Nowcast3D: Reliable precipitation nowcasting via gray-box learning

2025-11-06

AI Summary

Extreme precipitation nowcasting faces a fundamental trade-off between high spatiotemporal resolution and long lead times: numerical weather prediction (NWP) and deep learning simulations suffer from high computational cost and coarse resolution; extrapolation- and purely data-driven methods accumulate errors and produce over-smoothed forecasts; existing 2D radar models neglect vertical dynamical information. Method: We propose the first full 3D gray-box framework that directly models the 3D spatiotemporal evolution of volumetric radar reflectivity scans. It innovatively incorporates a vertically varying conservative advection field, spatially adaptive diffusion parameterization, and a Brownian-motion-inspired stochastic term—integrating physics-constrained neural operators, residual-based convective initiation modeling, and diffusion-type uncertainty estimation. Contribution/Results: End-to-end trained, it delivers high-accuracy 3-hour forecasts. In double-blind evaluation, it ranked first in 57% of cases according to scores from 160 meteorological experts, significantly improving forecast skill and credibility.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Extreme precipitation nowcasting demands high spatiotemporal fidelity and extended lead times, yet existing approaches remain limited. Numerical Weather Prediction (NWP) and its deep-learning emulations are too slow and coarse for rapidly evolving convection, while extrapolation and purely data-driven models suffer from error accumulation and excessive smoothing. Hybrid 2D radar-based methods discard crucial vertical information, preventing accurate reconstruction of height-dependent dynamics. We introduce a gray-box, fully three-dimensional nowcasting framework that directly processes volumetric radar reflectivity and couples physically constrained neural operators with datadriven learning. The model learns vertically varying 3D advection fields under a conservative advection operator, parameterizes spatially varying diffusion, and introduces a Brownian-motion--inspired stochastic term to represent unresolved motions. A residual branch captures small-scale convective initiation and microphysical variability, while a diffusion-based stochastic module estimates uncertainty. The framework achieves more accurate forecasts up to three-hour lead time across precipitation regimes and ranked first in 57% of cases in a blind evaluation by 160 meteorologists. By restoring full 3D dynamics with physical consistency, it offers a scalable and robust pathway for skillful and reliable nowcasting of extreme precipitation.

Problem

• Improves precipitation nowcasting accuracy and lead time for extreme weather events
• Addresses limitations of 2D methods by incorporating full 3D vertical dynamics
• Reduces error accumulation and excessive smoothing in existing nowcasting models

Innovation

• Gray-box 3D framework coupling physics with neural operators
• Learns vertical advection fields with conservative physical constraints
• Incorporates stochastic diffusion module for uncertainty estimation