Scientists have built a digital twin of Earth that can simulate the entire planet at a one-kilometer scale. It’s precise enough to track the birth of a thunderstorm or the movement of a single cloud over a mountain range. It’s also a sign of how far we’ve gone to understand what we’re doing to our own world.
The project, led by Daniel Klocke of the Max Planck Institute in Germany, uses a model called ICON—short for ICOsahedral Nonhydrostatic—which divides the planet into 672 million individual cells. Each cell represents a patch of air, water, or ground, and the model runs calculations on temperature, wind, and energy flow for every one of them.
At that scale, the system merges short-term weather forecasting with long-term climate modeling, a combination researchers once described as the “holy grail” of their field. Traditional models cover 40-kilometer stretches at best, which means small-scale weather—the kind that destroys coastlines and rearranges summers—gets blurred out of the picture.
The team ran the simulation on two of Europe’s most powerful supercomputers, JUPITER in Germany and Alps in Switzerland, both packed with Nvidia GH200 Grace Hopper chips. Each chip connects a CPU to a GPU, giving the model the range to process thunderstorms in seconds and ocean chemistry over decades. The result was staggering: 145 days of global weather simulated in a single day.
That kind of computation doesn’t come easy. The original climate code was written in Fortran, a relic from the Cold War era, so the researchers rewired it using a modern framework called Data-Centric Parallel Programming to handle the load. The final system relied on more than 20,000 superchips and nearly a trillion individual calculations.
It’s not something your local weather app can handle anytime soon. Supercomputers capable of that workload are usually busy doing more profitable things, like training chatbots or optimizing ads. But for scientists, this model shows what’s possible when we stop simulating shortcuts and start simulating the planet itself.
The team describes it as a test run for a more complete digital Earth—one that could, in theory, predict climate futures before we live through them. Whether anyone listens to what it says is another question entirely.
The post Scientists Built a Digital Earth That Can Predict Real Disasters appeared first on VICE.
