At around 00.20, a 120m-long barge carrying 2,200t of vinyl chloride (extremely flammable and carcinogenic gas) was damaged during locking. Navigation was stopped along a 10km stretch. A 400m security perimeter was set up and 20 local homes were evacuated, though the residents were able to return during the night. Firefighters rescued the 5 crew members. A crisis unit was set up at 03.00 and the civil security response organisation was triggered at 07.30. There was a major leak in one compartment, but this did not cause any risk of submersion. The wheelhouse, located under torn-off sections of the lock gate, damaged various gas pipes. At around 07.30, measurements revealed concentrations at 30% of the LEL. They were negative again 8 hours later. Preventive floating booms were put in place due to the 35,000l of fuel oil stored on the barge.
The day after the accident, the plugging measures put in place were no longer totally effective and some sources of leaks were inaccessible or unknown. 10m from the barge, the gas readings were negative. A specialist firm finalised the plugging 5 days after the accident. Atmospheric monitoring showed highly variable results during the plugging, but they were below the toxicological reference values (figures taken into account: 1,300µg/m³ for a 14-day period). The tanks without any leaks were unloaded onto a backup barge. Concrete was cast into the casings of the 2 damaged tanks and the fuel oil was pumped out. The leaking tanks were unloaded 2 weeks after the accident. A 550m security perimeter was put in place and neighbouring houses and a sawmill were evacuated.
The barge entered the lock without colliding and was moored. The gate featured 4 sections “stacked” on top of each other. Leaks appeared on the sides of the downstream gate, then, with the water at a height of 8m, 1 or even 2 middle sections of the gate gave way. The mooring lines did not withstand the draft of the barge, which tore off the top section of the gate (23t). Based on the operator’s investigations (examination of recovered parts, kinematic and mechanical studies, video-based photogrammetrical analysis), it was concluded that the “suspended” gate featuring 4 stacked sections was considered closed by the automated systems even though it was up to 40cm off the nominal situation (top section potentially correctly engaged but bottom section either only very slightly or not at all engaged). The presence of a floating object may have caused this blockage. The lock gate breach scenario, due to incorrect engagement of the gate and a closure failure, was not taken into account in the site’s safety report.