Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WSSM: Geographic-enhanced hierarchical state-space model for global station weather forecast

Songru Yang, Zili Liu, Zhenwei Shi, Zhengxia Zou

TL;DR

The paper tackles the challenge of global station weather forecasting with a focus on extreme-event prediction. It introduces Weather State-space Model (WSSM), a novel Mamba-based framework that embeds geographical context and multiscale time-frequency representations within a state-space paradigm. The authors propose three innovations: Geographical encoding to incorporate acquisition time and sensor location, a Hierarchical Bi-Mamba encoder to fuse multi-scale sequences, and a Time-frequency Bi-Mamba block to capture low- and high-frequency dynamics. Experiments on the Weather-5K dataset show that WSSM achieves state-of-the-art overall accuracy and substantially improves extreme-weather forecasting, outperforming Transformer-based methods in SEDI while remaining efficient through continuous-time modeling. This work paves the way for more accurate and scalable global station weather forecasting by integrating domain-specific spatio-temporal structure into a Mamba-based state-space model.

Abstract

Global Station Weather Forecasting (GSWF), a prominent meteorological research area, is pivotal in providing timely localized weather predictions. Despite the progress existing models have made in the overall accuracy of the GSWF, executing high-precision extreme event prediction still presents a substantial challenge. The recent emergence of state-space models, with their ability to efficiently capture continuous-time dynamics and latent states, offer potential solutions. However, early investigations indicated that Mamba underperforms in the context of GSWF, suggesting further adaptation and optimization. To tackle this problem, in this paper, we introduce Weather State-space Model (WSSM), a novel Mamba-based approach tailored for GSWF. Geographical knowledge is integrated in addition to the widely-used positional encoding to represent the absolute special-temporal position. The multi-scale time-frequency features are synthesized from coarse to fine to model the seasonal to extreme weather dynamic. Our method effectively improves the overall prediction accuracy and addresses the challenge of forecasting extreme weather events. The state-of-the-art results obtained on the Weather-5K subset underscore the efficacy of the WSSM

WSSM: Geographic-enhanced hierarchical state-space model for global station weather forecast

TL;DR

The paper tackles the challenge of global station weather forecasting with a focus on extreme-event prediction. It introduces Weather State-space Model (WSSM), a novel Mamba-based framework that embeds geographical context and multiscale time-frequency representations within a state-space paradigm. The authors propose three innovations: Geographical encoding to incorporate acquisition time and sensor location, a Hierarchical Bi-Mamba encoder to fuse multi-scale sequences, and a Time-frequency Bi-Mamba block to capture low- and high-frequency dynamics. Experiments on the Weather-5K dataset show that WSSM achieves state-of-the-art overall accuracy and substantially improves extreme-weather forecasting, outperforming Transformer-based methods in SEDI while remaining efficient through continuous-time modeling. This work paves the way for more accurate and scalable global station weather forecasting by integrating domain-specific spatio-temporal structure into a Mamba-based state-space model.

Abstract

Global Station Weather Forecasting (GSWF), a prominent meteorological research area, is pivotal in providing timely localized weather predictions. Despite the progress existing models have made in the overall accuracy of the GSWF, executing high-precision extreme event prediction still presents a substantial challenge. The recent emergence of state-space models, with their ability to efficiently capture continuous-time dynamics and latent states, offer potential solutions. However, early investigations indicated that Mamba underperforms in the context of GSWF, suggesting further adaptation and optimization. To tackle this problem, in this paper, we introduce Weather State-space Model (WSSM), a novel Mamba-based approach tailored for GSWF. Geographical knowledge is integrated in addition to the widely-used positional encoding to represent the absolute special-temporal position. The multi-scale time-frequency features are synthesized from coarse to fine to model the seasonal to extreme weather dynamic. Our method effectively improves the overall prediction accuracy and addresses the challenge of forecasting extreme weather events. The state-of-the-art results obtained on the Weather-5K subset underscore the efficacy of the WSSM
Paper Structure (12 sections, 3 equations, 1 figure, 2 tables)

This paper contains 12 sections, 3 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Overview
  4. Geographical encoding
  5. Hierarchical Bi-Mamba encoder
  6. Time-frequency Bi-Mamba block
  7. EXPERIMENTS
  8. Experimental Dataset and Settings
  9. Evaluation Protocol and Metrics
  10. Implementation Details
  11. Comparison with the State-of-the-Art
  12. CONCLUSION

Figures (1)

  • Figure 1: The overall framework of WSSM. (a) The Geographical Encoding and Hierarchical Patching process integrate geographical information into weather sequences with different time resolutions. (b) The Hierarchical Bi-Mamba Encoder uses a shared encoder to fuse hierarchical meteorological sequences. (c) The Time-frequency Bi-Mamba Block based on bidirectional Mamba adds a frequency branch and a multivariable branch to simultaneously encode the time-frequency-multivariable features.