TyphoFormer: Language-Augmented Transformer for Accurate Typhoon Track Forecasting

TL;DR

TyphoFormer tackles the challenge of typhoon track forecasting by injecting high-level meteorological semantics from LLM-generated textual prompts into a Transformer-based model. Through a Prompt-aware Gating Fusion mechanism, the framework fuses language-derived context with numerical trajectory data, feeding a unified encoder–decoder pipeline that predicts future coordinates. Experiments on the HURDAT2 dataset demonstrate consistent improvements over traditional baselines and state-of-the-art time-series models, particularly for long-range and complex tracks, with a MILTON case study illustrating enhanced spatial fidelity. The approach suggests a scalable path toward more interpretable and context-aware meteorological forecasting, leveraging language models to supply domain prior knowledge.

Abstract

Accurate typhoon track forecasting is crucial for early system warning and disaster response. While Transformer-based models have demonstrated strong performance in modeling the temporal dynamics of dense trajectories of humans and vehicles in smart cities, they usually lack access to broader contextual knowledge that enhances the forecasting reliability of sparse meteorological trajectories, such as typhoon tracks. To address this challenge, we propose TyphoFormer, a novel framework that incorporates natural language descriptions as auxiliary prompts to improve typhoon trajectory forecasting. For each time step, we use Large Language Model (LLM) to generate concise textual descriptions based on the numerical attributes recorded in the North Atlantic hurricane database. The language descriptions capture high-level meteorological semantics and are embedded as auxiliary special tokens prepended to the numerical time series input. By integrating both textual and sequential information within a unified Transformer encoder, TyphoFormer enables the model to leverage contextual cues that are otherwise inaccessible through numerical features alone. Extensive experiments are conducted on HURDAT2 benchmark, results show that TyphoFormer consistently outperforms other state-of-the-art baseline methods, particularly under challenging scenarios involving nonlinear path shifts and limited historical observations.

Figures (2)

• Figure 1: The overall model architecture of TyphoFormer.
• Figure 2: Trajectory forecasting performance among all the methods on hurricane MILTON 2024.