arXiv cs.AI· Siyu Gan, Dongsheng Luo, Kunxiaojia Yuan, Dongjin Song, Jingchao Ni·· 12 小时前AI 评分20
BridgeCast：用条件流匹配与外生变量将海浪预报桥接至再分析数据
BridgeCast: Bridging Ocean Wave Forecasts to Reanalysis via Flow Matching with Exogenous Variables
AI 导读
BridgeCast 是一个基于条件流匹配（CFM）的概率模型，将物理模型的海浪预报视为受损观测，通过连续时间生成过程把其分布桥接至再分析数据，实现偏差校正。模型采用 Transformer 架构进行时空建模，可引入外生大气变量，并支持常微分与随机微分方程两种推理形式。在真实数据集上，BridgeCast 在各区域和预报时效上均持续优于 SOTA 基线。
正文 · 原文
Abstract:Ocean wave forecasting is essential for maritime safety, offshore operations, and coastal resilience, yet remains challenging due to systematic biases in physics-based models. Physical models, while widely used, rely on approximations and parameterizations that limit their accuracy under complex ocean-atmosphere conditions. To enhance ocean wave forecasting, we propose BridgeCast, within a physics-AI hybrid framework for bias correction. BridgeCast is a probabilistic model based on conditional flow matching (CFM) that learns to transform physical model forecasts into reanalysis-like fields. It treats physical forecasts as corrupted observations and employs a continuous-time generative process to bridge their distribution toward that of reanalysis data. BridgeCast is parameterized by a Transformer-based architecture that enables spatiotemporal modeling, incorporation of exogenous atmospheric variables, and flexible inference via both ordinary and stochastic differential equation formulations. Extensive experiments on real-world datasets demonstrate that BridgeCast consistently outperforms state-of-the-art baselines across regions and forecast lead times.
| Comments: | 22 pages, including references and appendix |
| Subjects: | Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.03759 [physics.ao-ph] |
| (or arXiv:2610.03759v3 [physics.ao-ph] for this version) | |
| https://doi.org/10.48550/arXiv.2610.03759 arXiv-issued DOI via DataCite |
Submission history
From: Siyu Gan [view email]
[v1]
Sun, 27 Sep 2026 00:43:54 UTC (4,060 KB)
[v2]
Sat, 17 Oct 2026 18:42:51 UTC (4,105 KB)
[v3]
Tue, 20 Oct 2026 16:16:21 UTC (4,105 KB)
来源：arXiv cs.AI · arxiv.org