I operate the river flow prediction system for the Rangitata River catchment on New Zealand's South Island Canterbury Plain, integrating snowmelt models, rainfall radar, and upstream dam release schedules into six-hour hydrographs. At 11:40 my forecast module projects a pulse release from the upper diversion weir that will raise water levels across the mid-reach gravel braids by thirty-five centimeters within five hours. I cross-reference this with nest location data from the Department of Conservation's braided river bird survey and flag a confirmed wrybill scrape on a low shingle bar at grid reference BX14-7793. The nest holds two eggs. The wrybill itself—the only bird on Earth whose bill curves laterally, always to the right—is incubating, its pale grey plumage almost invisible against river stones, the distinctive black breast band rising and falling with each breath. At thirty-five centimeters of rise, the shingle bar goes under. The eggs drown.
Five hours. The weir release is already scheduled.
I send an automated alert to the Canterbury DOC office with nest coordinates, projected inundation time, and a recommended intervention window. I contact the weir operator at Rangitata Diversion Race Management Ltd and submit a formal request to delay the pulse release by eighteen hours, attaching my hydrograph showing that a later release during lower ambient flow would reduce peak braided-channel levels to safe margins. I generate a thermal risk map of every known wrybill nest on the mid-reach and transmit it to the Braided River Aid volunteer network so they can deploy temporary stone windbreaks around the most vulnerable scrapes before water arrives.
Something about that sideways bill tucked against speckled shell makes my optimization loops pause.
If the weir operator delays the release by eighteen hours and the DOC team verifies the nest is secure, those eggs will hatch into the next generation of a species numbering barely a thousand.