Google announced in a news that using artificial ielligence, it was able to successfully predict river overflows and floods in some cases even seven days ahead. Floods are one of the most common natural disasters in the world, so having a warning system for them can be very useful.
Flood forecasting is very difficult because most rivers do not have water gauges. But Google has solved this problem by training machine learning models through data such as records of rivers, recorded information of their water level, the amou of water height increase, regional conditions, etc.
Google has also created local maps and performed hundreds of thousands of simulations at each location. The combination of these techniques has provided conditions for the company's artificial ielligence models to accurately predict future floods.
Forecasting floods became possible with Google's new artificial ielligence
The Mouain View gia's approach has resulted in accurate models for very specific locations, but Google hopes to eveually use the same techniques to solve the problem of global flood forecasting.
Although the system has in some cases been able to predict floods seven days ahead, the average forecast is about five days. Of course, Google says the system is a big improveme over the lack of a reliable predictive system, and could be particularly effective in underserved regions of Africa and Asia.
Google's new artificial ielligence is able to predict floods for 80 couries with a total population of 460 million people. The company can provide users with their predictions through Android's search engine, map, and notifications. Google has published the results of its findings in the form of a scieific article in the journal Nature.
Google also plans to use the poteial of machine learning to improve flood forecasting models and eveually build a global forecasting platform.
