Technology
Google's AI flood warnings currently cover all of India and have extended to Bangladesh.
Google reports that its flood forecasting service, which uses machine learning to classify flood-prone areas and warn users before the floods arrive, currently spans all of India and has also extended to parts of Bangladesh.
The internet giant unveiled a platform for India's Patna area in 2018, but it says it has been steadily growing coverage in collaboration with local government. In June, it reached the mark of reaching all of India's worst flood-affected regions. The company says this ensures that about 200 million people in India and 40 million people in Bangladesh are already likely to access warnings from its flood forecasting system.
In addition to increasing scope, Google is developing more reliable predictions and improving how its notifications show on users' computers. The company says it's already sent over 30 million updates to Smartphone devices.
Google has long been involved in delivering alerts of natural hazards and national emergencies such as flooding, wildfires and earthquakes. Many of them are done by the Emergency Warning service. Only last month , the company unveiled a new service that transforms Android smartphones into a network of seismometers, using the accelerometers within phones and tablets to track earthquake movements and submit warnings to consumers.
However, in the case of flood forecasts, Google does not use details from consumer computers. Instead, it builds on a combination of historical and current data on weather, river flows and flood predictions, using machine learning to construct modern prediction models.
Google says it's playing with experimental templates that will offer much more precise warnings. Its new prediction model will double the lead time of its previous system, says the firm, while also supplying people with details on the depths of the floods. In more than 90 percent of cases, our predictions will provide a right water level within a 15-centimeter margin of error, says Google researchers.
A analysis of Google's predictions in the Ganges-Brahmaputra river basin, undertaken with scientists from Yale, showed that 70 per cent of the people who got the flood warning did so before the floodwaters arrived, and 65 per cent of the households who got the alarm took action. Even in places of low literacy, restricted schooling and high poverty, the majority of people act on the information they get, reports the researchers. So, early alerts are always worth the effort.
They noticed that issues with the use of mobile warnings persisted. The key problems are clearly lack of access to smartphones and lack of faith in technical alerts.
