|
||||||||||||
![]() |
||||||||||||
|
Home > TERATEC FORUM > Workshops
HPC and AI serving the Climate: Challenges and Opportunities
One of the main thrusts of climate and climate change research involves running numerous numerical simulations with climate models, typically thousands of years in the future, in order to estimate uncertainties. This has prompted the climate research community (as well as the weather forecasting community) to evaluate model emulation approaches using AI, with convincing initial results. However, these approaches can have limitations, such as a lack of consistency in the set of simulated variables, even if it is possible to introduce physical constraints. This observation has motivated hybrid modeling approaches, which involve replacing certain parts of a climate model with AI algorithms. These include the computationally-intensive parts, or those representing the effect on climate of phenomena not explicitly resolved by the model (parameterizations). These approaches are producing promising results, but at the cost of painstaking work to ensure the stability of hybrid models, which must be capable of simulating climate over thousands of years. In addition, AI methods can also be mobilized for equation discovery approaches to guide model development, or to deepen analyses of simulation or observation data on climate, its variability and extremes.
Register now and get your badge here >>>
|
|||||||||||
© Ter@tec - All rights reserved - Lawful mention |