The Field Guide › Paper
On 10 February 2009, an operational Iridium communications satellite was struck and destroyed by a long-defunct Russian satellite, the first collision between two intact spacecraft in the history of orbital flight. Kelso, the astrodynamicist whose own conjunction-screening system had been tracking both objects, uses the incident to make a point that runs against the grain of most disaster analysis in this map: nobody's warning was ignored, because there was no warning distinct enough to act on. His system, SOCRATES, had in fact flagged a close approach between these two satellites in every one of the fourteen daily reports issued in the week before the collision. But it never appeared on the Top Ten list of closest predicted approaches, and its ranking against every other predicted conjunction for the Iridium constellation that week averaged 64th out of over a thousand; on the day of the collision itself it ranked 152nd. Over that same week, more than eleven thousand close approaches were reported across the tracked satellite population, upward of 1,600 a day. Any single one of these has a low probability of becoming an actual collision, but across enough of them a collision becomes, in Kelso's own words, simply a matter of time. This is a different failure mode from the hierarchy-suppressed warnings running through the rest of this map's disasters. Nothing was silenced; the specific signal was structurally indistinguishable, using the tracking data available, from the routine background of thousands of other predicted close approaches that same week that came to nothing. The uncertainty inherent in the orbital data used for these predictions, not any failure of nerve or authority, was what buried the one conjunction that mattered inside noise that looked exactly like it. Kelso's own prescription follows from this directly: better screening requires better data, and better data requires satellite operators and even rival national space agencies to share their best orbital tracking with each other rather than rely on the coarser, publicly available catalogue that produced this collision's false negative. That is a commons argument rather than a communication one: the near-Earth orbital environment is a shared, degradable resource, space debris is an externality imposed on every other user of it, and the fix Kelso proposes is cooperative data-pooling across competitors and adversaries alike, not a better way of getting one operator to listen to its own instruments. The conjunction also stands as a precisely quantified instance of a weak signal doing everything a weak signal is supposed to do and still not being amplified in time: subtle, genuinely indicative of a larger problem, dependent on interpretation against everything else being monitored, obvious only in hindsight, and in principle a trigger for action, yet buried inside a haystack that had, if anything, been made larger by the very success of automated tracking at generating more candidate signals than any system could rank with confidence. For a corpus about psychological safety and organisational learning, this paper supplies the missing case in which the barrier to acting on a signal was not power, status, or dismissal, but literally being unable to tell the important signal apart from the ordinary ones at the moment it mattered. Its limits are that it is a focused technical analysis of tracking data and debris dynamics rather than an organisational or human-factors study, so its relevance to this corpus is by extension and analogy rather than by original design.