Google used AI to predict floods seven days in advance
Google has announced that it has successfully used artificial intelligence to accurately predict river floods. In some cases, the company has been able to predict floods as much as a week in advance.
Here's What We Know
Predicting floods is usually difficult due to the lack of water level sensors on most rivers. Google trained machine learning models on a variety of relevant data, including historical events, water level indicators, terrain, and more. Localised maps were then created and hundreds of thousands of simulations were run at each location.
This approach allowed the AI models to accurately predict upcoming floods. While individual cases gave a forecast 7 days in advance, the average advance was about 5 days. According to Google, this significantly improved the reliability of global flood forecasts compared to previous capabilities.
The technology currently provides accurate forecasts in 80 countries with a total population of 460 million people. The forecasts are available through Google search, Google Maps, Android notifications and the Flood Hub web app.
Google will continue to improve its machine-learning-based flood forecasting models with academic researchers. The company aims to create a global end-to-end platform to provide as much advance warning of impending floods as possible.
Source: Engadget
