EO-WM: a physically informed world model for probabilistic Earth-observation forecasting

arXiv·low signal

This arXiv paper (2026-06-25) frames satellite-based Earth-observation forecasting as a partially observed, weather-driven world-modeling problem, where weather is a conditioning signal and prediction stays uncertain due to sparse observations and unobserved land-surface states. EO-WM produces probabilistic forecasts rather than collapsing uncertainty as deterministic models do. The probabilistic-world-model framing is a transferable pattern for builders modeling partially observed physical systems.