FuXi-RTM: A Physics-Guided Prediction Framework with Radiative Transfer Modeling

2025-03-25

AI Summary

Current deep learning–based weather forecasting models lack physical constraints, leading to inaccurate radiation process modeling and high computational costs—thus limiting operational reliability. To address this, we propose the first hybrid forecasting framework that explicitly embeds physical processes: a differentiable deep learning radiative transfer model (DLRTM) is integrated as a fixed surrogate module into the FuXi meteorological foundation model’s backbone network. The framework employs end-to-end joint training with physics-informed loss constraints to simultaneously optimize physical consistency and forecast accuracy. Evaluated on a 5-year dataset across 3,320 variable–lead-time combinations, the method achieves statistically significant accuracy improvements in 88.51% of scenarios, with breakthrough performance gains in shortwave and longwave radiative flux prediction.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Similar to conventional video generation, current deep learning-based weather prediction frameworks often lack explicit physical constraints, leading to unphysical outputs that limit their reliability for operational forecasting. Among various physical processes requiring proper representation, radiation plays a fundamental role as it drives Earth's weather and climate systems. However, accurate simulation of radiative transfer processes remains challenging for traditional numerical weather prediction (NWP) models due to their inherent complexity and high computational costs. Here, we propose FuXi-RTM, a hybrid physics-guided deep learning framework designed to enhance weather forecast accuracy while enforcing physical consistency. FuXi-RTM integrates a primary forecasting model (FuXi) with a fixed deep learning-based radiative transfer model (DLRTM) surrogate that efficiently replaces conventional radiation parameterization schemes. This represents the first deep learning-based weather forecasting framework to explicitly incorporate physical process modeling. Evaluated over a comprehensive 5-year dataset, FuXi-RTM outperforms its unconstrained counterpart in 88.51% of 3320 variable and lead time combinations, with improvements in radiative flux predictions. By incorporating additional physical processes, FuXi-RTM paves the way for next-generation weather forecasting systems that are both accurate and physically consistent.

Problem

• Lack of physical constraints in deep learning weather prediction
• Challenges in accurate radiative transfer simulation
• Need for physically consistent and accurate weather forecasting

Innovation

• Hybrid physics-guided deep learning framework
• Integrates forecasting model with radiative surrogate
• Enhances accuracy with physical consistency