Bell 47G-2 fuel exhaustion accident at Sigean (F-GVIG)

Accident overview and flight sequence On 31 July 2018, at about 10:40 local time, a Bell 47G-2 helicopter, registration F-GVIG, operated by Société de travail aérien Giragri 17, was destroyed after a fuel exhaustion during a spraying flight. The pilot was the only person on board. The flight was a mosquito-larvae treatment operation. The pilot departed a temporary dirt-road base at about 06:15 and made repeated returns for refuelling and larvicide loading. For the final rotation, the aircraft was refuelled to a total of 80 litres, and the pilot requested at least 100 litres for a more distant sector near Marseillette. After treating a final parcel near Sigean, the helicopter was returning over Étang de Bages-Sigean. At about five kilometres from the base and roughly 100 feet above the water, the engine stopped. The pilot performed an autorotation onto the water surface; the helicopter sank, and the pilot evacuated from the bottom of the lake. ## Aircraft, fuel system, and evidence The Bell 47G-2 was a light three-seat helicopter with a two-blade main rotor and a Lycoming VO-435-A1 piston engine. It had two communicating fuel tanks with a total capacity of 162 litres, of which 155 litres were usable. A fuel-level indicator and a low-fuel warning light were installed; the light was intended to illuminate when 15 to 20 minutes of endurance remained. The aircraft also carried a Simplex spraying system with two larvicide tanks, two spray booms, and a dedicated hydraulic circuit. The maximum takeoff mass was 1,111 kg, and the empty mass of F-GVIG was 727 kg, including 78 kg for the Simplex equipment. A Trimble GNSS unit recorded position, altitude, and larvicide quantity; it showed that 199 litres, about 200 kg, had been sprayed during the final rotation. The helicopter had no fuel flowmeter. The wreckage was recovered from the lake and showed little overall damage. The main rotor was free to rotate and drove the rear transmission shaft; the freewheel was functional. The entire fuel circuit, including tanks, filters, and carburettor, was filled with water and showed no visible leak, although vent openings allowed water to enter and displace fuel while submerged, so the fuel quantity at the time of the accident could not be determined. Later examinations by the operator found a small amount of fuel in the carburettor float chamber and a functional electrical circuit, and the engine was restarted. The cable carrying the low-fuel signal to the warning light was found broken above the tank, but it could not be determined whether the break predated or resulted from the accident. The wreckage examination found no pre-accident malfunction that could explain the engine stop. ## Pilot, weather, and fuel management The pilot was 56 years old and held a valid commercial helicopter pilot licence (CPL(H)) obtained in 1991, with more than 3,000 flight hours, almost all on Bell 47 in agricultural work. He was employed under a three-month fixed-term contract that had begun about a month and a half before the accident; it was his second contract with the company. The automatic METAR from Perpignan–Rivesaltes, about 40 km away, at 10:30 reported variable wind of 2 kt, a temperature of 30 °C, and CAVOK conditions. The pilot and mechanic performed the pre-flight inspection, and the mechanic refuelled the helicopter on its trailer with 40 litres, bringing the total on board to 80 litres. Refuelling was done from a fuel tank on a follow-up truck using a distribution pistol without a totalising volucometer. The pilot recalled requesting two 40-litre refuelling stops after rotations, representing about 40 minutes of flight, to start new rotations with 80 litres. He stated the Bell’s consumption was generally 55 to 58 litres per hour and that he planned endurance using 60 litres per hour, with a 30- to 40-minute margin. For the final rotation, he requested at least 100 litres to obtain 1 hour 40 minutes of endurance, noting the mechanic usually added 10 litres more than requested. He tested the low-fuel warning lamp bulb before each takeoff. On the return from Marseillette, knowing he had four to five litres of larvicide left, the pilot decided to treat an additional parcel near Sigean that had not been completed earlier because of nearby tourists. GNSS data showed he began treating that parcel about 2 minutes 30 seconds before the accident. He made several passes before returning toward the base over Étang de Bages-Sigean. During the treatment, he looked at the low-fuel warning light and saw it was off. He reported flying at 350 feet above the water, then descending. At the end of the descent, he no longer heard the engine and saw engine and rotor rpm fall. He lowered collective pitch, tried to gain speed for autorotation, flared, and released the controls before water contact. He said sunlight reflections on the water made height estimation difficult. The sudden water impact shattered the windscreen, and he managed to detach and evacuate. The pilot said he sometimes checked remaining fuel at the end of flights by comparing it with flight time, but did not use the dashboard fuel-level indicator, which he considered unreliable, preferring a direct reading of the tank level. This allowed him to compare actual consumption with the theoretical 60 litres per hour. On the accident day, he had not detected unusual overconsumption. He sometimes checked total fuel level after refuelling but had not done so immediately before the accident flight because he remained on board to study the next job and prepare the flight. The mechanic was accustomed to this type of refuelling. The landing area was a dirt road with ruts from agricultural tractors, so the helicopter could be slightly nose-up or nose-down depending on its position. The pilot said the tank-level reading during the last refuelling could have been distorted by a nose-up attitude due to the ground slope, estimating the level could be overread by 20 to 30 litres in such a configuration. The company manager stated that F-GVIG consumed about 62 litres per hour on average and that at full load consumption could rise to 75 to 80 litres per hour.