On the 9th of February 2020, Utair flight 595 was coming in to land when it unexpectedly struck a snowbank. The aircraft, a Boeing 737-500 registration VQ-BPS, was a scheduled domestic passenger flight from Moscow to Usinsk. The captain was an experienced pilot with almost 18,000 hours, 7,600 of which were on this aircraft type. The first officer had 6,500 hours, 5,000 on type. There were 94 passengers on board.

The Interstate Aviation Committee (MAK) released their report on the 16th of March 2021. The final report is in Russian only; I have referred to the translations posted by AVHerald and b737.org.

When Utair flight 595 left Moscow, it was already clear that the conditions at Usinsk were not good. A Notice to Airmen (NOTAM) warned that the Instrument Landing System and approach lights for runway 13 were out of service. Local weather showed moderate wind blowing at 11 knots, gusting to strong winds of 20 knots. It was -21°C (-6°F) with a cloud ceiling at 450 metres (1,500 feet). Visibility was 1,100 metres (just over 3,500 feet): below the minimum needed for an approach.

The crew decided to take their chances, hoping that conditions would improve. Even though the ILS was out of service, they could perform a GNSS (RNAV) approach, with assistance from Air Traffic Control. This RNAV (“area navigation”) procedure uses the Global Navigation Satellite System (GNNS) to allow for a low visibility approach. The crew also discussed various options, including clear plans to divert if it wasn’t possible to land at Usinsk.

For the initial flight, the first officer was the Pilot Flying. As they descended, the flight crew checked the weather again. At this point, Usinsk had a visibility of 1,600 metres (5,250 feet) with the cloudbase remaining at 450 metres. They agreed to continue to Usinsk.

The Runway Visual Range (RVR) is a measurement of how much of the runway you can see from the three key positions. A runway has three sensors installed, one at each threshold and one in the centre, which measure the horizontal visibility, so that the flight crew know the true visibility for the full landing or take off. Usinsk RVR was given as 1,600 at the touchdown point, 1,300 at the midpoint and 1,400 at the end.

This was right at the minimum needed for landing. Air Traffic Control cleared the aircraft for the GNSS approach to runway 13.

The Boeing 735 intercepted the final approach in the cloud at 2,200 feet above ground level at a speed of 190 knots indicated airspeed. The controller reported that the RVR was now 1,800, 1,700 and 1,700 metres; the horizontal visibility was improving. The crew confirmed that they were continuing their approach for landing and lowered the landing gear.

The wind dropped from 40 knots to 14 knots. The aircraft descended through 1,530 feet above ground level, below the glide path of 1,570 feet for the RNAV approach.

Neither pilot seemed to notice that they were descending below the glidepath. RNAV approaches use the barometric altimeter. As a pressure instrument, high or low temperatures can cause variance in the altitude readings. Specifically, in cold weather, the barometric altimeter will indicate that the aircraft is higher than it is.

The air traffic controller cleared the flight to land as they descended through 1,000 feet above ground level.

At 700 feet, the flight crew had the runway in sight. They disconnected the autopilot and the autothrust as they descended through 620 feet.

The captain asked the first officer to recycle the flight directors, which involves turning them off and on again to clear the data, and to set the altitude target ready for going around.

The first officer confirmed that the flight directors were off and that the altitude target was set.

The captain was now the Pilot Flying. He commented something along the lines of “hmm, that whiteness, of course.” It seems likely that this was a reference to the difficult visual conditions.

The first officer, in his role as the Pilot Monitoring, belatedly called out that they were below the glide path and that they needed to correct this. The captain agreed.

The Extended Ground Proximity Warning System (EGPWS) sounded as they reached 100 feet above ground level. At 60 feet above ground level, the first officer commented again that they were still below the glide path.

The captain said nothing, although he did increase the engine thrust. This slowed the descent but did not gain the height they needed to return to the glide path.

The EGPWS sounded at 50 feet above the ground. The flight was now dangerously close to the ground.

At 30 feet, the warning system sounded again. They continued at 138 knots indicated airspeed.

Earlier, a snowplow had cleared the runway of snow, leaving behind a snowbank at about 32 metres (105 feet) before the runway threshold. The snowbank was 1.1 metres or 3½ feet high.

As they flew over the snowbank, the landing gear tore through the snow. The aircraft made it to the runway threshold but, as it touched down, the right main gear broke off completely and the left main gear crumpled and collapsed.

A passenger in the cabin happened to be filming out the window as they struck the snowbank and landed.

The captain called out for reverse thrust and speed brakes before realising that he had no control over the Boeing, which continued to skid over the runway for another 2,000 metres. It yawed to the right and then came to a halt.

Everyone took a deep breath.

The first officer spoke first, saying that they should evacuate the passengers immediately. The captain disagreed, believing that the aircraft was fine other than the nose wheel steering.

Then the cabin crew called the cockpit to say that they could see fuel leaking from the right wing. It was actually hydraulic fluid but this convinced the captain that they needed to evacuate the aircraft on the runway.

The passengers disembarked using slides and using the overwing exits. Thankfully, there were no injuries.

The Russian Interstate Aviation Committee (MAK) immediately opened an investigation. The airport released a statement that, at the time of the incident, they were operating normally and that conditions were fine for inbound traffic. They had not, they reported, received a single negative remark from any crew in the last two years.

The MAK released their final report a year later; however, only in Russian. For my quotations from the report, I have made use of translations from AVHerald and b737.org.

The accident of the Boeing 737-500 VQ-BPS occurred during the landing as result of collision of the aircraft with a snow bank of 1.1 meters height 32 meters ahead of the runway threshold (within the paved safety area of the runway), that caused damage to the main gear struts and subsequently their collapse while travelling down the runway.

Explaining that the crash happened because the aircraft collided into a snowbank is rather like explaining that the murder was the result of the victim being stabbed with a knife. Normally, the probable cause would focus on a primary action which was taken or not taken which directly contributed to the crash. The MAK listed these as causal factors.

The existence of contradictions in Federal Aviation Regulations in Airspace of the Russian Federation, the airline’s standard operating procedures and the operations manual of the aircraft with respect to temperature corrections to barometric altimeter readings at low temperatures of ambient air

The Usinsk aerodrome operator’s failure to comply with FAP-262 for aerodromes [which] required a paved safety area free of snow banks

The Usinsk aerodrome operator’s failure to eliminate shortcomings in the winter operations of the aerodrome that were noted in a Rosaviatsia commission inspection on Jan 22nd 2020

Lack of risk assessment by the airline of conducting baro-VNAV approaches in the presence of factors hampering such approaches (low ambient air temperatures, presence of snow cover on surfaces, drifting snow/snowstorm, significant changes in terrain profile before the runway, no approach guidance lights like PAPI) as well as lack of according recommendations to crews of how to conduct such approaches including the transition to visual approach, and lack of training of crews

Insufficient assessment of present hazards and threats during pre-landing preparations by the crew reaching an insufficiently substantiated decision to perform a RNAV (GNSS) approach under control of the autopilot in LNAV/VNAV without introduction of correction for the low ambient temperature which led to the approach flown below the glidepath

Performing a flight below extended glidepath after the autopilot was disconnected and transition to manual control without attempting to acquire the correct glide path

The PIC is likely to have a visual illusion of being high on the glide path due to the snow covered surface, blowing snow and the presence of a ravine directly in front of the runway’s paved surface in absence of PAPI lights, which led to incorrect estimation of the aircraft’s height, lack of timely call outs by the first officer leading to crossing the begin of paved surface of the runway significantly below the required height.

That’s a bit of a handful but it comes down to a few key points.

Most obviously: The snowbank shouldn’t have been there. Other winter operations issues at the Usinsk had been highlighted during an inspection the previous winter and were not rectified.

The crew should have considered the low temperature before deciding on the RNAV (GNSS) approach. As they approached Usinsk, it was -21°C (-6°F). As a result, the barometric altimeters may have shown them up to 280 feet higher than their actual altitude.

However, the airline training and procedures did not include VNAV approaches in freezing conditions. As a result, the flight crew had no references or recommendations as to what to do in such circumstances.

The first officer, distracted by other tasks and a slow handover, didn’t notice that they were beneath the glidepath until they were just 17 seconds away from the snow ridge. There were no references to radio altimeters in the translated report other than the EGPWS. Even so, once they had established that they were below the glide path, they should have prioritised regaining the glidepath or breaking off the unstabilised approach (or both!).

The captain had no visual indications to help him work out his height above the runway. Given the snow and visibility, it’s also possible that the captain suffered a visual illusion, leading him to feel confident that they were coming in at the right height. PAPI lights would have immediately alerted the captain to the fact that they were low; however, they were not working.

Without the temperature correction the aircraft would thus cross the runway threshold at an incorrect, less than required height, which, if the crew attempted to correct, required to unbalance the aircraft in visual flight without any instrument indications to verify the aircraft’s position related to the runway in limited time.

When the EGPWS sounded the 100, 50 etc. calls, there was no reduction in the rate of descent. The main gear collided with snow banks of 1.1 meters (3.6 feet) height 32 meters before the runway threshold. The aircraft should have crossed the runway threshold at 50 feet above the runway surface.

The crew’s initial decision to fly to Usinsk in hopes that the weather would improve was not unreasonable. They identified an alternate aerodrome which they could divert to if the weather were not good enough for the approach. During the briefing, they discussed the approach, the wind, the runway friction values, applying brakes directly after touch down for maximum braking and how much time they could afford to waiting at Usinsk before they would need to divert to their alternate.

It all seemed very professional, other than the decision to approach in RNAV(GNSS) mode. On top of all of this, it seems that the aircraft itself was not approved for this type of approach. The flight crew never discussed this and they never considered the risks associated with this approach at low temperatures.

As they descended, they verified the barometric altitude against the approach chart. However, in the final stages of the descent, the aircraft was 50 feet below the glide path. The featureless white of the snow-covered landscape gave no hints. The snow bank was only one metre high. If the snowbank hadn’t been there, would they have maintained that height above the ground to reach the runway? The answer to this depends on what the captain thought he saw, which is impossible to reconstruct.

The final decision that I would have listed as probable cause was the decision to make the RNAV approach despite the NOTAM. Even assuming that the weather would improve, they knew that visibility would be limited and that there were no approach lights on runway 13. The snow should not have been piled up on the approach path but for context, the snow was piled up to about the height of R2-D2. At a fair distance before the threshold (32 metres or 105 feet), it would not have seemed much of an obstacle to an inbound aircraft 50 feet above the ground. The lack of PAPI lights in poor visibility with temperatures so low that the barometric altimeters could not be trusted should have caused much more concern than it did.