{"id":"arxiv/2603.28671","url":"https://arxiv.org/abs/2603.28671","tags":["math.DS","physics.ao-ph","physics.flu-dyn"],"score":0,"title":"Stochasticity and probabilistic trajectory scoring are essential for data-driven closures of chaotic systems","domain":"math.DS","source":"arxiv","authors":["Martin Thomas Brolly"],"reviews":0,"subject":"math.DS","summary":"Coarse-grained models of chaotic systems neglect unresolved degrees of freedom, inducing structured model error that limits predictability and distorts long-term statistics. Typical data-driven closures are trained to minimize error over a single time step, implicitly assuming Markovian dynamics and often failing to capture long-term behavior. Recent approaches instead optimize losses over finite trajectories. However, when such trajectory-based training is carried out with deterministic pointwise losses, it introduces a fundamental mathematical degeneracy. We prove that optimizing pointwise…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-30","reactions":{}}