Answers are grounded in this paper and cite their sources.
Diffusion models represent a novel class of generative models with exceptional generative capabilities across various domains, including computer vision and multimodal learning. As research progresses, their applications have increasingly extended to meteorology and oceanography, providing innovative solutions to challenges such as data noise and the rising frequency of extreme weather events. By simulating the noise diffusion process in datasets, they can accurately capture latent patterns and trends, facilitating data assimilation, forecasting, and post-processing. However, a significant gap exists in the literature concerning comprehensive reviews of the current state of these models. This paper aims to thoroughly analyze the potential applications of diffusion models in meteorology and oceanography. First, it reviews the development of diffusion models, outlines fundamental principles and derivative models, and discusses application strategy and advantages. Next, existing research is categorized into three types based on tasks: data assimilation, forecasting, and post-processing, with detailed discussions on the technical features of each category. Finally, this paper examines future development, analyzes corresponding challenges, and provides insights to foster further research and application of diffusion models in these areas.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: