Guided Unconditional and Conditional Generative Models for Super-Resolution and Inference of Quasi-Geostrophic Turbulence
arXiv:2507.00719v3 Announce Type: replace-cross Abstract: Typically, numerical simulations of Earth systems are coarse, and Earth observations are sparse and gappy. We apply four generative diffusion modeling approaches to super-resolution and inference of forced two-dimensional quasi-geostrophic turbulence on the beta-plane from coarse, sparse, and gappy observations. Two guided approaches minimally adapt a pre-trained unconditional model: SDEdit modifies the initial condition, and Diffusion Posterior Sampling (DPS) modifies the reverse diffusion process score. Two conditional approaches, a v
The continuous advancements in generative AI models, specifically diffusion models, are enabling increasingly sophisticated applications in scientific domains like Earth systems modeling, where traditional computational methods are often limited.
This development indicates a significant leap in the ability to generate high-resolution data and make inferences from sparse environmental observations, critical for climate science, disaster prediction, and resource management.
Access to more detailed and accurate Earth system simulations and predictions becomes possible, improving understanding of complex natural phenomena and informing policy decisions.
- · Climate scientists
- · Environmental monitoring agencies
- · AI researchers
- · Computational fluid dynamics sector
- · Traditional coarse-resolution modeling approaches
- · Data collection methods reliant purely on high-density sensors
Improved resolution and inference capabilities for Earth system simulations using generative AI.
Enhanced accuracy in climate change projections and prediction of extreme weather events.
New frameworks for environmental policy and infrastructure planning based on higher-fidelity data models.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG