AI and the World's Hard Problems · Earth & climate science · arXiv · 2026-10-20
BridgeCast: Bridging Ocean Wave Forecasts to Reanalysis via Flow Matching with Exogenous Variables
Ocean wave forecasts from physics-based models carry systematic biases due to approximations in their parameterizations. The authors propose BridgeCast, a probabilistic physics-AI hybrid model based on conditional flow matching that treats physical forecasts as corrupted observations and transforms them toward reanalysis-like fields, with atmospheric variables as additional inputs. On real-world data, it consistently outperformed state-of-the-art baselines across regions and lead times.