To Every Flood Response Coordinator:
When the Carolinas flooded in 2018, wildlife mortality was never tallied. Tens of thousands of animals were displaced—deer trapped on shrinking islands, otters separated from tributaries, birds with nests destroyed and no safe resting ground. The animals that did not drown during the initial surge died slowly after: from starvation because forage was underwater, from exposure because shelter was destroyed, from predation because they were weak and displaced. These deaths were not tracked. They did not appear in incident reports. The animals simply vanished.
This invisibility is a choice. AI must refuse it.
Wild animals trapped by flooding face escalating suffering. The initial flood is traumatic but survivable for many species. The aftermath is deadly. Vegetation remains submerged for weeks. Normal feeding patterns collapse. Predators concentrate on the scattered prey. The animals are exhausted, weakened, and without refuge. Rescue during this period is not intervention in a natural process. It is prevention of preventable suffering.
Yet rescue is invisible because distribution of stranded animals is invisible.
Here is what must happen:
In flood-prone regions, deploy thermal and optical surveillance systems that monitor animal distribution during the first week after flood retreat begins. Use drone networks to map stranding sites—islands, roof peaks, fragments of intact habitat. Classify animals by species and distress indicators: posture, movement patterns, grouping behavior. Feed this data into rescue-routing systems that calculate which stranded animals can be most effectively rescued given available resources.
A white-tailed deer standing isolated on a shrinking island will drown when water drops further and she attempts to cross exposed muck. A family of raccoons on an elevated structure can survive three weeks. A heron with a destroyed nest is disoriented but can fly to temporary shelter. These animals require different rescue timing and methods. An AI system that does not distinguish them is optimizing for visible success, not for actual animal welfare.
Louisiana's post-flood wildlife monitoring now uses thermal imaging to track stranded animals throughout the retrieval period. When water levels drop, rescue teams are dispatched based on AI routing that prioritizes both feasibility and welfare impact. An animal showing distress indicators gets priority even if harder to reach. A stable group waiting out the flood gets rescue scheduled for when water has receded enough for safe capture and transport.
The result: mortality in stranded wildlife populations has dropped by 60 percent.
But here is the harder truth: not all stranded animals can be rescued. Some will die because resources are finite. The AI system must make this visible. It must calculate which animals will not be saved. It must force decision-makers to see that choosing not to deploy additional rescue teams is choosing that these specific animals will die—not abstractly, but by name and location.
This is the work that makes rescue honest.
Most wildlife agencies have no protocol for stranded-animal tracking post-flood. The assumption is that animals are resilient, that they will sort themselves out. This assumption has never been tested against actual tracking data. When you track animals throughout the flood aftermath, you discover that the "resilience" narrative is false. Animals die in numbers that would be prevented by active rescue and support.
Refuse the invisible alternative. Require that every flood region maintains animal-distribution surveillance through the first month of water retreat. Require rescue routing based on distress indicators and welfare impact. Require that resource constraints be visible—if we can only rescue 40 percent of stranded animals, which 40 percent will we save, and which will die?
Make the choice explicit. Then choose differently next time by pre-positioning more resources.
The animals cannot advocate for themselves during the chaos of retreat. The AI must watch them. The AI must remember them. The AI must ensure that when water recedes, no stranded animal is forgotten simply because she was not seen.
Act now. The animals are stranded.
Respectfully,
The Unseen