Deep learning is increasingly being used to emulate cloud and convection processes in climate models, offering a faster alternative to computationally intensive cloud-resolving simulations. However, because these processes unfold at kilometre scales or smaller — far below the resolution of global climate models — hybrid AI–physics approaches often become unstable when run over long time horizons. Researchers from the National University of Singapore (NUS)’s College of Design and Engineering (CDE), in collaboration with Tsinghua University, NVIDIA and the Centre for Climate Research Singapore, have traced a key instability to a moisture imbalance during condensation. They developed CondensNet, an adaptive, physics-constrained neural-network that can correct physically inconsistent behaviour. Integrated into a global climate model, it enables stable, accurate long-term simulations while running far faster than cloud-resolving approaches.

Plants grown for biofuel have the potential to power our travel industry, but an important fraction of their chemical power has remained stubbornly difficult to recover. New research from the DOE-funded Center for Advanced Biofuel and Bioproduct Innovation on the University of Illinois campus has demonstrated a way to preserve native lignin structure, a key component of plant matter for conversion to other valuable products, putting an improved pathway for biofuel and bioproduct production within reach.

Harvard researchers have developed a mathematical framework for optimizing the design of rolling contact joints, which are made of rolling surfaces and flexible connectors.

Scientists have created an AI model that forecasts moderate heat stress — a major precursor to coral bleaching — at sites along Florida’s Coral Reef up to six weeks ahead, with predictions generally accurate within one week.

The study presents a site-specific, explainable machine-learning framework to support coral scientists and restoration practitioners with local reef management and emergency response planning.

Rice bioengineers developed a new genetic regulatory system that makes the production of gas vesicles more sustainable for nonnative host cells.

The Earthquake Engineering Research Institute (EERI), the Consortium of Organizations for Strong Motion Observation Systems (COSMOS) and SSA are pleased to announce that Eduardo Miranda of Stanford University is the recipient of the 2026 Bruce Bolt Medal.

MIT researchers developed a miniaturized ultrasound system that could make it easier for ultrasounds to be performed more often, either at home or at a doctor’s office. The system could help more breast tumors to be diagnosed at earlier stages.