20,000 GPUs and a consumption of 20 megawatts: this is the project that wants to build a "digital twin" Earth to simulate the world’s weather in real time

On December 29, 1972, Edward Lorenz spoke at the annual meeting of the American Association for the Advancement of Science and uttered a phrase that has persecuted us ever since: «The flapping of a butterfly in Brazil can cause a tornado in Texas». A decade earlier, while trying to make sense of a handful of weather models, he had discovered that there are things that can behave unpredictably.

In other words, Lorenz had discovered the “essence of chaos” and had done it precisely with meteorology because, in short, time is complicated. So complicated that a team of researchers from ETH Zurich Is conviced that the only way to simulate the climatic dynamics and atmospheric variations of the Earth is creating a “digital twin” of her.

The mainstream of digital twins

The idea is not new and, in fact, already used in physiology and immunology to create “digital twins” of patients that help physicians simulate different treatments with highly accurate models updated in real time with dozens of clinical data. However, what the Swiss team wants to do goes much further.

The terrestrial twin it would be a much more sophisticated computer system. Not only because the goal (to simulate the climate and weather across the entire Earth), but because the amount of real-time data that the model would have to integrate is unmatched in anything that has been done so far in the field: researchers calculate that it would take 20,000 GPUs and it would consume 20 megawatts of power.

The project has the support of institutions such as the European Space Agency, the European Organization for the Exploitation of Meteorological Satellites or the European Center for Medium-Range Weather Forecasts and if it finally saw the light could become in a large-scale scientific project because it would involve the joint development of the hardware and algorithms necessary to carry out a very high-resolution simulation. That is to say, a golden opportunity to break the relative stagnation experienced by conventional processors.

However, it also harbors many uncertainties. The main one is that we don’t know if it will help. That is, yes, we can build the most complex and accurate model of Earth’s weather and meteorology, worse We don’t know if our predictions about him were really worth it.

or that Brazilian butterfly will continue to play with us. But I suppose that is precisely why it is a scientific project: because it opens doors for us that we do not even know exist.

Image | THIS

–
The news

20,000 GPUs and a consumption of 20 megawatts: this is the project that wants to build a “digital twin” of the Earth to simulate the world’s climate in real time

was originally published in

Engadget

for
Javier Jimenez

.