SciGroveBeta
Climate

TSSM: Triaxial State Space Model for Global Station Weather Forecasting with Temporal-Variable-Historical Modeling

Songru Yang, Zili Liu, Tao Han, Ben Fei, Fenghua Ling, Lei Bai, Chang Liu, Xiangyang Ji, Zhenwei Shi, Zhengxia Zou

Featured July 21, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

Predicting local weather is hard because it's chaotic, but this paper's new model looks at the same time of year from many past years to find hidden patterns, making forecasts much better, especially for big storms.

In depth
The paper introduces the Triaxial State Space Model (TSSM), a novel approach to Global Station Weather Forecasting that moves beyond short-term observations. It achieves this by reorganizing weather data into a temporal-variable-historical 3D tensor, allowing the model to learn from period-aligned historical patterns across years. This design significantly improves accuracy, especially for extreme events and long-horizon predictions, by leveraging long-term climatological context.

Key Takeaways

  • 1
    The paper proposes a triaxial data reorganization strategy, stacking multivariate weather series from previous years at the same month/day/hour into a tensor, which provides intrinsic historical context without extra inputs.
  • 2
    The Triaxial State Space Model (TSSM) employs dedicated T-Scan, V-Scan, and H-Scan modules to efficiently capture short-term temporal dynamics, inter-variable correlations, and long-term historical evolution, respectively.
  • 3
    TSSM demonstrates superior performance in long-horizon and iterative forecasting, significantly reducing error accumulation and exhibiting strong robustness to missing observations by leveraging historical support.

Conceptual Flow

HIGH LEVEL
1
Methodology: Learning from Past Years

Instead of just looking at recent days, the model looks at the same day and hour from many past years to find repeating patterns.

Today's Weather
Past Years' Weather
Find Matching Times
Combined Past & Present
2
Results: Better Storm and Long-Term Forecasts

By using past years' patterns, the model can predict normal weather more accurately and is much better at spotting big storms and forecasting far into the future.

Old Forecasts
New Forecasts
Compare Accuracy
Much Better Predictions
Spot More Storms