Nipping the Butterfly Effect in the Bud: Self-Output Fine-Tuning for Autoregressive Weather Prediction

2026-07-23

AI Summary

Long-term autoregressive weather forecasting suffers from rapid error accumulation due to input distribution shifts induced by initial-step predictions, severely limiting accuracy. This work identifies, for the first time, that the root cause lies in distributional mismatch during the early inference phase. To address this, we propose a plug-and-play Self-supervised Output Fine-Tuning (SOFT) strategy that calibrates the input distribution at the earliest stage via a self-supervised one-step prediction, thereby suppressing error propagation at its source. SOFT seamlessly integrates distribution alignment with autoregressive fine-tuning without modifying the model architecture. Experiments across multiple tasks demonstrate that SOFT significantly reduces both long-term prediction errors and distributional discrepancies, achieving state-of-the-art performance and validating the effectiveness of rethinking the foundational pipeline of deep learning–based weather forecasting.

Abstract

Long-horizon weather forecasting is a fundamental challenge in atmospheric science, for which autoregressive Deep Learning Weather Prediction (DLWP) has emerged as the primary paradigm. Although the autoregressive pipeline is highly scalable and flexible, its prediction errors grow rapidly over long forecasting horizons. In this work, we study this error growth phenomenon from both theoretical and empirical perspectives. Our analysis reveals that the growth is driven by a feedback loop between output errors and input distribution shifts. Specifically, the autoregressive process amplifies small initial output errors, which progressively corrupt subsequent input distributions, echoing the butterfly effect in atmospheric science and ultimately deteriorating forecasting accuracy over longer horizons. Furthermore, we show that this distributional shift originates at the earliest stage of inference, with out-of-distribution signatures detectable as early as the first autoregressive step. To mitigate this issue, we propose \textbf{Self-Output Fine-Tuning (SOFT)}, a plug-and-play strategy that leverages the model's own one-step predictions to calibrate the biased input distribution encountered at the first step. Extensive experiments demonstrate that, despite its simplicity, SOFT achieves state-of-the-art performance on long-horizon forecasting tasks and substantially reduces both prediction errors and distributional discrepancy. The success of SOFT highlights the importance of reexamining the fundamental pipeline of deep learning weather prediction, representing a critical pipeline advance for atmospheric science.

Problem

• autoregressive weather prediction
• error growth
• distribution shift
• butterfly effect
• long-horizon forecasting

Innovation

• Self-Output Fine-Tuning
• autoregressive weather prediction
• distribution shift
• error amplification
• long-horizon forecasting