This paper shows how AI can combine broad global weather forecasts with very detailed satellite images of the ground, allowing them to predict local conditions like temperature or wind speed much more precisely.
Earth observation embeddings are effective sub-grid descriptors for probabilistic weather downscaling · Pedro Sousa, Will Tebbutt, Sadiq Jaffer et al. · arXiv:2608.12271
A system to predict when radio waves from Earth will travel unusually far, interfering with sensitive radio telescopes. It uses standard weather forecasts to anticipate these atmospheric "ducts" that trap and guide radio signals, making them travel hundreds of kilometers further than usual.
Tropospheric Ducting Prediction based on GFS Model Data at Inyarrimanha Ilgari Bundara, the CSIRO Murchison Radio-astronomy Observatory · Balthasar Indermuehle, Hajime Suzuki · arXiv:2608.12861
This paper uses satellite measurements to track Ghana's underground water, employing a smart computer program to find unusual dry or wet periods that traditional methods might miss.
Unsupervised Detection of Groundwater Storage Anomalies in Ghana Using GRACE Satellite Data · George Yamoah Afrifa, Theophilus Ansah-Narh, Marcellin Atemkeng · arXiv:2608.10233
This paper rigorously evaluates new AI models, similar to those behind ChatGPT but designed for movement data, to see if they can reliably interpret our physical activity from wearables for health insights.
Foundation models for movement data: Are they ready for prime-time? · Alexander Bräuer, Benjamin Cauchi, Nils Strodthoff · arXiv:2608.13316
This paper tests how well new AI models, trained on today's climate, can predict how oceans would respond to very different natural conditions, like those 6,000 years ago. It finds these AI models can capture many large-scale changes, but they struggle with the full strength of these changes and the deep ocean's slow, internal evolution.
Paleoclimate Boundary Conditions as an Out-of-Sample Test for the Forced Response of Ocean Climate Emulators · Adam Subel, Laure Zanna · arXiv:2608.13494
This paper uses a special camera that sees many colors beyond what our eyes can (hyperspectral imaging) to predict how fresh fish are, day by day.
Few-Shot Ordinal Learning for Day-Wise Freshness Estimation with Hyperspectral Fish Images · Kazi Nabiul Alam, Pooneh Bagheri Zadeh, Akbar Sheikh-Akbari · arXiv:2608.12230
A new AI model that learns a comprehensive understanding of heart activity by simultaneously analyzing electrical, blood flow, and sound signals. Instead of treating each signal separately, this powerful model learns a shared "language" of the heart's function, even accounting for the natural time delays between these different events.
CardioState-JEPA: Delay-Aware Cross-Modal Learning of a Shared Cardiac Representation · Hamza Shafiq, Hung Manh Pham, Bin Zhu et al. · arXiv:2608.12944
This paper challenges a common computer modeling method that uses observed ocean temperatures to understand how Earth's warming response changes over time.
Statistical Noise and Missing Forcing Limit Estimates of Earth's Feedback from Prescribed Sea-Surface Temperature Simulations · Gergana Gyuleva, Reto Knutti, Robin Noyelle et al. · arXiv:2608.13219