Google used an artificial intelligence-powered weather forecasting tool to accurately predict floods up to a week before they occur.

In a new paper published in the scientific journal Nature,

Google researchers describe their system as an early warning system that can be used worldwide to mitigate the worst effects of floods, one of the most common natural disasters. [Predicting floods, especially river floods, is difficult due to a lack of resources and data on flood history, water levels, and topography. Google took the data that existed, used it to train machine learning models, and used them to make global forecasts.

The use of AI in weather forecasting is not new. However, the rise of more powerful processors and large data sets is enabling new scales of forecasting, including the new NVIDIA Digital Earth.

A small percentage of the world's rivers are equipped with stream gauges, which can warn of changes in the amount of water flowing through a river.

Combined with other data, such as local topography, settlements near rivers, weather forecasts, and past events on the same river, floods can be predicted.

Problems arise on rivers that are not equipped with such expensive technology or where historical data is not reliably available.

To solve this problem, Google researchers collected as much global data as they could access and used it to train new AI models.

The team launched the Flood Hub platform, which provided access to forecasts for 20 countries where flood forecasting had previously been difficult. The platform fed all locally available information into the AI system as prompts, allowing AI to fill in the gaps.

Flood Hub has since been expanded to 80 countries in Africa, Asia Pacific, Europe, and the Americas, covering more than 460 million people worldwide, especially those in flood-prone areas.

"Thanks to advances in our global AI-based models, access to flood forecasting in Africa is now comparable to that in Europe," the Google team declared.

"We have found that AI is helping to provide more accurate information on river floods up to seven days ahead," they explained.

"This has enabled us to provide flood forecasts in 80 countries where 460 million people live. Whenever possible, we also provide forecasts in Google Search, Google Maps, and Android notifications."

"We have also found that the forecasting system is more accurate than ever before," they explained.