![[2505.10271] RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2505.10271] RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours
Summary
RainPro-8 is a novel deep learning model designed for high-resolution rainfall probability forecasting over an 8-hour horizon, integrating multiple data sources for improved accuracy.
Why It Matters
Accurate precipitation forecasting is critical for various sectors, including agriculture, disaster management, and urban planning. RainPro-8's integration of diverse data sources and its efficient architecture could significantly enhance forecasting capabilities, addressing the limitations of existing models.
Key Takeaways
- RainPro-8 integrates radar, satellite, and numerical weather prediction data for enhanced forecasting.
- The model achieves superior accuracy compared to existing operational systems and deep learning models.
- Its compact architecture allows for efficient training and faster inference times.
- Robust uncertainty quantification is provided through consistent probabilistic maps.
- The model sets a new standard for high-resolution precipitation forecasting in Europe.
Computer Science > Machine Learning arXiv:2505.10271 (cs) [Submitted on 15 May 2025 (v1), last revised 16 Feb 2026 (this version, v3)] Title:RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours Authors:Rafael Pablos Sarabia, Joachim Nyborg, Morten Birk, Jeppe Liborius Sjørup, Anders Lillevang Vesterholt, Ira Assent View a PDF of the paper titled RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours, by Rafael Pablos Sarabia and 5 other authors View PDF HTML (experimental) Abstract:We present a deep learning model for high-resolution probabilistic precipitation forecasting over an 8-hour horizon in Europe, overcoming the limitations of radar-only deep learning models with short forecast lead times. Our model efficiently integrates multiple data sources - including radar, satellite, and physics-based numerical weather prediction (NWP) - while capturing long-range interactions, resulting in accurate forecasts with robust uncertainty quantification through consistent probabilistic maps. Featuring a compact architecture, it enables more efficient training and faster inference than existing models. Extensive experiments demonstrate that our model surpasses current operational NWP systems, extrapolation-based methods, and deep-learning nowcasting models, setting a new standard for high-resolution precipitation forecasting in Europe, ensuring a balance between accuracy, interpretability, and computational...
