.png){kind=icon}
.png){kind=icon}
Return
ST-TFFM-Observer: A Global Ionospheric TEC Prediction Model Based on Dynamic System Theory With Enhanced Time-Frequency Dependence Modeling
.png){kind=icon}
OA
AI
L
H
H
J
S
X
G
DOI：10.1029/2025SW004876.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
ST-Observer is a novel proposed spatiotemporal prediction model based on dynamic system theory, which has been proven to be superior to existing models in multiple spatiotemporal prediction tasks. However, its structural design results in weak time dependent modeling capabilities and a lack of frequency dependent modeling capabilities. This article innovatively designs a time-frequency fusion module TFFM and embeds it into ST-Observer to compensate for its shortcomings in time dependent modeling and frequency dependent modeling. Based on our improvements, we have designed a new TEC spatiotemporal prediction model ST-TFFM-Observer. The ablation experiment showed that the proposed TFFM module can reduce the RMSE $\text{RMSE}$ , MAPE $\text{MAPE}$ and n RMSE $n\text{RMSE}$ of the model by 10.89%, 17.46%, and 10.90% in low solar activity year, 5.62%, 12.02%, and 5.67% in high solar activity year. Then we compared the proposed ST-TFFM-Observer with 6 spatiotemporal prediction models and a product C1PG widely used in TEC prediction from multiple perspectives. All the comparison results have confirmed the superiority of the model proposed in this paper. The TFFM proposed in this article improves the predictive performance of TEC. The idea of time-frequency dependent modeling can also provide reference for deep learning modeling in other fields.
Journal
S
IF:
0
Papers:
115
Citations:
0
Organization
Cited Papers