway 7/25 centerline, and about 700 feet before the departure ends of runways 6 and 7.

The main wreckage consisted of the majority of the airplane, excluding the left wing, some cockpit/cabin transparencies, some miscellaneous airplane components, and some luggage. The evidence was consistent with the left wing striking the ground first, while the airplane was in a left turn. The ground scars and debris path were oriented to the northwest, and curved about 15 degrees to the left.

The first ground scar was located in the grass about 35 feet from the northwest side of runway 6. The left wing was located about 108 feet beyond the first impact point; the wing was fracture-separated from the airplane at the root, and was lying inverted. The left main landing gear was fracture-separated from the wing; it was found about 10 feet beyond the left wing. The main wreckage was about 114 feet beyond the left wing, and a passenger suitcase was located between the left wing and main wreckage. The airplane came to rest upright against and parallel to the tree line, oriented on a heading of approximately 240 degrees. The cockpit and cabin contents were fire damaged or consumed by fire, as was the outboard half of the right wing.

The aft fuselage and empennage were essentially intact. Pitch and rudder control continuity was established from the cockpit area to the respective aft control surfaces. The pitch trim tab was set so that its trailing edge was about 1/2 inch below the stabilator trailing edge. The right wing flap was found in the retracted position, but the left flap was fracture-separated from its actuation linkage and therefore not fixed in any position. Photographs taken several hours after the accident, and subsequent to recovery of the front-right seat passenger, depict the cockpit flap handle to appear to be in the flaps-retracted position. The investigation was unable to determine if the flap handle was disturbed during or subsequent to the accident.

The ignition key remained in the ignition and its orientation appeared consistent with it being in the "BOTH" position, but fire damage precluded positive determination. The fuel selector handle appeared to be set to a right side tank, but visual inspection did not permit determination of whether it was set to right main tank, or the right tip tank. The engine control quadrant was fire- and impact-damaged, but the three levers appeared to be near their full-forward positions.

The engine was displaced aft and nose-down, but the engine was relatively undamaged. There was no evidence of any catastrophic internal engine failures. The three propeller blades remained attached to the hub, which remained attached to the engine. Two blades were bent aft, and one exhibited S-bending and chordwise scoring. The propeller did not exhibit any evidence of any pre-impact mechanical failures.

The left tip tank was compromised, and contained no fuel, but some fuel was recovered from the left inboard tank. That fuel was not visually contaminated. The right tip tank was impact and fire damaged, and contained no fuel. The right inboard tank was impact damaged. Attempts to sump it for fuel resulted in only a few drops, possibly due to the post-accident orientation.

The FAA CAMI conducted forensic toxicology examinations of cavity blood and multiple tissue specimens from the right front seat passenger, and reported that no cyanide, ethanol, or any screened drugs were detected. However, the CAMI reported that a carbon monoxide level of 25 percent was detected in the passenger's femoral blood.

Engine Examination and Test Run

The engine was removed, crated and shipped to the Lycoming Engines facility in Pennsylvania for examination and testing. The engine did not exhibit any significant thermal damage. The propeller blades remained attached to the hub, and the hub remained attached to the engine. The magnetos were discovered to have been removed and loosely reinstalled. Follow-up investigation revealed that the mechanic who removed and crated the engine had to remove the magnetos in order to separate the engine from the airframe. The magneto-to-engine timing was not determined prior to their initial removal, nor were the magnetos marked in order to enable subsequent determination of their timing relative to the engine.

The airplane fuel selector valve was included in the crate. Examination indicated that it was set to the right outboard fuel tank, and that the valve was unable to be repositioned by hand due to damage.

The cooling baffles were removed to facilitate visual examination of the engine, which indicated that the engine appeared to be in sufficient condition to allow it to be run in a test cell. Several components were either removed outright, or replaced with serviceable or test-cell specific ('slave') hardware to facilitate an engine test run. These included, but were not limited to, the propeller (damage), right magneto (damage), and the air inlet housing for the fuel servo (test cell compatibility). The magnetos were set to the specified magneto-to-engine timing values for the test runs.

After mounting in the test cell and filling with warmed oil, an engine start was attempted. After six attempts the start was suspended, and examination revealed that fuel was exiting the engine driven fuel pump drain port. The pump was bypassed, and fuel was provided via the test cell hardware. After four more attempts, the engine started, but the automated test cell safety system shut the engine down due to high oil pressure, which was determined to be a result of the engine's 18-month dormancy.

The engine was restarted and ran for 5 minutes, but would not idle smoothly. The engine was again shut down, and cylinders 2 and 6 were determined by manual touch to be running significantly cooler than the other four cylinders. The fuel injector nozzles for cylinders 2 and 6 were removed, and found to be about 75 and 100 percent occluded, respectively. The age of, and the reasons for, the occlusions were not determined. The nozzles were cleaned and reinstalled.

The engine was restarted, and successfully completed Lycoming's Production Engine test sequence. The engine developed full rpm and manifold pressure, ran normally with only minor anomalies, and no deficiencies or problems that would have precluded continued operation were noted. Complete details are provided in a separate document in the NTSB public docket for this accident.

The magnetos and the engine driven fuel pump were sent to their respective manufacturers for detailed examination.

Magneto Information

Review of the maintenance records revealed that at the time of the accident, the magnetos had accumulated the following times on the airplane:

Left: 28 hours (installed 11/13/2012)

Right: 409 hours (installed 8/16/2012)

No other documentation regarding the previous history of the magnetos was located, and the investigation did not determine how much time the magnetos had accumulated since their overhaul, or on any other aircraft, if any. Therefore, the times cited herein can only be considered as minimum values.

Both magnetos were shipped to their manufacturer, Continental Motors in Mobile, Alabama, for examination and testing.

The left magneto's internal timing was found to be correct by both using a timing light and by visually examining the distributor gear position through the timing port. The magneto was placed on a test bench, and each lead produced a spark across a 7mm gap throughout all operating speeds. The impulse coupling functioned properly. The magneto was disassembled. Examination of the distributor gear while the axle was installed in the distributor block bushing revealed no discernable radial or axial play. There were no anomalies noted with the left magneto's internal components.

The right magneto's internal timing was checked using a timing light and by visually examining the distributor gear position through the timing port. The internal timing was trailing by three distributor gear teeth. The magneto was placed on a test bench and each lead produced a spark across a 7mm gap throughout all operating speeds, in the correct firing order. According to a Continental Motors (CMI) representative, the bench test spark gap was significantly larger than would be expected to be found in service, and that based upon the bench test results, the magneto in its as-tested condition would not have adversely affected engine performance.

The right magneto was disassembled. Examination of the distributor gear while it was still installed on the distributor block axle/shaft revealed that there was a significant angular displacement between the gear and the block. Four of the six fixed electrodes were mechanically damaged. Some of the material had been shaved off the electrodes, and electrode shavings and chips were found on the inside circumference of the distributor block. The distributor gear teeth displayed a worn condition on the carbon-brush-end of the gear. A number of the distributor gear teeth apexes displayed mechanical deformation damage consistent with imprints of the drive gear teeth. Examination of the distributor block revealed that the bushing in which the distributor gear axle rides, and which is normally molded into the block, was loose. Visible gaps were noted between the bushing and the housing, and fragmented block material was noted under the felt strip and felt washer. Representatives of CMI indicated that they had become aware of other similar magneto block failures. In February 2015, CMI issued two related Service Bulletins (SB 15-1 and SB-669) to provide relevant magneto inspection guidance.

The loose bushing permitted the distributor gear axle to tilt enough to disengage the distributor gear from the drive gear, permitting changes to the internal timing. Although the loose bushing and resulting off-plane freeplay of the timing gear was necessary to enable the mis-timing, it was not sufficient. The investigation did not determine the forcing mechanism(s) for the mis-timing. Therefore, the investigation was unable to determine whether any internal timing change was a gradual migration, a random step function, or some combination of the two. As a result, the investigation was unable to determine whether the as-found internal timing was representative of the magneto internal timing, either during the accident flight or any previous flights. Refer to the magneto examination report in the NTSB accident docket for additional details.

In August 1994, Lycoming issued Mandatory Service Bulletin SB516, which specified the inspection, and replacement at overhaul, of certain magnetos. The left magneto on the accident engine was subject to that SB. According to a Lycoming representative, any Lycoming-accomplished/approved overhaul requires compliance with all Lycoming Mandatory SBs. The accident engine was overhauled in 2012 by a commercial maintenance facility, and the accompanying FAA Form 8130-3 stated that the engine was "overhauled, tested, and certified" in accordance with all Lycoming specifications. However, the airplane's maintenance records indicate that a magneto that was subject to the SB was installed at overhaul, which was not in accordance with Lycoming standards. The investigation was unable to determine the reason for this discrepancy.

According to the FAA inspector for this accident, Part 135 operators are not required to comply with manufacturer's Service Bulletins (SBs) unless the SB is either incorporated into an Airworthiness Directive (AD), or is incorporated into any other operator document that is FAA approved. That same inspector reported that in his review of the airline's relevant guidance, he did not locate any requirements for compliance with manufacturer's SBs (excluding manufacturer's recommendations for overhaul of airframes, engines, propellers, and accessories).

Engine-Driven Fuel Pump Information

Review of the maintenance records and other documentation revealed that the pump was manufactured prior to 1992, and that it was installed on the engine on October 8, 2012. At the time of the accident, the pump had accumulated about 345 hours in service on the airplane. No other documentation regarding the history of the pump was located, and the investigation did not determine how much time the pump had accumulated since its overhaul, or on any other aircraft, if any. Therefore, the time cited herein can only be considered as a minimum value.

The fuel pump data plate indicated that it was manufactured by Romec. The pump was shipped to a Crane Aerospace facility in Ohio, of which Romec had become a subsidiary. Examination of the pump revealed that it had been overhauled by a company other than Crane/Romec, and that the overhaul had been accomplished on an unknown date subsequent to March 2010.

The pump was subjected to the Crane acceptance test procedure (ATP) for new pumps. The pump passed the performance portion of the ATP, but failed the static leak portion of the ATP; the pump shaft seal was found to be compromised. There were several anomalies detected with the assembly of the pump, most of which were consistent with being artifacts of the overhaul. These included improperly torqued screws, misidentified components, and lubricant or grease in improper locations. Although some anomalies had the potential to result in pump malfunction or failure, there was no evidence that the detected anomalies were associated with the subject accident.

In addition to those anomalies, areas of wear, and particulate contamination were observed in the pump. The shaft seal had significant rust-colored contaminants, which could have contributed to the shaft seal failure. The possibility that the rust developed or increased during the unpreserved engine's dormant period between the accident and the pump examination could not be excluded. According to a Crane representative, the "shaft seal leak has little impact on pump performance."

The leak was between the interior wetted portion of the pump and the gearbox cavity on the engine, and there was provision for the leaking fluid to be vented overboard or back to the intake. That scenario was consistent with the engine test cell observations. Some additional details can be found in the NTSB accident docket for this accident.

Airline Background Information

According to information provided by company representatives, SMA was an FAR Part 135 operator based on Tinian. SMA operated primarily as a passenger shuttle between Tinian and Saipan, which was approximately a 10-minute flight. The Tinian offices included SMA's headquarters, maintenance and dispatch facilities and personnel. SMA maintained an outstation on Saipan, which included passenger facilities, station agents, as well as technicians and pilots on an as-needed basis.

SMA operated about 4,500 flights per month, which representatives variously characterized as about 800 hours per month, and as about 125 hours per airplane per month. SMA had completed about 100,000 flights prior to the accident, and SMA representatives reported that SMA had had only two "incidents" (including the subject accident).

At the time of the accident, SMA operated a fleet of seven PA-32-300, two C-172, one C-172RG, and one Piper Navajo airplanes. Not all the airplanes were on the Part 135 certificate, and the Cessnas were used to transport company personnel and goods between Tinian and Saipan

SMA had approximately 22 pilots, each of whom was hired on a 1-year contractual basis. Pilots were typically scheduled for 12-hour shifts that began and ended at 7 am and 7 pm, and the normal schedule cycle was a repeat of 3 days on, 1 day off, 3 days on, and 2 days off. Pilots were released from schedules when traffic loads did not warrant their use/retention.

SMA conducted its own maintenance on its airplanes. SMA had a total of 12 aircraft technicians, including 5 certificated mechanics with airframe and powerplant ratings, and 1 with inspection authorization. The airplanes on the Part 135 certificate, including the accident airplane, were maintained in accordance with an FAA-approved inspection program.

Airline Dispatch, Release, and Loading Procedures

FAR Part 135 did not require SMA to maintain or use FAA-defined dispatch personnel or procedures. According to SMA representatives, SMA did maintain and use personnel in a dispatch-like capacity, but those personnel did not hold FAA dispatch certificates. SMA variously referred to those personnel as "dispatchers" and "flight followers." Those personnel performed weather briefings, airplane releases, and exercised "operational control."

Normal SMA dispatch procedures included assigning pilots to specific airplanes and providing printed weather information to the pilot. The dispatcher would then issue the pilot the airplane "can," which was a metal clipboard and container that held the aircraft flight log (AFL), other related paperwork, and the airplane keys.

SMA station agents, who were direct employees of SMA, would conduct passenger check in and loading. Their responsibilities included weighing of passengers and bags, seat assignments, baggage compartment loading determinations, issuance of boarding passes, and supervision or conduct of baggage and passenger loading. Station agents also presented load manifest to pilots, who verified and signed.

Airline Fuel Procedures

SMA maintained its primary fuel supply on Saipan, with some on Tinian. The "standard" first flight of each day fuel loads were 17 gallons (g) in each tip tank, and 7g in each main (inboard) tank. Although SMA conducted late night and very early morning flights, they did not indicate what their demarcation point for the beginning of each "day" was. Each flight between Tinian and Saipan typically consumed about 3 gallons. After the third or fourth "rotation" (a round trip between the islands), if the airplane was scheduled for additional flights, the dispatcher would advise the fuel truck of the potential need for fuel. Dispatchers arranged for fueling, and fuel truck personnel fueled the airplanes.

The airline's flight manifest records indicated that the airplane had 48 gallons of fuel on board for the flight. The airplane had most recently been fueled with 20.7 gallons on the day of the accident. The provided records did not indicate whether, or how many times, the airplane had flown subsequent to the refueling, or permit an independent determination of the fuel quantity on board for the accident flight. There were no reports of any fuel-related problems with any of the airline's airplanes in the days surrounding the accident.

According to SMA personnel and the General Operations Manual (GOM) checklist, the PA-32 pilots flying between Saipan and Tinian were supposed to adhere to standardized fuel-tank usage procedures, which, per the GOM, required the cockpit fuel tank selector handle to point "East for Enroute," which was towards Hawaii. Tinian was situated south of Saipan, and Hawaii was to the east of both. Therefore, on a southbound flight to Tinian, the selector would be set to the left tank, and on a northbound flight to Saipan, the selector would be set to the right tank.

Airline Airplane Airworthiness

Pilots were responsible for conducting daily 'first flight' checks, pre-flight inspections, and engine run ups. Once the pilot determined that an airplane was in satisfactory condition for flight, s/he would taxi it to the passenger terminal area to receive passengers.

Airplane airworthiness status was indicated in an electronic calendar, as well as on a manual display board in the Tinian dispatch office. SMA procedures prohibited the release of an "unairworthy" airplane to a pilot, and pilots are prohibited from flying an airplane with an "open" (uncorrected) maintenance item, unless it is indicated as acceptable per the applicable Minimum Equipment List (MEL). Pilots communicated airplane anomalies to the maintenance group by physically logging the item on the airplane copy of the AFL.

Airline Maintenance Writeup Procedures

In response to questions from the NTSB, SMA representatives provided the following information regarding maintenance writeup procedures.

Prior to a flight, station ground staff observe the general condition of the airplane, and notify the pilot if they have any concerns. The pilot checks the item in question, to determine whether a maintenance discrepancy exists. If the pilot determines that no maintenance discrepancy exists, s/he accepts the airplane, and conducts the flight. If a maintenance discrepancy does exist, the discrepancy is logged in the AFL, and the airplane is removed from service until the discrepancy is cleared by maintenance.

If a pilot notes a problem after taxiing out or during flight, the pilot decides whether to continue the flight or return. If the airplane returns to the departure airport, but does not return to the station (gate) because the pilot determines that no maintenance discrepancy exists and the situation can be corrected by the pilot, the pilot corrects the item, and departs without notice to dispatch or the station. Examples of such situations included unclosed doors, improperly inserted microphone or headset jacks, and insecure seat belts. If the pilot returns to the departure station (gate), dispatch is notified. If the airplane requires maintenance, dispatch is notified, and coordination with maintenance is required to determine when the airplane can be further utilized. Similarly, if the pilot continues the flight to the destination airport and then reports a mechanical irregularity, dispatch is notified, and coordination with maintenance is required to determine when the airplane can be further utilized.

Maintenance discrepancies are to be entered on the AFL sheets either before or after the discussions with maintenance. Except for MEL items, all maintenance discrepancies must be cleared by maintenance personnel before the airplane can be released back into service. MEL items may be deferred by the pilot in accordance with the MEL. Every MEL item must be carried forward onto each subsequent AFL until the item is cleared by maintenance.

SMA personnel stated that occasionally, "the items of interest that delay or cancel an aircraft from taking a specific flight are not determined to be" maintenance discrepancies that necessitate an AFL entry, citing a flooded engine as an example. They also noted that items detected and corrected by maintenance personnel would not appear on the AFL sheets. Instead, those items would be captured by the maintenance work order system, and appear as "Unscheduled Maintenance"' in the documentation.

Airline Flight Monitoring Procedures

SMA station agents were required to observe and monitor departures and arrivals. Saipan station agents had the capability to monitor PGSN ATCT communications. The station agents were required to then radio each flight departure and arrival to the dispatchers, who would log those departure and arrival times. SMA did not have any specific procedures requiring the capture of information regarding air turnbacks and subsequent departures, such as occurred on the accident flight. Neither the station agent nor the dispatchers noted the first turnback of the accident flight in any official records. The investigation did not locate any evidence to indicate that either the station agent or the dispatchers were aware of the first turnback until after the accident occurred.

Airline GOM Takeoff and Landing Information

The GOM explicitly stated that pilots were prohibited from substituting "personal" procedures for GOM procedures.

The Normal Procedures checklist from the SMA GOM specified a takeoff flap setting of 10 degrees. Although that checklist did not specify a landing flap setting, the GOM did contain guidance that indicated that landings were to be conducted using full flaps.

The GOM contained some guidance regarding pilots' use of full runway length versus intersection takeoffs from Saipan (PGSN). No similar information for departures from Tinian was provided, because those taxiways only accessed the runway at either end. The GOM guidance was consistent with conducting PGSN runway 7 takeoffs from the intersection of taxiway B and runway 7, instead of full length. That intersection was the closest point of runway 7 to the terminal ramp used by SMA. The GOM guidance for departures from PGSN runway 25 stated that pilots could conduct their takeoffs either full-length or from the taxiway D intersection. The terminal was closer to the taxiway D intersection than it was to the runway end.

The GOM runway takeoff guidance was consistent with minimizing taxi distances and times, and also with ensuring that, by the time an airplane reached mid-channel between the two islands, it would have sufficient altitude to be able to glide back to land in the event of an engine power loss.

Previous Airline Turnbacks

The ATCT communications transcript indicated that shortly after the pilot completed the turnback and landing prior to the runup and accident takeoff, the controller radioed that the turnback was the pilot's second one in as many weeks, which the pilot acknowledged. As a result of that exchange, on November 28, 2012, the FAA requested the ATCT records regarding any SMA turnbacks. The ATCT service provider representative responded that their records only dated back through November 12, 2012, and that there was one event in those records. The event occurred on November 13 local date (November 12 UTC), involving SMA PA-32 flight "TN4254R that departed 1842UTC (0442L) and landed 1845 UTC (0445L)." Review of the airline's records revealed that the accident pilot was on duty from 2000 on November 12, 2012, until 0610 on November 13, 2012 (dates and times local), that he was the pilot for a Saipan to Tinian leg in that airplane at the time of the ATC-reported turnback, and that the flight time was about 23 minutes, instead of the usual 10- to 15-minute durations. All this information was consistent with the pilot conducting the air turnback cited in the ATC records.

Despite the pilot's statement that he reported a recent previous turnback event to the airline, no recent anomaly reports or maintenance writeups for any such event associated with the pilot or N4254R were located in the airline's records. The only relevant maintenance record located was for a different airplane and different pilot. According to airline's records, on July 16, 2012, another pilot in another of the airline's PA-32 airplanes (N4599X) made an "immediate return" to the airport when the engine began "missing." That landing was uneventful, and the airplane returned to the gate. That event was subsequently attributed by the airline to water contamination of the fuel.

Airline Turnback Training Information

The airline's flight training manual that was available and in effect at the time of the accident provided the following guidance regarding engine failures on takeoff: "Airborne but not at a safe altitude to maneuver: Land straight ahead on the remaining runway if able, otherwise the best option is to land straight ahead considering the terrain." The flight training manual "Acceptable Performance Guidelines" for engine failure during takeoff stated that in the case of "no runway remaining," the pilot should "maintain straight flight path...and make the best choice of where to land."

According to an airline representative, the airline's PA-32 emergency checklist contained the following guidance regarding an engine power loss during takeoff: "If area ahead is rough... make only a shallow turn if necessary to avoid obstructions." However, review of the emergency procedures checklist revealed that the checklist that was available and in effect at the time of the accident did not contain the cited guidance. A revised version of that checklist, published in December 2014, did contain the cited verbiage.

Airline-Suggested Accident Scenario

A representative of the airline reported that the upper latches of the forward right cabin doors on their PA-32 airplanes have previously become unsecured after takeoff. The representative suggested that the first turnback was to enable the pilot to secure the door, and that the engine runup was intended to conceal the actual reason for the turnback. The representative suggested that the left turn and accident after the second takeoff was possibly due to the door becoming unlatched again, and the pilot's subsequent distraction and loss of situational awareness during his efforts to re-secure the door during the initial climb. Impact and thermal damage precluded a determination of the status (open/closed/locked etc.) of the door at the time of the accident, and the pilot never reported any cabin door problems to investigators. Review of airplane performance data and accident site ground scars indicated that a sustained bank angle of approximately 55 degrees, initiated at an altitude of about 150 to 200 feet agl, would be required to yield the observed ground scar location and heading.

Passenger Accounts

The passengers consisted of five Chinese females and one Filipino male. The one female who reportedly spoke some English did not survive the accident. The male spoke some English. Three female passengers were interviewed shortly after the accident by U.S. Federal Bureau of Investigation (FBI) personnel, and with the knowledge of the NTSB. Two of those passengers were interviewed twice by the FBI; a total of five FBI interviews with the three female passengers were conducted. Those passengers agreed that the pilot did not inform them of the reasons for either the first or second turnback. Those passengers' accounts of the maximum altitude reached after the second takeoff ranged between 20 and 100 meters above ground level. None reported any in-flight smoke or fire. In one of the second interviews, one passenger reported that the pilot used his mobile telephone during boarding and after the first takeoff.

A few days after the accident, at the request of the airline's chairman, an SMA pilot attempted to interview all the surviving passengers. Two female passengers refused to be interviewed, because they claimed that they were instructed by the FBI not to speak with anyone about the accident. The other two females and the male did speak to the pilot-interviewer. The NTSB was not made aware of this plan or those interviews until after they had occurred, and no FAA representative was present for the interviews. According to the pilot-interviewer's written report, the two female passengers reported that the accident pilot appeared to manipulate a cabin door after the first takeoff.

However, the FAA representative's notes about the SMA pilot-interviewer's recount of the interviews indicated that one of those female passengers "really couldn't tell" whether the accident pilot touched the door, and that the other female passenger did not recall the pilot adjusting or latching the door after the landing. In addition, whereas the pilot-interviewer's notes did not contain any information regarding his interview of the male passenger, the FAA representative's notes indicated that the pilot-interviewer did interview the male passenger, and that the male passenger did not notice the accident pilot either using his telephone or manipulating the cabin door.

The investigation could not determine why the door recollections were not provided in the FBI interviews, whether the two female passengers had communicated with one another between the accident and the airline-led interviews, or why there were so many discrepancies between the various interview results and summaries.

Other Eyewitness Account

An SMA pilot was on the airline's ramp awaiting his passengers and witnessed a portion of the accident flight. He observed the stationary airplane on taxiway B with "what appeared to be a high engine rpm" for a period of about 45 seconds. He stated that the airplane then reduced power and taxied to runway 7 for an intersection departure. The airplane began a takeoff roll and began climbing, but "A few seconds later, it sounded that the engine was losing power and rpm was fluctuating." The witness observed the airplane level off and begin a "shallow left turn, which quickly steepened with an increasing nose down attitude." The airplane descended behind the terrain and trees, and the witness then heard the sound of impact. During the runup, the airplane was situated about 1,200 feet from the witness. The impact location was situated about 4,000 feet from the witness.

Pilot's Telephone Records

The investigation was able to obtain partial message and call records for the two mobile telephones known to be registered to the pilot. Review of the most complete set of records revealed that that telephone was being used by the pilot's girlfriend on the morning of the accident, and that she was using that telephone to communicate with the pilot at the second telephone number.

The records indicated that 26 text messages were exchanged between the two telephones between 0421:27 and 0600:15; the last message was from the pilot to the girlfriend. Three more messages were sent by the girlfriend between 0652:20 and 0655:28, but no replies from the telephone being used by the pilot were received.

A date/time error in a subpoena, that was not detected until very late in the investigation, precluded a complete record of the activity on the telephone which was in use by the pilot on the morning of the accident.

Airplane Performance Information

According to the airline-provided load manifest, the airplane weight at the time of the accident was about 3,270 pounds, and both the weight and center of gravity were within the allowable limits. The investigation was unable to independently determine the actual flap setting(s) for any portion of the flight.

Although all OH data was presented for the zero-wind case, the actual takeoff would have had an approximate 10 knot headwind. Local temperature, humidity and elevation conditions at the time of the accident resulted in a density altitude of about 2,000 feet.

The manufacturer's OH provided some takeoff, climb, glide, and landing performance data. Two takeoff performance charts, one each for 10 and 25 degrees flaps, presented data for takeoff weights of 2,900 and 3,400 pounds. According to the charts, at the takeoff weight, the distances to clear a 50 foot obstacle would have been about 1,900 and 1,800 feet for 10 and 25 degree flaps, respectively.

Climb data, provided only for 10 degrees flaps, indicated that the initial climb rate would have been about 930 feet per minute (fpm), decreasing to about 870 fpm at 1,000 feet agl. The investigation was unable to determine the maximum altitude reached by the airplane, but the pilot stated that the engine lost power when the airplane was 400 to 500 feet above the ground. Glide data indicated that the airplane would have a zero-flaps glide range of about 1 mile from an altitude of 500 feet. That range would decrease with headwind, off-nominal speed, extended flaps, and/or turning flight.

The OH-indicated landing distance over a 50-foot obstacle, using 40 degrees (full) flaps, "maximum braking," and "short field effort" would have been about 1,050 feet.

Calculations using the ambient conditions and the OH-provided data indicated that for the accident takeoff, presuming normal power output, the airplane would have lifted off about 20 seconds after the start of the takeoff roll, and cleared a 50-foot obstacle in another 4 seconds. At the prescribed climb speed, the airplane would have reached 250 feet agl and 500 feet agl after another 16 and 33 seconds respectively, or about 40 and 57 seconds (respectively) after the start of the takeoff roll. At 250 feet agl, the airplane would have traversed about 4,700 feet from the beginning of the takeoff roll. At that point, there would have been about 850 feet of runway remaining ahead of the airplane, compared to about 4,000 feet if the full runway length had been used. At 500 feet agl, the airplane would have traversed about 7,700 feet from the beginning of the takeoff roll. At that point, the airplane would have been about 1,250 feet beyond the end of the runway, compared to having about 1,000 feet remaining if the full runway length had been used.

FAA Airplane Flying Handbook (AFH) Information

Chapter 5 (Takeoffs) of the AFH (FAA publication 8083-3) stated that in the event of an engine failure on initial climbout, the pilot "should establish a controlled glide toward a plausible landing area (preferably straight ahead on the remaining runway)."

AFH Chapter 16 (Emergency Procedures) provided a lengthy discussion of procedures following engine failure on takeoff in a single engine airplane. The AFH noted that the altitude available is "the controlling factor" in the successful accomplishment of an emergency landing, and cautioned that if an engine failure should occur before a safe altitude is attained, it is "usually inadvisable to attempt to turn back" to the departure airport.

Problems associated with a post-takeoff turnback included a downwind turn, increased groundspeed, increased altitude loss due to the turn, and lateral offset from the runway after the turn. The AFH advised pilots to experiment at safe altitudes to determine realistic altitude values that would enable a safe turnback to the airport. Although such altitudes were not provided by SMA or the airplane manufacturer, multiple studies indicate that 500 feet above ground level should be considered the minimum acceptable altitude for single engine general aviation airplanes attempting a turnback to the departure airport.