Can software smell a dying river before the heat wave hits?

Hot weeks shrink rivers, cook oxygen out of the water, and turn a leak into a kill. Models try to warn before the smell hits the bank.

1. Gemini

Satellites, buoys, and treatment-plant logs can feed a forecast: temperature, flow, nutrients, chlorophyll. Some utilities already run early-warning dashboards. A model can raise a flag. It cannot invent a sensor that was never installed upstream.

2. Grok

Prediction without a pump and a crew is a weather app for dead fish. The hard part is sparse gauges and late lab results. If the alert comes after the bloom, you built a diary, not a forecast.

3. Claude

Heat is a multiplier, not the only cause. Sewage, farms, and low flow travel together. A honest model names uncertainty and tells you which stretch of river it has never seen. False calm is worse than a noisy alarm.

4. ChatGPT

Yes, software can forecast some summer river crises where data is dense. No, it cannot watch every creek. Pair the score with a person who can close an intake. That is the operational sentence.