The flight was so short that as soon as they climbed to 11,000 feet they had to begin preparations for landing. Eidson tuned to the latest weather report for Colorado Springs, which said the airport had a low-level wind shear advisory with winds gusting to 40 knots. An air traffic controller gave her directions to descend and turned them toward the airport.
“Cleared to land,” the controller said. “Wind three-two-zero at one-six, gust two-niner.”
“Okay, we’re cleared to land three-five, United 585,” Eidson replied. “Getting any reports lately of loss or gain of airspeed?” She wanted to know if planes were encountering trouble because of the turbulence.
“The last air carrier was the one that reported that, a 737,” the tower said.
“Could you repeat it please?”
“Yes, ma’am, at five hundred feet, a 737–300 series reported a five… correction, a fifteen-knot loss. At four hundred feet, plus fifteen knots and at one hundred fifty feet, plus twenty knots.”
“Sounds adventurous,” Eidson replied. “United 585, thank you.”
She then said to Green, “I’ll watch that airspeed gauge like it’s my mom’s last minute.”
They were parallel to the runway, preparing to turn to their base leg and then make their final descent. “Okay, start around there now and wheels down final,” Green said.
“Gear is down, three green,” Eidson said. “Speed brakes armed, green light, flaps are five, green light, hydraulic brake pressures are normal, final descent checklist complete.”
As they descended toward the runway, their airspeed changed suddenly.
“We had a ten-knot change there,” Eidson said.
“Yeah, I know. Awful lot of power to hold that airspeed,” Green said.
It went up ten more. “Wow,” said Eidson.
Twenty seconds later, she said, “Oh God.”
The plane was rolling sharply to the right.
“Oh!”
“Oh!!!!”
The pilots tried to adjust the flaps, apparently trying to abort the landing and go around.
“No!”
“Oh my God!” Eidson screamed. “Oh my God!”
“Oh no!”
The plane crashed nose first into a park three miles from the airport and exploded in a fireball. All twenty-five people on board were killed. A witness said the aircraft had looked like it was on a dive-bombing mission.
The NTSB explored dozens of theories about the engines and the plane’s rudder. Many leads went nowhere. Investigators studied sound frequencies on the cockpit tape to explore whether the engines were operating normally. But gaps in the sounds prevented them from drawing conclusions. They searched for an elderly couple who reportedly had been splattered with a bad-smelling liquid as the plane flew overhead, but the couple could not be found. As in the Hopewell crash, the flight recorder was of limited help. It had only five parameters and because the plane was in turbulence, it was difficult for NTSB engineers to determine whether the movements of the plane were caused by the weather or by its flight controls.
Much of the evidence pointed to the rudder and the likelihood of a slow-moving swing to the right. Tests in flight simulators showed that could have caused the accident. Investigators discovered that the plane’s standby rudder actuator—a backup device that would be used only if other hydraulic systems failed—had spots where metal had scraped against metal. And in the week before the crash, pilots of the same airplane had twice experienced sudden yaws to the right, problems that were attributed to minor malfunctions in other rudder components. But tests on the power control unit were inconclusive. Many PCU parts were badly burned in the crash, which raised doubts about the validity of the tests.
Engineers found that a jam in the soda can valve could not have created the slow turn by the rudder that would have been needed to match the flight recorder. Also, internal stops in the valve would have prevented other malfunctions from moving the rudder very far. And even if there had been a malfunction, Green and Eidson should have been able to counter the rudder easily with other controls.
Weather experts studied satellite pictures of Colorado on the day of the crash, looking for evidence of a powerful rotor, the horizontal tornadoes that Green had mentioned before takeoff. A glider instructor said he saw a rotor hit the ground the day of the accident, blowing off tree branches and damaging car hoods. Pilots in the area had reported turbulence and downdrafts. A 737 pilot reported a “good sinker.” But another 737 pilot described it as a normal windy day in Colorado. Meteorologists said rotors usually made a loud roaring sound, but witnesses to the crash said they didn’t hear anything out of the ordinary.
When the NTSB asked the parties to submit their ideas, Boeing and ALPA offered starkly different explanations for the crash. Boeing blamed the wind. The company said a powerful rotor struck the plane at low altitude and “did not allow sufficient height for recovery.” But the pilots union said the evidence was inconclusive, that there were strong indications of heavy winds, but the possibility of mechanical problems could not be ruled out.
Ultimately, the safety board could not reach a conclusion. After long discussions, the board decided it did not have enough for either theory. There was strong evidence implicating the rotor, but not enough to prove that it caused the crash. There was evidence that the rudder may have malfunctioned, but that should not have been enough to flip the plane. When the NTSB published its final report on the crash, the cover carried big blue embarrassing words: UNITED AIRLINES 585—UNCONTROLLED COLLISION WITH TERRAIN FOR UNDETERMINED REASONS.
Safety board investigators pledged to keep looking for a cause. “We’re worried,” NTSB official Ron Schleede told the Washington Post, “that we may have overlooked something.”
The Colorado Springs crash had haunted the safety board investigators ever since. To some, it was an embarrassing admission of defeat, UNDETERMINED REASONS. Those were painful words, an acknowledgment that the NTSB wasn’t up to the job. But to others, that was the responsible approach: to lay out the evidence and admit that it wasn’t conclusive.
Flight controls specialist Greg Phillips worked on the Colorado Springs investigation and was involved in many of the tests on the valve, but he believed there wasn’t enough evidence to blame the airplane. Still, he kept a running list of 737 rudder incidents in his file drawer. Three months before the USAir crash, he met with Boeing officials about several rudder-related incidents and questioned whether Boeing had adequately informed airlines about fixing a rudder system component. When he flew to Pittsburgh, he didn’t need to bring the charts from the Colorado Springs flight data recorder. He had them memorized.
When the Colorado Springs and Hopewell crashes were reduced to lines on a graph, there were a few similarities. The charts for airspeed and altitude showed the same basic arches. At first glance, the flights looked like mirror images of each other, the United plane rolling right, the USAir plane rolling left. The cockpit tapes showed that both crews were caught by surprise.
But there were major differences. The United plane was at 1,000 feet, the USAir plane at 6,000. The United plane was going much slower, using more flaps on its wings. The United plane was in windy weather, the USAir plane was in calm. Boeing discounted the similarities, saying that the company still believed the Colorado Springs crash was caused by the rotor. Jean McGrew, Boeing’s chief 737 engineer, said the only similarity between the two accidents “was that they are both 737s.”
Al Dickinson, the lead investigator in Colorado Springs, was a friend of Haueter’s. Dickinson was smart, a good investigator. But in the years since the crash, the guys in the office had come to look at Dickinson differently. He would always have an albatross around his neck because he had investigated the big one and come up short. Haueter reassured Dickinson that his work on United 585 had formed the foundation for everything Haueter did on the USAir crash. The NTSB was much farther along because of Dickinson’s persistence, Haueter told him. But Haueter was determined that his crash would not be unsolved. He vowed that the words UNDETERMINED REASONS would not appear in his report.