Alitalia Flight 404: “Catch the glide!”

8 Apr 22 9 Comments

On the 14th of November 1990, Alitalia flight AZ404 crashed into high ground upon approach to Zürich. The accident was investigated by the Federal Aircraft Accidents Inquiry Board in Switzerland and was supported by the United States National Transport and Safety Board. One aspect that makes it interesting is that the two boards came to slightly different conclusions as to what happened that night.

I-ATJA photographed by Michel Gilland at Charles de Gaulle airport in Paris.(GNU Free Documentation License 1.2)

The aircraft was a McDonnell Douglas USA DC-9-32 with the registration I-ATJA. Alitalia had acquired the DC-9 along with a number of other aircraft from Aero Trasporti Italiani (ATI). The ATI aircraft were equipped with King navigation receivers whereas the original Alitalia aircraft had Collins navigation receivers. The two types of navigation receivers were interchangeable and, over time, the units became intermixed.

The flight crew had flown in to Milan-Linate together the day before the accident flight. They spent the next 15 hours resting at a local hotel.

The captain held an ATPL with 1,200 hours of military flying experience. He made captain on the DC-9-30 in March 1988, two years before the crash. He had been flying with Alitalia for about six months. His total flying time was 10,193 flight hours, of which 3,194 were on type.

The first officer had received his pilot training from Alitalia, where he’d been employed as a DC-9-32 co-pilot since July 1989. He had a total of 831 hours, of which 621 were on type.

They were rostered to fly from Milan to Frankfurt and back followed by another return flight from Milan to Zürich and back.

The DC-9-32 arrived from Düsseldorf at 09:27 that morning. The flight crew who flew the aircraft from Düsseldorf to Milan left two notes in the technical logbook.

VHF NAV 2: In Radio Selector position Radio 2 VHF-NAV 2 does not give a TO-FROM indication on HSI 2. In position Approach no TO-FROM indication on HSI-2

CAT II Simul. Appr. At 200 feet the autopilot had a tendency to fly under the glide path then to return to it followed by an accentuated “dive”. The autopilot was switched off and we continued manually

The aircraft VHF NAV receivers were labeled VHF NAV 1 and VHF NAV 2. They are a critical piece of navigation equipment used to receive signals from VOR stations (Very High Frequency Omnidirectional Range). Simply put, the receiver on the aircraft uses the variable signal and the reference signal from the VOR station to establish where the aircraft is in reference to the station.

The accident flight crew checked in at 13:00 for the scheduled departure to Frankfurt. The passenger flight departed at 14:07 and arrived at Frankfurt on schedule. While at Frankfurt, a ground floodlight was changed. There was no further maintenance.

Once back at Milian, the captain spoke to a maintenance crew member to say that something was wrong with the VHF-NAV navigation receiver. He had experienced the same problem: there was no TO-FROM indicator, just as the previous flight crew had written in the logbook. There was one difference: he had experienced the failure in Radio 1. He did not make log the issue in the Technical Logbook.

The Alitalia technicians replaced both VHF-NAV receivers, installing a King receiver in the No. 1 navigation system and a Collins receiver in the No. 2 navigation system. The technicians conducted a self-test on the systems and checked the equipment in navigation mode. However, there was no way to test the systems for ILS (instrument landing system) glideslope signals on the ground. They asked the crew to do a Simulated Cat II approach at their next stop, Zürich, so that the aircraft would regain its full CAT II status.

A CAT II approach in this context is a Category II approach. It is needed in bad weather because the decision height at the point at which you must either have the runway in sight or abort the approach is less than 200 feet (but not less than 100 feet) and allows for a safe landing for runway visual ranges between 1,200 and 1,800 feet. This is to say, this precision instrument approach allows you to land safely in bad conditions with a lot less visibility than is needed for a CAT I approach, which requires a decision height of at least 200 feet and a runway visual range of at least 1,800 feet.

The crew were being asked to do a simulated CAT II approach so that they could verify that the new receivers were working as intended in conditions with enough visibility that they were not reliant on the new receivers for a safe approach.

They needed to do the simulated CAT II approach at Zürich because the forecast for Milan-Linate was such that they might require a CAT II approach and landing for the return flight. If they did not test the receivers at Zürich, then when they returned to Milan, they might need to divert if the weather closed in.

For the flight to Zürich, the first officer was the Pilot Flying and the captain was the Pilot Monitoring (although in 1990, those terms were not yet commonly used). At 18:36, the DC-9 departed from runway 36R for the flight to Zürich.

The flight climbed to FL200 (20,000 feet). Once established in the cruise, the flight crew listened to the Zürich VOLMET which provides weather information for aircraft in flight. The captain predicted that they would be landing on runway 28, as the surface wind at Zürich was 8 knots from 240. They listened to the ATIS, a recorded broadcast of the airport’s current information, which stated that the runway in use was runway 14. Despite this, the captain continued to talk about a right-hand circling approach for runway 28. He then talked the first officer through a left-hand circling approach for the same runway.

The captain then had the first officer work out that if the QNH was 1019 hPa (hectoPascals) then the QFE was 970 hPa. The QNH is the height above sea level, so the altimeter with a pressure setting of 1019 hPa showed the aircraft’s height above sea level. QFE is the elevation above the airfield, so they have changed the pressure setting so that the altimeter will show their height above the runway. It saves a bit of mental maths, as when landing, your altitude above mean sea level is not as important as knowing how far away the ground is.

As they entered the descent for Zürich, they discussed the approach procedure for runway 14. The first officer stated the Outer Marker height, but he named the Outer Marker height for runway 16 instead of the correct height for runway 14.

The captain did not correct him.

On an ILS (instrument landing system), the Outer Marker is a beam at the point where an aircraft will generally enter the glide slope for the instrument landing. It is the beginning of the final approach to the runway and usually is located around four to seven nautical miles from the runway threshold.

If the navigation systems are correctly configured, a tone will sound as the aircraft passes the outer marker and a flashing blue light will annunciate on the panel. The Outer Marker height that the first officer referred to is the height that the aircraft should be above the ground as it passes the Outer Marker.

After they discussed setting the navigation aids, the two pilots went over the correct procedure in the case of a communications failure.

The air traffic controller instructed the crew that after following radar vectors (that is, following the headings that the controller was about to give them), they should fly an ILS approach to runway 14.

The first officer said, “We perform a CAT II,” as they needed to test the new receivers. The captain agreed. Then the first officer tried to verify the decision height; based on his statements, he was clearly still looking at the approach chart for runway 16.

As they descended to FL90 (9,000 feet on the standard pressure setting), the captain commented that the air traffic controller was bringing them in high.

The controller cleared flight 404 to descend to FL60 and given a heading of 325°

VHF NAV 1 was tuned to the Trasadinging VOR and VHF NAV 2 was tuned to the Kloten VOR. At 19:04, the captain explained to the first officer, “the outer marker is at 1,200 feet [above the ground], it can be verified by 3.8 [nautical miles] from Kloten.”

ATC gave the flight crew a new heading which the captain confirmed. After they had identified the ILS for runway 14, the controller gave them the approach clearance for runway 14, a new heading of 110° and clearance to descend to 4,000 feet on the QNH.

His intent was that the flight crew would remain at 4,000 feet until they entered the glide slope.

The captain confirmed the clearance but he read back the heading as 120° instead of 110°. He then told the first officer that they had the approach clearance with a cleared altitude of 4,000 feet.

The first officer said “radio approach” as they descended past 5,000 feet above mean sea level. One of the two pilots, it’s not clear which, asked the other whether he’d seen the glide path indication.

When the question was made, the aircraft was descending through 4,700 feet and they had not yet reached the Outer Marker.

The next question from the first officer is unclear on the recording, it sounded like “Do you have the glide?”

The captain responded with “On one…”, clearly referring to Radio 1 on VHF-NAV 1.

The first offer said, “I don’t have it.”

The captain said, “Good, so let’s do it on one.”

“Radio one,” confirmed the first officer.

From this conversation, we can assume that the crew had selected Radio APP (approach) on the selectors ready to intercept the ILS, however, they were seeing conflicting glide slope indications on the two horizontal situation indicators. We know from the flight data recorder that the glide slope comparitor warning light was illuminated. The first officer, who was looking at HSI 2, clearly did not have an indication. The captain, looking at HSI 1, saw that they were almost on the glide path. The first officer’s instruments must be wrong.

The first officer switched his instruments to the signal coming from VHF-NAV Receiver 1. Now both flight crew saw a centred indication on all of the instruments, exactly what they were expecting to see.

By now, the aircraft had passed through the localiser and was east of it. The captain said, “Capture LOC capture glide path capture… so we are on the localiser, a little off track, but…”

Clearly, the captain considered the flight to be established on the localiser, that is, he believed that they intercepted the localiser beam and could continue to descendon the ILS.

Neither pilot noticed that they were still 11.5 nautical miles from the threshold, too far out to receive glide slope indication. They were also 1,200 feet below the height of the glide path.

When the investigators reconstructed the flight’s approach to Zürich, they found that at this point of the approach, at 11.5 miles out, the glideslope needles on all four instruments were in the fully UP position, which meant they were out of sight. At this distance and altitude, there was no reason why the captain should have seen a glideslope indication.

The first officer set the altimeter pressure to 970 hPa so that the altimeter would show the height above the ground. An Altitude Exit Alert sounded as the aircraft descended through 3,700 feet above mean sea level. The Altitude Exit Alert is to warn the flight crew if the altitude deviates from the selected altitude by more than 250 feet. Having changed the pressure, the altitude would have changed quite dramatically.

The captain cancelled the warning by pressing 5,000 feet (the altitude for a go-around) on the Altitude Preselect. He said, “There’s another one [aircraft, Finnair 863] in front, quite close. You can reduce even further to 150 [knots], otherwise we’ll end up with a go around.”

He did not want to come in to land too close behind Finnair 863, because they would have to go around if the aircraft was still on the runway when they were ready to land, so he was advising the first officer to slow down, increasing the distance between the aircraft.

Another discussion took place, this time about possible icing. Then the flaps were set to 25°, which set off the No Landing Gear horn, which sounds to alert the crew if they appear to be in a landing configuration without extending the landing gear.

The captain said, “Outer Marker check is at 1,250 feet”. At this point, the aircraft was 1,600 feet above the elevation of the runway threshold.

Once the flaps finished extending to 25°, one of the pilots set the flaps to 50°. They descended through the Outer Marker height of 1,250 feet above the ground, except that they still hadn’t reached the Outer Marker. The captain said “Bravo.” The sounds of switching were recorded in the cockpit.

At eight nautical miles out, the captain said, “3.8….almost 4 miles.”

“Haven’t we passed it?” The aircraft was seven miles out and now just 670 feet above the runway elevation.

The captain said, “No, no. It hasn’t changed yet.”

At 6.6 nautical miles from the threshold, the captain said, “Oh, it shows 7.”

At that moment, the Zürich arrivals controller asked the crew to change frequency to Zürich Tower.

The captain acknowledged the call but didn’t change the frequency, instead saying to the first officer, “That doesn’t make sense to me.”

“Nor to me,” said the first officer.

Suddenly, the captain shouted, “Pull, pull, pull, pull!” The sound of the autopilot disconnect could be heard.

They were 1,250 feet below the glideslope and 500 feet above the ground below them. However, neither pilot could see that they were flying straight towards Stadlerberg mountain, a 2,090 foot mountain about 11 kilometers from the airport. The mountain would have blocked out the lights from the runway still over five miles away. All the flight crew knew for sure was that something wasn’t right.

The first officer called out, “Go around.”

An aviation accident has many factors: a number of small decisions and actions that contribute to the final disaster. Sometimes, there is a single clear moment in the chain of events where the last possibility to save the aircraft is lost.

This was that moment.

Going around was the right choice. The captain had just said outright that what his instruments were showing didn’t make sense to him. But instead of supporting his first officer in the go around, he dismissed the idea.

“No, no, no, no. Catch the glide!” The captain knew they were not on the glideslope but wanted the first officer to recover the approach.

The flight data recorder shows the pitch of the aircraft changed from -2° (nose down) to +5.4° (nose up). At the same time, the thrust was increased from 1.3 to 1.7 EPR (engine pressure ratio).

The sink rate was reduced from 1,100 feet/minute to 190 feet/minute. The aircraft was in level flight and still descending, just much slower. The captain expected the first officer to level out until they intercepted the glideslope again and then continue the descent.

Eleven seconds passed. The captain said, “Can you hold it?”

“Yes,” said the first officer.

The Radio Altimeter sounded a warning that they were only two hundred feet above ground level. As it sounded, the captain said, “Hold on, let’s try to–”

His final words were cut off as the DC-9 flew into the trees of Stadlerberg mountain. The right wing dropped. The aircraft turned violently to the right and began to invert, ripping the right engine from the fuselage. The tail disintegrated as the cockpit and cabin crashed into the hill. The fuel pouring from the broken wings ignited and spread over the slope. The fire spread over the woods, burning until the evening of the following day. There was no chance of survival.

Firefighter at the scene of the crash the following evening. (Source unknown)

The last data from the flight data recorder showed that the flight was at an altitude of 1,660 feet above sea level. They were 5.2 nautical miles from the runway.

Next week, we’ll look at the investigation.

Category: Accident Reports,


  • Well, it sounds plain enough at this point. Weird avionics failure, PiC is making assumptions about position, captain isn’t double-checking PiC, both are assuming that everything is fine so they don’t even notice the lack of marker tones or potentially weird DME indications. (I don’t see any mention of DME, perhaps they didn’t have it?) And then captain tries to take over at the last moment.

    But there’s clearly a lot more to learn.

  • At that moment, the Zürich arrivals controller asked the crew to change frequency to Zürich Tower.

    Nowadays, I’d expect ATC to see the aircraft altitude on their secondary radar, and knowing their airspace, the controller should know the aircraft is much too low to be safe, and say something. It’ll be interesting to see if the mode S tech back in 1990 provided the same information.

    Today, the TAWS would generate a warning; but obviously this still requires that the aircraft’s position is correctly known. Flying in the Swiss Alps in the dark not knowing that you don’t know your position is always a recipe for disaster. Thankfully, navigation systems have improved a lot.

  • Technical dweebie bit – the engine wasn’t on the wing, and therefore could not have been ripped from.

  • If I can get GPS driving directions 100 ft. before I have to negotiate a turn on my phone, how is it that a modern aircraft can be some 11.5 miles off with no warning that a glideslope indicator is WRONG?
    And using the same logic, why isn’t the pressure (hPa) setting of the altimeter computer controlled using GPS, or at the very least computer verified?
    If it hasn’t been done already – it would be an extremely simple matter to create a database holding data for every airport on the planet, that planes could download before / in flight, to get assistance e.g. automatic settings, or double-checking for human error?
    It astounds me that this is not already being done.

  • I have not finished reading and must do some work first, but two remarks:
    Many (nowadays most) operators prefer to use QNH, where the reference is sea level. As Sylvia states, it requires a bit of mental gymnastics, but especially in mountainous terrain, it gives the pilots a better reference to the obstacles, like well, mountains.
    QFE uses the height above the airport as reference. Easier whilst on the approach, but it does of course not relate to the surrounding terrain.
    Outer marker: No, this is not, as suggested, the height (QFE) or altitude (QNH) at which the glide slope is intercepted.
    Typically, this happens at 2000 feet above the airport elevation. Ideally, the light is level when on the LLZ (localizer, the beam aligned with the runway). This allows the crew to get the aircraft “in the groove”: reduce speed, select the (initial) flap ands slat settings and, when the glideslope is indicating on the ILS or flight director, extend the landing gear when it shows a little (usually one “dot”) below the slope, and final flap setting when “on the glide”. Often ATC will ask to keep the speed up so that following traffic and be slotted in behind, but in this situation the approach is stable or very nearly so and will not require many adjustments.
    The OM (outer marker) triggers a bleeping sound and a blue flashing light in the cockpit. The distance from the runway is selected to coincide with approximately 1200 feet above the touchdown zone when on the glide slope. After passing this marker, ATC can no longer request the crew “to keep the speed up”, the crew can (and must) ensure that the aircraft’s configuration and Vref (final approach reference speed) are all set and correct.
    OK, I must leave it, I will read the rest as soon as I have some time. Maybe with more comments, who knows?
    Who would have thought that a nearly 79 year old is still asked to work on a regular basis?
    Love you and leave you (for now).

    • Regarding the QFE/QNH, it doesn’t actually get much treatment in the report. I think the point is that it was one of many things that the captain was showing/quizzing the first officer on.

      I’ve taken to sticking to height/altitude based on the report’s usage, as opposed to what I was taught, as it doesn’t seem to be consistent.

      Still more work for you, Rudy! It looks like I’ve still not got the OM right. I had been relying on Cliff for that bit! This is the kind of reference that led me to making that statement:

      Outer Marker: Identifies glideslope intercept or the Final Approach Fix (light flashes blue)

      Is this wrong or have I misunderstood?

  • Reading this casually, I find it difficult to make sense of the sequence of events that led to this crash.
    Maybe it would help if I get hold of an approach chart of that particular period?

    For Jon: Entry in a tech log:
    Pilot wrote: Left engine missing.
    Engineer’s reply: Engine found on the port wing after a brief search.

    Switching from QNH to QFE after mentally calculating the QFE during an initial approach adds unnecessary workload, especially at night, in a mountainous area. The crew were on their third leg of the day and fatigue could have crept in.
    Especially as there are high obstacles in the area, sometimes referred to as “cumulus granitus” a change to QFE would not seem to make sense. And anyway, ATC would give it on request.
    On approach charts (Jeppesen), the altitudes or heights, like initial approach altitude, OM, and other relevant positions or fixes, are given both in QFE and QNH.
    Most airlines have strict rules regarding QNH or QFE. The chosen setting is prescribed in the company’s manuals. So why was there a switch from QNH to QFE? And if this was SOP, why did it seem to have given an extra workload?
    No, there are many things here that in my mind do not add up

Post a comment:

Your email address will not be published. Required fields are marked *


This site uses Akismet to reduce spam. Learn how your comment data is processed.