SRNDiff: Short-term Rainfall Nowcasting with Condition Diffusion Model

2024-02-21

arXiv.org

AI Summary

To address the insufficient physical plausibility and structural distortion of generated samples in 0–2-hour nowcasting precipitation prediction, this paper proposes the first end-to-end radar-based precipitation forecasting method built upon a conditional diffusion model. The method introduces a multi-scale independent U-Net conditional encoder coupled with a synchronized conditional decoder, enabling fine-grained and stable denoising generation driven jointly by historical radar reflectivity and rain gauge observations. Compared to state-of-the-art GAN-based approaches, our method achieves significant improvements in generation quality (measured by PSNR and SSIM), precipitation structure fidelity, and physical consistency, while exhibiting greater training stability and efficiency. Extensive experiments demonstrate that the framework substantially enhances the modeling capability for short-term precipitation spatial–temporal distributions, establishing a novel paradigm for generative modeling in meteorological forecasting.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Diffusion models are widely used in image generation because they can generate high-quality and realistic samples. This is in contrast to generative adversarial networks (GANs) and variational autoencoders (VAEs), which have some limitations in terms of image quality.We introduce the diffusion model to the precipitation forecasting task and propose a short-term precipitation nowcasting with condition diffusion model based on historical observational data, which is referred to as SRNDiff. By incorporating an additional conditional decoder module in the denoising process, SRNDiff achieves end-to-end conditional rainfall prediction. SRNDiff is composed of two networks: a denoising network and a conditional Encoder network. The conditional network is composed of multiple independent UNet networks. These networks extract conditional feature maps at different resolutions, providing accurate conditional information that guides the diffusion model for conditional generation.SRNDiff surpasses GANs in terms of prediction accuracy, although it requires more computational resources.The SRNDiff model exhibits higher stability and efficiency during training than GANs-based approaches, and generates high-quality precipitation distribution samples that better reflect future actual precipitation conditions. This fully validates the advantages and potential of diffusion models in precipitation forecasting, providing new insights for enhancing rainfall prediction.

Problem

• Develops conditional diffusion model for rainfall nowcasting
• Improves prediction accuracy over GANs-based approaches
• Generates high-quality precipitation distribution samples

Innovation

• Conditional diffusion model for rainfall nowcasting
• Additional conditional decoder in denoising process
• Multiple independent UNet networks extract features