We all check the weather before leaving our houses. We have to make sure that we are prepared for really sunny or rainy conditions. We also need to know how the temperature will change throughout the whole day.
Unfortunately even though modern meteorology has come a long way, the complexity of using hundreds of data points and methods to predict the weather makes accurate forecasts short-handed. Good thing Artificial Intelligence has come into play to improve the existing ways of weather forecasting. Google DeepMind recently announced a system named GenCast. The AI powered program is set to make things a lot better.
The program was trained on four decades worth of historical data through 2018, taken from The European Centre for Medium-Range Weather Forecasts’ (ECMWF) historical archives. The ECMWF is considered as the most accurate weather-prediction service in the world. Its archival data from its years of service includes wind speed, temperature, and pressure at different altitudes. All these data points went into the training program.
After the AI model’s training. It was tasked to predict the weather for fifteen day periods in 2019. GenCast’s prediction was then compared to the actual weather that occurred in that year, GenCast hit a 97.2% accuracy rate which surpasses that of ECMWF Prediction. When they had to step it up and narrowed down the window for prediction time to just thirty six hours, GenCast did even better by being 99.8% more accurate than ECMWF, with this rate GenCast, outshines Google’s previous weather prediction program known as GraphCast.
The new system generates over 50 predictions about potential weather conditions. It bases these predictions on current trends. Then, the system combines these predictions to form its forecast. Google claims that using one of the company’s TPU v5 AI processors, the system can produce a fifteen day forecast in just eight minutes, compared to the several hours required by supercomputers to achieve the same result.
Google says that, beyond helping you choose the right outfit for the weather, GenCast could save lives. It forecasts the trajectory of severe weather events. These events are becoming more common due to climate change several days in advance. The company also notes that the data generated by GenCast could support renewable energy initiatives, such as identifying wind patterns to optimize the placement and efficiency of wind farms.
You can’t yet download an app powered by GenCast. However, Google is making the data from its testing and real-time forecasts publicly available. The company is encouraging researchers, meteorologists, and other organizations to leverage this technology. As a result, don’t be surprised if your current weather app becomes more accurate in the near future.
More information behind the research done in GenCast has been publish in the journal Nature
Comments 3