.jpg)
During the take-off roll on Runway 06 at Edinburgh, just before rotation, both pilots heard a bang from the nosewheel. About nine minutes later, climbing through FL220, the crew saw a GPS PRIMARY annunciation and a downgrade to CAT 3 SINGLE. The MCDU showed IRS 2 deviation as dashes, and IRS 1 and 3 drifting from 5 nm up to 28 nm. Both autopilots and auto thrust then disconnected. There were no ECAM alerts or overhead fault lights.
The crew declared a PAN and levelled at FL320. Because the flight was now non-RVSM, ATC asked them to squawk 7700 and descend to FL290. No QRH checklist exactly matched the symptoms. A GPS waypoint check was wrong, and a VOR cross-check showed a 5 nm split between the two NDs. The crew switched to ROSE VOR and continued to Heathrow on radar vectors.
In the descent, AP and A/THR came back. The crew used them but planned to hand-fly the final approach and briefed a go-around. AP and A/THR dropped out again at 4,200 ft. At about 1,500 ft an AUTO FLT RUD TRV LIM 1 alert appeared. The ECAM actions couldn’t be completed by 1,000 ft, so the crew went around as briefed. The second approach and landing were uneventful. IR faults appeared on ECAM after landing.
Vertical shock loads from the nose gear reached the avionics rack, which sits just aft of and above the NLG bay. These loads caused severe drift in the IRS.
The CVR recorded rhythmic thumps consistent with the nosewheel striking centreline lights. It then recorded two larger impulses about one second apart at roughly 118 kt groundspeed. At the same time, the NLG squat switch flicked between air and ground, meaning the leg extended fully despite the aircraft being on the runway.
The FDR accelerometer showed nothing unusual, because it is mounted near the centre of gravity rather than the nose.
The exact source of the shock wasn’t found. The runway had several repair patches on or near the centreline just before the rotation point, but none was judged capable of causing it on its own. Post-flight, the ADIRUs and nose gear were checked and found serviceable.
The Northrop Grumman ADIRU part-number series fitted (465020-030x-0316/318) is known to be sensitive to shock loads. Earlier safety actions focused on NLG servicing. They included a new LANCE servicing tool, still not approved but expected in service during 2026. Airbus is taking no further action, since the loads exceed certification requirements and the gear had been correctly maintained. Edinburgh’s runway was fully resurfaced, completed in April 2026.
A runway surface issue turned into a whole-flight automation and navigation problem, with no ECAM guidance and no matching QRH procedure. The crew handled it well. They used conservative automation, briefed a go-around in advance, and committed to it when the ECAM actions couldn’t be finished.
Lessons From This Event
Fall back on the golden rules when the checklist runs out. When nothing matches, priorities matter more than procedures. Fly the aircraft at the most appropriate level of automation and be ready to go to none. Navigate, know where you are and where the terrain and traffic are. Communicate, with each other, with ATC, and with the cabin. Manage, and only then, troubleshoot.
The G-EUPS crew followed this order. They levelled off, got radar vectors, and declared early, all before they had a diagnosis.
Use structured decision-making that works without a checklist. Tools like T-DODAR or FOR-DEC or the pneumonic used by your company as a framework when there's no procedure to follow.
Time - how much do we actually have?
Diagnose - what do we know, and what's an assumption?
Options - what are they, and what's the safest?
Decide, Assign, Review - and keep reviewing.
The "Review" step is critical. In G-EUPS, the automation came back during the descent. The crew used it but didn't trust it. They planned for it to fail again, and it did.
Keep the basics sharp. Unfamiliar failures strip you back to fundamentals. Raw data and conventional navigation. The G-EUPS crew used VOR and radar vectors when GPS and IRS couldn't be trusted. Manual flying skills, especially in degraded modes. Cross-checking sources and asking which one to believe when they disagree.
Know the aircraft, not just the procedures. Understanding how systems connect can help reason through failures. In this case that the IRS data feeds the FMGC and the FAC, so IRS drift could take out AP, A/THR and the rudder travel limiter. Systems knowledge turns a confusing cascade into a story that makes sense. Extending the slats and flaps should restore the full rudder authority for landing.
Briefing the "what if it gets worse" before it gets worse lowers workload when it matters. The G-EUPS crew planned to hand-fly the final approach and briefed a go-around in advance. When the ECAM couldn't be finished by 1,000 ft, the decision was already made.
Training
Every bulletin like G-EUPS is a free scenario from somebody else's bad day. Use them in classroom CRM, build them into sim sessions.
Training competencies not events is the thinking behind Evidence-Based Training (EBT) and Competency-Based Training and Assessment (CBTA). The sim is used to build the nine ICAO competencies, such as problem solving, situation awareness, workload management and manual flying. Ticking off a list of failures is not the goal. The scenario is just the vehicle. Varying the scenarios including "unknown" failures with ambiguous symptoms, where ECAM, or a checklist only partly fits. Debrief the thinking, not just the outcome.
Train startle and surprise deliberately. The first 10–30 seconds of the unexpected are where performance can be compromised. Physically pause, hands off, breathe, look at the basic instruments. Use a simple verbal anchor, something like "Fly. Where are we? What's working?". Expose crews to surprise in a safe environment so the reaction isn't new on the day.
Culture
Crews need to feel empowered to deviate from the plan, declare early and divert when nothing fits, without fearing being second-guessed later. Organisations should also make clear that a crew facing something genuinely unprecedented is expected to think, not just comply.
We can't write a checklist for everything. What we can do is train crews to stay calm, fly first, build a picture, keep it simple, and give themselves an exit. G-EUPS is a good example of that working.
Photo source: Jonathan Palombo, CC BY 2.0, via Wikimedia Commons