Epistemological Chicken

Accident investigators are now almost certain that birds (more specifically, Canada geese) were responsible for the ditching of US Airways Flight 1549 in the Hudson River, shortly after its takeoff from LaGuardia Airport last January. For these experts, who maintain a keen sense of history, Flight 1549’s “miracle landing” was all the more miraculous in light of an analogous incident in 1960, when a similarly sized airliner, Eastern Airlines Flight 375, struck a flock of starlings as it left Boston. In both cases, birds damaged the aircraft engines, forcing the pilots to ditch into the water. There the similarity ends, however. For unlike Flight 1549’s controlled and casualty-free touchdown, Flight 375 yawed hard to the left at three hundred feet and plunged headlong into the shallow green water of Winthrop Bay, killing sixty-two passengers and crew.

Birds are a well-recognized aviation problem. The Federal Aviation Administration (FAA) estimates that “bird-strikes” (as they are known) cost the US aviation industry $600 million every year in damage. Meanwhile, the US Air Force—whose lower-flying aircraft share more airspace with birds—reports that birds kill two crew members every three to five years and down two of their aircraft annually. Even NASA has recorded a bird-strike, on 26 July 2005, as Discovery left the launch pad.

Civil aircraft, unlike Air Force jets, generally fly higher than birds, so most collisions happen near airports during takeoffs and landings. Ground staff employ elaborate ruses to keep birds away from runways, ranging from plastic hawks and rubber snakes to distress calls played over loudspeakers, but their success is limited. Natural selection has yet to endow birds with an aversion to airports, and most become inured even to sophisticated attempts at intimidation. Scarecrows are claimed as nesting places and loudspeakers double as popular perches.

The onus, therefore, is on the aircraft and especially its engines: they must be bird-resilient. This is no mean challenge. Protective grilles over engine mouths are an engineering dead-end. Any grille strong and dense enough to withstand birds at high speeds would occlude the turbines and run the risk of smashing into the blades. Instead, the manufacturers design the engine interiors to be tough. Modern turbojet blades are as “tough as nails,” as they say, although that hardly begins to do them justice; modern engines would smash nails like matchsticks. Each blade costs tens of thousands of dollars and represents the vanguard of materials science. Their metal elements are “grown” as a single crystal: a delicate and esoteric art.

For safety’s sake, all new engine designs must demonstrate their bird-gulping powers in a series of standardized “type-certification” tests drafted and overseen by the FAA. These tests look straightforward enough: they emulate a strike and measure the engine’s ability to “take a bird.” The procedure is simple. First the engineers firmly mount an enormous engine on an outdoor test stand. Then they gradually open the throttle, urging it to maximum climbing speed where it bellows and blows like an uncorked hurricane. The giant fan-blades spin faster and faster until their tips are moving at close to the speed of sound and the engine’s bowels are hot enough to melt steel. And then, into the mouth of this Brobdingnagian blender-furnace, this magnificent technological colossus, the engineers, holding their breath and crossing their fingers, cannon-launch an unplucked four-pound chicken.

The “chicken test”, as it is sometimes known (although any four-pound bird will do), is just one of a series of bird-strike tests any new engine must pass to earn FAA certification. Along with the single four-pound bird (recently raised to an eight-pound bird for very large engines), an engine must also swallow a volley of eight one-and-a-half pound birds, fired in quick succession, and a further volley of sixteen smaller birds of three ounces each.

If the turbine disintegrates or catches fire when the chick hits the fan, if the “pilot” cannot shut it down afterwards, or if blade fragments tear through the engine housing, then the engine fails its exam and slinks back to the drawing-board. The tests with smaller birds are more demanding still: engines fail these tests if their output is reduced by more than twenty-five percent.

These criteria may seem straightforward, but in practice they involve complex and nuanced judgments that can lead to engines being retested in the event of an occurrence deemed “unrepresentative” by the FAA—if, for instance, an engine “fails” the initial test because two birds strike the same fan blade, or “passes” it because the birds do not strike designated points in the engine. More intractable than ambiguities in the pass/fail criteria, however, are questions about what passing the tests actually implies. We might imagine that a successful test on the ground shows that an engine will safely ingest birds in flight, but incidents such as those at New York and Boston suggest otherwise. The truth is that it is very difficult to extrapolate confidently from engineering tests.

Lab tests get engineers close to the real world, but not all the way. Tests are artificial by definition: they simulate real situations in a controlled way. To control is to simplify, however, and laboratories can never fully replicate the real world in every tiny detail. An engine on a wing, for instance, is slightly different from an engine on a test frame, and a white engine is slightly different from an engine painted blue; maybe these are not significant differences (such that they will affect the test), but, with a multitude of such tiny disparities, who is to say that one will not prove fateful? Failures often happen for unthought-of reasons.

Bird-strike tests should be read as a delicate bargain between representativeness and practicality. Take, for example, the choice of bird. In many respects, the FAA selects birds with a careful eye for realism. It prefers them freshly killed, for instance, because frozen birds might contain dense ice particles if under-thawed, or be dehydrated if over-thawed. The concern for authenticity only extends so far, however. The regulator stipulates the masses of the birds, for example, but not the species. Chickens get fired into engines because they are cheap and available, but they rarely fall afoul of jet engines outside laboratories. The tests use them as substitutes for other species of a similar mass, such as ducks and gulls. Yet different species with the same mass still have varying shapes, volumes, and densities, and this could have consequences. When an American Airlines flight swallowed a great-crested cormorant in 2004, for instance, American’s spokesperson justified the extensive damage by explaining that “a cormorant is chunkier, meatier, and has more bones than a looser, watery bird.” Their claim had some foundation. Engineers have found that small changes in bird volume or density can affect what they vividly refer to as its “slice mass,” which, in turn, can significantly alter its digestibility.

Another active question mark hangs over the volleys of eight “medium” and sixteen “small” birds. They are supposed to represent flocks, but critics debate whether a series of birds hurtling one after another in neat succession adequately represents the way a flock of birds actually flies. Critics also note that many small birds flock in very large numbers and aircraft sometimes ingest many more than the sixteen stipulated in the test—as was the case with the MD-80 transport aircraft that left 430 dead starlings on a runway in Dallas, or the US Airways B-737 that left over 200 gulls on the tarmac in Daytona Beach. Furthermore, the tests do not simulate volleys of “large” birds at all, although geese, swans, and storks all flock in large numbers, especially around migration time.

These doubts are illustrative, but hardly comprehensive. Experts debate many other aspects of bird-strike tests: the sizes of the birds, and the speed at which they hit the engine; whether to attach automatic surge-recovery systems; the extent to which manicured test engines are good proxies for engines in service; and much else. Few of these questions are straightforward, and, in most cases, the FAA has good arguments for its choices and judgments. Judgments are never unchallengeable, however, and there is no such thing as the “perfect” or “unambiguous” bird-strike test.

When confronting this uncertainty, it should be remembered that civil aircraft routinely inhale birds with barely a hiccup, and that the safety record of modern air travel is indisputable. At the same time, given their technical requirements, modern engines can never be fully resilient in the face of flocks of large birds like Canada geese, and a tiny number of forced landings is probably an acceptable trade for affordable air travel. We do not respond to car accidents by demanding that everyone drive a tank.

In a wider sense, we should understand that bird-strike tests are entirely typical in being uncertain and contested. All technological tests share this quality, and everything we know about the functioning of any complex machine is tainted by an unbridgeable “epistemic gap” between the laboratory and the world it represents. Epistemologists of science and technology have long known this, sometimes referring to it as “the problem of relevance,” and counsel against the idea that we can “know” complex systems (even after we have protracted experience with them). Regulators may speak reassuringly of rigorous tests and objective analyses, but the truth is that we do not know our machines as well as we think we do, and this, at minimum, is something we should know.

Footage of chicken tests can be seen at tinyurl.com/lu67f3, tinyurl.com/m2tyws, and tinyurl.com/nfomcz [link defunct—Eds.].