Google researchers use AI technology to predict floods in India, potentially preventing loss of life

The rapid growth of artificial intelligence (AI) has sparked concerns about its potential for both positive and negative impacts. Despite fears about the risks associated with this technology, there is no slowing down its progress. Google is leading the way in using AI for the benefit of humanity, as seen in its partnership with Apollo Hospitals in India to aid in early disease detection. Researchers are now turning their attention to predicting floods using AI technologies, with the goal of potentially saving lives.

Google researchers use artificial intelligence technology to predict floods

In a research paper published in the journal Nature, Google researchers shed light on how they are harnessing artificial intelligence to work toward the goal of flood forecasting. Machine learning (ML) is used to create hydrological simulation models that can be transferred to basins that are not yet measured. To do this, the researchers say they will use version 4 of the Global Flood Awareness System (GloFAS), which is used by the European Center for Medium-Range Weather Forecasts (ECMWF). Comparing GloFAS to the latest river forecasting models developed by researchers reveals that it not only provides accurate forecasts earlier, but also covers “bigger and more impactful events”.

Yossi Matias, Director of Technology and Research, and Gray Nearing, Research Scientist, Google Research, said: “The flood forecasting initiative is part of our adaptation and resilience efforts and reflects Google’s commitment to combating climate change while helping global communities become more resilient. We believe AI and ML will continue to play a critical role in promoting science and research towards climate action.”

The AI model used in the study predicts daily flow over a 7-day forecast horizon using long-short-term memory (LSTM) networks. The researchers say this model was trained on a sample of 5,680 flowmeters using “random k-fold cross-validation.” This means the Flood Hub system can deliver river forecasts up to 7 days in advance, covering rivers in 80 countries.

In their multi-year journey, the researchers have “advanced research over the years hand-in-hand by building a real-time operational flood forecasting system that provides alerts via Google search, Maps, Android notifications and Flood Hub.”