analiza aiqbench
SDIR addresses a critical trade-off in precipitation nowcasting by decoupling low-frequency structural extraction from high-frequency texture refinement to eliminate both over-smoothing and ungrounded hallucinations. By enforcing physically consistent spectral distributions through a specialized loss function, the framework achieves superior spatial accuracy while maintaining the fidelity of diffusion-based models. This deterministic approach offers a robust path forward for reliable, high-resolution operational forecasting in disaster mitigation scenarios.
Learning to Refine: Spectral-Decoupled Iterative Refinement Framework for Precipitation Nowcasting
arXiv:2606.02661v1 Announce Type: cross Abstract: Accurate precipitation nowcasting is vital for disaster mitigation, but deep learning methods face a key trade-off: regression models produce over-smoothed, spectrally decaying predictions that blur convective details and violate turbulence power laws; diffusion models generate realistic yet unanchored hallucinations lacking physical grounding. We propose Spectral-Decoupled Iterative Refinement (SDIR), a deterministic framework that reformulates nowcasting as progressive frequency-decoupled refinement. SDIR first extracts a stable low-frequency synoptic skeleton, then iteratively refines high-frequency textures under physical constraints, eliminating both blurring and hallucinations. It features a dual-path design: the Synoptic Frequency-Guided Former (SFG-Former) with Scale-Adaptive Transformers for global structure, and the Fourier Residual Refiner (FR-Refiner) with Scale-Conditioned Fourier Neural Operators for fine residuals. A Physically Consistent Power Spectral Density (PCPSD) loss with dynamic masking enforces a turbulence-consistent spectral distribution. Experiments on three benchmarks show SDIR significantly outperforms SOTA methods in spatial accuracy while achieving spectral fidelity competitive with diffusion-based methods, enabling reliable high-resolution operational nowcasting. Code link: https://github.com/RuntimeWarning/SDIR.
Abstract:Accurate precipitation nowcasting is vital for disaster mitigation, but deep learning methods face a key trade-off: regression models produce over-smoothed, spectrally decaying predictions that blur convective details and violate turbulence power laws; diffusion models generate realistic yet unanchored hallucinations lacking physical grounding. We propose Spectral-Decoupled Iterative Refinement (SDIR), a deterministic framework that reformulates nowcasting as progressive frequency-decoupled refinement. SDIR first extracts a stable low-frequency synoptic skeleton, then iteratively refines high-frequency textures under physical constraints, eliminating both blurring and hallucinations. It features a dual-path design: the Synoptic Frequency-Guided Former (SFG-Former) with Scale-Adaptive Transformers for global structure, and the Fourier Residual Refiner (FR-Refiner) with Scale-Conditioned Fourier Neural Operators for fine residuals. A Physically Consistent Power Spectral Density (PCPSD) loss with dynamic masking enforces a turbulence-consistent spectral distribution. Experiments on three benchmarks show SDIR significantly outperforms SOTA methods in spatial accuracy while achieving spectral fidelity competitive with diffusion-based methods, enabling reliable high-resolution operational nowcasting. Code link: this https URL.
Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2606.02661 [eess.IV]
(or arXiv:2606.02661v1 [eess.IV] for this version)
https://doi.org/10.48550/arXiv.2606.02661 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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 · 258 cuvinte
Subiecte conexe
Iterative Refinement Diffusion for Super-Resolved Data Assimilation of Multiscale Physical Systems
arXiv · 19 aug. 2026
Physics-Guided Dual Decoding and Spectral Supervision for Global 3D Hydrometeor Prediction
arXiv · 9 iun. 2026
SemMSA: Latent Semantic-Aided Robust Multimodal Sentiment Analysis with Incomplete Data
arXiv · 25 sept. 2026
Beyond Simulation: Retain-and-Repair Neural Operators for Real-World Adaptation
arXiv · 1 oct. 2026