On the Integration of Spatial-Temporal Knowledge: A Lightweight Approach to Atmospheric Time Series Forecasting

TL;DR

This paper introduces STELLA, a lightweight atmospheric time series forecasting model that leverages spatial-temporal position embedding to effectively capture correlations without complex Transformer architectures, achieving superior results with minimal parameters.

Contribution

The paper presents STELLA, a novel lightweight ATSF model that uses spatial-temporal position embedding and MLPs, reducing complexity while maintaining high accuracy.

Findings

STELLA outperforms advanced methods on five datasets.

It requires only 10k parameters and one hour of training.

Spatial-temporal position embedding effectively models correlations without attention.

Abstract

Transformers have gained attention in atmospheric time series forecasting (ATSF) for their ability to capture global spatial-temporal correlations. However, their complex architectures lead to excessive parameter counts and extended training times, limiting their scalability to large-scale forecasting. In this paper, we revisit ATSF from a theoretical perspective of atmospheric dynamics and uncover a key insight: spatial-temporal position embedding (STPE) can inherently model spatial-temporal correlations even without attention mechanisms. Its effectiveness arises from the integration of geographical coordinates and temporal features, which are intrinsically linked to atmospheric dynamics. Based on this, we propose STELLA, a Spatial-Temporal knowledge Embedded Lightweight modeL for ASTF, utilizing only STPE and an MLP architecture in place of Transformer layers. With 10k parameters and one hour of training, STELLA achieves superior performance on five datasets compared to other advanced methods. The paper emphasizes the effectiveness of spatial-temporal knowledge integration over complex architectures, providing novel insights for ATSF. The code is available at https://github.com/GestaltCogTeam/STELLA.