Tag
This paper presents a multimodal auto-regressive transformer surrogate that models variable well operations and geological uncertainty for geological carbon storage, achieving accurate predictions and enabling uncertainty quantification via MCMC data assimilation.
Proposes KL-DNN, a scalable operator learning framework that uses Karhunen-Loève expansions to handle large-scale PDE problems, achieving lower errors and two-order-of-magnitude speedup over DeepONet on a 3D carbon storage problem.
A study maps global arbuscular mycorrhizal fungal networks, finding their threads can stretch far beyond Earth and that they are threatened by agricultural conversion, with grassland ecosystems holding 40% of the world's biomass but being poorly protected.