How AI is Improving Access to Reliable Flood Forecasts
Simon Storm Frigon
? CEO of cdrg+RedTeam? | ? My mission is to be the "Diversified Entrepreneur"??. 3rd Gen. Entrepreneur | Rebuilding Environments by Design since 1955 ??
Did you know that floods are the most common type of natural disaster worldwide, resulting in roughly $50 billion in annual financial damages. Shockingly, 1.5 billion people, or 19% of the global population, are exposed to significant risks from severe flood events.
The good news is, Google has been using AI to predict floods up to seven days in advance with great success. Further, upgrading early warning systems can make accurate and timely information accessible to these populations, saving thousands of lives per year.
Therefore if floods are the most common natural disaster worldwide, any early warning system is good news. Google's approach involved training machine learning models with relevant data, including historical events, river level readings, elevation and terrain readings, and more.
After generating localized maps and running "hundreds of thousands" of simulations in each location, the models accurately predicted upcoming floods. This technology allowed Google to offer reliable forecasting to residents of 80 countries, with a total population of 460 million.
The company made these forecasts available in Google Search, Google Maps, and via Android notifications, as well as through its proprietary Flood Hub web app.
What's next? Google will continue to explore the potential of machine learning to create better flood forecasting models and has partnered with academic researchers to fine-tune the AI-driven approach. The company hopes this will eventually result in a global end-to-end flood forecasting platform.