The first step after determining the debris’ location is to call in the mathematicians. Based on all the data available—the aircraft’s last known position, route of flight, altitude, prevailing winds, sea currents, ocean depth, and so on—a probability is assigned to each variable, and a distribution map of probable locations on the sea floor is generated. Searchers can then deploy their underwater assets to scour the vastness of the deep, working back and forth along grid lines laid out in the areas of maximum probability.

There’s a deep problem inherent in this approach, however, and it’s that the probabilities are themselves only guesses. Searchers are uncertain even as to the extent of their own uncertainty. In the case of Air France 447, the set of base-set assumptions turned out to be wrong, and the first two search seasons scoured thousands of square miles in vain.

What turned the tide for AF447 searchers, in the end, was better math and better undersea technology. A recalculation of the location probabilities using a different mathematical approach led to the redrawing of the search grids much closer to the site of the plane’s disappearance. And a new type of autonomous undersea vehicle—a robot sub, in other words—became available for the first time. Called Remus 6000, these subs were able to navigate on their own along precise grid lines, ascending and diving to match the contours of the undersea terrain. On April 3, 2011, less than a week after the refined search began, one of the three submersibles deployed in the search returned to its mother ship bearing images of a debris field scattered across an abyssal plain. AF447 had been found. A month later another type of unmanned submersible brought the black boxes to the surface.