![]() |
ИСТИНА |
Войти в систему Регистрация |
ИПМех РАН |
||
Weather and climate sector has been for long one of major consumers of supercomputing hardware and programming techniques. Due to complexity of the Earth system, this domain is "supercomputing" by nature. Since the first von Neumann machines, the paradigm of developing the weather&climate prediction codes has based on mathematical physics and a solid belief in the Moore law. In latest decades, the additional improvement of the forecast quality costs significant coordinated efforts in physical parameterizations, data assimilation, ensemble forecasting, parallel computing. The recent breakthrough in the AI-driven weather forecast promises the further progress in fidelity of predictions and even the drop of computational costs. Can AI replace the hydrodynamic models? What are the perspectives for AI in producing climate projections? How the niche of physics-based models is changing? We will try to touch these questions of the crucial research domain with rapidly changing landscape.