Google, in a new paper published in the science journal Nature, showcased how its Machine Learning (ML) technology can be used to forecast incoming floods.

For reference, floods are the most common type of natural disaster and cause more than $50 billion USD (roughly $67.6 billion CAD) in annual damages worldwide. Forecasting floods in advance can save lives and money, and according to Google, its AI-based technology can reportedly do so up to seven days in advance.

To predict floods, models normally use publically available weather data like precipitation and physical watershed information. The models are also calibrated with data from streamflow gauges. However, not all basins have the gauge installed, “and it’s challenging for hydrological simulation and forecasting to provide predictions in basins that lack this infrastructure,” wrote Google.

According to Google, ML helps solve this problem by allowing a single model to be trained on all available river data and applied to ungauged basins where no data are available. This allows models to be trained globally and predict any river location.

Google has forged partnerships with international aid organizations, including the Centre for Humanitarian Data and the Red Cross, to deliver forecasts that can be directly acted upon in times of need.

You can read more about the effort here.

Image credit: Shutterstock

Distributed also on: Google