PnP-Corrector: A Universal Correction Framework for Coupled Spatiotemporal Forecasting

TL;DR

The paper introduces PnP-Corrector, a universal correction framework that improves long-term coupled spatiotemporal forecasts by decoupling physical simulation from error correction, significantly reducing prediction errors.

Contribution

It proposes a novel plug-and-play correction framework with a dedicated correction agent and an efficient backbone model, enhancing stability and accuracy in coupled forecasting systems.

Findings

Reduces prediction error by 29% in 300-day global ocean-atmosphere forecasts.

Significantly improves long-term stability and accuracy of coupled models.

Outperforms state-of-the-art models on key metrics.

Abstract

Coupled spatiotemporal forecasting is important for predicting the future evolution of multiple interacting dynamical systems, such as in climate models. However, existing methods are severely constrained by the persistent bottleneck of compounding errors. In coupled systems, errors from each subsystem simulator propagate and amplify one another, a phenomenon we term Reciprocal Error Amplification, leading to a rapid collapse of long-range predictions. To address this challenge, we propose a universal framework called PnP-Corrector (Plug-and-Play Corrector). The core idea of our framework is to decouple the physical simulation from the error correction process: it freezes pre-trained physics simulation engines and exclusively trains a correction agent to proactively counteract the systematic biases emerging from the coupled system. Furthermore, we design an efficient predictive model architecture, DSLCast, to serve as the backbone of this framework. Extensive experiments demonstrate that our method significantly enhances the long-term stability and accuracy of coupled forecasting systems. For instance, in the challenging task of a 300-day global ocean-atmosphere coupled forecast, our PnP-Corrector framework reduces the prediction error of the baseline model by 29% and surpasses state-of-the-art models on several key metrics.