Scientists in America have developed a “digital twin” of Earth that can predict the weather faster than traditional services. The system, created by Nvidia, uses graphics processing units (GPUs) to carry out high-speed mathematical equations, enabling tasks like machine learning. The platform, known as Earth-2, utilizes machine learning technologies and a supercomputer to create a virtual replica of Earth’s climate.
Earth-2 allows users to run simulations of weather and climate across the planet, with a high level of detail down to 2 km. The AI technology is trained on global data sets to estimate climate changes over the past 50 years. The system quickly generates potential outcomes and calculates the probability of weather conditions in specific locations, providing real-time predictions within seconds.
In places like Taiwan, where severe weather is common, the Earth-2 system can make a significant difference in predicting and planning for extreme weather events. The supercomputer can process vast amounts of data to build accurate predictive simulations, helping cities prepare for events like heat waves in the future.
The Earth-2 system is already being tested by Taiwan’s weather agency to forecast and plan for extreme weather events this year. Its ability to quickly sift through data and provide real-time predictions could save lives and prevent damage in the face of severe weather events.