Researchers have developed FNO-MoE, a novel Fourier neural operator with a mixture of experts, designed to significantly accelerate the simulation of atmospheric chemistry. This model can predict the full near-explicit chemical state over 30-minute intervals in a single call, a task that is computationally prohibitive for traditional Earth-system models. FNO-MoE also guides its own training by selecting the most informative simulations, leading to a more accurate emulator than random sampling methods. The developed emulator demonstrates high fidelity in reproducing the evolution of thousands of species and runs orders of magnitude faster than existing numerical solvers, paving the way for integrating near-explicit chemistry into Earth-system models. AI
IMPACT This development could enable more accurate and faster climate modeling by integrating complex atmospheric chemistry simulations.
RANK_REASON This is a research paper describing a new model and methodology. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
