Tiny traces of melted rock are helping scientists uncover the Colorado Plateau's explosive past. These small clues offer a glimpse into the geological processes that took place deep beneath the Earth's surface millions of years ago. 

Researchers at Jeonbuk National University, South Korea, have explored a promising approach to producing hydrogen while removing carbon dioxide from the atmosphere. By combining biomass-based hydrogen production with carbon capture and storage, the study shows how generating electricity on-site can improve environmental performance. The findings highlight the importance of clean energy, efficient production, and sustainable biomass sourcing in developing carbon-negative hydrogen technologies that could support a cleaner energy future. 

A new review in Science China Earth Sciences examines advances and challenges in reconstructing temperatures from soil microbial lipids across arid and semi-arid regions. Published in Chinese and English, it brings together Academician Fahu Chen’s team at the Institute of Tibetan Plateau Research, Chinese Academy of Sciences, and collaborators from the National Research Center for Geoanalysis, Lanzhou University and the University of Pittsburgh. Drawing on evidence from the Pan-Third Pole, particularly the Tibetan Plateau and arid Central Asia, the review highlights the potential of combining bacterial and archaeal lipids to improve terrestrial temperature reconstructions.

How do octopuses remember what they learn? A new study reveals a surprising “molecular switch” in the octopus brain that helps preserve memories by keeping connections between nerve cells strengthened over time, without the need to produce new proteins. The discovery sheds new light on one of nature’s greatest mysteries: how creatures with vastly different brains evolved remarkably sophisticated learning and memory abilities.

Foundation models now excel at single Earth-science tasks, from weather forecasting to flood mapping, but a review in The Innovation Reviews argues that task-level wins do not yet add up to trustworthy Earth system science. Examining six fields — geology, geophysics, atmosphere, ocean, geography, and the integrated Earth system — the authors find that model capability is shaped jointly by data conditions and physics, so lessons from one field do not transfer automatically. They identify five cross-cutting barriers, including the AI-ready Earth data dilemma, physical inconsistency, and deficient scientific validation, and propose six coordinated paths, from a unified Earth representation engine to physics-embedded architectures and human-AI discovery interfaces. The goal is reusable model capability that becomes testable scientific capability.