Learning Coupled Earth System Dynamics with GraphDOP

2025-10-23

AI Summary

Explicit modeling of coupled dynamics across Earth’s multi-sphere system—namely the atmosphere, ocean, land surface, and cryosphere—faces bottlenecks including high computational cost and strong reliance on parameterizations. To address this, we propose GraphDOP: the first end-to-end data-driven graph neural network framework that implicitly learns cross-sphere coupling dynamics from heterogeneous satellite and in-situ observations. It constructs a unified graph structure over multi-sphere observational data and embeds them into a shared latent space, obviating explicit coupling interfaces required by traditional numerical models. This paradigm shift bypasses conventional multi-model coupling architectures and substantially enhances representation capacity for nonlinear, cross-domain interactions. Experiments demonstrate that GraphDOP successfully reproduces key coupled events—including Arctic sea-ice rapid freeze-up, hurricane Ian–induced oceanic cold wakes, and the 2022 European extreme heatwave—validating its effectiveness and generalizability in complex Earth system dynamics simulation.

Technology Category

Machine Learning: Graph-based Machine LearningComputer Vision: Diffusion Models for VisionReasoning under Uncertainty: Graphical Models

Application Category

Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphsSystems and Infrastructure for Web, Mobile and WoT: Web applications in cross-disciplinary domains and verticals such as mixed reality, smart cities, and digital healthSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Interactions between different components of the Earth System (e.g. ocean, atmosphere, land and cryosphere) are a crucial driver of global weather patterns. Modern Numerical Weather Prediction (NWP) systems typically run separate models of the different components, explicitly coupled across their interfaces to additionally model exchanges between the different components. Accurately representing these coupled interactions remains a major scientific and technical challenge of weather forecasting. GraphDOP is a graph-based machine learning model that learns to forecast weather directly from raw satellite and in-situ observations, without reliance on reanalysis products or traditional physics-based NWP models. GraphDOP simultaneously embeds information from diverse observation sources spanning the full Earth system into a shared latent space. This enables predictions that implicitly capture cross-domain interactions in a single model without the need for any explicit coupling. Here we present a selection of case studies which illustrate the capability of GraphDOP to forecast events where coupled processes play a particularly key role. These include rapid sea-ice freezing in the Arctic, mixing-induced ocean surface cooling during Hurricane Ian and the severe European heat wave of 2022. The results suggest that learning directly from Earth System observations can successfully characterise and propagate cross-component interactions, offering a promising path towards physically consistent end-to-end data-driven Earth System prediction with a single model.

Problem

• Modeling coupled Earth system interactions across ocean, atmosphere, land, and cryosphere
• Overcoming explicit coupling challenges in traditional weather prediction systems
• Capturing cross-domain interactions directly from observations without physics-based models

Innovation

• GraphDOP learns weather forecasting from raw observations
• It embeds diverse Earth system data into shared latent space
• Model implicitly captures cross-domain interactions without explicit coupling