At three seconds past 0847 on 8 January 2003, the captain of Air Midwest flight 5481 said two words into the cockpit voice recorder: “help me.” The Beech 1900D had just left runway 18R at Charlotte and the nose would not come down. Ten seconds later the flight data recorder caught a pitch attitude of 54 degrees nose up. At 0847:28 the recording ended as the aeroplane struck a US Airways maintenance hangar on airport property. Two crew and nineteen passengers were killed.

The load manifest they had worked from said the aircraft weighed 17,028 pounds with its centre of gravity at 37.8 per cent of the mean aerodynamic chord. Both numbers sat comfortably inside the Beech 1900D envelope. Neither was true. When the National Transportation Safety Board rebuilt the aeroplane from the wreckage and the recorder data, it came out at 17,700 pounds with the centre of gravity at 45.5 per cent MAC — over the certified weight limit, and five and a half per cent of chord behind the aft boundary the type certificate allows.

Every passenger has watched a ground crew argue about which bags go where and assumed it was a problem of fitting them in. It is not. It is a problem of where the aeroplane balances, and the margin is far narrower than anyone outside the industry imagines.

The Envelope Is Narrower Than It Looks

An aeroplane in cruise is a balancing act with three forces in it. The wing makes lift, acting through a point roughly a quarter of the way back along the chord. Gravity pulls down through the centre of gravity. And the tailplane makes a third force — on most conventional airliners a download, pushing the tail towards the ground — to hold the whole arrangement in equilibrium. Move the centre of gravity and you change how hard the tail has to work.

The certified envelope is the set of weight and balance combinations where that arrangement still behaves. It has two edges and they protect against entirely different things.

El forward limit is a control-power limit. The further forward the centre of gravity, the larger the nose-down moment the tail has to overcome, and at some point the elevator simply runs out of travel — typically at the worst moment, rotating for takeoff or flaring to land at low speed. The aft limit is a stability limit. As the centre of gravity moves back towards the wing’s aerodynamic centre, the aeroplane becomes progressively less willing to return to trimmed flight after a disturbance. Stick forces lighten, the stall margin shrinks, and eventually a pitch-up starts that the elevator cannot arrest. Both limits are hard. Neither announces itself to the crew.

Air Midwest 5481: The Paperwork Was Inside the Envelope

Under the Air Midwest weight and balance program in use that morning, flight 5481 was legal. The program put the aeroplane at 17,028 pounds against a limit of 17,120, and at 37.8 per cent MAC against an aft limit of 40. Nothing on the manifest asked anyone a question.

The Safety Board calculated the real figures three separate ways — by rebuilding the aircraft from component weights, by re-running the airline’s own program, and by analysing the takeoff ground roll from the flight data recorder. The answer was 17,700 pounds, give or take 200, and 45.5 per cent MAC, give or take two. The gap between the paper aeroplane and the real one was most of the difference between flying and not.

What that did to the controls is the detail worth keeping. The investigation also found the elevator control system had been rigged incorrectly during maintenance at Huntington, West Virginia, which left the available travel running from about 20 degrees nose-up to only about 7 degrees nose-down. At the weight and balance the aeroplane was actually at, between 9 and 10 degrees of nose-down elevator were needed to establish a controlled climb. The crew did not run out of skill. They ran out of elevator, by about three degrees, in an aeroplane that should have had twenty.

Two independent failures, neither individually fatal, lined up. That is the whole of it, and it is why the Board’s probable cause names both.

The Board was unambiguous about how the two faults combined, and about who else was in the chain.

The Numbers Everyone Assumed

Buried in the list of contributing factors is an item that reads oddly until you understand how airlines weigh people: the Federal Aviation Administration’s average weight assumptions in its weight and balance program guidance at the time of the accident. Nobody weighs passengers. An operator of multiengine turbine aircraft with five or more seats can apply for authority to use standard average weights instead, and then the load sheet is arithmetic performed on assumptions.

Which makes the assumptions load-bearing. The current guidance, Advisory Circular 120-27F, handles this in a way that surprises people who go looking for the number: it does not publish one. The example table in the AC is printed with the weight columns blank. Operators derive their own standard averages from Centers for Disease Control and National Health and Nutrition Examination Survey data, add a clothing allowance of five pounds in summer and ten in winter, and submit the result to the FAA for approval into their operations specifications. Summer weights may be used from 1 May to 31 October, winter weights from 1 November to 30 April, and an operator with seasonal variation that wants a single year-round figure is told to use the winter one.

The same AC sets the housekeeping that keeps the empty aeroplane honest. Basic empty weight and centre of gravity have to be re-established whenever cumulative logged changes exceed half of one per cent of maximum landing weight, or whenever the centre of gravity has drifted by half of one per cent of the mean aerodynamic chord. Half a per cent. That is the tolerance on an empty aeroplane before anyone has loaded a single bag.

Charlotte is the case study, but it is not the only shape this failure takes. Two other investigations show the other two ways the sum can come out wrong: when the form itself is built incorrectly, and when the load refuses to stay where it was put.

Fine Air 101: When the Form Itself Is Wrong

Six years before Charlotte, a Douglas DC-8-61 freighter lifted off runway 27R at Miami and pitched up far beyond anything the crew wanted. Fine Air flight 101 crashed shortly after takeoff on 7 August 1997, killing its three flight crew and a security guard on board, and a motorist on the ground.

The Safety Board found the cargo had not been loaded the way the load plan specified, which is the headline. The more unsettling finding is what it discovered about the paperwork itself. Fine Air’s printed load sheet form said DC-8-61 at the top, but the fuel distribution scale on it had been built from DC-8-62 and -63 data. The same form listed the maximum allowable weight for pallet position 18 as 6,088 pounds when the correct figure was 3,780 — so that position went out 1,247 pounds over its limit on the accident flight, with nothing on the form to say so.

The stabilizer was set to 2.4 units nose-up. The Board calculated that the correct setting for the weight involved was 0.9, which makes the trim wrong by at least 1.5 units in the nose-up direction before the aeroplane ever moved. On a DC-8 at rotation, that is a pitch-up the crew has seconds to catch. The report is honest about the limit of what could be established: the actual centre of gravity could not be determined, so the severity of the control problem could not be either.

National Airlines 102: When the Load Moves in Flight

The third case is the one that haunts loadmasters, because the aeroplane was loaded, the paperwork was done, and the centre of gravity moved anyway.

On 29 April 2013 a Boeing 747-400 freighter took off from Bagram with five mine-resistant ambush-protected vehicles on the main deck — two 12-ton M-ATVs and three 18-ton Cougars. It was the first time National Airlines had attempted to carry five of them. They were too large for standard unit load devices, so they rode on centreline-loaded floating pallets, held by tie-down straps. The aeroplane climbed steeply after rotation and then descended in a manner the Safety Board described as consistent with an aerodynamic stall. All seven people on board were killed.

The investigation found strong evidence that at least one vehicle, the rear M-ATV, broke loose and travelled aft into the tail. It damaged hydraulic systems one and two, and it damaged components of the horizontal stabilizer drive mechanism — the same jackscrew assembly that, on a different Boeing, is the subject of an airworthiness directive published this week. After that there was no pitch control left to recover with. A shifting load is not a weight and balance error in the usual sense. It is a restraint failure that becomes one, in flight, at the worst possible moment.

The cockpit voice recording and the flight data from Bagram have been reconstructed in detail, and the sequence is short enough to watch in full.

What Actually Protects You

Read the three reports together and a pattern comes out that is not about arithmetic at all. In every case the number on the paper was reasonable. In Charlotte the program said the aeroplane was legal. In Miami the form was internally consistent and wrong. At Bagram the load plan was followed and the restraint was not up to the job. The failure was never that someone could not add up.

What protects a flight is the chain that keeps the paper and the aeroplane in agreement: a weight and balance program that reflects the operation it is actually used in, average weights that match the people actually boarding, empty weights re-established before they drift half a per cent, cargo restrained to a standard that was designed for the load being carried, and an operator exercising real control over the loading rather than delegating it to a handling agent and hoping. Three of those four appear by name in the contributing factors of the three accidents above.

The centre of gravity is not a number on a form. It is a physical place inside the aeroplane, and it moves every time something is put in or taken out. The form is only ever a claim about where it is. Thirty-three people died in these three accidents because, on those three days, the claim was wrong.

Sources: NTSB Aircraft Accident Report AAR-04/01, Loss of Pitch Control During Takeoff, Air Midwest Flight 5481; NTSB AAR-98/02, Fine Airlines Flight 101; NTSB AAR-15/01, National Airlines Flight 102; FAA Advisory Circular 120-27F, Aircraft Weight and Balance Control.