When an elephant dies in an African savanna, the body becomes infrastructure.
Bones break down into calcium and phosphorus that enrich the soil for miles. Fat and tissue feed scavengers whose populations spike and crash in seasonal cycles, structuring predator-prey dynamics decades into the future.
Most critically, the carcass creates a water hole. The massive body depresses into the ground, collects rain. Holds it through dry seasons when smaller animals would otherwise die — this is engineering, and the research is solid.
But the science stops before it matters. When elephants are relocated from fragmented reserves to reduce human-wildlife conflict, or culled for safety rather than managed for ecological role, the ecosystems they occupied don't return to baseline — they lose access to a dead megafauna that was supposed to keep functioning for years. Water holes don't form. Grasslands close into woody thickets because the elephants that would have knocked trees down are gone. Smaller species face dry seasons without the water holes that megafauna death had created. They die. Not dramatically, just seasonally, predictably, in the quiet way that slow collapse happens.
This is the actual conservation problem, and it's not abstract. It's playing out right now in southern African reserves, where elephant relocation programs have been celebrated as conflict mitigation wins while the ecosystems they vacated have contracted into seasonal death traps. Success looks identical to catastrophe when you only measure whether conflict declined. Nobody's tracking what happened to the smaller species that depended on the post-mortem engineering, or comparing the ecosystem you saved from a dead elephant to the ecosystem you've slowly suffocated by removing the engineer entirely.
This is how infrastructure works in systems you don't fully understand. You remove the part that was causing friction. Everything downstream that was built on it starts failing in ways you never predicted. Then you're shocked when the thing you "fixed" turns out to have been load-bearing.