Texas

On February 23, 2019, at 1239 central standard time, Atlas Air Inc. (Atlas) flight 3591, a Boeing 767-375BCF, N1217A, was destroyed after it rapidly descended from an altitude of about 6,000 ft mean sea level (msl) and crashed into a shallow, muddy marsh area of Trinity Bay, Texas, about 41 miles east-southeast of George Bush Intercontinental/Houston Airport (IAH), Houston, Texas. The captain, first officer (FO), and a nonrevenue pilot riding in the jumpseat died. Atlas operated the airplane as a Title 14 Code of Federal Regulations Part 121 domestic cargo flight for Amazon.com Services LLC, and an instrument flight rules flight plan was filed. The flight departed from Miami International Airport (MIA), Miami, Florida, about 1033 (1133 eastern standard time) and was destined for IAH. The accident flight’s departure from MIA, en route cruise, and initial descent toward IAH were uneventful. As the flight descended toward the airport, the flight crew extended the speedbrakes, lowered the slats, and began setting up the flight management computer for the approach. The FO was the pilot flying, the captain was the pilot monitoring, and the autopilot and autothrottle were engaged and remained engaged for the remainder of the flight. Analysis of the available weather information determined that, about 1238:25, the airplane was beginning to penetrate the leading edge of a cold front, within which associated windshear and instrument meteorological conditions (as the flight continued) were likely. Flight data recorder data indicated that, during the time, aircraft load factors consistent with the airplane encountering light turbulence were recorded and, at 1238:31, the airplane’s go-around mode was activated. At the time, the accident flight was about 40 miles from IAH and descending through about 6,300 ft msl toward the target altitude of 3,000 ft msl. This location and phase of flight were inconsistent with any scenario in which a pilot would intentionally select go-around mode, and neither pilot made a go-around callout to indicate intentional activation. Within seconds of go-around mode activation, manual elevator control inputs overrode the autopilot and eventually forced the airplane into a steep dive from which the crew did not recover. Only 32 seconds elapsed between the go-around mode activation and the airplane’s ground impact.

The pilot was conducting a personal cross-country flight with one passenger in his twin-engine airplane. There was no record that the pilot received a weather briefing before the accident flight. While en route to the destination, the pilot was in contact with air traffic control and received visual flight rules flightfollowing services. About 18 miles from the destination airport, the radar service was terminated, as is typical in this geographic region due to insufficient radio and radar coverage below 7,000 ft. The airplane was heading northeast at 4,900 ft mean sea level (msl) (about 2,200 ft above ground level [agl]). About 4 minutes later, radar coverage resumed, and the airplane was 6 miles west of the airport at 4,100 ft msl (1,400 ft agl) and climbing to the north. The airplane climbed through 6,000 ft msl (3,300 ft agl), then began a shallow left turn and climbed to 6,600 ft msl (3,800 ft agl), then began to descend while continuing the shallow left turn ; the last radar data point showed the airplane was about 20 nm northwest of the airport, 5,100 ft msl (2,350 ft agl) on a southwest heading. The final recorded data was about 13 miles northwest of the accident site. A witness near the destination airport heard the pilot on the radio. He reported that the pilot asked about the cloud height and the witness responded that the clouds were 800 to 1,000 ft agl. In his final radio call, the pilot told the witness, "Ok, see you in a little bit." The witness did not see the airplane in the air. The airplane impacted terrain in a slightly nose-low and wings-level attitude with no evidence of forward movement, and a postimpact fire destroyed a majority of the wreckage. The damage to the airplane was consistent with a relatively flat spin to the left at the time of impact. A postaccident examination did not reveal any preimpact mechanical malfunctions or anomalies that would have precluded normal operation. A detailed examination of the cockpit instruments and other portions of the wreckage was not possible due to the fire damage. A cold front had advanced from the northeast and instrument meteorological conditions prevailed across the region surrounding the accident site and the destination airport; the cloud ceilings were 400 ft to 900 ft above ground level. The airplane likely experienced wind shear below 3,000 ft, and there was likely icing in the clouds. While moderate icing conditions were forecast for the accident site, about the time of the accident, investigators were unable to determine the amount and severity of icing the flight may have experienced. The weather conditions had deteriorated over the previous 1 to 2 hours. The conditions at the destination airport had been clear about 2 hours before accident, and visual flight rules conditions about 1 hour before accident, when the pilot departed. Based on the available evidence it is likely that the pilot was unable to maintain control of the airplane, which resulted in an aerodynamic stall and spin into terrain.

The pilot, copilot, flight attendant, and six passengers departed on a corporate flight to a private airstrip. After leveling off at flight level 280, the flight crew checked the weather conditions at nearby airports. Based on the weather information that they had, the pilot planned for a visual approach to the runway. As the airplane neared the destination, the pilot flew over the runway and entered a left downwind visual traffic pattern to check if any animals were on the runway and what the windsock on the airstrip indicated. The pilot stated that they did not see the windsock as they passed over the runway. The pilot reported that there were turbulence and wind gusts from the hills below and to the west. When the airplane was over the runway about 50 ft above ground level (agl), the pilot reduced the engine power to idle. The pilot reported the airplane then encountered wind shear; the airspeed dropped rapidly, and the airplane was "forced down" to the runway. A representative at the airstrip reported that the airplane hit hard on landing. The pilot unlocked the thrust reversers, applied brakes, and reached to deploy the ground spoilers. As he deployed the thrust reversers, the pilot said it felt like the right landing gear collapsed. He applied full left rudder and aileron, but the airplane continued to veer to the right. The pilot tried using the tiller to steer to the left but got no response. The airplane left the side of the runway and went into the grass, which resulted in substantial damage; the right main landing gear was broken aft and collapsed under the right wing. Postaccident examinations of the airplane revealed no preimpact mechanical malfunctions or failures that would have precluded normal operation. A review of weather conditions showed surface winds out of the north to northwest at 15 kts, with some gusts up to 20 kts. There was potential for turbulence and wind shear below 5,000 ft, but there were no direct observations. The area forecast about 30 minutes after the accident called for northwesterly winds at 10 to 17 kts with a few higher gusts in the afternoon for the general area. Data from an onboard enhanced ground proximity warning system (EGPWS) revealed that the crew received a terrain alert just before the airplane crossed the runway threshold. At the time the airplane was over the runway threshold, it was 48 ft agl and in a 1,391 ft per minute rate of descent. The airplane impacted the runway 3 seconds later. Given the pilot's account, the weather information for the area, and the data from the airplane's EGPWS, it is likely that the airplane encountered wind shear while transitioning from approach to landing.

According to the copilot, before takeoff, he and the pilot had briefed that the copilot would conduct the takeoff for the planned cross-country flight and be the pilot flying and that the pilot would be the pilot monitoring. The accident flight was the copilot's first takeoff in the accident airplane with it at or near its maximum gross weight. The pilot reported that he taxied the airplane onto the runway and locked the tailwheel in place and that the copilot then took over the controls. About 13 seconds after the start of the takeoff roll, the airplane veered slightly right, and the copilot counteracted with left rudder input. The airplane then swerved left, and shortly after the pilot took control of the airplane. The airplane briefly became airborne; the pilot stated that he knew the airplane was slow as he tried to ease it back over to the runway and set it back down. Subsequently, he felt the shudder “of a stall,” and the airplane rolled left and impacted the ground, the right main landing gear collapsed, and the left wing struck the ground. After the airplane came to a stop, a postimpact fire ensued. All the airplane occupants egressed through the aft left door. Postaccident examination of the airplane revealed no evidence of any mechanical malfunctions or failures with the flight controls or tailwheel. Both outboard portions of the of the aluminum shear pin within the tailwheel strut assembly were sheared off, consistent with side load forces on the tailwheel during the impact sequence. The copilot obtained his pilot-in-command type rating and his checkout for the accident airplane about 2 months and 2 weeks before the accident, respectively. The copilot had conducted two flights in the accident airplane with a unit instructor before the accident. The instructor reported that, during these flights, he noted that the copilot had directional control issues; made "lazy inputs, similar to those for small airplanes"; tended to go to the right first; and seemed to overcorrect to the left by leaving control inputs in for too long. He added that, after the checkout was completed, the copilot could take off and land without assistance; however, he had some concern about the his reaction time to a divergence of heading on the ground. Given the evidence, it is likely the copilot failed to maintain directional control during the initial takeoff roll. It is also likely that, if the pilot, who had more experience in the airplane, had monitored the copilot's takeoff more closely and taken remedial action sooner, he may have been able to correct the loss of directional control before the airplane became briefly airborne and subsequently experienced an aerodynamic stall.

The commercial pilot and passenger, who held a student certificate, departed runway 18R for a local flight in a multi-engine airplane. The pilot held a flight instructor certificate for single-engine airplane. Just after takeoff, the tower controller reported to the pilot that smoke was coming from the left side of the airplane. The pilot acknowledged, stating that they were going to "fix it," and then entered a left downwind for runway 18R, adding that they didn't need any assistance. The controller subsequently cleared the airplane to land on runway 18L, which the pilot acknowledged. Two witnesses reported seeing the smoke come from the left engine. Still images taken from airport security video show the airplane before making the turn to land with white smoke trailing and the landing gear down. The airplane was then seen in a steep left turn to final approach exceeding 90° of bank, before it impacted terrain, just short of the runway in a near vertical attitude. A postcrash fire ensued. The examination of the wreckage found that the left engine's propeller was not being driven by the engine at the time of impact. The left propeller was not in the feathered position and the landing gear was found extended. The damage to the right engine propeller blades was consistent with the engine operating at high power at impact. The examination of the airframe and engines revealed no evidence of preimpact anomalies; however, the examinations were limited by impact and fire damage which precluded examination of the hoses and lines associated with the engines. The white smoke observed from the left side of the airplane was likely the result of an oil leak which allowed oil to reach the hot exterior surfaces of the engine; however, this could not be verified due to damage to the engine. There was no evidence of oil starvation for either engine. Both the extended landing gear and non-feathered left propeller would have increased the drag on the airplane. Because the pilot's operating procedures for an engine failure in a climb call for feathering the affected engine and raising the landing gear until certain of making the field, it is unlikely the pilot followed the applicable checklists in response to the situation. Further, the change from landing on runway 18R to 18L also reduced the radius of the turn and increased the required angle of bank. The increased left banked turn, the right engine operating at a high-power setting, and the airplane's increased drag likely decreased the airplane airspeed below the airplane's minimum controllable airspeed (Vmc), which resulted in a loss of control.

The pilot in the multi-engine, retractable landing gear airplane reported that, during an instrument flight rules cross-country flight, about 5,000 ft above mean sea level, the left engine surged several times and he performed an emergency engine shutdown. Shortly afterward, the right engine lost power. During the emergency descent, the airplane struck treetops, and landed hard in a field with the landing gear retracted. The airplane sustained substantial damage to both wings, the engine mounts, and the lower fuselage. The pilot reported that he had requested 200 gallons of fuel from his home airport fixed base operator, but they did not fuel the airplane. The pilot did not check the fuel quantity during his preflight inspection. According to the Federal Aviation Administration Airplane Flying Handbook, Chapter 2, page 2-7, pilots must always positively confirm the fuel quantity by visually inspecting the fuel level in each tank. The pilot reported that there were no mechanical malfunctions or failures with the airplane that would have precluded normal operation.

The airline transport rated pilot and passenger departed on a cross-country business flight in a twin-engine, turbo-propeller-equipped airplane in day, visual meteorological conditions. Shortly after takeoff, the airplane banked left, descended, and impacted terrain about 1/2 mile from the end of the runway. There was not a post-crash fire and fuel was present on site. A postaccident airframe examination did not reveal any anomalies that would have precluded normal operation. Examination of the left engine found signatures consistent with the engine producing power at impact. Examination of the right engine revealed rotational scoring on the compressor turbine disc/blades, and rotational scoring on the upstream side of the power vane and baffle, which indicated that the compressor section was rotating at impact; however, the lack of rotational scoring on the power turbine disc assembly, indicated the engine was not producing power at impact. Testing of the right engine's fuel control unit, fuel pump, propeller governor, and overspeed governor did not reveal any abnormities that would have accounted for the loss of power. The reason for the loss of right engine power could not be determined based on the available information.

The pilot and two medical crew members departed on an air ambulance flight in night instrument meteorological conditions to pick up a patient. After departure, the local air traffic controller observed the airplane's primary radar target with an incorrect transponder code in a right turn and climbing through 4,400 ft mean sea level (msl), which was 800 ft above ground level (agl). The controller instructed the pilot to reset the transponder to the correct code, and the airplane leveled off between 4,400 ft and 4,600 ft msl for about 30 seconds. The controller then confirmed that the airplane was being tracked on radar with the correct transponder code; the airplane resumed its climb at a rate of about 6,000 ft per minute (fpm) to 6,000 ft msl. The pilot changed frequencies as instructed, then contacted departure control and reported "with you at 6,000 [ft msl]" and the departure controller radar-identified the airplane. About 1 minute later, the departure controller advised the pilot that he was no longer receiving the airplane's transponder; the pilot did not respond, and there were no further recorded transmissions from the pilot. Radar data showed the airplane descending rapidly at a rate that reached 17,000 fpm. Surveillance video from a nearby truck stop recorded lights from the airplane descending at an angle of about 45° followed by an explosion. The airplane impacted a pasture about 1.5 nautical miles south of the airport, and a post impact fire ensued. All major components of the airplane were located within the debris field. Ground scars at the accident site and damage to the airplane indicated that the airplane was in a steep, nose-low and wings-level attitude at the time of impact. The airplane's steep descent and its impact attitude are consistent with a loss of control. An airplane performance study based on radar data and simulations determined that, during the climb to 6,000 ft and about 37 seconds before impact, the airplane achieved a peak pitch angle of about 23°, after which the pitch angle decreased steadily to an estimated -42° at impact. As the pitch angle decreased, the roll angle increased steadily to the left, reaching an estimated -76° at impact. The performance study revealed that the airplane could fly the accident flight trajectory without experiencing an aerodynamic stall. The apparent pitch and roll angles, which represent the attitude a pilot would "feel" the airplane to be in based on his vestibular and kinesthetic perception of the components of the load factor vector in his own body coordinate system, were calculated. The apparent pitch angle ranged from 0° to 15° as the real pitch angle steadily decreased to -42°, and the apparent roll angle ranged from 0° to -4° as the real roll angle increased to -78°. This suggests that even when the airplane was in a steeply banked descent, conditions were present that could have produced a somatogravic illusion of level flight and resulted in spatial disorientation of the pilot. Analysis of the performance study and the airplane's flight track revealed that the pilot executed several non-standard actions during the departure to include: excessive pitch and roll angles, rapid climb, unexpected level-offs, and non-standard ATC communications. In addition to the non-standard actions, the pilot's limited recent flight experience in night IFR conditions, and moderate turbulence would have been conducive to the onset of spatial disorientation. The pilot's failure to set the correct transponder code before departure, his non-standard departure maneuvering, and his apparent confusion regarding his altitude indicate a mental state not at peak acuity, further increasing the chances of spatial disorientation. A post accident examination of the flight control system did not reveal evidence of any preimpact anomalies that would have prevented normal operation. The engine exhibited rotational signatures indicative of engine operation during impact, and an examination did not reveal any preimpact anomalies that would have precluded normal engine operation. The damage to the propeller hub and blades indicated that the propeller was operating under high power in the normal range of operation at time of impact. Review of recorded data recovered from airplane's attitude and heading reference unit did not reveal any faults with the airplane's attitude and heading reference system (AHRS) during the accident flight, and there were no maintenance logbook entries indicating any previous electronic attitude director indicator (EADI) or AHRS malfunctions. Therefore, it is unlikely that erroneous attitude information was displayed on the EADI that could have misled the pilot concerning the actual attitude of the airplane. A light bulb filament analysis of the airplane's central advisory display unit (CADU) revealed that the "autopilot disengage" caution indicator was likely illuminated at impact, and the "autopilot trim" warning indicator was likely not illuminated. A filament analysis of the autopilot mode controller revealed that the "autopilot," "yaw damper," and "altitude hold" indicators were likely not illuminated at impact. The status of the "trim" warning indicator on the autopilot mode controller could not be determined because the filaments of the indicator's bulbs were missing. However, since the CADU's "autopilot trim" warning indicator was likely not illuminated, the mode controller's "trim" warning indicator was also likely not illuminated at impact. Exemplar airplane testing revealed that the "autopilot disengage" caution indicator would only illuminate if the autopilot had been engaged and then disconnected. It would not illuminate if the autopilot was off without being previously engaged nor would it illuminate if the pilot attempted and failed to engage the autopilot by pressing the "autopilot" push button on the mode controller. Since the "autopilot disengage" caution indicator would remain illuminated for 30 seconds after the autopilot was disengaged and was likely illuminated at impact, it is likely that the autopilot had been engaged at some point during the flight and disengaged within 30 seconds of the impact; the pilot was reporting to ATC at 6,000 ft about 30 seconds before impact and then the rapid descent began. The airplane was not equipped with a recording device that would have recorded the operational status of the autopilot, and the investigation could not determine the precise times at which autopilot engagement and disengagement occurred. However, these times can be estimated as follows:
- The pilot likely engaged the autopilot after the airplane climbed through 1,000 ft agl about 46 seconds after takeoff, because this was the recommended minimum autopilot engagement altitude that he was taught.
- According to the airplane performance study, the airplane's acceleration exceeded the autopilot's limit load factor of +1.6 g about 9 seconds before impact. If it was engaged at this time, the autopilot would have automatically disengaged.
- The roll angle data from the performance study were consistent with engagement of the autopilot between two points:
1) about 31 seconds before impact, during climb, when the bank angle, which had stabilized for a few seconds, started to increase again and
2) about 9 seconds before impact, during descent, at which time the autopilot would have automatically disengaged. Since the autopilot would have reduced the bank angle as soon as it was engaged and there is no evidence of the bank angle reducing significantly between these two points, it is likely that the autopilot was engaged closer to the latter point than the former. Engagement of the autopilot shortly before the latter point would have left little time for the autopilot to reduce the bank angle before it would have disengaged automatically due to exceedance of the normal load factor limit. Therefore, it is likely that the pilot engaged the autopilot a few seconds before it automatically disconnected about 9 seconds before impact. The operator reported that the airplane had experienced repeated, unexpected, in-flight autopilot disconnects, and, two days before the accident, the chief pilot recorded a video of the autopilot disconnecting during a flight. Exemplar airplane testing and maintenance information revealed that, during the flight in which the video was recorded, the autopilot's pitch trim adapter likely experienced a momentary loss of power for undetermined reasons, which resulted in the sequence of events observed in the video. It is possible that the autopilot disconnected during the accident flight due to the pitch trim adapter experiencing a loss of power, which would have to have occurred between 30 and 9 seconds before impact. A post accident weather analysis revealed that the airplane was operating in an environment requiring instruments to navigate, but it could not be determined if the airplane was in cloud when the loss of control occurred. The sustained surface wind was from the north at 21 knots with gusts up to 28 knots, and moderate turbulence existed. The presence of the moderate turbulence could have contributed to the controllability of the airplane and the pilot's inability to recognize the airplane's attitude and the autopilot's operational status.

While conducting a post maintenance test flight in visual flight rules conditions, the private pilot of the multi-engine airplane reported an oil leak to air traffic control. The controller provided vectors for the pilot to enter a right base leg for a landing to the south at the nearest airport, about 7 miles away. The pilot turned toward the airport but indicated that he did not have the airport in sight. Further, while maneuvering toward the airport, the pilot reported that the engine was "dead," and he still did not see the airport. The final radar data point recorded the airplane's position about 3.5 miles west-northwest of the approach end of the runway; the wreckage site was located about 4 miles northeast of the runway, indicating that the pilot flew past the airport rather than turning onto a final approach for landing. The reason that the pilot did not see the runway during the approach to the alternate airport, given that the airplane was operating in visual conditions and the controller was issuing guidance information, could not be determined. Regardless, the pilot did not execute a precautionary landing in a timely manner and lost control of the airplane. Examination of the airplane's left engine revealed that the No. 2 connecting rod was broken. The connecting rod bearings exhibited signs of heat distress and discoloration consistent with a lack of lubrication. The engine's oil pump was intact, and the gears were wet with oil. Based on the available evidence, the engine failure was the result of oil starvation; however, examination could not identify the reason for the starvation.

The pilot executed an instrument approach and landing in heavy rain. The airplane touched down about 21 knots above the applicable landing reference speed, which was consistent with an unstabilized approach. The airplane touched down near the displaced runway threshold about 128 kts, and both wing ground spoilers automatically deployed. The pilot reported that the airplane touched down “solidly,” and he started braking promptly, but the airplane did not slow down. The main wheels initially spun up; however, both wheel speeds subsequently decayed consistent with hydroplaning in the heavy rain conditions. When the wheel speeds did not recover, the brake control unit advised the flight crew of an anti-skid failure; the pilot recalled an anti-skid CAS message displayed at some point during the landing. The pilot subsequently activated the emergency brake system and the wheel speeds decayed. The airplane ultimately overran the departure end of the runway about 60 kts, crossed an airport perimeter road, and encountered a small creek before coming to rest. The wings had separated from, and were located immediately adjacent to, the fuselage. The pilot reported light to moderate rain began on final approach. Weather data and surveillance images indicated that heavy rain and limited visibility prevailed at the airport during the landing. Thunderstorms were active in the vicinity and the rainfall rate at the time of the accident landing was between 4.2 and 6.0 inches per hour. About 4 minutes before the accident, a surface observation recorded the visibility as 3 miles. However, 3 minutes later, the observed visibility had decreased to 3/8 mile. A review of the available information indicated that the tower controller advised the pilot of changing wind conditions and of better weather west of the airport but did not update the pilot regarding visibility along the final approach course or precipitation at the airport. The pilot stated that the rain started 2 to 3 minutes before he landed and commented that it was not the heaviest rain that he had ever landed in. The pilot was using the multifunction display and a tablet for weather radar, which showed green and yellow returns indicating light to moderate rain during the approach. He chose not to turn on the airplane’s onboard weather radar because the other two sources were not indicating severe weather. The runway exhibited skid marks beginning about 1,500 ft from the departure end and each main tire had one patch of reverted rubber wear consistent with reverted rubber hydroplaning. The main landing gear remained extended and both tires remained pressurized. The tire pressures corresponded to a minimum dynamic hydroplaning speed of about 115 kts. The airplane flight manual noted that, in the case of an antiskid failure, the main brakes are to be applied progressively and brake pressure is to be modulated as required. The emergency brake is to be used in the event of a brake failure; however, the pilot activated the emergency brake when the main brakes still functioned; although, without anti-skid protection.