analiza aiqbench
This study demonstrates that adding lightweight temporal conditioning layers to existing deep learning architectures significantly enhances the accuracy of high-intensity rainfall nowcasting, particularly by capturing seasonal variability. The proposed TA-SmaAt-UNet model achieves these improvements with minimal parameter overhead, confirming that simple, physically motivated context encoding can boost the reliability of radar-based precipitation forecasts. For practitioners in environmental AI, this offers a practical pathway to integrate broader meteorological context into motion-focused models without incurring heavy computational costs.
Temporal Context Conditioning for Seasonality-Aware Precipitation Nowcasting of High-Intensity Rainfall
arXiv:2606.09959v1 Announce Type: new Abstract: Precipitation nowcasting is increasingly being approached with deep learning models that learn directly from recent radar observations. Although such models can efficiently capture short-term precipitation motion, they often lack broader contextual information about the meteorological conditions under which rainfall develops. This paper investigates whether lightweight temporal context can improve radar-based nowcasting, particularly for high-intensity rainfall. We propose the Time-Aware Small-Attention U-Net (TA-SmaAt-UNet), which extends the core SmaAt-UNet model with temporal conditioning layers that use cyclical encodings of time-of-day and time-of-year to modulate intermediate feature representations. Experiments on KNMI radar precipitation data show that temporal conditioning is most beneficial for rare, high-intensity precipitation events, while also improving the representation of seasonal variability and predicted rainfall-intensity distributions. A layer conductance analysis further indicates that the added temporal conditioning layers are actively used by the model despite their small parameter cost. These findings suggest that simple, physically motivated temporal context can improve the realism and reliability of deep learning-based precipitation nowcasts. The implementation of our models and training setup is available on \href{https://github.com/gijsvn/TA-SmaAt-UNet}{GitHub}.
Abstract:Precipitation nowcasting is increasingly being approached with deep learning models that learn directly from recent radar observations. Although such models can efficiently capture short-term precipitation motion, they often lack broader contextual information about the meteorological conditions under which rainfall develops. This paper investigates whether lightweight temporal context can improve radar-based nowcasting, particularly for high-intensity rainfall. We propose the Time-Aware Small-Attention U-Net (TA-SmaAt-UNet), which extends the core SmaAt-UNet model with temporal conditioning layers that use cyclical encodings of time-of-day and time-of-year to modulate intermediate feature representations. Experiments on KNMI radar precipitation data show that temporal conditioning is most beneficial for rare, high-intensity precipitation events, while also improving the representation of seasonal variability and predicted rainfall-intensity distributions. A layer conductance analysis further indicates that the added temporal conditioning layers are actively used by the model despite their small parameter cost. These findings suggest that simple, physically motivated temporal context can improve the realism and reliability of deep learning-based precipitation nowcasts. The implementation of our models and training setup is available on \href{this https URL}{GitHub}.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2606.09959 [cs.LG]
(or arXiv:2606.09959v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.09959 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 267 cuvinte
Subiecte conexe
Quantifying Event Impacts on Time Series via Multiscale Contrastive Learning
arXiv · 21 aug. 2026
HyperPatch: Sequential Knowledge Editing Under n-ary Structural Drift
arXiv · 3 iun. 2026
Presentation: Building Evals for AI Adoption: from Principles to Practice
InfoQ · 29 mai 2026
Neural Video Compression Based on Deformable Temporal Alignment and Difference-aware Fusion
arXiv · 5 sept. 2026