Google’s AI-based flood detection & alert system to expand its coveragehttps://indianexpress.com/article/technology/google-flood-detection-alert-system-ganga-india-rivers-monsoon-5723415/

Google’s AI-based flood detection & alert system to expand its coverage

Under its pilot project, Google had partnered with India’s Central Water Commission to collect data for its models that it claims predict the impact of floods more accurately.

The Competition Commission of India (CCI) last year started looking into the complaint, which is similar to one Google faced in Europe that resulted in a 4.34 billion euro ( billion) fine on the company. (Image source: Reuters)

Taking forward a pilot project in Patna that started last September, US technology major Google will expand its artificial intelligence-based flood detection and alerting system for the upcoming monsoon season that will cover areas along the Ganga and Brahmaputra rivers. Under its pilot project, Google had partnered with India’s Central Water Commission to collect data for its models that it claims predict the impact of floods more accurately.

“Every year, they affect up to 230 million people across the world — more than storms and earthquakes combined. Twenty per cent of flood fatalities happen in India alone … and without consistent accurate warning systems, people are prone to ignore warnings and be unprepared. That’s especially detrimental in areas hit with annual monsoons … the expanded area will cover millions of people living along the Ganges and Brahmaputra river areas.

“Not only are we increasing the area of coverage, but we are also better forecasting where the floods will hit hardest. Through a new version of our public alerts, people can better understand whether they will be affected,” Jeff Dean, Google’s head of AI division, said earlier this week at the company’s annual developer conference in San Francisco, California.

According to the Indian Meteorological Department (IMD), monsoons in India during 2019 will be near-normal. Further, as per the department’s monsoon forecast, rainfall will be well distributed across the country. The south-west monsoon makes its onset over India around May-end.

Advertising

Dean explained that Google’s model simulates water behaviour across the floodplain, showing the areas that are likely to be affected.

The model combines thousands of satellite images to create high resolution elevation maps to figure out the height of the ground.

It then uses neural networks to correct the terrain, making it more accurate, after which it simply applies physics to simulate how flooding will happen.

However, effective riverine flood forecasting at scale is hindered by a number of factors, with the biggest factor being the need to rely on human calibration in current methodology, limited amount of data for a specific location, and computational difficulty of building continent level models that are sufficiently accurate.

Google said it has collaborated with the government to collect up-to-date stream gauge data and accordingly sends out real-time updates.

During the pilot project, Google exhibited, via Public Alerts, a map that included areas designated as ‘high risk’, ‘medium risk’ and ‘low risk’ and across these different alert types after this year’s heavy post-monsoon rains, which Google says recorded “high accuracy metrics”.

The pilot used an operational hydro-dynamic model, with the explicit goal of preparing the ground for integrating machine learning (ML) models into the process. Alerts were then sent out to individuals in the catchment area in the form of maps and Android notifications. Patna was chosen for the pilot as the city, and its surrounding area, has a large population and has one of the most frequent occurrences of severe riverine flooding.

“After we sent out the alerts, this allowed us to look into how people experience and interact with the alerts we send. One example of this is that we’ve realised that many people who receive these alerts prefer a stronger emphasis on text that describes the same information our maps show,” Sella Nevo from Google’s Research and Machine Intelligence team, had said.

In the early stages of the Patna pilot, where a lot of time was spent understanding the needs of people in the area, as well as the river morphology and conditions on the ground, a big learning for Google was that people who received these alerts preferred a stronger emphasis on text that described the same information that Google’s maps showed.

“This allowed us to be relatively confident that our pilot will already have a significant impact, and will also be informative as we prepare to scale our efforts up … Our alerts, though limited in scope for our first pilot, achieved results we’re happy with,” Nevo had said.