Overview of the investigation

The occurrence

On 18 February 2025, the ATSB was notified of and subsequently commenced an investigation into a level crossing collision between a road vehicle and a track machine consist at Dalby, Queensland. 

Just before midday, a self-propelled track machine consist was being operated between Dalby and Meandarra. This rail vehicle, number ZG86 and operated by Queensland Rail, consisted of a split head ballast tamper and a ballast regulator. It departed the Angle Siding in Dalby Yard at around 1152 local time. After several minutes of shunting to reach the western yard limit,[1] the track machine consist was granted a proceed authority[2] to depart Dalby Yard at 1200.

The track machine consist then travelled north-west through a dual-track active level crossing with Cunningham Street, and a single-track active level crossing with the Bunya Highway. A third active level crossing with Nicholson Street was located around 450 m further along the line. The Nicholson Street level crossing had a different layout to nearby level crossings, as it consisted of dual single-track active level crossings 30 m apart in the same crossing square (Figure 1).

Figure 1: Layout of the Nicholson Street level crossing

Source: Google Earth, annotated by the ATSB

To reach Meandarra, the track machine consist had to travel on the Glenmorgan branch line. Just after it turned out onto the branch line, the flashing red warning lights for the branch line crossing at Nicholson Street automatically activated. The track machine consist then entered the Nicholson Street level crossing at 1204. 

At the same time, a road vehicle was also approaching the Nicholson Street level crossing from the south‑west. The driver of the road vehicle reported they were experienced and familiar with the area and the Nicholson Street level crossing.

The most recent Australian Level Crossing Assessment Model assessment from November 2024 noted that the minimum warning time for road users of an approaching train on the branch line was 23 seconds. The speed limit for rail vehicles through the Nicholson Street level crossing was 25 km/h. The assessment also reported that the level crossing and its associated controls were visible to road users from up to 500 m away.

As the track machine consist approached the Nicholson Street level crossing, it was travelling at an average speed of around 18 km/h. At this lower speed, the available warning time to road users to slow down and stop once the level crossing lights activated increased to over 30 seconds.

The track machine consist was equipped with forward‑facing video cameras and equipment that recorded its position The onboard video from the track machine captured the road vehicle as it approached the level crossing, and showed that the vehicle was not slowed sufficiently to stop at the stop line adjacent the flashing lights.

From the video evidence, the vehicle was calculated to have been travelling near the speed limit of 60 km/h as it approached and crossed the level crossing stop line. 

At this moment, the track machine consist was already halfway across Nicholson Street and about to pass in front of the road vehicle. The video showed the road vehicle slowing down significantly after crossing the stop line, however, it did not stop and consequently collided with the track machine.

The front of the road vehicle sustained significant damage from impacting the front left side of the track machine. The split head tamper at the front of the consist sustained minor damage and the operating crew were uninjured. The driver of the road vehicle was also uninjured.

Level crossing information

The split level crossings at Nicholson Street were a consequence of the track layout. A branch line to Glenmorgan deviated from the mainline around halfway between the Bunya Highway and Nicholson Street level crossings, and both tracks then crossed Nicholson Street separately. 

The mainline crossing on Nicholson Street had flashing warning lights and half-boom barriers as active protection, whereas the branch line crossing only had flashing warning lights and did not have booms.

The difference in active controls between the mainline and branch line crossings was likely due to:

• the proximity of the side streets which intersected with Nicholson Street near the level crossing
• the lower frequency of trains on the branch line compared to the mainline
• the direct traffic control[3] safeworking system, which prevented more than one train movement in any direction across Nicholson Street at any one time. 

As a result, only one of the 2 level crossings would activate with the presence of a train, and the other crossing would remain open for road traffic.

ATSB observations

Following the collision, the ATSB spoke with the involved parties, reviewed operational information from the rail infrastructure manager, reviewed recorded data and video recordings from the track machine consist, and assessed the design, condition, and operation of the level crossing. The ATSB’s review of the evidence collected identified that:

• For undetermined reasons, the road vehicle driver did not respond to the active controls as they approached the crossing.
• The road vehicle driver braked heavily upon seeing the rail vehicle but was unable to stop in time.
• The rail vehicle driver did not mishandle or overspeed the track machine consist.
• There were no mechanical factors which influenced the operation of either vehicle.
• There were no environmental factors which influenced the operation of either vehicle.
• The lights on the front of the track machine consist were turned on before the consist departed the siding.
• The horn on the track machine consist was sounded when it entered each level crossing. 

The ATSB’s analysis of the Nicholson Street level crossing found that:

• The level crossing operated normally prior to the collision.
• The warning lights for the Glenmorgan branch line were activated and visible at the time of the collision.
• The design and operation of the level crossing eliminated the risk of short stacking[4] and road vehicle encroachment, with the 2 level crossings unable to be activated simultaneously.
• The active controls at the level crossing were compliant with the relevant Australian Standards.
• There was no evidence of the level crossing being poorly maintained.
• There were no obstructions to visibility or sightlines along the road corridor, which prevented road users from seeing, interpreting, and responding to the active level crossing controls.

Reasons for the discontinuation

Based on a review of the available evidence, the ATSB considered it was unlikely that further investigation would identify any systemic safety issues or important safety lessons. Consequently, the ATSB has discontinued this investigation. 

The evidence collected during this investigation remains available to be used in future investigations or safety studies. The ATSB will also monitor for any similar occurrences that may indicate a need to undertake a further safety investigation.

Purpose of safety investigations The objective of a safety investigation is to enhance transport safety. This is done through:  identifying safety issues and facilitating safety action to address those issues providing information about occurrences and their associated safety factors to facilitate learning within the transport industry. It is not a function of the ATSB to apportion blame or provide a means for determining liability. At the same time, an investigation report must include factual material of sufficient weight to support the analysis and findings. At all times the ATSB endeavours to balance the use of material that could imply adverse comment with the need to properly explain what happened, and why, in a fair and unbiased manner. The ATSB does not investigate for the purpose of taking administrative, regulatory or criminal action. Terminology An explanation of terminology used in ATSB investigation reports is available here. This includes terms such as occurrence, contributing factor, other factor that increased risk, and safety issue. Publishing information Released in accordance with section 25 of the Transport Safety Investigation Act 2003 Published by: Australian Transport Safety Bureau © Commonwealth of Australia 2025 Ownership of intellectual property rights in this publication Unless otherwise noted, copyright (and any other intellectual property rights, if any) in this report publication is owned by the Commonwealth of Australia. Creative Commons licence With the exception of the Commonwealth Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this report is licensed under a Creative Commons Attribution 4.0 International licence. The CC BY 4.0 licence enables you to distribute, remix, adapt, and build upon our material in any medium or format, so long as attribution is given to the Australian Transport Safety Bureau.  Copyright in material obtained from other agencies, private individuals or organisations, belongs to those agencies, individuals or organisations. Where you wish to use their material, you will need to contact them directly.

Investigation summary

What happened

On 12 February 2025, Alliance Airlines Embraer E190, VH-UYO, was operating Qantas flight QF1888 from Cairns, Queensland to Darwin, Northern Territory. At 1634 local time, passing the initial approach fix for the instrument landing system (ILS) approach to Darwin Airport’s runway 29, the auto‑flight system approach mode unexpectedly disarmed and reverted to basic flight director modes. The aircraft then deviated right and then left of the ILS course, before intercepting the lateral course at about the final approach fix. 

Passing 1,000 ft above aerodrome elevation, the aircraft was above the glideslope, at a high rate of descent and high airspeed. The flight crew elected to continue the approach, as the aircraft was then in visual meteorological conditions. Passing 500 ft, the flight crew assessed that the aircraft was stabilised, although still too fast. The pilot monitoring subsequently identified that the flaps were not in the landing configuration and selected the correct position. The flight crew continued the approach and conducted an uneventful landing. 

What the ATSB found

The ATSB found that on crossing the initial approach fix for the ILS approach, due either to a system synchronisation issue or the pilot flying inadvertently disarming the approach mode, the aircraft’s auto‑flight system reverted to roll and flight path angle modes. 

Following the unexpected mode change, the pilot flying did not reengage approach mode or disconnect the autopilot. This likely contributed to the aircraft deviating outside the required lateral tolerance of the approach below the minimum safe altitude while in instrument meteorological conditions. 

Additionally, the ATSB found that the flight crew did not discontinue the approach when the aircraft was unstable at the 1,000 ft stabilisation height as they incorrectly assessed that they could continue to 500 ft in visual meteorological conditions with multiple stabilised approach criteria unmet.

In the limited time available to stabilise the aircraft by 500 ft, the flight crew incorrectly assessed that the aircraft was stable and continued the approach, unaware that the pilot monitoring had inadvertently selected an incorrect flap configuration. 

Finally, the ATSB found that Alliance Airlines' standard operating procedures were unclear about the criteria for continuing an unstable instrument approach to 500 ft when aircraft entered visual conditions.

What has been done as a result

Following this incident, Alliance Airlines issued an operations notice ‘to improve clarity and compliance’ with the stabilised approach criteria. The notice detailed the stabilised approach policy. It also amended the stabilisation height such that for 3‑dimensional and 2‑dimensional instrument approaches, and straight‑in visual approaches, the stabilised criteria were to be met by 1,000 ft above aerodrome elevation. The 500 ft stabilisation height applied only to a visual circuit or circling manoeuvre approaches. The notice reminded flight crew of Alliance’s ‘non punitive go‑around policy’ and required all unstable approaches to be reported. Finally, Alliance Airlines conducted a flight data review of unstable approaches over the previous 6 months operations to identify similar occurrences.

Safety message

The Flight Safety Foundation’s (FSF) Reducing the risk of runway excursions report found that, in the 16 years to 2009, the most common accident was a runway excursion, accounting for 33% of all aircraft accidents. The highest risk factor for runway excursions was identified as an unstable approach. Further, an FSF survey (Normalization of Deviance) identified that only 3–4% of approaches were unstable but that in over 97% of those, the flight crews did not conduct a go-around. It stated: 

Noncompliance with standard operating procedures (SOPs) — especially tolerance of unstabilized approaches — is a serious impediment to further reduction of accident risk. 

Guidance from the International Air Transport Association for preventing unstable approaches stated that pilots must be trained to understand the risks of an unstable approach, because an unstable approach can be completed successfully, which may reinforce bad practice.

Additionally, this incident highlights how important continuous attention to automatic flight system modes displayed on the primary flight display is to the maintenance of situation awareness. 

This incident also illustrates the need for effective flight crew monitoring. The Flight Safety Foundation identified that monitoring can be improved by standard operating procedures, increased emphasis and practice, and stated:

One of the most important aspects of a safe flight operation is the requirement for crewmembers to carefully monitor the aircraft’s flight path and systems, as well as actively cross-check each other’s actions.

The investigation

The occurrence

On the afternoon of 12 February 2025, Alliance Airlines (Alliance) Embraer ERJ 190‑100 IGW (E190), registered VH‑UYO, was operating Qantas flight QF1888 from Cairns, Queensland to Darwin, Northern Territory. On board were 2 flight crew, 2 cabin crew and 49 passengers. The captain was the pilot flying (PF), and the first officer was the pilot monitoring (PM).[1] 

After departing Cairns, the aircraft climbed to cruise at flight level (FL) 340.[2] En route the flight crew obtained air traffic control (ATC) clearances, first to deviate up to 30 NM (56 km) left, and later up to 50 NM (93 km) right of the planned route to avoid hazardous weather. At 1614 Darwin local time the flight crew received clearance to descend to FL 120. Just over 2 minutes later they requested a further clearance to deviate up to 60 NM (111 km) right to avoid weather. 

At 1619 the flight crew requested, and received, a clearance to deviate up to 70 NM (130 km) right of route and to track direct to waypoint LAPAR, once clear of the weather. About 2 minutes later, descending through FL 190, the aircraft turned left from a position about 30 NM (56 km) right of the planned route, to track 60 NM (111 km) direct to LAPAR (Figure 1).

Figure 1: VH-UYO recorded flight data showing weather diversion and tracking to LAPAR

Source: Google Earth overlaid with FlightRadar24 data, annotated by the ATSB

At 1622 the PM contacted Darwin Approach ATC, advised they were descending to FL 120, had received automatic terminal information service (ATIS) X‑ray (X), and were tracking direct to LAPAR. ATIS X included advice of:

• the expectation of an instrument approach
• wet runways
• wind from 340° at 15 kt, with a maximum 15 kt crosswind on runway 29
• visibility of 2,000 m
• showers of rain
• scattered[3] cloud 1,200 ft above aerodrome elevation. 

The PM advised the approach controller when the aircraft was approaching FL 120, and received further clearance to descend to 9,000 ft and, 2 minutes later, to 7,000 ft. At 1629, as the aircraft descended through about 8,000 ft, the controller requested a reduction to ‘minimum clean speed’, as by radar VH‑UYO was showing a groundspeed of 270 kt, which exceeded the 250 kt maximum indicated airspeed below 10,000 ft. Although the aircraft’s airspeed at that time was 250 kt, the flight crew actioned the request to reduce speed, advised that they were approaching 7,000 ft, and were then cleared to descend to 5,000 ft.

At 1631, the approach controller cleared the flight crew to descend to 3,000 ft and conduct the instrument landing system (ILS)[4] -Z approach to runway 29. LAPAR was the initial approach fix[5] for the ILS and was aligned with the runway centreline (Figure 2).

Figure 2: VH-UYO flight path (green/blue) and ILS‑Z localiser path (black) showing key events (blue arrows)

Source: FlightRadar24 data overlaid on Google Earth, annotated by the ATSB

The PM selected flap 1 as the aircraft descended through about 4,600 ft. About 30 seconds later, the approach controller instructed the PM to contact the tower controller when leaving 3,000 ft. In preparation for the ILS, the PF then pressed the approach (APP) pushbutton on the aircraft’s guidance panel, arming the approach mode. This also armed flight director (FD) localiser (LOC) lateral and glideslope (GS) vertical modes. With approach mode armed, when the aircraft intercepted the localiser (at LAPAR), LOC should become the active lateral mode and when it subsequently intercepted the glideslope, GS should become the active vertical mode.

As LAPAR was a ‘fly-by’ (rather than a ‘flyover’) waypoint, the aircraft’s flight guidance and control system (FGCS) pre‑empted the turn and passed by LAPAR at 1634:51, at 183 kt airspeed and at the selected altitude of 3,000 ft. The selected altitude was then wound down to 2,200 ft and 2 seconds later, the FGCS captured the localiser and the recorded flight data showed LOC and flight path angle (FPA) modes became active. One second later, the lateral mode reverted to the basic FD ROLL mode (Figure 2 No 1 and Figure 3).

Figure 3: Flight mode annunciation display showing localiser (LOC) capture followed by reversion to basic modes (ROLL/FPA)

Recorded flight data showed LOC mode became active and then disarmed within 1 second. The active mode was recorded every second, but the armed mode was recorded every 2 seconds. Hence at 1634:56, the recorded data showed LOC as both the armed and active mode. Source: Embraer animation of recorded flight data, annotated by the ATSB

The FO recalled being very surprised seeing the ROLL/FPA modes. In those modes, the aircraft captured and maintained the roll angle and flight path angle it was in at the time of activation. At that time, the aircraft was:

• aligned with the localiser
• at a flight path angle of approximately 0°
• half a dot above the glideslope
• banked 20° right.

After intercepting the localiser, the ILS frequency became the active navigation source for the aircraft’s primary flight display (PFD), for the remainder of the flight. As such, lateral deviation from the localiser and vertical deviation from the 3° glideslope would be depicted on the PFD with 2 dots in each direction (left‑right/up‑down), with 2.5 dots representing full‑scale (or greater) deviation. Localiser deviation was also depicted by a course deviation indicator (CDI) and dots either side of the course on the compass instrument on the lower part of the PFD (Figure 4).

Figure 4: Compass instrument course deviation indicator aligned with the localiser at 1634:56

Source: Embraer animation of recorded flight data, annotated by the ATSB

Following activation of the ROLL/FPA modes, the aircraft deviated right of the localiser, maintaining approximately 20° roll for about 20 seconds, as it descended. After 14 seconds, the PM selected the landing gear down and the PF moved the heading bug and selected heading (HDG) mode to command the aircraft to turn left towards the localiser. 

The CDI exceeded half scale deflection (1.5 deviation dots) at 1635:16. The aircraft was then outside the lateral flight tolerance for the ILS and below the 10 NM minimum safe altitude of 3,000 ft.   

At 1635:30, the PM contacted the Darwin aerodrome (tower) controller and advised that they were ‘just slightly right of the localiser and re‑intercepting’ (Figure 2 No 2). The controller responded with the instruction to ‘maintain 2,000 [ft] until glidepath interception’. The PM read back ‘maintain 2,000’, but not ‘until glidepath interception’. The controller then stated: ‘once you’ve intercepted the glidepath, cleared the ILS’. The PM did not respond. 

The controller later advised the ATSB that 2,000 ft was the highest minimum vector altitude around Darwin Airport, which assured terrain separation, and there was no conflicting traffic. The controller reported that although the PM did not complete the readback, they had read back the safe altitude. The controller assessed that the flight crew were ‘working really hard’ to get back onto the ILS and would let ATC know when they wanted further descent or commenced a missed approach. The captain reported that they wanted to get re‑established on the localiser so they could conduct the published missed approach under automation if required. On reaching the maximum deviation to the right of the localiser, the aircraft was (Figure 5):

• banked 27° left
• at full scale localiser deflection
• one dot above the glideslope
• at 2,470 ft above mean sea level (AMSL)
• descending at 1,254 ft/minute (fpm). 

Figure 5: Primary flight display at 1635:30 showing full‑scale localiser deviation (localiser left of the aircraft)

Source: Embraer animation of recorded flight data, annotated by the ATSB

Flap 2 was selected at 1635:55, at about 2,200 ft AMSL. The active lateral mode automatically changed from HDG to LOC mode one second later, but because the APP mode was not armed, GS mode did not become active although the aircraft was within one dot of the glideslope. The aircraft then passed through the localiser 52° off the runway heading, before entering a right turn, as the FGCS commanded re‑interception of the localiser course. 

At 1636:22, 177 kt airspeed and 2,000 ft, the PF advised the controller that ‘Qantas 1888 is established’ (Figure 2 No 3). Established was defined as being within half full‑scale deviation of the specified track.[6] The aircraft was then (Figure 6):

• at full-scale localiser deviation
• banked 33° right
• nearly 2 dots above the glideslope. 

Figure 6: Primary flight display at 1636:22 showing full‑scale localiser deviation (localiser right of the aircraft) and nearly 2 dots glideslope deviation (glideslope below the aircraft) 

Source: Embraer animation of recorded flight data, annotated by the ATSB

The PM then requested further descent, to which the controller replied, ‘cleared ILS runway 29’, and just below 2,000 ft, the PM selected flap 3. 

At 1636:51, the aircraft passed the final approach fix (FAF) TOROT (Figure 2). At the FAF, the PM was required to call out ‘FAF, height checked, missed approach altitude set’. At that time:

• the FAF procedure height was 1,330 ft and the aircraft was at 1,695 ft (AMSL)
• the aircraft was at full‑scale deviation above the glideslope and descending at 1,331 fpm
• the set altitude was 800 ft, and the missed approach altitude was 3,000 ft.

After the FAF, Alliance’s standard operating procedures required the aircraft to be within one dot of the localiser and glideslope. The procedures also stated that the aircraft should regain the 3° profile no later than 1,500 ft above aerodrome elevation. However, passing 1,500 ft, the aircraft was full‑scale deviation above the glideslope and descending at over 1,500 fpm. 

Passing about 1,100 ft AMSL, the PM intended to select flap 5, but inadvertently selected flap 4. Flap 4 had the same flap and slat extension as flap 5, but flap 4 was a take‑off configuration not a landing configuration (see the section titled Flap configuration).

At 1637:16, the aircraft passed 1,000 ft radio altitude,[7] which coincided with Alliance’s stabilisation height of 1,000 ft above aerodrome level (AAL) for conducting an instrument approach. Contrary to the stabilised approach criteria, the:

• airspeed was 162 kt, 29 kt above approach speed (V_AP) – faster than the permitted V_AP + 10
• flap setting was 4 instead of flap 5 – not in the landing configuration
• before landing checks had not been completed
• aircraft was 1.4 dots above the glideslope – not within the allowable 1 dot of the glideslope
• descent rate was 1,582 fpm – higher than the allowable rate of descent than 1,000 fpm.

Providing all other stabilisation criteria were met, Alliance permitted the airspeed to be higher than V_AP + 10 until 500 ft in visual meteorological conditions (VMC)[8] by day. 

The captain recalled that the aircraft entered VMC at about 2,500 ft. However, the FO reported that just prior to 1,000 ft, they were ready to call out ‘unstable’ approach, when the captain stated that they were now ‘visual’ and could therefore continue the approach and be stabilised by 500 ft. The captain later reported that the lower stabilisation height of 500 ft in VMC was the approved procedure at their previous company. The FO reported being uncertain about Alliance’s policy and deferred to the captain. 

Alliance procedures stated that below 1,000 ft AAL, the descent rate ‘shall not normally’ exceed 1,000 fpm. The descent rate exceeded 1,000 fpm until 1637:29, when passing 723 ft radio altitude, with the PF arming the APP mode 3 seconds later and LOC/GS becoming the active modes. The selected altitude was then set to the missed approach altitude of 3,000 ft. 

At 1637:42 and 500 ft radio altitude, contrary to the stabilisation criteria in VMC, the:

• airspeed was 17 kt above V_AP, with a maximum of V_AP + 10 permitted
• aircraft was not in the landing configuration (flap 4 was a take‑off setting). 

The PM observed that although slightly fast, the speed was trending down, and the flight crew reported thinking the aircraft met the stabilised approach criteria at 500 ft. The PM reported having completed the before landing checklist (gear and flaps) at about 800 ft, but at 411 ft radio altitude, identified that the flap lever was in the flap 4 detent and quickly moved it to the flap 5 position. The flap transitioned to the flap 5 position at 345 ft radio altitude.

The aircraft decelerated to the target approach speed V_AP (133 kt) at 264 ft radio altitude. The captain disconnected the autopilot at 212 ft radio altitude and manually flew the aircraft to an uneventful landing at 1638:32.

Context

Flight crew information 

The captain and first officer each held an air transport pilot licence (aeroplane), Embraer E190 type rating and a class 1 aviation medical certificate. 

The captain had a total flying time of 7,100 hours, 4,000 of which were on Embraer E190 aircraft, and 150 of those were accrued in the last 90 days. The captain had previously flown Embraer E120 aircraft for about 6 years and E190 aircraft for another 4–5 years with a different operator. As a direct entry captain to Alliance, the captain conducted 6 sectors and then a line check before operating for about 2 years with Alliance on the E190 aircraft. The captain was a training captain at their base and had conducted some of the first officer’s training. 

The first officer had a total flying time of 16,325 hours, 274 of which were on the Embraer E190, and 130 of those were accrued in the last 90 days. The FO commenced ground training for the Embraer E190 type rating with Alliance in May 2024 and started operating for Alliance in late September 2024. 

The captain reported having slept 7 hours the previous night, 6 hours the night before that, and assessed their fatigue as ‘Okay, somewhat fresh’. The FO reporting having slept 8.5 hours on each of the previous 2 nights and assessed their fatigue as ‘Very lively, responsive, but not at peak’.[9] There was no evidence that fatigue was a factor in this occurrence. 

The captain assessed their workload during the approach as 8/10. Their workload was increased by diverting around weather and a thunderstorm building in the vicinity of the airport, flying the descent in manual speed mode and the approach in manual modes. The FO assessed their workload during the approach as 10/10, in attempting to achieve the stabilised approach criteria. 

Approach mode disarming

VH-UYO had Load 27 Honeywell flight management system (FMS) software, which used flight director (FD) modes and non‑FMS navigation for the ILS approach. Provided the ILS approach was part of the flight plan and NAV 1 and 2 were in AUTO FMS mode, the system would transition from FMS navigation to ILS approach through the preview mode at 150 NM. The recorded flight data showed that both NAV 1 and 2 frequencies were selected to the Darwin runway 29 ILS‑Z, and the PM reported that preview mode selected as expected.

The transition from FMS to FD navigation could also be done manually by pressing the VOR/LOC (V/L) button on the guidance panel. The V/L button selects VOR or LOC as the primary navigation source for the on‑side PFD and toggles between VOR/LOC1 and VOR/LOC2. The recorded data showed that the primary navigation source was FMS1 until it changed to VOR/LOC1 at 1634:57, less than one second after LOC became active before dropping out, and remained VOR/LOC1 until after landing. 

Embraer analysed the recorded flight data and identified that when the auto‑flight control system (AFCS) captured LOC mode, the display systems had not transitioned to the LOC navigation source. In that situation, the localiser capture is invalidated, and the autopilot reverts to ROLL/FPA. Although the navigation source only recorded every 4 seconds, Embraer advised that this was the most likely reason for the mode reversion. Honeywell advised that as the recorded data rate was less than the system update rate, there was insufficient information to assess whether a synchronisation issue existed. However, Honeywell will assess whether aircraft operators have reported any similar events.  

Pressing the approach (APP) pushbutton arms approach mode if it is not armed or active, and disengages or disarms approach mode if it is armed or active. In interview several weeks after the incident, the captain reported that they pushed the APP button once to arm the approach mode after receiving ATC clearance for the ILS, and probably inadvertently pushed it again, disarming the mode, on receiving a second clearance for the ILS.

However, ATC recordings showed that the second ILS clearance was received about 50 seconds after the mode reversion and a third clearance 40 seconds later. The captain also reported having re‑armed approach mode as soon as they selected heading mode after the reversion (at about 2,700 ft), but a review of flight data identified that the approach mode was not re‑armed until 620 ft AAL. 

The approach mode armed status was recorded every 2 seconds in the flight data and was armed 1 second before LOC captured and not armed (just less than) 1 second after LOC capture. Therefore, the possible second press of the APP button would have to have occurred within 1 second of the LOC capture. 

Mode reversion

The E190 Aircraft Operations Manual described roll hold (ROLL) mode as the basic lateral mode. Depending on the bank angle at the moment of ROLL activation, the autopilot (AP) will maintain the following bank angles until another lateral mode is selected: 

• bank angle at 6° or below: AP levels the wings
• bank angle above 6° and below 35°: AP holds the present bank angle
• bank angle at or above 35°: AP maintains a bank angle of 35°

Relevant to this occurrence, roll mode is activated:

• when there is no lateral mode active and a vertical mode is selected
• by deselecting an active lateral mode. 

Roll mode is deactivated when another lateral mode is activated.

Flight path angle (FPA) is the basic vertical mode, except during take‑off. The FPA can be used for vertical navigation by selecting a higher or lower altitude and then pressing the FPA button. 

Alliance’s procedures required changes in lateral and vertical engaged modes to be announced by the pilots:

• the pilot who changes the flight mode checks on the flight mode annunciator (FMA) and verbally confirms the selected mode
• the other pilot verbally confirms the new flight mode. 

The procedures also stated that:

When the aircraft does not perform as expected, the autopilot must be disconnected and manual flight promptly established. 

Flap configuration

The E190 has slats on the leading edge of the wing and flaps on the trailing edge. There are 7 slat/flap control lever positions: up, 1–5 and full. Flap 4 is a take‑off configuration, and normal landing configurations are flap 5 or full. In flap 4 and flap 5 positions, the flaps extend to 20° and slats to 25°. Although the physical positioning of the flaps and slats is identical, the flight system logic triggers different warnings based on the selected position as follows: 

• Retard mode reduces the thrust levers to idle during flare on landing. This only occurs when the slat/flap lever position is at 5 or full and landing gear is down.
• Enhanced ground proximity warning system (EGPWS) mode 2 – excessive terrain closure rate operates mode 2A when flaps are not in the landing position and 2B with flaps down and in the landing position. Violations of the alert envelopes produce aural ‘TERRAIN TERRAIN’ and ‘PULL UP’ alerts. Mode 2B has a desensitized alert envelope with a reduced upper height limit and a higher closure rate when:
  • flaps are in the landing position
  • on an ILS, the aircraft is within +/- 2 dots of the localiser and glideslope
  • within 5 NM and 3,500 ft of the runway (and terrain awareness is functioning)
  • during the first 60 seconds after take‑off.
• EGPWS mode 3 alerts if a decrease in altitude occurs immediately after take‑off or during a go‑around when the flaps are not in a landing configuration (‘DON’T SINK’ aural and GND PROX PFD alerts).
• EGPWS mode 4 unsafe terrain clearance depends on radio altitude, airspeed, landing gear and flap position. In mode 4B, with landing gear down and flaps not in the landing configuration, pilots hear a ‘TOO LOW, FLAPS’ alert and see a GND PROX alert on the PFD, when below 159 kt airspeed and 245 ft radio altitude. Mode 4C operates when the landing gear or flaps are not in the landing configuration and will trigger a ‘TOO LOW, TERRAIN’ aural alert. 

On this occasion, had the flight crew continued towards landing with flap 4 configured, they would likely have received any or all of ‘TERRAIN TERRAIN’, ‘PULL UP’, ‘TOO LOW FLAPS’, and ‘TOO LOW TERRAIN’ aural alerts. 

Threshold and approach speeds      

The aircraft’s reference speed (V_REF) is 1.3 times the stalling speed (V_S) and is the minimum recommended speed at 50 ft over the threshold. It is used for landing distance calculation and is the speed to which the airspeed reference bugs are set before an approach. The approach speed (V_AP) exceeds the reference speed to account for gusts or windshear. Alliance’s E190 standard procedures defined V_AP = V_REF = + 1/2 steady headwind component + gust increment, limited to a minimum of V_REF + 5 kt and maximum V_REF + 20 kt. 

Meteorological conditions 

Forecast 

The Bureau of Meteorology (BoM) graphical area forecast encompassing the Darwin area, issued at 1359 and valid between 1430–2030 local time, stated:

• visibility greater than 10 km, scattered cumulus/stratocumulus clouds with bases 4,000 ft above mean sea level and tops above 10,000 ft, and bases from 2,000 ft over sea to 30 NM (56 km) inland
• 1,000 m visibility in scattered heavy showers of rain, occasional towering cumulus clouds with bases at 2,000 ft and tops above 10,000 ft
• 500 m visibility in occasional heavy thunderstorms with rain, occasional cumulonimbus clouds with bases from 2,000 ft and tops above 10,000 ft, broken stratus between 500 and 1,500 ft
• moderate turbulence below 400 ft in smoke and thermals over land until 1830. 

There was a tropical cyclone off the northern coast of Western Australia but no SIGMET[10] for the Northern Territory. 

The Darwin Airport forecast (TAF)[11] issued at 0851 was valid for 30 hours from 0930 on 12 February until 1530 on 13 February. Two subsequent TAF amendments were issued before the occurrence, one at 1148 and another at 1452. All TAF cloud heights are above aerodrome elevation. All had the same forecast for:

• wind from 340° (True) at 12 kt
• visibility greater than 10 km
• light showers of rain
• scattered cloud at 2,000 ft. 

For the period of the aircraft’s approach, 1614–1638, the forecast and amendments included (intermittent) periods of up to 30 minutes of:

• visibility reducing to 1,000 m in heavy showers of rain
• broken cloud at 500 ft
• scattered towering cumulus with bases at 2,000 ft.

Forecast to occur from 2330 on 11 April to 1330 on 12 April, there was a 30% probability of thunderstorms for periods up to 1 hour with:

• variable direction winds of 15 kt winds gusting to 30 kt
• visibility reducing to 500 m in thunderstorms and heavy rain
• broken cloud at 400 ft
• scattered cumulonimbus clouds with bases at 2,000 ft. 

A third TAF amendment was issued at 1642 (4 minutes after the aircraft landed), which included an additional intermittent period lasting up to 30 minutes of: 

• visibility reducing to 500 m in thunderstorms and heavy rain
• broken cloud at 400 ft
• scattered cumulonimbus clouds with bases at 2,000 ft. 

Observations 

The following weather observations at Darwin Airport occurred during the time of the aircraft’s approach.

At 1617, a special report of meteorological conditions (SPECI)[12] was issued, which included: 

• 3,000 m visibility
• thunderstorms in the vicinity (lightning detected outside the 8 km radius of the airport)
• few cloud at 1,500 ft
• broken cloud at 2,600 ft
• few cumulonimbus clouds at 2,000 ft.

At 1630, a SPECI was issued that included: 

• 4,000 m visibility
• heavy showers of rain
• few cloud at 1,500 ft
• scattered cloud at 1,900 ft
• broken cloud at 4,000 ft
• detail that 2 mm of rain had fallen in the previous 10 minutes.

At 1637, one minute prior to landing, a SPECI was issued that included:

• 8,000 m visibility
• thunderstorms and rain
• scattered cloud at 1,300 ft
• broken cloud at 3,600 ft
• few cumulonimbus clouds with bases at 2,000 ft
• advice that 0.6 mm of rain had fallen in the previous 10 minutes.

Radar imagery

The BoM Darwin/Berrimah radar was located 7 km south‑east of Darwin Airport. 

Radar images showed moderate rain in the Darwin Airport area at 1614, that reduced to light to no rain around the time of the approach. The associated cloud/cell was stationary, and according to the BoM, it was not unusual to have a stationary rain/storm sitting over the airport in the wet season. The BoM information about that radar included: 

Heavy rain over the radar site will cause attenuation of all signals. Path attenuation also occurs when the radar beam passes through an intense thunderstorm cell; the returned signal from cells further along that path will be reduced...it may 'undershoot' high level storms and rain echoes may appear less intense than actual rainfall rate.

The BoM advised that due to attenuation, the light rainfall indicating on the radar image at 1634, may have been less intense than the actual rainfall rate. 

Webcam

The BoM provided webcam images for each minute from 1634–1638 facing north and east. 

At 1638, the aircraft would have been due east of the webcam on final approach, at about 300 ft, probably the light visible in the east view at 1638 (Figure 7).

Figure 7: BoM webcam image facing east at 1638 and probable aircraft light 

Source: Bureau of Meteorology, annotated by the ATSB

The cell with heavier rain to the north was evident in the north view at 1638 (Figure 8).

Figure 8: BoM webcam image facing north at 1638 

Source: Bureau of Meteorology

Figure 9 shows satellite imagery of cloud cover at Darwin Airport at 1640, and lightning strikes that occurred in the 10 minutes prior (red) and during the period 10–30 minutes prior (orange).

Figure 9: Cloud cover at 1640 and recorded lightning strikes 

Source: Satellite image originally processed by the Bureau of Meteorology from the geostationary meteorological satellite Himawari-9 operated by the Japan Meteorological Agency. Lightning data sourced from Weather Zone Lightning Network

Automatic terminal information service 

The Royal Australian Air Force’s Darwin Airport automatic terminal information service (ATIS)[13] X‑ray (X) was issued at 1610 and included advice to expect an instrument approach and that runways 29 and 36 were active for arrivals, with runway 29 in use for departures. The runways were wet, with surface condition code 555. ATIS X‑ray also detailed the following weather conditions:

• wind from 340° at 15 kt, maximum 15 kt crosswind on runway 29
• visibility 2 km
• showers of rain
• cloud scattered at 1,200 ft
• temperature 30°C
• QNH 1007.[14]

ATIS Yankee (Y) came into effect at 1624, with the only change being a temperature reduction to 27°C. 

Based on the forecast and observed weather conditions at the time of the approach, the flight crew could not have anticipated VMC or expected to remain clear of cloud until below about 1,200 ft.

Recorded data

Flight data from the aircraft’s quick access recorder (QAR) for the incident flight was analysed by the ATSB and Embraer. 

Key parameters in the vertical flight path, depicted in Figure 10, identified that between LAPAR and about 600 ft AAL, there were:

• 7 vertical mode changes (purple)
• 11 selected altitude changes (cyan)
• significant variations in flight path angle (orange)
• significant variations in vertical speed (pink) and rates of descent exceeding 1,600 fpm including below 1,000 ft AAL.

Figure 10: QAR data including key vertical flight path parameters 

Source: Alliance Airlines QAR data, analysed by the ATSB

Although cockpit localiser and glideslope indications showed full‑scale deviation of 2.5 dots on the primary flight display and compass instrument, the recorded QAR data captured up to 5 dots deviation. Key parameters depicted in Figure 11 showed: 

• airspeed (red) remained above the target approach speed of 133 kt (V_AP) (purple) until 1638 (deviation in kt – green)
• 5 dot left and right deviations from the localiser course between LAPAR and TOROT
• 3 dot deviation above the glideslope approaching TOROT.

Figure 11: QAR data depicting speeds and key lateral flight path parameters 

Source: Alliance Airlines QAR data, analysed by the ATSB

Approach briefing

Alliance procedures required flight crew to conduct a briefing before commencing an approach. It detailed the requirements for an instrument approach briefing and a visual approach briefing. There was no policy on re‑briefing during an instrument approach if a transition to a visual approach occurred. Alliance advised that the expectation was for flight crew to continue an instrument approach even if visual reference was established. 

Descent below lowest/minimum safe altitude 

Alliance’s Operations Policy and Procedures Manual (OPPM) required all flights to be planned and conducted under instrument flight rules (IFR).[15] Company policy allowed the use of visual approaches and departures, but pilots were not permitted to downgrade to visual flight rules.

The OPPM section 7.4.11.1 stated that operation below the lowest/minimum safe altitude (LSALT/MSA) was permissible only when:

• under radar control;

• in accordance with a published [distance measuring equipment] DME arrival instrument approach or holding procedure;

• when necessary during climbing after departure from an airport, or

• when flying in VMC by day. 

CASR Part 91 Plain English Guide stated that in accordance with sub‑regulation 91.305 Minimum heights – IFR flights, aircraft must not be flown below the lowest/minimum safe altitude except when taking off or landing in VMC by day, or in accordance with: 

• a published visual or instrument approach or departure procedure
• an air traffic control clearance. 

Instrument landing system approach 

Darwin Airport runway 29 ILS-Z approach is depicted in Figure 12.

Figure 12: Darwin ILS-Z approach 

Source: Jeppesen

The Aeronautical Information Publication (AIP)[16] stated that:

Unless authorised to make a visual approach, an IFR flight must conform to the published instrument approach procedure nominated by ATC.

During an instrument approach, flight crew were required to maintain the aircraft’s flight path within certain flight tolerances. For an ILS approach the AIP stated:

Pilots must conform to the following flight tolerances:

a) To ensure obstacle clearance, both [localiser/Ground based augmentation system (GBAS) landing system] LOC/GLS final approach course and glideslope should be maintained within half scale deflection (or equivalent on expanded scale).

b) If, at any time during the approach after the [final approach point] FAP, the LOC/GLS final approach or glideslope indicates full scale deflection, a missed approach should be commenced.

Although the AIP wording combined the terms ‘must’ and ‘should’, Airservices Australia confirmed that aircraft were required to comply with the vertical guidance of the glideslope when conducting an ILS approach, and that descent outside this vertical guidance could be safety critical. 

Alliance’s Standard Operating Procedures Manual (SOPM) detailed the procedure for conducting an ILS approach. This included:

The standard profile assumes the aircraft will approach 3,000 ft AFE [above field elevation] with Flap 1 selected and will be configured with landing flap with checklists complete prior to stabilization altitude. 

Standard calls were documented in the Alliance SOPM, which stated: 

Deviation calls are to be made if the listed deviation limit is exceeded and no corrective action has been observed. Upon acknowledgement of a deviation, corrective action must be taken.

Table 1 is an extract of the standard calls relevant to this approach. 

Table 1: Selected standard calls 

Source: Alliance Airlines 

Alliance’s OPPM section Instrument approaches included:

Any time the PNF [pilot not flying]/PM calls deviations from 'on slope' the PF should make corrections to avoid flight path excursions towards full scale. 

The PNF/PM should continue slope deviation calls until the glideslope indicator stops moving toward full scale and whenever the indicator is at full scale.

…

When continuing the approach, continually cross-check visual profile indications against glideslope profile indications down to 100 ft AGL…Duties of the PNF/PM apply on all instrument approaches through to 100 ft above threshold height, even if visual flight conditions are encountered before reaching the minimum.

The same section stated:

During a visual approach using the ILS, the glideslope calls do not need to be given.

Visual approaches

The term ‘visual’ was used by: 

• ATC to instruct a pilot
• a pilot to accept responsibility

to see and avoid obstacles while operating below the minimum vector altitude or minimum/lowest safe altitude. 

Alliance permitted flight crew to conduct visual approaches by day and night in accordance with AIP requirements. The AIP required an IFR flight to conform to the published instrument approach procedure unless they were authorised by ATC to make a visual approach. By day, ATC could authorise an IFR aeroplane to conduct a visual approach when:

• the aircraft is within 30 NM of the aerodrome
• the pilot has established and can continue flight to the aerodrome with continuous visual reference to the ground or water
• visibility along the flight path is not less than 5,000 m or the aerodrome is in sight. 

If these conditions existed, the AIP stated:

the pilot need not commence or may discontinue the approved instrument approach procedure to that aerodrome…

In controlled airspace, an ATC clearance was required to conduct a visual approach. The pilot was required to report ‘visual’ to signify to ATC that the visual approach requirements could be met and maintained as part of any request for a visual approach. The pilot was then required to maintain the track or heading authorised by ATC until (by day) within 5 NM (9 km) of the aerodrome. 

The Alliance OPPM stated that in VMC on a visual approach, the aircraft must join the circuit on the upwind, crosswind or downwind leg, or make a straight-in approach after establishing on final approach by 5 NM. During this occurrence, when 5 NM from the airport, the aircraft was about 1 km left of the extended runway centreline.

The required visual approach callouts at 500 ft stabilisation height were:

• if stabilised criteria satisfied: PM verifies or calls out ‘500 STABLE’
• otherwise: PM verifies or calls out ‘500 NOT STABLE’ and the PF initiates a go‑around. 

The flight crew had not briefed for a visual approach, advised ATC they were visual or received clearance to conduct a visual approach. 

Go-around and discontinued approach 

The Alliance OPPM included Non punitive go around policies, and listed conditions which, if encountered, the PF should consider carrying out a go‑around. The conditions included those that could lead to stabilised approach requirements not being adhered to. The Alliance SOPM distinguished between a discontinued approach and a go‑around in which take‑off/go‑around (TOGA) thrust was applied.

SOPM 5.20.4 Discontinued approach included: 

During the initial phase of the approach, there may be a situation where the approach needs to be discontinued. If the aircraft is at or near the missed approach altitude, far from the missed approach point and not fully configured for landing, a go around procedure may not be appropriate. The go around can lead to an excess thrust that may result in overshooting the missed approach altitude. For this situation, a discontinued approach is recommended.

NOTE

❖ A Go-around should be conducted:

o In Day VMC below 500FT AFE,

o In [instrument meteorological conditions] IMC[17] or at night below 1000ft AFE.

SOPM 5.20.3 Go-around included: 

No approach should be initiated unless the prevailing conditions have been understood and the crew found that landing is acceptable without undue risk. Philosophically all approaches should be treated as approaches followed by missed approaches, and landing should be treated as the alternate procedure. This mindset depends on a good approach briefing, on the knowledge of the missed approach procedure and on proper programming of the FMS.

Alliance advised that in this incident, a discontinued approach at 3,000 ft (when the approach mode disarmed) would have been appropriate. 

Stabilised approach criteria

A stabilised approach is one where an aircraft maintains a constant angle descent to the runway while other key flight parameters such as airspeed and aircraft configuration are controlled within specific ranges. An approach is stable when all the stabilisation criteria specified by the operator are met.

According to an International Air Transport Association report Unstable approaches: risk mitigation policies, procedures and best practices (IATA, 2017), historical commercial aviation accident data indicated that many accidents occurred during the approach and landing phases of flight. Frequent contributing factors were an unstable approach and subsequent failure to initiate a go‑around. Failure to maintain a stable approach could result in landing too fast or too far down the runway, a hard landing, runway excursion, loss of control, or collision with terrain. The report also highlighted the importance of callouts in enhancing situational awareness and encouraging rapid error correction. 

The report described an operator’s minimum stabilisation height as a ‘gate’ at which if the aircraft was not stable on the approach path in the landing configuration, a go‑around must be executed. Additionally, although many operators have one gate for IMC and a lower gate for VMC, variations in stabilisation heights between operators, approach types and meteorological conditions (VMC/IMC) could cause confusion. As a result, many airlines implemented a single set of criteria and one gate for a particular approach type, such as 1,000 ft for an instrument approach, and 500 ft for visual circuit or circling approaches. Having a single gate also makes it easier for an operator to track compliance using flight data monitoring programs.

The Flight Safety Foundation’s Approach and landing accident reduction briefing note 7.1 – Stabilized Approach described the benefits of a stabilised approach as:

• increasing the flight crew’s situational awareness of the:
  • horizontal
  • vertical
  • airspeed
  • energy state
• more time and attention for monitoring communications, weather and aircraft systems
• more time for monitoring by the PM
• defined criteria to support land or go‑around decision‑making
• consistent landing performance. 

Alliance’s standard operating procedures required all flights conducting instrument approaches to be stabilised by 1,000 ft above aerodrome level, and an immediate go‑around was required for any approach that did not meet the following stabilised approach criteria: 

a) the correct flight path;

b) only small changes in heading/pitch are required to maintain the correct flight path;

c) the aircraft speed is not more than V_APP + 10 knots indicated airspeed and not less than V_REF;

d) the aircraft is in the correct landing configuration;

e) sink rate is no greater than 1,000 feet per minute

f) thrust or power setting is appropriate for the aircraft configuration;

g) all briefings and checklists have been completed;

h) specific types of approaches are stabilized if they also fulfil the following

i. instrument landing system (ILS) approaches must be flown within one dot of the glideslope and localizer

ii. a Category II or Category III ILS approach must be flown within the expanded localizer band

i) unique approach procedures or abnormal conditions requiring a deviation from the above elements of a stabilized approach require a special briefing to have been completed prior to beginning the approach.

• Note 1: A momentary excursion is permitted for points (c) & (e). A momentary excursion is defined as a deviation lasting only a few seconds and where every indication is that it will return to the stabilised criteria as listed in points (c) & (e).

• Note 2: Where the nominal descent path for a particular approach requires a descent rate greater than 1000 fpm. This is only permitted when expected rates of descent have been briefed prior to the approach being commenced. 

Stabilized Heights 

All flights shall meet all of the above stabilized approach criteria by 1,000 feet above aerodrome level except under the following circumstances: 

Visual approach: 

Speed may be higher than VAPP + 10, provided it is within limits and expected to reduce to Vapp+10 or below by no later than 500ft AAL.

Note 3: Visual conditions as defined by Jeppesen AUS or AIP - the pilot has established and can continue flight to the airport with continuous visual reference to the ground or water; and visibility along the flight path is not less than 5000m.

…[followed by Visual circuit, Circling approach and RNP-AR approach]

An approach that does not meet, or subsequently exhibits sustained deviations outside of these criteria requires an immediate go-around.

Alliance advised that the Note 3 under the Visual approach heading was supposed to clarify that the term ‘visual approach’ and the associated policy was not only for the situation where flight crew were cleared for a visual approach, but included when they encountered visual conditions during an instrument approach. 

Alliance SOPM section Intercepting glideslope from above, included:

Several different situations, such as ATC restriction, may lead to a glideslope interception from above. If that happens, the pilots must take the appropriate actions to guarantee a stabilized approach. If the stabilized approach criteria are not met, the PF must initiate a Go Around. The approach must be stable before reaching 1000 ft AGL (IMC), 500 ft (VMC), or other altitude in accordance with company policies. 

Alliance’s SOPM (5.16.7) Stabilised approach stated that the aircraft must meet the stabilised approach criteria in the OPPM, and included: 

For a 3D instrument approach or a visual approach (except for a visual circuit or circling approach) the aircraft should be established on profile by 3000ft AFE. If, due to operational circumstances the aircraft is not on a 3° profile, an acceptable flight path should be maintained to regain profile no later than 1500ft AFE. Descent rate limits are outlined in the OPPM.

The OPPM specified ‘Descent rate limits’: 

The following values for the rate of descent below the transition altitude shall not normally be exceeded:

• 3000 fpm down to an altitude of 3000 ft above aerodrome level (AAL).

• 2000 fpm down to an altitude of 2000 ft AAL transitioning to 1000 ft AAL

• 1000 fpm below 1000 feet AAL

The SOPM (5.16.7) Stabilised approach also stated: 

On the normal profile, the aircraft should approach 3,000 ft AFE with Flap 1 selected. Flap 2 should be selected leaving 3000ft. The landing gear should be selected down and flap 3 selected no later than 2000ft AFE. Final landing flap should be selected such that all stable approach criteria can be satisfied.

Alliance advised that in the previous 4 years, almost all their approach incidents occurred following selection of flap 2 below the prescribed 3,000 ft. They reported that delayed selection of flap 2 ‘then compresses everything’.

Guidance material for Part 121.200 from CASR Part 121 Acceptable means of compliance/guidance material - Australian air transport operations—larger aeroplanes included discontinuing an approach to continue in VMC as one situation that reduced the likelihood of a stabilised approach and should be avoided where not operationally necessary.

Crew coordination 

The captain reported being aware when the aircraft was outside half scale deflection of the localiser and more than half scale deflection above the glideslope, which was why they advised ATC and were cleared to maintain the minimum vector altitude. The captain could not recall whether they had advised ATC they were visual but assessed that they could continue the unstable approach below 1,000 ft as they were in visual conditions. The captain reported that there were no callouts from the PM and if they were not stable at 500 ft, the captain would have expected an ‘unstable’ callout from the PM and would have conducted a go‑around. 

The FO (PM) reported that normally they would aim to be fully configured and stable by 1,500 ft and the ‘absolute latest’ by 1,000 ft for an instrument approach. In this case they were not, which is why the captain called ‘visual’ and said they would use the visual approach ‘gate’ of 500 ft. The FO reported that in that split second (passing 1,000 ft), they agreed with the captain to continue because they were visual, although the aircraft was above the glideslope and ‘slightly’ fast, the speed was trending down, and the FO thought that the aircraft was fully configured for landing. The FO assessed that the aircraft was stable at 500 ft, unaware that the flaps were incorrectly configured, and had they called ‘unstable’ at 500 ft, the captain would have been obliged to conduct a go‑around.  

The FO also reported that the en route diversions and storms near the airport may have resulted in perceived time pressure to continue the landing. They had sufficient fuel to discontinue the approach and make a second attempt. 

The cockpit voice recording was not retained for the investigation. The flight crew could not recall whether the PF called out mode changes after the mode reversion. The PM reported calling distance and heights for the profile, but did not make speed, slope or track deviation calls because the PF was taking corrective action. 

The Alliance OPPM detailed task sharing and the responsibilities of the PF and PM and noted that this was particularly important in high workload phases of flight, including approach and landing. It also stated that the PM should:

query the PF actions that are not understood or considered inappropriate. He/she should also demonstrate assertiveness and express advocacy to share any concern on the flight progress. 

Alliance OPPM defined authority gradient as ‘the relative authority of air crew in the chain of command … if the gradient is too steep, air crew may be unwilling or unable to express their beliefs to those in higher authority’. 

Similar occurrences 

Unstable approach involving Embraer 190, VH‑UZI, about 4 km north‑east of Brisbane Airport, Queensland, on 9 May 2024 (AO‑2024‑030)

During approach to Brisbane Airport, the aircraft automated ILS flight mode unexpectedly disengaged. The flight crew focused on troubleshooting the unexpected change and recapturing the ILS flight director mode, rather than conducting a go‑around. During that time, the flight crew did not effectively monitor the aircraft's flight path, and the aircraft exceeded the stabilised approach criterion of one dot glideslope deviation. 

After recognising that the aircraft was low, the captain increased the aircraft pitch, resulting in an enhanced ground proximity warning system (EGPWS) glideslope warning. The flight crew did not perform the required terrain avoidance manoeuvre, instead continued the approach. The captain arrested the aircraft’s descent and re‑established the aircraft on the glidepath, before continuing the approach and landing.

Alliance took safety action in response to that occurrence including issuing an operational notice to remind flight crew of the stabilised approach criteria and go‑around requirements. 

Descent below glideslope involving Fairchild SA227, VH‑VEU, about 17 km north‑east of Brisbane Airport, Queensland, on 2 July 2024 (AO‑2024-040)

During the ILS approach, the aircraft descended below the 3° glideslope, triggering an air traffic control ATC minimum safe altitude warning about 8 NM (14 km) from the runway. ATC advised the crew that they were observed below the glideslope, however the aircraft continued descent below the glideslope until 3 NM (6 km) when the descent rate was reduced. The aircraft then passed above the glideslope before the rate of descent increased again and subsequently the glideslope was re‑intercepted from above 1 NM (2 km) from the runway at 500 ft. The aircraft then followed a stabilised flight path to landing.

The investigation found that the pilot monitoring was not monitoring the glideslope and did not challenge the pilot flying to correct the deviation and reduce the aircraft’s descent rate. Additionally, the operator's standard operating procedures contained areas of inconsistency when an aircraft entered visual conditions during an instrument approach, and that the AIP was unclear as to whether pilots were required to comply with precision approach flight tolerances.

The aircraft operator subsequently amended its standard operating procedures as follows:

• the instrument approach procedure has been updated:
  • the approach brief now requires discussion of expectations if visual conditions arise
  • the statement that ‘during a visual approach using the ILS, the glideslope calls do not need to be given’ has been removed
  • a requirement to make callouts using reference to visual slope indications has been added.
• a note has been added to the visual approach procedures stating that crew require a clearance to discontinue an instrument approach in controlled airspace
• increased focus on pilot monitoring skills during:
  • proficiency checks, which will now include standard instrument departure and standard arrival routes
  • line training for new flight crew
• addition of a pilot monitoring sector to the annual line check.

Safety analysis

Mode reversion

Nearing the initial approach fix (LAPAR) for the Darwin Airport runway 29 instrument landing system (ILS)‑Z approach, the aircraft’s approach, localiser and glideslope modes were armed, with the correct ILS frequency set as the armed navigation source. As the aircraft flew by LAPAR, the localiser mode engaged for one second before the auto‑flight system reverted to lateral roll and vertical flight path angle modes. 

It is possible that this occurred due to the captain inadvertently pressing the approach pushbutton at the same time the localiser captured. The captain recalled that this occurred because they received a second ATC clearance for the ILS, however, at that time, they had only been cleared for the ILS once. They received a second clearance about 50 seconds later and a third clearance 40 seconds after that. The pushbutton was not recorded in the data but pressing it when approach mode was armed would disarm approach mode consistent with the mode reversion. 

Embraer's analysis of the recorded data found that the ILS navigation display source had not engaged at the time localiser mode became active. Therefore, it was most likely that this non‑synchronisation led the flight control and guidance system to invalidate the capture and revert to basic modes. The avionics manufacturer, Honeywell, advised that as the recorded data did not update as frequently as the system status, it was not possible to determine whether a synchronisation issue existed.  

The mode reversion resulted in the aircraft maintaining the 20° right bank and approximately level flight path angle that were present at the time of the mode reversion. 

Automation and approach continuation

Alliance’s procedures stated that if the auto‑flight system was not doing what the pilot expected, they were to disconnect the autopilot and manually fly the aircraft. However, following the unexpected mode change, the captain, who was the pilot flying (PF), did not re‑engage approach mode or disconnect the autopilot until about 200 ft above the aerodrome elevation, in accordance with normal landing procedures. This resulted in the aircraft deviating beyond the permitted tolerance from the localiser course and above the glideslope.

It took less than 3 minutes from the mode reversion at 3,000 ft, to the aircraft being established on the ILS and in the landing configuration, about 400 ft above aerodrome elevation. During that time, there were several changes to the lateral and vertical modes, selected heading and altitude, and significant variations in vertical speed, as the PF manipulated the auto‑flight system to recapture the localiser and glideslope. 

There were also several triggers for the flight crew to discontinue the approach. The flight crew reported that their en route weather deviations and thunderstorms in the vicinity of the airport may have led to perceived pressure to continue the approach, but that there was adequate fuel on board to discontinue and conduct a second approach. 

Once outside the lateral tolerance of the approach, the aircraft was operating contrary to the air traffic control (ATC) clearance and descended below the minimum safe altitude, likely while in instrument meteorological conditions (IMC). The pilot monitoring (PM) reported that they did not call out the course deviations, as the PF was actively working to correct them. Instead, after descending about 500 ft, the PM advised ATC they were re‑intercepting the localiser and was assigned the minimum vector altitude to ensure terrain separation. As a result, the aircraft then maintained altitude, while the glideslope diverged beneath the flight path. 

The PM did not read back the ATC clearance to descend once they had intercepted the glideslope, consistent with being focused on monitoring tasks and their self‑assessed 10/10 workload during the approach. The PM selected flap 2 as the aircraft captured the localiser, about 800 ft below the normal (but not mandatory) 3,000 ft flap 2 selection height. Alliance reported that this late selection had a similar effect to other recent approach incidents, where the associated delayed speed reduction reduced the time available to stabilise the approach. 

As the aircraft passed the final approach fix (FAF) and then descended through 1,500 ft, it was established on the localiser but 3 dots above the ILS glideslope. Alliance’s procedures required the ILS to be flown within 1 dot of the glideslope, and stated that aircraft ‘should be’ on the 3° glideslope no later than 1,500 ft. 

The PF reported stating to the PM that they were visual at about 2,500 ft, but the PM reported this occurred just as the aircraft approached 1,000 ft. Without a cockpit voice recording, it could not be determined when visual flight conditions were established. 

At 1,000 ft, which was the stabilisation height for an instrument approach, the PF incorrectly assessed that although several stabilised criteria were not met, as they were then in visual conditions, they could continue (unstable) to 500 ft. Although this had been the procedure at the PF’s former company, Alliance required all stabilised approach criteria, other than airspeed, to be met at 1,000 ft. The PM was sufficiently unsure of Alliance’s policy to defer to the captain’s experience, as the senior base training captain who had conducted some of the PM’s training. 

At 500 ft, the PM assessed that the stabilised approach criteria were met as, although the airspeed was still slightly fast, it was trending down, and they continued the approach. Although the flight crew reported having completed the before landing checks prior to 500 ft, shortly afterwards, the PM identified that they had inadvertently selected a take‑off flap setting (passing about 1,100 ft) instead of a landing setting, which they had not detected earlier. The PM rectified the flap setting and an uneventful landing followed. 

Although the landing was uneventful, had the incorrect flap setting not been rectified, thrust retard mode would not have engaged during the landing flare, and EGPWS warnings would have been triggered by the use of different tolerances than with the aircraft configured for landing. The use of selected heading and selected altitude to drive the aircraft’s trajectory meant that these may not have been set correctly had a go‑around been initiated either by the flight crew or ATC. 

A coupled ILS approach should require minimal flight crew intervention other than monitoring. Continuing an approach using inappropriate modes increased the likelihood of an unstable approach. International research showed that unstable approaches and failure to initiate a go-around could result in landing too fast or too far down the runway, a hard landing, runway excursion, loss of control, or collision with terrain. Although visual conditions reduce the risk of collision with terrain, they do not mitigate against landing incidents resulting from poor energy management. 

Alliance stabilisation heights

Alliance provided stabilised approach criteria consistent with international and Civil Aviation Safety Authority guidance, including airspeed, flight path and energy management parameters, to reduce the risk of landing accidents. However, Alliance’s minimum stabilisation heights at which these criteria were required to be met, varied between approach types and meteorological conditions, and were not clearly documented. 

Alliance specified a stabilisation height of 1,000 ft for instrument approaches, but under the heading Visual approach, permitted a 500 ft stabilisation height for an instrument approach in VMC, provided only the airspeed exceeded the stable speed criterion, and all other stabilised approach criteria were met. Additionally, elsewhere in the procedures for intercepting a glideslope, it stated that an approach must be stable before reaching 1,000 ft in IMC, 500 ft in VMC ‘or other altitude in accordance with company policies’. Alliance did not have a policy for transitioning from an instrument approach to a visual approach and advised that its expectation was for flight crew to continue an instrument approach even when entering VMC.

In this incident, the lack of clarity regarding which of the stabilised approach criteria were not required before continuing to 500 ft, resulted in the flight crew incorrectly assessing that they could continue to 500 ft in VMC while the criteria were not met for:

• speed
• rate of descent
• glideslope
• landing configuration
• before landing checks complete.  

International Air Transport Association guidance stated that using a single stabilisation height for one type of approach, regardless of weather conditions, can reduce confusion and make it easier for operators to track stabilised approach compliance using flight data monitoring. Clear stabilisation heights also support flight crew decision-making to initiate a go‑around. 

Findings

From the evidence available, the following findings are made with respect to the unstable approaching involving Embraer E190, VH-UYO, near Darwin Airport, Northern Territory, on 15 February 2025.

Contributing factors

• On crossing the initial approach fix for the instrument landing system (ILS) approach, due either to a system synchronisation issue or the pilot flying inadvertently disarming the approach mode, the aircraft’s auto‑flight system reverted to roll and flight path angle modes.
• Following the unexpected mode change, the pilot flying did not re‑engage approach mode or disconnect the autopilot. This likely contributed to the aircraft deviating outside the required lateral tolerance of the approach below the minimum safe altitude while in instrument meteorological conditions.
• The flight crew did not discontinue the approach when the aircraft was unstable at the 1,000 ft stabilisation height as they incorrectly assessed that they could continue to 500 ft in visual meteorological conditions, with multiple stabilised approach criteria unmet.
• In the limited time available to stabilise the aircraft by 500 ft, the flight crew incorrectly assessed that the aircraft was stable and continued the approach, unaware that the pilot monitoring had inadvertently selected an incorrect flap configuration.
• Alliance Airlines' standard operating procedures were unclear about the criteria for continuing an unstable instrument approach to 500 ft when aircraft entered visual conditions.

Safety actions

Safety action by Alliance Airlines 

Alliance Airlines issued an operations notice ‘to improve clarity and compliance’ with the stabilised approach criteria. It detailed the stabilised approach policy. It also amended the stabilisation height such that for 3‑dimensional and 2‑dimensional instrument approaches, and straight‑in visual approaches, the stabilised criteria were to be met by 1,000 ft above aerodrome elevation.

The 500 ft stabilisation height applied only to visual circuit or circling manoeuvre approaches. The notice reminded flight crew of Alliance’s ‘non punitive go around policy’ and required all unstable approaches to be reported.

Alliance Airlines also conducted a flight data review of unstable approaches over the previous 6 months to identify similar occurrences.

Sources and submissions

Sources of information

The sources of information during the investigation included:

• the flight crew and air traffic controllers
• Alliance Airlines
• Airservices Australia
• Bureau of Meteorology
• Embraer
• Honeywell
• recorded flight data.  

References

Civil Aviation Safety Authority (2023, December 20). CASR Part 121 Acceptable means of compliance/guidance material - Australian air transport operations—larger aeroplanes. Retrieved May 15, 2025, from Part 121 of CASR Australian air transport operations - larger aeroplanes | Civil Aviation Safety Authority

Civil Aviation Safety Authority (2025, January). CASR Part 91 Plain English Guide version 4.2. Retrieved May 15, 2025, from Part 91 plain English guide version 4.2 

Flight Safety Foundation (2000). Approach and landing accident reduction Briefing Note 7.1: Stabilized Approach. Retrieved May 15, 2025, from FSF ALAR Briefing Note 7.1: Stabilized Approach 

Flight Safety Foundation (2015, June 8). Normalization of Deviance. Retrieved May 15, 2025, from Normalization of Deviance - Flight Safety Foundation

International Air Transport Association (2017). Unstable approaches: risk mitigation policies, procedures and best practices. Retrieved May 15, 2025, from Unstable approaches: risk mitigation policies, procedures and best practices 

Submissions

Under section 26 of the Transport Safety Investigation Act 2003, the ATSB may provide a draft report, on a confidential basis, to any person whom the ATSB considers appropriate. That section allows a person receiving a draft report to make submissions to the ATSB about the draft report. 

A draft of this report was provided to the following directly involved parties:

• the flight crew and air traffic controllers
• Alliance Airlines
• Airservices Australia
• Civil Aviation Safety Authority
• Bureau of Meteorology
• Embraer
• Honeywell
• United States National Transportation Safety Board
• Brazil Aviation Accident Investigation and Prevention Center.

Submissions were received from:

• Airservices Australia
• Alliance Airlines
• the first officer
• Embraer.

The submissions were reviewed and, where considered appropriate, the text of the report was amended accordingly.

Purpose of safety investigations The objective of a safety investigation is to enhance transport safety. This is done through:  identifying safety issues and facilitating safety action to address those issues providing information about occurrences and their associated safety factors to facilitate learning within the transport industry. It is not a function of the ATSB to apportion blame or provide a means for determining liability. At the same time, an investigation report must include factual material of sufficient weight to support the analysis and findings. At all times the ATSB endeavours to balance the use of material that could imply adverse comment with the need to properly explain what happened, and why, in a fair and unbiased manner. The ATSB does not investigate for the purpose of taking administrative, regulatory or criminal action. Terminology An explanation of terminology used in ATSB investigation reports is available here. This includes terms such as occurrence, contributing factor, other factor that increased risk, and safety issue. Publishing information Released in accordance with section 25 of the Transport Safety Investigation Act 2003 Published by: Australian Transport Safety Bureau © Commonwealth of Australia 2025 Ownership of intellectual property rights in this publication Unless otherwise noted, copyright (and any other intellectual property rights, if any) in this report publication is owned by the Commonwealth of Australia. Creative Commons licence With the exception of the Commonwealth Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this report is licensed under a Creative Commons Attribution 4.0 International licence. The CC BY 4.0 licence enables you to distribute, remix, adapt, and build upon our material in any medium or format, so long as attribution is given to the Australian Transport Safety Bureau.  Copyright in material obtained from other agencies, private individuals or organisations, belongs to those agencies, individuals or organisations. Where you wish to use their material, you will need to contact them directly.

Investigation summary

What happened

On the morning of 25 February 2025, an Agusta A109E helicopter was conducting a marine pilot transfer operation on the inbound bulk carrier Star Coral at Blossom Bank pilot boarding ground, about 200 km north‑east of Mackay, Queensland.

At 0901 local time, during take‑off from the ship with 2 pilots on board, the helicopter developed severe vibrations. The pilots discontinued the take-off but their attempts to recover control of the helicopter were unsuccessful. The helicopter came to rest in an upright position on the helideck, having spun more than 90° counterclockwise from its initial heading, and sustaining substantial damage. The pilots and ship’s crew were unharmed.

What the ATSB found

The investigation did not identify any airworthiness issues with the helicopter and it was considered that the loss of control was not attributable to a mechanical issue. 

The ATSB found that the vibration was likely the result of the helicopter entering ground resonance, a phenomenon that dissipates when airborne, while it was in the process of departing from the ship. The discontinuation of the take‑off, after the onset of the vibration, probably resulted in the loss of control and subsequent damage to the helicopter.

What has been done as a result

The operator has added new guidelines on ground resonance to its procedures. The guidelines include procedures for recognising and recovering from ground resonance and feature case studies and video resources for training purposes. 

The operator has also developed an updated procedure for training and checking flight briefings that will include confirming the roles of each pilot, procedures for transferring aircraft control between pilots, and actions to be followed in the event of an actual emergency.

Safety message

The occurrence highlights the dangers of ground resonance, a potentially catastrophic phenomenon that can occur in helicopters with fully articulated rotor systems. Typically, the onset of ground resonance is sudden and if the pilot does not take immediate corrective action, a loss of control can rapidly occur. 

The occurrence also highlights the importance of proper coordination between a helicopter’s pilots when responding to abnormal or emergency situations. This is particularly pertinent for situations where the pilot flying is not the pilot in command. Ideally, the pilots’ individual roles and responsibilities for emergency response and flying duties should be well established prior to the flight. 

The investigation

The occurrence

At about 0730 local time on 25 February 2025, the 229 m bulk carrier Star Coral arrived at the Blossom Bank pilot boarding ground, about 200 km north‑east of Mackay, Queensland (Figure 1). The ship waited to embark a coastal marine pilot by helicopter for its inbound transit of the Great Barrier Reef via Hydrographers Passage.[1] It was in ballast and bound for Hay Point to load coal. 

Figure 1: Blossom Bank pilot boarding ground and Hydrographers Passage

Source: Australian Hydrographic Office, annotated by the ATSB

Meanwhile, at Mackay Airport, a twin‑engine Agusta A109E helicopter, operated by Flyon Helicopters and registered VH‑XUM (XUM), with 2 pilots on board, embarked the marine pilot scheduled to conduct the ship’s pilotage. The marine pilot transfer (MPT) flight to Star Coral was the first scheduled for the helicopter and its pilots that day. These flights were normally conducted as a single‑pilot operation. However, on this occasion, the pilot flying, a pilot recently engaged by the operator under its ‘in‑command‑under supervision’ (ICUS)[2] program, was under the supervision of a company check pilot (pilot supervising). 

The pilots’ plan was to transfer Star Coral’s marine pilot and then proceed to a nearby outbound ship to collect its marine pilot for return to Mackay.

At 0759, the helicopter departed Mackay Airport under the control of the pilot flying. En route, the pilots established communication with Star Coral’s master via VHF[3] radio. The master advised that the ship was rolling about 3° on its inbound heading due to a 2 m south‑easterly swell. Subsequently, the pilots requested the master to reposition the ship on a heading[4] of 270° to reduce rolling. At 0853, the pilot flying landed the helicopter on the ship’s helideck, situated on the number 5 cargo hold hatch cover (Figure 2). The marine pilot exited the helicopter and proceeded to the ship’s bridge. 

Figure 2: Landing position of VH-XUM aboard Star Coral

This figure is a representation of the helicopter’s orientation relative to the wind during the take‑off. Source: Flyon Helicopters and Star Coral, annotated by the ATSB

Meanwhile, the helicopter remained on the helideck at flight idle[5] while its pilots radioed the outbound ship’s pilot to coordinate the transfer. After some discussion, the pilots elected to keep the helicopter on the deck of Star Coral until the outbound ship had departed the compulsory pilotage area.

After about 5 minutes, as the 2 ships were about to pass each other, the helicopter pilots began conducting their pre‑take‑off checks. The pilots observed a 20 to 28 knot headwind (relative to the helicopter) and noted that the ship was rolling less than 2°. The pilot flying conducted a brief for a performance category 1[6] take‑off, which involved establishing the helicopter in a hover 35 ft above deck height before departing. Both pilots later recalled that everything seemed normal as the take‑off checks were completed. 

At about 0900, the pilot flying raised the collective[7] and observed the engine torques increasing through 50%. The pilot flying recalled the aircraft became light on its oleos as though it was ‘right at the point of lifting off’. Meanwhile, the pilot supervising was observing the outbound ship passing. A few seconds later, both pilots felt a sudden and substantial vibration. 

The pilot supervising immediately looked down at the controls and recalled that the pilot flying was holding the cyclic[8] in an abnormally aft position. Concerned that the main rotor might have struck the tail boom, the pilot supervising decided to assume control of the helicopter and took hold of the cyclic and collective unannounced. Meanwhile, the pilot flying was still attempting to lift off, unaware of the pilot supervising’s decision to take control. The pilot supervising recalled that the pilot flying had centred the cyclic and ‘must have’ lowered the collective by the time the pilot supervising took hold of the controls. In contrast, the pilot flying stated that the pilot supervising rapidly lowered the collective after the vibration started, causing the aircraft to descend from being light on its oleos and bounce heavily on the helideck.

Moments later, the cyclic became uncontrollable as the vibrations suddenly worsened into a violent, vertical oscillation of the airframe. The pilot supervising tried to stabilise the helicopter but was unable to control the cyclic movement. Subsequently, the pilot supervising elected to shut down the engines. 

The pilot supervising initially struggled to reach the engine mode switches (located on the centre console) due to the severe vibrations but subsequently managed to shut down engine number 2. The vibrations slightly eased and moments later, they were able to also shut down engine number 1. The vibration dissipated and the helicopter came to rest in an upright position on the helideck, having spun more than 90° counterclockwise from its initial heading. The sequence, from the attempted take‑off to shut‑down occurred within a period of about one minute.

Soon after, the pilots exited the wreckage and inspected the damage. The tail rotor was separated from the helicopter and had come to rest on the main deck between cargo hatches 4 and 5. Items of debris, including main rotor fragments, laid scattered on the deck along with some hydraulic fluid pooled beneath the substantially damaged fuselage (Figure 3).  

Figure 3: Helicopter wreckage 

Source: Star Coral

Apart from a thumb sprain to the pilot supervising and some bruising to both pilots’ upper leg areas, where they had been struck by the cyclic, neither were significantly injured and no‑one on board Star Coral was injured.

Context

Helicopter information

The helicopter was an Agusta A109 E variant, manufactured in 2006 and issued serial number 11684. It was registered in Australia in 2006 and began services under the operator’s Air Operator’s Certificate (AOC) in 2023. 

The Agusta A109E is a multipurpose helicopter equipped with 2 Pratt & Whitney PW206‑C turbine engines. It has a fully articulated 4‑blade main rotor system, a 2‑blade tail rotor and retractable tricycle landing gear. Able to carry up to 7 occupants, it has a maximum allowable take‑off weight of 2,850 kg. 

The helicopter was able to perform flight performance class 1 operations by adherence to Category A procedures[9]. While the helicopter was normally operated from the right crew seat, it was fitted with dual controls. A left seat‑approved pilot in command (PIC) was permitted to occupy either seat during training flights. Each set of controls could not be operated independent of the other. 

The helicopter’s wreckage was recovered from the ship 2 days after the incident and transported to a secure hangar at Mackay Airport. Prior to its removal, photographs of the wreckage and the accident area were taken. There were no indications that the main rotor or tail rotor had struck any part of the ship during the accident. 

Based on its inspections, the operator advised that no engine faults or exceedance alarms had been recorded by the helicopter’s electronic engine management systems. Additionally, no faults or defects had been reported by any of XUM’s pilots or maintainers leading up to the occurrence flight. 

Post-accident activities

There was no recorded flight data available to determine the flight control inputs and their effect on the motion of the helicopter during the occurrence.[10] The pilots’ accounts, a witness statement from the master of Star Coral and photographs of the wreckage were the main sources of evidence. 

The ATSB also sought the manufacturer’s input for this occurrence. The manufacturer advised that its preliminary assessment of the available evidence suggested that the helicopter damage appeared consistent with a ground resonance phenomenon (see the section titled Ground resonance).

The licenced maintenance organisation for XUM carried out an examination of the wreckage at the Mackay hangar. On advice from the manufacturer, the examination included inspection of specific components commonly associated with ground resonance. These included main rotor dampers, landing gear struts and tyres. The operator advised the ATSB that the inspection did not identify any airworthiness issues that may have contributed to the occurrence. The operator did not provide the inspection report or findings to the manufacturer for its assessment.

Pilot flying 

The pilot flying obtained a New Zealand commercial helicopter licence (CPL) in 2011 and started flying commercially in 2014. They converted their CPL over to an Australian CPL in 2016 and held a grade 2 flight instructor rating and a class 1 aviation medical certificate. They had experience flying both single and twin-engine helicopters in various operations. Prior to joining the operator’s in‑command‑under‑supervision (ICUS) program in September 2024, they had no previous experience on the A109E, or with marine pilot transfers (MPT). 

Under the ICUS program, the pilot was required to accrue 200 hours on the A109E before they could be assessed to fly the helicopter unsupervised on daytime VFR[11] MPT operations. At the time of the occurrence, the pilot had completed the operator’s training requirements and accrued around 50 hours flight time on the A109E. They had also been cleared to conduct unsupervised MPT operations on single‑engine Eurocopter AS350 helicopters.

Pilot supervising

The pilot supervising was the operator’s head of flying operations and held an air transport pilot (helicopter) licence, issued in 2014, and a class 1 aviation medical certificate. They were approved under the operator’s training and checking system to conduct check and supervision flights on the A109E.

The pilot supervising had been flying helicopters for 26 years in various operations and had accumulated over 10,000 hours flying time, including 3,800 hours in the A109E. They first started MPT operations in 2007 and commenced working with the operator in December 2016.   

Star Coral

Star Coral was built in 2009 by Jansu Newyangzi Shipbuilding, China, registered in The Bahamas and classed with Bureau Veritas. The ship was owned by Panormos Shipping, The Bahamas, and managed and operated by Charterwell Maritime, Greece. 

At the time of the occurrence, the 229 m ship had a mean draught of 6.51 m and the helideck height was about 18 m above the waterline.   

In a written witness statement, the master reported that:

• shortly after the helicopter started to take off, it began to pound on the helideck before it spun and the tail rotor separated

• during the sequence, the helicopter became airborne for no more than 2 seconds.

Ground resonance

Ground resonance can be defined as a vibration of large amplitude resulting from a forced or self‑induced vibration of a helicopter in contact with the ground.[12] The phenomenon is normally associated with helicopters equipped with fully articulated main rotor systems consisting of 3 or more rotor blades. It is more common on helicopters with sprung landing gear than those with skids. Typically, ground resonance occurs during landing, take‑off and ground manoeuvres.[13]

In fully articulated rotor systems, drag hinges allow each blade to advance or lag in the plane of rotation to compensate for the stresses caused by the acceleration and deceleration of the rotor hub. Such rotor systems are typically fitted with lead‑lag dampers to limit the extent of this movement and help prevent excessive vibrations. However, if for any reason one or more of the blades assumes a dragged position different to the others, the blades will move out of phase and the rotor will become imbalanced, transmitting an oscillation throughout the entire airframe.[14]

The risk of ground resonance arises when the unbalanced forces in the rotor system cause the fuselage to oscillate on its landing gear at or near its natural frequency. Ground resonance will occur if the helicopter’s damping systems are unable to compensate for the oscillation.[15] Unless corrective action is taken, the amplitude of the oscillation will increase until the helicopter becomes uncontrollable.[16] Ground resonance can also be induced when the helicopter is in light contact with the ground, if the landing gear oscillation frequency is in sympathy with the rotor head vibration.[17]

Ground resonance is commonly precipitated by the helicopter making hard or asymmetric contact with the ground, landing on a slope or sudden control movements by the pilot.[18] It can also result from other factors such as improper blade balancing and tracking, or damage to any of the blades.[19] Hard contact with the ground by some part of the landing gear when the main rotor is in an unbalanced state can further aggravate the condition.[20]

Additionally, improper maintenance of the helicopter’s main rotor and fuselage damping systems, or incorrect tyre pressures, can induce or worsen ground resonance.[21]

Flight control inputs that may induce ground resonance typically involve sudden control movements or a mishandling of the cyclic that causes the fuselage to bounce.[22]

The helicopter manufacturer advised that the application of certain cyclic commands, such as extreme aft cyclic input, could theoretically reduce the main rotor damper effectiveness in respect to the damping action on the blades’ regressive lead‑lag dynamic.

Recovery technique

The onset of ground resonance can be recognised by a rocking motion or oscillation of the fuselage while on the ground.[23] The United States Federal Aviation Administration (FAA) Helicopter Handbook[24] documented 2 widely accepted recovery techniques: 

• if the condition arises when there is insufficient rotor speed for take‑off, the only option is to lower the collective to reduce the pitch of the blades. The rotor rpm[25] should also be reduced as soon as possible.[26]

• If the rotor speed is in the normal operating range for flight, the Helicopter Handbook recommends lifting the helicopter off the ground to allow the rotor blades to rephase themselves automatically. 

Additionally, the FAA cautioned that:

If a pilot lifts off and allows the helicopter to firmly re‑contact the surface before the blades are realigned, a second shock could move the blades again and aggravate the already unbalanced condition. This could lead to a violent, uncontrollable oscillation.  

In practice, a pilot experiencing ground resonance typically has seconds to identify the condition and take corrective action. 

Similar occurrences

The ATSB reviewed several investigation reports relating to previous A109E accidents attributed to ground resonance. The incidents reviewed occurred outside of Australia between 2006 and 2025 and the contributing factors were found to be operational. Technical factors which may have caused or exacerbated ground resonance were not identified. 

Details of the previous incidents bear similarity to the occurrence involving XUM, particularly in respect to subsequent damage to the helicopter (Figure 4). 

Figure 4: Previous occurrences of ground resonance involving the Agusta A109E

Source: Leonardo Helicopters

Flight manual procedures

The A109E rotorcraft flight manual (RFM) listed fault conditions and corrective actions for emergencies and malfunctions that might occur during take‑off.

The RFM included the caution below for ground resonance within the normal flight procedure for take‑off. This was not part of the emergency and malfunction procedures.

The RFM procedure for ground resonance was consistent with recovery techniques published by the FAA. The RFM reference to the helicopter being ‘free of ground resonance’ was intended to indicate  that, like all helicopters, the A109E was designed and certified to applicable standards so that the rotor and fuselage systems do not vibrate at the same frequency under normal conditions.   

Operator procedures

As an AOC holder, the operator maintained a CASA‑approved[27] operations manual/exposition[28] to promulgate general policy and standardised procedures for MPTs on the A109E. The version of the operations manual current at the time of the occurrence was issued by the operator in November 2023.

Ground resonance

The operator’s normal procedures and emergency checklists for the A109E were derived from the RFM and did not contain any procedures related to ground resonance. 

Crew coordination in response to abnormal situations

While MPT flights were predominantly conducted by a single pilot, the helicopter was certified for operations with either a single pilot or 2 pilots. In either case, the normal procedure and emergency checklists remained the same, except that 2‑pilot checklist procedures were to be based on challenge and response. 

Normal handover and takeover procedures provided that:

In the case where the pilot flying (PF) is not the PIC and the PIC determines that the PF is not maintaining adequate control of the aircraft, the PIC may elect to take control, in which case they will signal their intention by saying ‘I have control’ upon which the PF will immediately relinquish control and the roles will reverse.

In abnormal or emergency situations, the PIC was responsible for ensuring the aircraft was flown and kept under control. The operations manual emphasised the importance of cockpit resource management (CRM) standards throughout the situation, in accordance with the below procedure:

Note: In the above procedures PM stands for ‘pilot monitoring’, NR refers to main rotor speed and IAS means indicated airspeed. 

In the context of rapidly escalating emergencies such as ground resonance, pilots have limited time to perform the procedure. 

Pilot in command responsibility during training flights

As the holder of a certificate that authorised air transport and aerial work operations, the operator was required to have in place a training and checking system (TACS). A training and checking manual (TACM) sets out policies and procedures for conducting training flights. It provided that a check pilot supervising ICUS training was to be the PIC. Check pilots were to ensure that pilots involved in training exercises were made aware of who was acting as the PIC through proper handover of control procedures. 

While an ICUS pilot might be considered the PIC for flight‑time logging purposes, the pilot supervising was deemed the PIC and responsible for the safety of the flight. The TACM stated that in the event of an actual emergency during flight training:

If the flight examiner or check pilot deems it necessary to take physical control of the aircraft at any stage after the occurrence of the emergency, then they shall do so in accordance with the hand‑over and take‑over procedures specified in the Operations Manual - Hand over and take‑over procedures.

The flight examiner or check pilot must be prepared and ready to assume physical control of the aircraft at any stage, particularly during critical manoeuvres such as during take‑off and landing.

As such, beyond the normal handover of control procedures, there were no special provisions in the TACM for the allocation of PIC responsibility and PF duties during ICUS flights. 

Briefings

For 2‑pilot operations or training flights, the operator’s procedures did not require pilots to brief who would assume PF duties in the event of an abnormal or emergency situation during critical phases of flight. 

Operational limits

Under the operator’s operations manual, the A109E was permitted to conduct daytime MPT operations up to a wind strength of 30 knots, with a maximum crosswind of 20 knots. The operational limit for ship’s pitch was 4° up and 2° down while the maximum permissible roll was 4°. The manufacturer did not have input into these operator‑defined limits. 

The pilots reported that the conditions at the time of the occurrence (20‍–‍28 knot headwind, 2° roll and minimal pitching) were within the operator’s limits for MPTs.

Safety analysis

Prior to the accident, VH‑XUM (XUM) made an uneventful landing on Star Coral and remained on the deck for several minutes without incident. There was no evidence that the helicopter was operating abnormally or experienced any instability during this period. 

Examination of the accident site did not reveal any evidence to suggest that the occurrence resulted from the main rotor or tail rotor striking the ship. Star Coral’s master reported that the tail rotor separated after the helicopter started contacting on the deck, indicating that contact with the tail boom by the main rotor was a consequential rather than causative factor. 

In that context, it is most likely that the helicopter encountered ground resonance. Assessment of the damage to the helicopter following the occurrence revealed significant similarities to that seen in previous A109E incidents attributed to this phenomenon. 

It is well established that ground resonance only arises when the helicopter is in contact with the ground. Both pilots asserted that the helicopter did not become airborne prior to the vibrations while the master reported that it became airborne for about 2 seconds. However, it is more likely this occurred after the vibration worsened and the helicopter started rebounding on the helideck. 

The exact cause of the vibration could not be determined. The possibility of causative operational factors such as flight control inputs or environmental factors could not be ruled in or out. 

Similarly, while the operator’s post‑accident inspection of the helicopter (including examination of its rotor and fuselage damping systems) did not reveal any apparent defects, causative technical factors could not be discounted.

However, the sudden lowering of the collective after the onset of the vibration likely aggravated the situation. The helicopter was almost certainly light on its oleos when the vibration began. Therefore, a sudden lowering of the collective would have caused the helicopter to come down firmly on the helideck. The United States Federal Aviation Administration (FAA) Helicopter Handbook describes that such an impact when the rotor is already in an unbalanced state can cause the rotor blades to move further out of phase, resulting in violent uncontrollable oscillations. This description is consistent with the occurrence sequence described by the pilots and the master. 

The pilots’ accounts of who lowered the collective differed. The recollection of the pilot flying that their intention was to lift the helicopter off the deck in response to the vibration was not consistent with a lowering of the collective. In contrast, the pilot supervising did not immediately identify the source of the vibration and later shut down the engines, believing the main rotor may have struck the tail boom. In this context, lowering of the collective would be a natural and expected response. Therefore, it is most likely that the pilot supervising lowered the collective while the pilot flying was attempting to lift the helicopter off the helideck.

In isolation, the immediate responses taken by each pilot following the sudden onset of the significant vibration were understandable. However, since the helicopter’s rotor speed was in the normal operating flight range, continuation of the take‑off would probably have resulted in the vibration dissipating (as detailed in the FAA Helicopter Handbook).

The operator had adequate procedures for responding to abnormal and emergency situations. However, the rapidly escalating nature of this occurrence meant that there was virtually no time to implement them. There was no requirement for the pilots to conduct a pre‑flight or pre‑take‑off brief about who would assume flying duties in the event of an emergency on take‑off. Therefore, the normal procedures for handover and takeover of control were assumed to apply.

However, the time between observing the vibrations and the loss of control severely limited the time available for a formal transfer of control between the pilots. As a result, neither of these procedures were followed and each pilot responded to the situation separately. 

Findings

From the evidence available, the following findings are made with respect to the loss of control during marine pilot transfer operations, involving an Agusta A109E, VH‑XUM and bulk carrier Star Coral, about 200 km north‑east of Mackay, Queensland, on 25 February 2025.

Contributing factors

• During take‑off, the helicopter likely experienced ground resonance, resulting in the rapid onset of significant vertical oscillations through the airframe.
• Discontinuing the take‑off after the onset of the vibration, with the rotor speed in the flight range, probably resulted in the loss of control and substantial damage to the helicopter.

Safety actions

Safety action by Flyon Helicopters 

Following this occurrence, the helicopter’s operator, Flyon Helicopters, established ground resonance guidelines for its pilots. Forming part of its exposition, the guidelines were purposed to raise awareness of ground resonance and provide information about how to recognise and respond to the phenomenon. They included response procedures and featured case studies and video resources. The procedures were to be implemented into the operator’s training framework for new and current pilots. 

Flyon Helicopters advised the ATSB that it also planned to implement an additional briefing procedure in its training and checking manual (TACM). The briefing is to be conducted by the training or checking pilot prior to any training or checking flight. It will include:

• the objectives and scope of the flight, including the intended lesson plan or sequence
• the training/checking outcomes
• the roles of each pilot, including the allocation of aircraft command responsibility
• procedures for transferring aircraft control between pilots
• actions to be followed in the event of an actual emergency
• procedures to be used in the simulation of emergencies
• procedures for the conduct of unusual operations
• the method to be used to simulate instrument flight conditions, if required
• human factors/non‑technical stills and threat and error management.

Sources and submissions

Sources of information

The sources of information during the investigation included:

• the pilots and operator of VH-XUM
• the master and manager of Star Coral
• the helicopter manufacturer, Leonardo Helicopters

References

Lemmens Y, Troncone E, Dutré S, Olbrechts T. (2012). Identification of Helicopter Ground Resonance with Multi-body Simulation, 28^th International Congress of the Aeronautical Sciences

United Kingdom Ministry of Defence, AP3456 Central Flying School Manual of Flying Vol 12 - Helicopters

Salini S N, Haradev G S, Ranjith M. (2020). Ground Resonance: Nonlinear Modelling and Analysis, 6th Conference on Advances in Control and Optimization of Dynamical Systems (ACODS), India

United States Federal Aviation Administration. (2019). Helicopter Flying Handbook

Schafer J. (1980). Helicopter Maintenance

Submissions

Under section 26 of the Transport Safety Investigation Act 2003, the ATSB may provide a draft report, on a confidential basis, to any person whom the ATSB considers appropriate. That section allows a person receiving a draft report to make submissions to the ATSB about the draft report. 

A draft of this report was provided to the following directly involved parties:

• the pilots and operator of VH-XUM
• the master and manager of Star Coral
• the ship’s flag State administration, The Bahamas
• the helicopter manufacturer, Leonardo Helicopters
• Agenzia Nazionale per la Sicurezza del Volo (ANSV)
• Civil Aviation Safety Authority
• Australian Maritime Safety Authority 

Submissions were received from:

• the pilots of VH-XUM
• the ship’s flag State administration, The Bahamas
• the helicopter manufacturer, Leonardo Helicopters
• Agenzia Nazionale per la Sicurezza del Volo (ANSV)

The submissions were reviewed and, where considered appropriate, the text of the report was amended accordingly.

Purpose of safety investigations The objective of a safety investigation is to enhance transport safety. This is done through:  identifying safety issues and facilitating safety action to address those issues providing information about occurrences and their associated safety factors to facilitate learning within the transport industry. It is not a function of the ATSB to apportion blame or provide a means for determining liability. At the same time, an investigation report must include factual material of sufficient weight to support the analysis and findings. At all times the ATSB endeavours to balance the use of material that could imply adverse comment with the need to properly explain what happened, and why, in a fair and unbiased manner. The ATSB does not investigate for the purpose of taking administrative, regulatory or criminal action. Terminology An explanation of terminology used in ATSB investigation reports is available here. This includes terms such as occurrence, contributing factor, other factor that increased risk, and safety issue. Publishing information Released in accordance with section 25 of the Transport Safety Investigation Act 2003 Published by: Australian Transport Safety Bureau © Commonwealth of Australia 2025  Ownership of intellectual property rights in this publication Unless otherwise noted, copyright (and any other intellectual property rights, if any) in this report publication is owned by the Commonwealth of Australia. Creative Commons licence With the exception of the Commonwealth Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this report is licensed under a Creative Commons Attribution 4.0 International licence. The CC BY 4.0 licence enables you to distribute, remix, adapt, and build upon our material in any medium or format, so long as attribution is given to the Australian Transport Safety Bureau.  Copyright in material obtained from other agencies, private individuals or organisations, belongs to those agencies, individuals or organisations. Where you wish to use their material, you will need to contact them directly.