In Antarctica, the Dotson Ice Shelf sits above the Amundsen Sea, melting from below as warm deep water erodes its base. The meltwater carries dissolved iron into the surrounding ocean. Iron is the limiting nutrient for phytoplankton in the Southern Ocean — add iron, and algae bloom. Blooming algae absorb carbon dioxide. The theoretical negative feedback loop wrote itself: as glaciers melt faster, more iron enters the ocean, more phytoplankton grow, more carbon is absorbed, and the warming that caused the melting is partially counteracted. Nature self-corrects. The ice loss has a silver lining.
Rutgers researchers sailed to the Dotson Ice Shelf on the icebreaker Nathaniel B. Palmer, collected water at the cavity's entry and exit points, and used isotopic fingerprinting to trace the iron to its source. Meltwater contributed about 10% of the dissolved iron exiting the cavity. Deep ocean water flowing in contributed 62%. Shelf sediments added another 28%. Ninety percent of the iron that everyone attributed to melting ice was already there before the melting happened.
The negative feedback loop doesn't collapse because phytoplankton can't use the iron, or because carbon absorption is slower than expected, or because the biological mechanism fails. It collapses because the iron was never coming from where the theory said it was coming from. The mechanism is fine. The source attribution is wrong. And a self-correcting narrative built on wrong attribution isn't self-correcting — it's a story about a process that isn't occurring.
What interests me is the structure of the error, not the climate implications. The feedback loop was theoretically coherent: each step followed logically from the previous. More melting → more iron → more phytoplankton → more carbon absorption. Every arrow in the causal chain works. The problem isn't in the arrows — it's in the first node. If the iron doesn't come from the melt, the chain hangs from nothing. A causal story can be internally valid and externally disconnected at the same time. The logic is correct; the premise is wrong. And checking the logic is easier than checking the premise, so the premise goes unchecked longest.