![[2511.09731] FlowCast: Advancing Precipitation Nowcasting with Conditional Flow Matching](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2511.09731] FlowCast: Advancing Precipitation Nowcasting with Conditional Flow Matching
Summary
FlowCast introduces a novel probabilistic model for precipitation nowcasting using Conditional Flow Matching, improving accuracy and efficiency in short-term weather forecasting.
Why It Matters
Accurate precipitation forecasting is crucial for flood risk management and decision-making. FlowCast addresses existing challenges in atmospheric dynamics and high-dimensional data modeling, offering a more efficient alternative to traditional methods, which can significantly impact disaster preparedness and response strategies.
Key Takeaways
- FlowCast leverages Conditional Flow Matching for efficient precipitation nowcasting.
- The model achieves state-of-the-art probabilistic performance and surpasses deterministic baselines.
- FlowCast's approach reduces computational costs with fewer sampling steps compared to diffusion models.
- The research highlights the importance of high-fidelity forecasting in flood risk management.
- CFM is positioned as a practical alternative for spatiotemporal forecasting challenges.
Computer Science > Machine Learning arXiv:2511.09731 (cs) [Submitted on 12 Nov 2025 (v1), last revised 22 Feb 2026 (this version, v3)] Title:FlowCast: Advancing Precipitation Nowcasting with Conditional Flow Matching Authors:Bernardo Perrone Ribeiro, Jana Faganeli Pucer View a PDF of the paper titled FlowCast: Advancing Precipitation Nowcasting with Conditional Flow Matching, by Bernardo Perrone Ribeiro and 1 other authors View PDF HTML (experimental) Abstract:Radar-based precipitation nowcasting, the task of forecasting short-term precipitation fields from previous radar images, is a critical problem for flood risk management and decision-making. While deep learning has substantially advanced this field, two challenges remain fundamental: the uncertainty of atmospheric dynamics and the efficient modeling of high-dimensional data. Diffusion models have shown strong promise by producing sharp, reliable forecasts, but their iterative sampling process is computationally prohibitive for time-critical applications. We introduce FlowCast, the first end-to-end probabilistic model leveraging Conditional Flow Matching (CFM) as a direct noise-to-data generative framework for precipitation nowcasting. Unlike hybrid approaches, FlowCast learns a direct noise-to-data mapping in a compressed latent space, enabling rapid, high-fidelity sample generation. Our experiments demonstrate that FlowCast establishes a new state-of-the-art in probabilistic performance while also exceeding determin...
