LaDCast: A Latent Diffusion Model for Medium-Range Ensemble Weather Forecasting

2025-06-10

AI Summary

High-accuracy probabilistic forecasting and efficient uncertainty quantification remain mutually exclusive challenges in mesoscale ensemble weather prediction. Method: We propose the first latent-space diffusion model tailored for global mesoscale forecasting. It introduces latent-variable diffusion modeling to this domain; incorporates GeoRoPE—a spherical geometry-aware positional encoding scheme; employs a dual-stream attention mechanism for efficient conditional modeling; and integrates a VAE-based latent-space compression with a Transformer-based diffusion framework and sinusoidal time embeddings. Results: The model achieves probabilistic forecast skill comparable to ECMWF’s IFS-ENS, with superior accuracy in tracking extreme-weather phenomena such as cyclones. Crucially, it attains equivalent uncertainty representation without explicit initial-condition perturbations. Computational and memory costs are reduced by one to two orders of magnitude, enabling real-time kilometer-scale hourly forecasting.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Accurate probabilistic weather forecasting demands both high accuracy and efficient uncertainty quantification, challenges that overburden both ensemble numerical weather prediction (NWP) and recent machine-learning methods. We introduce LaDCast, the first global latent-diffusion framework for medium-range ensemble forecasting, which generates hourly ensemble forecasts entirely in a learned latent space. An autoencoder compresses high-dimensional ERA5 reanalysis fields into a compact representation, and a transformer-based diffusion model produces sequential latent updates with arbitrary hour initialization. The model incorporates Geometric Rotary Position Embedding (GeoRoPE) to account for the Earth's spherical geometry, a dual-stream attention mechanism for efficient conditioning, and sinusoidal temporal embeddings to capture seasonal patterns. LaDCast achieves deterministic and probabilistic skill close to that of the European Centre for Medium-Range Forecast IFS-ENS, without any explicit perturbations. Notably, LaDCast demonstrates superior performance in tracking rare extreme events such as cyclones, capturing their trajectories more accurately than established models. By operating in latent space, LaDCast reduces storage and compute by orders of magnitude, demonstrating a practical path toward forecasting at kilometer-scale resolution in real time. We open-source our code and models and provide the training and evaluation pipelines at: https://github.com/tonyzyl/ladcast.

Problem

• Develops LaDCast for medium-range ensemble weather forecasting
• Improves accuracy and uncertainty quantification in weather predictions
• Reduces computational costs by operating in latent space

Innovation

• Latent diffusion model for ensemble weather forecasting
• Autoencoder compresses high-dimensional weather data
• Transformer-based diffusion with Earth's geometry embedding