Adaptive ensemble method improves global weather forecast accuracy
TaiJi: State-Conditional Adaptive Combination of Machine-Learning and Physics-Based Global Weather Forecasts
Computational Engineering, Finance, and Science
Summary
Weather forecasts often come from different computer models that each have strengths and weaknesses depending on the weather element or location. The authors present TaiJi, a system that smartly combines these models by learning when and where to trust each one, based on recent weather patterns. This combination improves the accuracy of forecasts for up to 10 days, including during extreme weather events. TaiJi uses much less computing power than training big individual models and builds on existing weather prediction systems instead of replacing them.
What this means in practice
- •For weather forecasting centers: Combine existing machine-learning and physics-based weather models to improve forecast accuracy without high computational costs.
- •For disaster management teams: Obtain more accurate predictions of extreme weather events like tropical cyclones by using TaiJi’s adaptive combination of forecasts.
Authors
Jiale Wang, Yining Li, Guihua Wang
Abstract
Data-driven machine-learning weather models now rival, and in some respects surpass, operational numerical weather prediction (NWP), yet no single model dominates across variables, pressure levels, lead times, or regions, and the marginal returns from developing ever-larger individual models are diminishing. We present TaiJi (after the Chinese concept of yin-yang harmony, the complementarity of opposites), a spatiotemporal adaptive ensemble framework that recasts optimal model-weight combination as a learned prediction problem: a lightweight convolutional neural network, conditioned on the recent atmospheric state, predicts at every grid point and lead time a set of affine combination weights and an additive residual for the constituent forecasts. The combiner is trained end-to-end over Pangu-Weather, GraphCast, FuXi, IFS-HRES, and the IFS-ENS ensemble mean, an explicitly hybrid set of ML and NWP constituents, on two consumer-grade RTX 4090 GPUs in about 8 hours, at least an order of magnitude below the training cost of any of its ML constituents. Under the WeatherBench 2 protocol (2020 test year), TaiJi outperforms the strongest baseline for each variable and lead time across the eight core variables and lead times of up to 10 days, under both RMSE and ACC, with a near-100% win rate, and this advantage is preserved across extreme scenarios, including tropical-cyclone tracks and heavy precipitation. We further show that the learned weights vary coherently with latitude, season and lead time rather than following a single fixed rule. TaiJi thus offers a computationally inexpensive route to improving global forecast skill that builds on, rather than replaces, the substantial investment already made in existing forecast models.