Skillful Precipitation Nowcasting using Deep Generative Models of Radar

2021/04/02 by Suman Ravuri, Karel Lenc, Matthew Willson +20 · 53 citations
Computer Science · Earth and Planetary Sciences · Environmental Science · #Flood Risk Assessment and Management #Meteorological Phenomena and Simulations #Precipitation Measurement and Analysis #cs.LG

paper · pdf · doi:10.1038/s41586-021-03854-z

46 pages, 17 figures, 2 tables

arxiv created 2021/04/02 · openalex publication_date 2021/09/29 · arxiv updated 2021/11/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/30

Abstract

Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socio-economic needs of many sectors reliant on weather-dependent decision-making. State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-based wind estimates, and struggle to capture important non-linear events such as convective initiations. Recently introduced deep learning methods use radar to directly predict future rain rates, free of physical constraints. While they accurately predict low-intensity rainfall, their operational utility is limited because their lack of constraints produces blurry nowcasts at longer lead times, yielding poor performance on more rare medium-to-heavy rain events. To address these challenges, we present a Deep Generative Model for the probabilistic nowcasting of precipitation from radar. Our model produces realistic and spatio-temporally consistent predictions over regions up to 1536 km x 1280 km and with lead times from 5-90 min ahead. In a systematic evaluation by more than fifty expert forecasters from the Met Office, our generative model ranked first for its accuracy and usefulness in 88% of cases against two competitive methods, demonstrating its decision-making value and ability to provide physical insight to real-world experts. When verified quantitatively, these nowcasts are skillful without resorting to blurring. We show that generative nowcasting can provide probabilistic predictions that improve forecast value and support operational utility, and at resolutions and lead times where alternative methods struggle.

Citations

• Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
• THE WORLD METEOROLOGICAL ORGANIZATION
• Scale-Selective Verification of Rainfall Accumulations from High-Resolution Forecasts of Convective Events
• The Critical Success Index as an Indicator of Warning Skill
• On Calibration of Modern Neural Networks
• Self-Attention Generative Adversarial Networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Adversarial Video Generation on Complex Datasets
• MetNet: A Neural Weather Model for Precipitation Forecasting
• Generative Adversarial Nets
• Exploiting Generative AI to Scale up Intelligent Tutoring Systems

Cited by

• Learning skillful medium-range global weather forecasting
• Postprocessing East African Rainfall Forecasts Using a Generative Machine Learning Model
• Challenges in data-driven geospatial modeling for environmental research and practice
• Selective Learning for Deep Time Series Forecasting
• A PDE-Informed Latent Diffusion Model for 2-m Temperature Downscaling
• Enhanced forecasting of bird nocturnal migration intensity in relation to previous days and synoptic weather patterns
• HPC-Driven Modeling with ML-Based Surrogates for Magnon-Photon Dynamics in Hybrid Quantum Systems
• Hierarchical Graph Networks for Accurate Weather Forecasting via Lightweight Training
• SynCast: Synergizing Contradictions in Precipitation Nowcasting via Diffusion Sequential Preference Optimization
• RainDiff: End-to-end Precipitation Nowcasting Via Token-wise Attention Diffusion
• An Operational Deep Learning System for Satellite-Based High-Resolution Global Nowcasting
• Leveraging Teleconnections with Physics-Informed Graph Attention Networks for Long-Range Extreme Rainfall Forecasting in Thailand
• DAWP: A framework for global observation forecasting via Data Assimilation and Weather Prediction in satellite observation space
• Simple and Robust Forecasting of Spatiotemporally Correlated Small Earth Data with A Tabular Foundation Model
• SimCast: Enhancing Precipitation Nowcasting with Short-to-Long Term Knowledge Distillation
• BlockGPT: Spatio-Temporal Modelling of Rainfall via Frame-Level Autoregression
• Probability calibration for precipitation nowcasting
• ClimSat – A diffusion autoencoder model for climate-conditional satellite image editing
• A Weather Foundation Model for the Power Grid
• Learning Regional Monsoon Patterns with a Multimodal Attention U-Net
• Forecasting Land Art Under Climate Scenarios
• Geostationary Satellite–Based Proxy Radar Observations: Expanding Coverage for Storm Tracking
• Knowledge-Guided Adaptive Mixture of Experts for Precipitation Prediction
• Accurate medium-range global weather forecasting with 3D neural networks
• From Noise to Precision: A Diffusion-Driven Approach to Zero-Inflated Precipitation Prediction
• IndiaWeatherBench: A Dataset and Benchmark for Data-Driven Regional Weather Forecasting over India
• RadarQA: Multi-modal Quality Analysis of Weather Radar Forecasts
• Observation-guided Interpolation Using Graph Neural Networks for High-Resolution Nowcasting in Switzerland
• MeteorPred: A Meteorological Multimodal Large Model and Dataset for Severe Weather Event Prediction
• Minimal Convolutional RNNs Accelerate Spatiotemporal Learning
• Physics-constrained generative machine learning-based high-resolution downscaling of Greenland's surface mass balance and surface temperature
• Wasserstein GAN-Based Precipitation Downscaling with Optimal Transport for Enhancing Perceptual Realism
• Do Echo Top Heights Improve Deep Learning Nowcasts?
• Generative Modeling of Weights: Generalization or Memorization?
• A Comprehensive Survey on Deep Learning Solutions for 3D Flood Mapping
• PeakWeather: MeteoSwiss Weather Station Measurements for Spatiotemporal Deep Learning
• End-to-End Probabilistic Framework for Learning with Hard Constraints
• Beyond Attention: Learning Spatio-Temporal Dynamics with Emergent Interpretable Topologies
• Black Hole Weather Forecasting with Deep Learning: A Pilot Study
• QWRF-Net: A Quantum-Wavelet Framework with Rectified Flow for Short-Term Precipitation Nowcasting
• From Proxies to Fields: Spatiotemporal Reconstruction of Global Radiation from Sparse Sensor Sequences
• Skilful nowcasting of extreme precipitation with NowcastNet
• How to use score-based diffusion in earth system science: A satellite nowcasting example
• Predicting Beyond Training Data via Extrapolation versus Translocation: AI Weather Models and Dubai's Unprecedented 2024 Rainfall
• RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours
• Physically constrained generative adversarial networks for improving precipitation fields from Earth system models
• Structural Causal Discovery and Predictive Sufficiency in High-Dimensional Dynamical Systems
• LLMs can construct powerful representations and streamline sample-efficient supervised learning
• Axial-UNet: A Neural Weather Model for Precipitation Nowcasting
• Sensoformer: Robust Sim-to-Real Inference on Variable-Geometry Sensor Sets via Physics-Structured Randomization
• AI-deploying organizations are key to addressing ‘perfect storm’ of AI risks
• A Survey of Large Language Model-Powered Spatial Intelligence Across Scales: Advances in Embodied Agents, Smart Cities, and Earth Science
• Physical Scales Matter: The Role of Receptive Fields and Advection in Satellite-Based Thunderstorm Nowcasting with Convolutional Neural Networks
• Temporal Dynamic Embedding for Irregularly Sampled Time Series

Related

• Machine Learning for Precipitation Nowcasting from Radar Images
• RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours
• A review of radar-based nowcasting of precipitation and applicable machine learning techniques
• An Operational Deep Learning System for Satellite-Based High-Resolution Global Nowcasting
• Code for "GPTCast: a weather language model for precipitation nowcasting"