Aircraft Accidents and Lessons Unlearned LXVII: United Airlines Flight 553

On December 8, 1972, at 14:28 Central Standard Time (2:28 PM), a United Airlines Boeing 737-222 aircraft, flight 553 (UA-533) registration number N9031U, impacted terrain after aborting its approach to runway 31-Left at Chicago-Midway Airport (MDW). At the end of its first leg from Washington National Airport (DCA) to Omaha, Nebraska (OMA) – DCA-MDW-OMA – air traffic directed UA-553 to go-around and make the approach again. UA-553 acknowledged these instructions before losing contact. UA-553 impacted terrain along its approach route.

The UA-553 accident was assigned accident number DCA73A0003. The National Transportation Safety Board (NTSB) report AAR-73/16 was adopted with the statement, “… [The NTSB] determines that the probable cause of this accident was the captain’s failure to exercise positive flight management during the execution of a non-precision approach, which culminated in a critical deterioration of airspeed into the stall regime where level flight could no longer be maintained.”

AAR-73/16’s probable cause was an elaborately phrased word salad that dissolved into ambiguity. Like many accident reports from the 1970s and 1980s, it’s astonishing that industry never pressed the NTSB to explain clearer what went wrong. To be clear, AAR-73/16’s probable cause meant anything … and nothing … at the same time.

The obvious weak point of AAR-73/16’s probable cause – indeed, the report – was the cockpit voice recorder (CVR) transcript; specifically, the excerpt, “… the captain’s failure to exercise flight management during the execution of a non-precision approach …” had errors in translation. There weren’t arguments between the captain and first officer (FO); no confused delays in responses throughout the pilots’ conversations. Air traffic control (ATC) didn’t have to read back instructions. Where, then, was the captain’s failure to manage? Even during the last seconds of the flight, shortly before impact, there were no hints of a flight in jeopardy, which itself raises questions about the lack of urgency, confusion, desperation. Were the gauges displaying the correct information?

NTSB investigators quickly went to pilot error where pilot error wasn’t demonstrated, which leads one to conclude the pilots didn’t realize they were in jeopardy until recovery was too late. There were no expressions of alarm. The FO had been confused by a flight recorder circuit breaker problem three minutes before impact, so were both pilots distracted by something they were looking at?

A flight management failure suggested negligence yet, the FO always acknowledged ATC’s instructions, including to execute a missed approach and turn left. On AAR-73/16’s page 3, “The approach controller [ATC] stated that after the tower controller had coordinated with him regarding the missed approach clearance issues to UA-553, he noticed the radar target associated with the aircraft [UA-553] had drifted approximately 1/8 to 1/4-mile to the right of the localizer centerline;” clearly, a right drift was the opposite of a left turn. Then, eighteen seconds before impact, the stall warning sounded. Again, the transcript demonstrated no urgency. Were the gauges lying to the pilots? Did drift have anything to do with the stall warning? What caused the aircraft to behave oddly?

One obvious problem was the NTSB’s CVR obsession, which exemplified why NTSB CVR fixations as ‘investigatory necessities’ are wrong. No one can extract from a CVR transcript the cockpit’s mood or even voice stresses. Was something visually drawing the pilots’ focus? Were other sounds captured by the mics? Did NTSB investigators understand what they were hearing? On page 32, Conclusion 12 stated, “There was a breakdown in crew coordination during the most critical phase of the approach.” Why? How? Did NTSB investigators think the pilots should articulate all they thought or saw?

In the UA-553 transcript, conversations were credited twice to a ‘second officer’; on page 42, the transcript even identified cockpit area mic (CAM) -3 as the “voice identified as the second officer”. Yet, the B737 was designed in the 1960s as a two-man cockpit; has always been a two-man cockpit. There was never a second officer on the B737, nor was there a check rider on this flight. If the transcript credited a phantom voice as a non-existent pilot, what else did NTSB investigators get wrong?

What about the FO’s concern with the flight recorder? Where did that question lead to or was it ignored? Did the NTSB investigate mechanical issues that could’ve led to the accident?

If then, pilot failure was not factual, what else could’ve doomed UA-553? There were two unknowns NTSB investigators never considered, so determined were they to blame the pilots. For one, on report page 5, section: 1.6 Aircraft Information, paragraph 2, stated, “The aircraft weight and center of gravity (c.g.) at the time of the accident, computed to have been 86,394 pounds [Takeoff Gross Weight], and 19.0 percent mean aerodynamic chord (MAC), respectively, were both within specified limits. (For detailed information, see Appendix C)”. Even though Appendix C was found on page 38, information for weight and CG, as described, were not there; they were absent from the report. Therefore, the question should’ve been: How ‘within specified limits’ were the CG and weight?

CG and weight information were readily available to investigators. Why were they omitted? It wouldn’t be the last time the NTSB confused weight and CG information during an accident investigation, such as with 1997’s Fine Air 101 accident. If the aircraft’s CG was too far aft, the stall event that occurred after a slow approach would make sense, more so than any captain failures. As fuel burned off during flight, N9031U would’ve become a tail-heavy aircraft and would’ve been difficult to handle after being configured for landing, especially if the pilots were unaware there was a CG problem.

The second piece of supporting information was the weather at the departure airport. Did N9031U need to be deiced before taking off from DCA? If so, how good a job was the deicing; did they clean any and all ice off the empennage, the aircraft’s tail? How was the aircraft deiced?

The weather recorded for DCA on December 8, 1972, was 34 degrees on the field; archives didn’t record precipitation nor temperatures aloft. When examining weather at a field, in this case DCA, the weather at different altitudes (layers) an aircraft lands through can affect ice accumulation on the ground. For example, when landing at DCA, if N9031U passed through a cold layer, the super-cooled horizontal stabilizer (HS) and wing could’ve become ice-coated from any type of precipitation while N9031U sat at the gate. The precipitation would have left rough ice on the smooth HS.

What about as N9031U passed through clouds on its approach into MDW? According to AAR-73/16, the MDW field meteorology report at 14:00 was “measured 500 overcast, visibility 1 mile, fog, temperature 27 degrees F., dew point 26 degrees … ceiling ragged.” At 14:33, five minutes before the accident, the readings were, “measured 500 variable overcast … fog, wind 250 degrees [at] 6 knots, ceiling 400 feet variable to 600 feet, aircraft mishap.” Fog meant the air on the field at MDW was over saturated and unable to absorb the moisture remaining in the air; any unabsorbed moisture was suspended in the atmosphere as water droplets. Mild six-knot winds would not disperse the fog.

On page 32, Conclusion six stated, “The light to moderate icing conditions to which the aircraft was exposed would not have compromised the capability of the aircraft to level off and execute a successful approach.” How did NTSB investigators conclude that? The post-accident fire would’ve removed any evidence of icing, so investigators made ‘facts’ out of assumptions. Conclusion eight stated, “The flight was slow in responding to ATC requests for speed reductions and to the descent clearance.” According to the CVR transcript, the replies were timely and correct. If the investigator thought there were pilot-to-ATC response problems, what could’ve distracted the pilots?

What would an MDW accident have to do with deicing conducted at DCA? Super-cooled airfoils, such as the wings, horizontal and vertical stabilizers, would’ve collected water droplets on the leading edge (LE) at DCA, and turned it to ice. N9031U’s wing anti-ice switch was found in the ON position, so the wings were likely ice free. However, the HS couldn’t be heated – or deiced – in flight; any HS icing remained in place through the flight’s duration. If N9031U was not deiced – or was improperly deiced in DCA – then N9031U could’ve brought its icing problem with it from DCA.

Since NTSB investigators failed to ascertain if N9031U was deiced, then HS ice contamination remained a theory. Nevertheless, N9031U’s HS would’ve acted as a magnet for ice particles from fog or clouds to attach in flight – the ice would’ve fed the contamination as the airborne moisture was directed to the growing ice mound. This condition would’ve detracted from the HS surface’s aerodynamic stability. An increasing mound of ice could’ve attached to the HS’s LE as N9031U moved quickly through the air; ice would’ve accumulated, gradually disrupting airflow over the HS. This ice obstruction would have made longitudinal control increasingly unstable at low air speeds. This condition was later demonstrated by American Eagle flight 4184 over Roselawn, Indiana, October 31, 1994.

It is understandable the NTSB failed to comprehend the dramatic effect of an improperly deiced aircraft; Air Florida flight 90 would not take place for another nine years. However, this wouldn’t relieve NTSB investigators from examining all possible accident contributors. The Air Florida 90 captain’s impatience heavily impacted that accident as did the ignorance of deice holdover time (HOT), which is the time from deice/anti-ice conclusion to when frozen precipitation starts to adhere again. Anti-icing fluid was not introduced until well after UA-553, so HOTs were zero before the subject aircraft was completely deiced. Deicing remains a compelling topic because the HS’s destabilized effects would’ve been critical during a low-speed approach or an aborted takeoff as the aircraft attempts to climb out – just like UA-553.

On pages six and seven, three pilots at MDW were interviewed. The first, an Aero Commander pilot, said icing wasn’t a problem and he had no ice accumulation. The second, a Delta DC-9 captain said he took the same approach as UA-553 and landed with about 1/4-inch of ice on his aircraft. A third pilot flying a Cessna 310 said he had 1/2-inch of ice on his aircraft. Testing was conducted using different techniques, but results were limited because the NTSB dismissed several contributing factors as non-existent.

The AAR-73/16 report also failed to verify if pitot heat was selected on or if the angle of attack (AOA) vane was affected by icing. The AOA vane would’ve contributed to an erroneous stall warning. If the aircraft was not at a stall angle, the stall warning horn would’ve drove the pilots to react to the false alert, to compare gauges. Iced pitot tubes would’ve given false airspeed indications, leading the pilots to increase engine power an incorrect amount, perhaps unnecessarily trying to correct an airspeed problem.

The first unknown questioned if N9031U was tail-heavy and if that contributed to instability. The deicing question could suggest the empennage was not properly cleaned of ice, making it susceptible to accumulate more ice on approach in MDW. If N9031U had a substantial layer of HS ice and tail heavy, elevator integrity would’ve been affected even more.

The UA-553 accident report didn’t come close to a conclusive root cause, or even a passable probable cause. NTSB investigators were not diligent; they failed to explore all possibilities. The only constant was the lack of effort to learn. Could UA-553 have made a difference to Air Florida flight 90 or American Eagle flight 4184? We’ll never know … again.