Researchers say AI could help predict floods. 

A Google Research team has been testing an artificial intelligence (AI) model capable of predicting floods with unprecedented accuracy. 

The technology could particularly benefit regions with scarce monitoring infrastructure.

Traditionally, flood predictions rely heavily on gauges installed along rivers. However, in many developing nations, the sparse distribution of these gauges severely hampers accurate forecasting. 

Addressing this challenge, the Google team's AI model was trained on data from 5,680 gauges to predict daily streamflow in gauge-absent areas. 

The model's predictions were then benchmarked against those of the Global Flood Awareness System (GloFAS), a leading software in global flood prediction.

The AI demonstrated the ability to predict floods five days in advance with reliability comparable to, if not surpassing, GloFAS's same-day forecasts. 

Additionally, in long-term predictions, the AI's five-year forecasts matched or exceeded the accuracy of GloFAS's one-year forecasts. 

Grey Nearing and his colleagues, the minds behind this AI model, have expressed optimism about its potential to enhance flood warning systems, especially in ungauged basins. 

By providing more timely and reliable forecasts, the model could substantially improve preparedness for both minor and major flood events, offering a lifeline to vulnerable communities in developing countries.

More details are accessible here.