In the artificial intelligent (AI) era, pure data-driven weather and climate models are gradually catching up with, and even surpassing, traditional numerical models. However, significant challenges remain in current deep learning models, such as low physical consistency and suboptimal prediction of divergent winds. These limitations hinder the predictive capabilities for complex weather and climate phenomena, including precipitation. A promising approach to address these challenges is to combine physics, atmospheric dynamics, and deep learning models.
A research team led by Prof. HUANG Gang from the Institute of Atmospheric Physics (IAP) of the Chinese Academy of Sciences has made strides in improving precipitation forecasting using a novel approach.
Using
EarthLab-a new Earth System Science Numerical Simulator Facility developed by IAP, the team employed data and computational power to improve the precipitation forecasting capabilities of numerical models.
They focused on coupling physical variables through graph neural networks (GNNs) to introduce physical constraints and improve the accuracy of precipitation forecasts.
To address the difficulties in precipitation forecasting, especially for heavy rainfall events, the team first investigated the drivers and mechanisms behind precipitation. They used the omega equation and water vapor equations for variable selection and constructed a variable coupling graph.
The omega and water vapor equations describe vertical motion and water vapor changes, both of which are critical factors influencing precipitation. The graph network abstracted these equations into a network structure that represents the nonlinear combinations of fundamental physical quantities and the relationships between key precipitation factors.
Considering the impact of climate factors on weather scales, especially systematic differences in model errors under different climate backgrounds, the study incorporated sparse data, such as seasonality, El Niño Southern Oscillation, and initialization time, using entity embedding techniques to calibrate the model.
In addition, they localized the ChebNet graph neural network for precipitation, maintaining its effectiveness while significantly reducing computational complexity by avoiding global operations.
The comparison results of the proposed models, omega-GNN and omega-EGNN, with numerical models showed a significant improvement in precipitation prediction capabilities across various categories. The performance of these models surpassed that of established physically unconstrained deep learning models, such as
U-NET and
3D-CNN.
Furthermore, ensemble forecasting, achieved through ten perturbations of all deep learning models, demonstrated the superior consistency and forecasting ability of the physics-constrained omega-GNN and omega-EGNN models, especially for heavy rainfall events.
"We have accumulated substantial expertise in climate dynamics, and in recent years, we have been exploring ways to improve weather and climate prediction using AI, and have won awards in related competitions. In the AI era, the integration of physics is a major challenge with various approaches and perspectives. Our team, drawing on atmospheric and climate dynamics considerations, has experimented with applying soft constraints to models from a physical coupling perspective, aiming to contribute incremental information to relevant fields," said Prof. HUANG Gang, corresponding author of the study.
A general diagram of the omega-GNN model (Image by IAP)
