DiffScale: Continuous Downscaling and Bias Correction of Subseasonal Wind Speed Forecasts using Diffusion Models

2025-03-31

AI Summary

This study addresses the dual challenges of low spatial resolution and substantial systematic biases in subseasonal-to-seasonal (S2S) wind speed forecasting. We propose an end-to-end, continuous-space super-resolution and bias correction framework based on a classifier-free guided diffusion model. Our method introduces, for the first time, a general-purpose super-resolution architecture supporting arbitrary scaling factors—enabling zero-shot adaptation to diverse target resolutions and forecast lead times without retraining. By integrating meteorological priors into a physics-informed conditional generation process—and eliminating autoregressive modeling—it ensures spatiotemporal consistency. Evaluated on downscaling ECMWF S2S forecasts to high-resolution ERA5 reference fields, our approach significantly improves Week-3 wind speed prediction accuracy, outperforming all existing baseline models across key metrics. This work establishes a new paradigm for S2S renewable energy forecasting: high-fidelity, multi-scale, and strongly generalizable.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Semi-Supervised Learning
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Renewable resources are strongly dependent on local and large-scale weather situations. Skillful subseasonal to seasonal (S2S) forecasts -- beyond two weeks and up to two months -- can offer significant socioeconomic advantages to the energy sector. This study aims to enhance wind speed predictions using a diffusion model with classifier-free guidance to downscale S2S forecasts of surface wind speed. We propose DiffScale, a diffusion model that super-resolves spatial information for continuous downscaling factors and lead times. Leveraging weather priors as guidance for the generative process of diffusion models, we adopt the perspective of conditional probabilities on sampling super-resolved S2S forecasts. We aim to directly estimate the density associated with the target S2S forecasts at different spatial resolutions and lead times without auto-regression or sequence prediction, resulting in an efficient and flexible model. Synthetic experiments were designed to super-resolve wind speed S2S forecasts from the European Center for Medium-Range Weather Forecast (ECMWF) from a coarse resolution to a finer resolution of ERA5 reanalysis data, which serves as a high-resolution target. The innovative aspect of DiffScale lies in its flexibility to downscale arbitrary scaling factors, enabling it to generalize across various grid resolutions and lead times -without retraining the model- while correcting model errors, making it a versatile tool for improving S2S wind speed forecasts. We achieve a significant improvement in prediction quality, outperforming baselines up to week 3.

Problem

• Enhance wind speed predictions using diffusion models
• Downscale subseasonal forecasts to finer spatial resolutions
• Correct model errors without retraining for flexibility

Innovation

• Diffusion model for continuous downscaling wind forecasts
• Classifier-free guidance enhances spatial super-resolution
• Flexible downscaling without retraining across resolutions